I have opened this bench as a result of listener196 putting my paper's title "Bucking Coils produce Energy Gain" on the shoutbox. Member polyrhythm then put a link to scribd.com where you can view this paper. I don't know how this paper got into scibd.com but looking at it reminded me of things I did years ago and have forgotten. In fact it led me to discover a folder on a memory stick that I used to get data from an old computer I had for 20 years when I updated to a new one. The folder name is "Bucking Coils" and I had completely forgotten of its existence. It goes back to 2012 where I initially wrote a paper looking at Osamu Ide's OU converter. There was a member here, called EMJunkie if I remember correctly, who was very much into bucking coils. After some disagreements he went off and set up his own forum. It seems his bench has been deleted and I may have posted stuff there.
The thing of interest here is that a pair of identical coils that are physically separated on a transformer core (like small coils at diametrically opposite positions on a ring core) can be connected in series aiding or series opposing (bucking) to produce a total inductance value that is either four times the individual inductance value or close to zero if there is zero magnetic propagation time between the two coils. That statement in italics is important because we know that there is a finite time for magnetic signals to travel along the core. Transformer theory generally ignores this. When you do the math for sinewave signals you find that the finite propagation time not only affects the inductance value of the series connected coils but also introduces a resistance value. This resistance value is independent of the actual resistance of the coils, it is generated by the presence of the magnetic delay. For aiding coils the resistance is positive, and for bucking coils the resistance is negative. The math demonstrating this is quite trivial, yet in all my years of studying EM theory I have never come across this.
Negative resistance is a source of energy. The fact that a passive device can produce energy is a huge thing to contemplate and contemporary science does not recognize its existence. I attach a file that I wrote in 2017 that discusses this and suggests experiments to investigate it.
This paper also touches on another possibility for an overunity heater using the Curie point. We wind a coil using resistance wire to heat the core to its Curie point using pulses of current. A fan carries heat away so the core temperature cycles above and below that temperature. With the right circuitry we gain energy. It would require an array of small cores.
Smudge
Edit. Attachment deleted because it has become corrupted.
If the coil were made of wire as elastic as rubber and conduct as copper. When current flows, it would stretch, increasing in diameter.and it perform mechanical work.
The magnetic energy of the rubber coil will also increase. In this case, we only need to increase the current in the coil with increasing diameter. O0
Quote from: chief kolbacict on 2026.03.27, 13:10:05
If the coil were made of wire as elastic as rubber and conduct as copper.
Why not made out of spring instead of rubber, ...like this ?:
(https://www.overunityresearch.com/index.php?action=dlattach;topic=4525.0;attach=53958)
Quote from: chief kolbacict on 2026.03.27, 13:10:05
If the coil were made of wire as elastic as rubber and conduct as copper. When current flows, it would stretch, increasing in diameter.and it perform mechanical work.
The magnetic energy of the rubber coil will also increase. In this case, we only need to increase the current in the coil with increasing diameter. O0
And that current is feeding a voltage from the coil demanding energy from the current source that exactly matches the mechanical work performed.
Quote from: Smudge on 2026.03.27, 14:29:58
that exactly matches the mechanical work performed.
True. But what if we stretch the plates of a parametric capacitor using mechanical work?
The capacitor increases its energy and will supply current to the coil.
The only problem is the size.
Hi Smudge,
Have you considered running your paper by Copilot AI?
It will provide a vigorous review of the concept and mathematics.
I have revisited a number of past projects using this technique to confirm the negative results I obtained and to see if the AI could suggest any modifications.
Regards L192
Would you say negative resistance is a vacuum and positive resistance is a pressure?
Quote from: listener192 on 2026.03.31, 15:11:19
Hi Smudge,
Have you considered running your paper by Copilot AI?
It will provide a vigorous review of the concept and mathematics.
I have revisited a number of past projects using this technique to confirm the negative results I obtained and to see if the AI could suggest any modifications.
Regards L192
In my limited experience of using AI you get different answers when you phrase your question in different ways. It doesn't find fault in the math predicitng negative resistance but goes on to say that any energy gained from this will obey CoE, resuting in energy being supplied, but doesn't give the math to show that. Of course that may be true, converting parts of the core into delay lines demands RF current to flow in those extra windings that could add an induced positive resistance. A more detailed math analysis might show this but that is now beyond my capability. IMO an experiment would be a quicker way of discovering this, that too is beyond me.
Smudge
Quote from: Smudge on 2026.04.01, 07:45:54
In my limited experience of using AI you get different answers when you phrase your question in different ways.
That has been my experience, too.
Comparing different AI models I found that Claude Opus is least susceptible to leading questions.
Have you had any experience with it ?
Quote from: Verpies on 2026.04.01, 08:01:42
That has been my experience, too.
Comparing different AI models I found that Claude Opus is least susceptible to leading questions.
Have you had any experience with it ?
No. I haven't set out to try AI, it has been imposed on me by my search engines. When I ask Google and Microsoft Edge the same question I get different answers.
Quote from: Smudge on 2026.04.02, 06:46:46
No. I haven't set out to try AI, it has been imposed on me by my search engines.
Would you like to if it did not mean installing any new programs on your computer ?
Quote from: Verpies on 2026.04.02, 07:52:28
Would you like to if it did not mean installing any new programs on your computer ?
I have no desire to compare different AI programs, I prefer to use what's left in my aging brain cells looking into possible OU systems. I am surprised that a simple passive system of ring core, coils and capacitors offers negative resistance when driven at the right frequency. No one has challenged the math, and no one has offered to carry out experiments. :(
What would be the minimum size core to experiment with?
I have Ferroxocube TN36/23/15-3R1 cores.
Quote from: Smudge on 2026.04.08, 06:38:50
No one has challenged the math, and no one has offered to carry out experiments. :(
Perhaps that is because your "
Acessing Hidden Magnetic energy.pdf" attachment is corrupted.
Don't be confused by the content of your own file cache.
I downloaded Smudge's pdf file on March 25 when he attached it to his Reply on that day. The file opened as usually a pdf file should.
Now I downloaded it again but now this newly downloaded file does not open, an error message appears:
Error happened at opening the file. The root object is missing or invalid.
I zipped the pdf file which opens for me since I downloaded it on March 25 and attached below. Could someone check whether it opens correctly after unzipping?
Also, I attached the pdf file I downloaded on March 25. If it gets corrupted after you download it, then this needs some investigation?
Gyula
EDIT I downloaded the pdf file I just uploaded below and it opens correctly. However, the pdf file in Smudge's very first post he had attached it does not open for me if I download. Something damaged it ? and since when ?
The PDF I downloaded the day posted and again yesterday, although I wish I could get hold of a core to try some experiments.
Quote from: unimmortal on 2026.04.09, 22:08:11
The PDF I downloaded the day posted and again yesterday, although I wish I could get hold of a core to try some experiments.
Clear your file cache, redownload and report.
I tried downloading my original attachment and also Gyula's two attachments and they were all corrupted. and failed to open. As a precaution I have removed the original. I now need to look at other attachments to see whether the whole site has been infected or just my file.
Smudge
OK, I have tried various downloads of files from other benches and they all work OK. It seems my pdf attachment on my first post on this bench has been corrupted here in this OUR site. My original in my computer is OK. So here it is again for anyone to see.
Smudge
Smudge, the file is ok, it opens correctly.
Gyula
Thanks Gyula, I did check that a download worked OK. What worries me is who managed to create the original corruption and why. Is it connected to the outage that Verpies just sorted? Had I inadvertently put the cat among the pigeons by posting something that some high powered organisation doesn't want people to see? I am looking for guidance as to whether this bench should be made private before the site reopens to visitors.
The strange thing is that both the pdf file in my zipfile and the separate pdf file I uploaded last night reports damaged file for me now when I download them. Last night they were correctly open when I checked them after the uploading.
The pdf file I downloaded on March 25 from Smudge's 1st post still opens fine when I open them from my hard drive. Also, the pdf in the zipped file and the separate pdf file I uploaded last night work works ok when I open them from my hard drive.
Gyula
@Smudge. About the subject matter of your paper. I find the first arrangement with a ferromagnetic substance transition through the Curie temperature compelling but impractical because thermal transitions are slow. I find difficult to accept the other one because while I accept that the magnetization of a ferromagnetic material has a finite propagation speed, I don't accept that the near H field has the same. Near-field H is not to be conflated with the magnetic field in far-field EM radiation which does have a finite propagation speed, but in this radiation the E and H fields are in-phase while in near-field they are not.
P.S.
The attached paper also distinguishes between these two regimes of magnetic fields.
Quote from: Verpies on 2026.04.10, 13:13:58
@Smudge. About the subject matter of your paper. I find the first arrangement with a ferromagnetic substance transition through the Curie temperature compelling but impractical because thermal transitions are slow. I find difficult to accept the other one because while I accept that the magnetization of a ferromagnetic material has a finite propagation speed, I don't accept that the near H field has the same. Near-field H is not to be conflated with the magnetic field in far-field EM radiation which does have a finite propagation speed, but in this radiation the E and H fields are in-phase while in near-field they are not.
P.S.
The attached paper also distinguishes between these two types of magnetic fields.
I understand E and H fields not being in phase in the near field, I spent much of my life working with near fields for proximity sensors. But with respect I don't see where this applies to the magnetization wave within a length of permeable core material where the E field of interest is external to the material. I would see it applying to a coil fully immersed within permeable material (like a coil being immersed in air but we have replaced the air) where radiation from the coil in all directions through the material would obey near and far field laws. The ring core is not that situation. I would point out that magnetic propagation delay is accepted as producing a loss in the guise of magnetic viscosity; where is the evidence for where that lost energy goes, or do we just assume it goes as heat or some form of radiation? Is it just possible that science has missed a trick here, that it disappears into some other quantum sink (zero-point energy?) and we can change things so as to receive energy from that sink? Surely that possibility needs exploring?
Quote from: Smudge on 2026.04.10, 15:26:27
But with respect I don't see where this applies to the magnetization wave within a length of permeable core material
The speed of H field in vacuum is not the same as the speed of the magnetization wave within a length of permeable core material.
Quote from: Verpies on 2026.04.10, 21:50:47
The speed of H field in vacuum is not the same as the speed of the magnetization wave within a length of permeable core material.
Very true and we have means for reducing the speed within the material even more so as to make use of that speed at practical frequencies. That time delay applied to the two bucking coils offers induced negative resistance, or when connected as non-bucking it becomes induced positive resistance.
Smudge,
i happen to have a T107/65/18 grade 3F4 (https://www.ferroxcube.com/upload/media/product/file/MDS/3f4.pdf) (measured / calculated µr=770) toroid from an earlier delay line experiment.
I put on 2x 6 turns Litz wire coils diametrically opposite to each other which measure each 52µH and when connected in series opposing/bucking, they measure 16.5µH.
With a 100pF capacitor in parallel, it resonates at 3.770MHz.
I tried to measure / calculate the Q-factor, but get different outcomes, but mostly around 14 which i find rather low.
What will be the best way to measure / calculate the Q-factor?
(https://www.overunityresearch.com/index.php?action=dlattach;topic=4477.0;attach=55740;image)
Itsu
@Smudge
Itsu's equipment is capable of making these two and one-port measurements up to the GHz range with the following accuracy:
(https://www.overunityresearch.com/index.php?action=dlattach;topic=3924.0;attach=35778)
He can also pulse the DUT with a nanosecond pulse in the kilovolt range and observe the response on the oscilloscope, e.g.: see here (https://www.overunityresearch.com/index.php?topic=4642.msg114944#msg114944).
Quote from: Itsu on 2026.04.11, 19:30:15
Smudge,
i happen to have a T107/65/18 grade 3F4 (https://www.ferroxcube.com/upload/media/product/file/MDS/3f4.pdf) (measured / calculated µr=770) toroid from an earlier delay line experiment.
I put on 2x 6 turns Litz wire coils diametrically opposite to each other which measure each 52µH and when connected in series opposing/bucking, they measure 16.5µH.
With a 100pF capacitor in parallel, it resonates at 3.770MHz.
I tried to measure / calculate the Q-factor, but get different outcomes, but mostly around 14 which i find rather low.
What will be the best way to measure / calculate the Q-factor?
(https://www.overunityresearch.com/index.php?action=dlattach;topic=4477.0;attach=55740;image)
Itsu
I would be inclined to loosely couple the RF signal source to the LC circuit, perhaps by a single turn on a small ferrite rod placed within the electrical loop of L connected to C. Sweep the input frequency through the resonant frequency to measure the response, then use the -3dB bandwith against the resonant frequency to get the Q.
Q = frequency/bandwidth.
Thanks, that's one of the methods i used which resulted in a Q = 9.666. (3770 / 390 where Fres. = 3770, F2=4000 and F1=3610).
My concern is the loading of the DUT by the scope probe as it was mentioned in the PDF to try avoiding this too much.
As my used probes are specified as 10Mohm impedance, thats only at DC, so at the 4MHz frequency of the DUT it is decreased to about 4k, see:
Link removed as it was to private thread, copied the text and pictures below.
Quotei was reading in an earlier post that the TPP0500B probe was used as a load (10Meg / 3.9pF), but after looking at this video: https://youtu.be/Pk7pMguQDy4?t=374 i understand that those values are only correct at DC or very low frequencies.
So i measured my two probes the same way as in the video using the RF spring for ground and measured the following using my nanoVNA:
The frequency range of the nanoVNA was from 10kHz to 60MHz and the marker was set to around 1MHz:
TPP0500B probe (spec: 10Meg / 3.9pF) measured at 1.059MHz 25.1K / 5.9pF
P6139B probe (spec: 10Meg / 8pF) measured at 1.059MHz 16.3K / 9.2pF
See below VNA output traces which show that the capacitance stays fairly stable (but somewhat higher as the specs) right after the start range, but the impedance quickly drops considerably between start and 5MHz range to only a fraction of the specified 10Meg.
(https://www.overunityresearch.com/index.php?action=dlattach;topic=3691.0;attach=55748;image)
(https://www.overunityresearch.com/index.php?action=dlattach;topic=3691.0;attach=55750;image)
Itsu
Hi Itsu,
The inductive reactance of the two bucking coils is XL = 391 Ohm (rounded) at 3.77 MHz when their L = 16.5 uH. The quality factor, Q is XL / r so even if the two coils
DC resistance is say r = 3 Ohm in their series connection (a rough estimate), Q should be around 130. (I did not consider core and skin losses.)
Perhaps use a single bipolar transistor like the schematic shows below. Measure the resonant voltage across the 100 pF capacitor by the 10x scope probe and divide it by
the input voltage of the signal generator driving the base. This method mimics a classical Q meter concept, the transistor replaces the wide band matching transformer.
You can check the input voltage across the emitter resistor too, it should be 0.9 - 0.95 times the input generator voltage and can use it for the calculation.
Gyula
Itsu, use 200-220 Ohm emitter resistor instead of the 1 kOhm to have higher emitter current so the virtual emitter resistance (26 mV / IE ) should be lower. This way the coil Q will better approach the real value.
Insure the roughly 6 V DC voltage across the emitter resistance, this means 27 - 30 mA emitter current.
10 - 20 mV input voltage from the signal generator is enough to feed in and the voltage across the 100 pF capacitor may be around 1 V or higher if the loaded Q is around 100. On loaded Q I mean the virtual emitter resistance
which appears in series with the coils+100 pF plus the coil losses of course.
I wonder what is the DC resistance of the two coils in series? Less than 1 - 2 Ohm?
Gyula
Hi Gyula,
thanks for your info, i will take a look at your suggested circuit.
As far as i can measure with my Fluke DMM, the DC resistance of the 2 coils in series is 0.1 Ohm (0.2 Ohm of the DMM probes alone, 0.3 Ohm including the 2 series coils).
By the way, if i use this calculator: https://www.omnicalculator.com/physics/rlc-impedance with "RLC in parallel", R=0.1, L=16.5uH, C=100pF and f=3770 i get as result among other: Q=0.000246183.
Itsu
Quote from: Itsu on 2026.04.12, 10:09:10
My concern is the loading of the DUT by the scope probe
You have a FET probe that loads the DUT much less ...also, there is always this (https://youtu.be/rzo4Ntxqu1E?list=PLvOlSehNtuHvIDfW3x2p4BY6l4RYgfBJE&t=990) with a larger resistor.
Quote from: Itsu on 2026.04.12, 18:32:52
As far as i can measure with my Fluke DMM, the DC resistance of the 2 coils in series is 0.1 Ohm (0.2 Ohm of the DMM probes alone, 0.3 Ohm including the 2 series coils).
By the way, if i use this calculator: https://www.omnicalculator.com/physics/rlc-impedance with "RLC in parallel", R=0.1, L=16.5uH, C=100pF and f=3770 i get as result among other: Q=0.000246183.
But the 0.1 Ohm is not in parallel, it is in series with the L and your frequency was 3.77MHz not 3.77KHz. Try the calculator with RLC in series and you should get Q=4. And that would apply to L (with its series R) in parallel with C and no shunt R present across C. The online calculator doesn't allow you to input practical values for loss resistors within the L or C.
You previously reported a measured Q of about 10 which is quite low, but the experiment adding phase delay along the core should show increase of this Q indicating induced negative resistance. Of interest is how far that Q can go, does it tend towards infinity (self oscillations) as you add more and more delay? That Q figure derived from frequency/bandwidth should be good enough for this exercise.
Smudge
Smudge,
QuoteBut the 0.1 Ohm is not in parallel, it is in series with the L and your frequency was 3.77MHz not 3.77KHz. Try the calculator with RLC in series and you should get Q=4. And that would apply to L (with its series R) in parallel with C and no shunt R present across C. The online calculator doesn't allow you to input practical values for loss resistors within the L or C.
If i use "RLC in series" on that calculator i indeed get a Q=4 (by the way, i inputted 3770kHz for f, so it was 3.77MHz).
QuoteYou previously reported a measured Q of about 10 which is quite low, but the experiment adding phase delay along the core should show increase of this Q indicating induced negative resistance. Of interest is how far that Q can go, does it tend towards infinity (self oscillations) as you add more and more delay? That Q figure derived from frequency/bandwidth should be good enough for this exercise.
OK, i will use the Q derived from the frequency/bandwidth method (or ring down method which gives similar results) using the several probe solutions (Gyula's single bipolar transistor schematic and my FET probe) suggested.
Then start adding more delay and see what is going to happen.
Itsu
I used 3 probes to measure / calculate the Q of my "parallel LCR" (16.5uH, 100pF and 0.1 Ohm) as shown earlier, by using the "frequency / bandwidth" method.
Scope TDS-3054B:
probe P6139B (passive) 10x (10MOhm / 8pF) Q=13
probe P6202 (active) 10x (10Mohm / 2pF) Q=11
Scope MDO-3054:
probe TPP0500B (passive) 10x (10MOhm / 3.9pF) Q=11
Surprising outcome as the probe with the highest load (P6139B) gives the highest Q (13)
Gyula his bipolar transistor probe schematic was also tested on the above device, but in "series LCR" configuration (had to run the scope channel in AC coupling)
Bipolar transistor 2n2222 (active) 11.2V input, 3.5V DC across emitter resistor (220 Ohm), 55mVpp input from FG, 565mVpp across 100pF cap. Q=10.7
So i think this is enough to have a baseline Q (13) to see if when adding the delays it will increase.
Itsu
How does the VNA sweep look like ?
Well, i spend a few hours yesterday using the nanoVNA using several methods (shunt measurement S11 and shunt-thru measurement S21) but could not make any sensible (to me) readings out of it.
There are Q-factor graphs which i could show, but the figures presented did not make any sense to me (very low dip at resonance frequency in the 0.0024 range (same as the calculator results i linked earlier).
The resonance peak at 3.77Mhz was clearly visible, but i could not translate it into a valid Q-factor somehow (probably doing something wrong).
I tried both in parallel resonance as in series resonance setup.
Now i know it should be around Q=13 i could give it another try.
Below is a sweep from 10kHz to 10MHz using the "shunt-thru measurement" with the DUT in a series setup.
The Quality Factor S11 graph does not make sense to me (Q=0.017).
(https://www.overunityresearch.com/index.php?action=dlattach;topic=3691.0;attach=55772;image)
Itsu
Hi Itsu,
Thanks for doing the test. The emitter current was 3.5V / 220 Ohm = 16 mA this gives 26 mV / 16 mA = 1.6 Ohm virtual emitter resistance which appears in series with the coil, too high resistance.
IF you agree, would you reduce the 220 Ohm emitter resistor to 120 Ohm, and reduce the top base bias resistor R1 to 2.2 to 4.7 kOhm? This way the emitter voltage hence the emitter current should increase to 7 - 8 V at least
so the emitter current would be in the 60-70 mA range. This way the virtual emitter resistance reduces to around 0.3 - 0.4 Ohm. So the original coil Q should increase to much closer to the real Q the coil may have.
Gyula
Quote from: Itsu on 2026.04.13, 19:31:01
Below is a sweep from 10kHz to 10MHz using the "shunt-thru measurement" with the DUT in a series setup.
So why does the VNA display S11 ?
Shunt-through and series VNA measurements are 2-port measurements so the VNA should display "S21" for them.
Let's keep this organized - there are four S21 measurement possible:
1) Shunt-through S21 measurement of a parallel LC DUT
2) Shunt-through S21 measurement of a series LC DUT
3) Series S21 measurement of a parallel LC DUT
4) Series S21 measurement of a series LC DUT
(https://www.overunityresearch.com/index.php?action=dlattach;topic=3924.0;attach=35778)
(https://www.overunityresearch.com/index.php?action=dlattach;topic=3926.0;attach=41174)
S21 calibration fixture
Well, the nanoVNA itself can display the S21 Q-factor graph, but the nanoVNA PC App only has the S11 Q-factor Graph to display.
As both graphs look similar i toke the nanoVNA PC app to copy it here.
I did use the 2) Shunt-through S21 measurement of a series LC DUT in the measurement shown.
Itsu
Quote from: gyula on 2026.04.13, 20:11:09
Hi Itsu,
Thanks for doing the test. The emitter current was 3.5V / 220 Ohm = 16 mA this gives 26 mV / 16 mA = 1.6 Ohm virtual emitter resistance which appears in series with the coil, too high resistance.
IF you agree, would you reduce the 220 Ohm emitter resistor to 120 Ohm, and reduce the top base bias resistor R1 to 2.2 to 4.7 kOhm? This way the emitter voltage hence the emitter current should increase to 7 - 8 V at least
so the emitter current would be in the 60-70 mA range. This way the virtual emitter resistance reduces to around 0.3 - 0.4 Ohm. So the original coil Q should increase to much closer to the real Q the coil may have.
Gyula
Gyula,
i changed the resistors (120 Ohm and 3.3K now) which brought the emitter voltage to 8V
The Q now measured is 11.4
Itsu
I see, sorry and thanks.
Gyula,
i used the "frequency / bandwidth" method to measure / calculate the Q using your schematic, but i understand i need to calculate the Q using "the resonant voltage across the 100 pF capacitor by the 10x scope probe and divide it by
the input voltage of the signal generator driving the base".
If i do that i have 5467mV / 526mV = 10.4 as the Q, see screenshot:
(https://www.overunityresearch.com/index.php?action=dlattach;topic=3691.0;attach=55780;image)
Yellow = voltage across 100pF capacitor
Blue = input voltage FG (different voltage scale for clarity)
So the Q is very similar as the "frequency / bandwidth" method.
Itsu
Hi Itsu,
Thanks for these. I assume the generator voltage across the 120 Ohm emitter resistor is pretty close to 526 mVpp. Generally, emitter followers has a 'gain' of 0.9- 0.95, but this 'improve' the calculated Q only a little.
Unfortunately, a much higher emitter current would be needed to reduce the virtual emitter resistance. If we choose say IE = 1 Amper emitter current, the 26 mV / IE gives 0.026 Ohm, this is much less than the coils DC resistance of 0.1 Ohm.
This would need say a RE = 10 Ohm and >10 W power rated resistor and a 2N3055 or similar transistor, R1 would be say 1 kOhm. IF you agree and assign the time, maybe we can get a more correct Q value.
(IF the power transistor has a low hFE value, use a small transistor in Darlington connection to the power transistor, R1 would be say 100 kOhm in this case to bias the base of the small transistor.)
Gyula
Quote from: gyula on 2026.04.14, 11:46:21
Hi Itsu,
Thanks for these. I assume the generator voltage across the 120 Ohm emitter resistor is pretty close to 526 mVpp. Generally, emitter followers has a 'gain' of 0.9- 0.95, but this 'improve' the calculated Q only a little.
Unfortunately, a much higher emitter current would be needed to reduce the virtual emitter resistance. If we choose say IE = 1 Amper emitter current, the 26 mV / IE gives 0.026 Ohm, this is much less than the coils DC resistance of 0.1 Ohm.
This would need say a RE = 10 Ohm and >10 W power rated resistor and a 2N3055 or similar transistor, R1 would be say 1 kOhm. IF you agree and assign the time, maybe we can get a more correct Q value.
(IF the power transistor has a low hFE value, use a small transistor in Darlington connection to the power transistor, R1 would be say 100 kOhm in this case to bias the base of the small transistor.)
Gyula
Gyula,
QuoteI assume the generator voltage across the 120 Ohm emitter resistor is pretty close to 526 mVpp. Generally, emitter followers has a 'gain' of 0.9- 0.95, but this 'improve' the calculated Q only a little.
yes, the generator voltage across the 120 Ohm is 518mV (with DC offset, so need to set the channel to AC coupled).
QuoteUnfortunately, a much higher emitter current would be needed to reduce the virtual emitter resistance. If we choose say IE = 1 Amper emitter current, the 26 mV / IE gives 0.026 Ohm, this is much less than the coils DC resistance of 0.1 Ohm.
This would need say a RE = 10 Ohm and >10 W power rated resistor and a 2N3055 or similar transistor, R1 would be say 1 kOhm. IF you agree and assign the time, maybe we can get a more correct Q value.
I do have the 2N3055, and a 10W 10 Ohm resistor, so i can set this up.
I also do want to make some nanoVNA measurements to see if i can make some sensible Q measurements with that.
Meanwhile, i have added the 2 extra delay line coils (21 turns each) around the toroid, and plan to solder some (21 each) 22pF smd capacitors to each winding when done with the base Q factor measurements:
(https://www.overunityresearch.com/index.php?action=dlattach;topic=3691.0;attach=55782;image)
Itsu
Thanks. Probably you would need to use a Darlington configuration to surely drive the base of the 2N3055 with say the 2N2222 (join the two collectors on the +11V rail, connect the emitter of the 2N2222 to the base of the 2N3055. R1 is between the base of the 2N2222 and the +11V rail, input generator goes to the base of the 2N2222 via the coupling cap. The DC voltage across the 10 Ohm should be around 10V. All this is in case the hFE of the 2N3055 proves to be too low at 1 A current. Try to use the 2N3055 alone first, with R1 in the range between 470 Ohm to 1 k.
Gyula
I removed the last post as there is something wrong with the 2N3055 somehow.
Itsu
Hi Itsu,
Yes I agree. Just was about to tell to check it. 8)
Thanks,
Gyula
Gyula,
I had the 2N3055 too good isolated from its heat sink, so good that also the 12V was isolated from the collector :-[
Anyway, running with the 2N3055 alone, using first 1K for R1 (7.5V across Re), then 470 Ohm for R1 (8.6V across Re, so better), but neither was able to produce output across the 100pF capacitor (800mVpp in, 800mVpp out).
No matter how i cranked up the FG signal (it seems to be clamped to 800mVpp).
When including the 2N2222 (darlington style), R1 to 100K, still at 12V input i get a distorted signal at the base of the 2N3055 and almost no output of the 2N3055.
I have to crank up the FG input lot to see this output jump up and at the same time see the voltage across Re jump to 10V, but the output is way to low, see screenshot.
Blue is distorted base signal of the 2N3055, yellow the signal across the 100pF cap.
(https://www.overunityresearch.com/index.php?action=dlattach;topic=3691.0;attach=55794;image)
LTspice sim added how i have it configured right now.
Itsu
Thanks. Try to put a few kOhm (1, 2.2 3.3) resistor across the base and emitter of the 2N3055. Without the LC circuit across the emitter resistor, try to achieve a normal operation i.e. a correct sine wave across the 10 Ohm (while the DC voltage is around 8 - 9 V at least) for the Darlington emitter follower like you had with the single 2N2222. Maybe with other power transistor like BD139 or TIP31 etc you may get better results. Though their transition frequency is around 3-4 MHz too.
Gyula
Just thought of LTSpice... :) will play with it, thanks for the circuit file.
The resonant frequency of the 100 pF - 16.7 uH is 3894.6 kHz, I modified it in the asc file, attached. I run the simulation with 20 V DC instead of the 12 V, the AC emitter voltage seems improving.
See the attached asc file. Try to run the breadboarded circuit from 20 V (or vary the level) of the PS. (The emitter resistor has around 1 A and 10 V now in the simulator)
Will be back tomorrow evening.
Edit, I assume that due to L and C component tolerances, the slightly differing resonant frequency is not an issue because you fine tune the signal generator for maximum peak to peak voltage across the 100 pF and the probe's capacitance may pull the frequency down a little.
Gyula
Gyula,
Thanks for the info, i will play around with the input voltage.
I see you also changed over the R1 and R2 resistors (from your original posted circuit in post #30) R1 is now 33K and R2 is 100K.
One question: how to measure the Q in the "darlington situation"?
Do we still use the "the input voltage of the signal generator driving the base" and if yes, which base? (the 2N2222 base, the 2N3055 base or the FG voltage itself).
If i use the 2N2222 base voltage signal (in your simulation) i have 11.0 to 12.1 = 1.1Vpp and have an output resonance voltage across the 100pF capacitor of 7.05 to 14 = 7Vpp meaning a Q of 7 / 1.1 = 6.36.
Is this the correct way?
Itsu
Hi Itsu,
I corrected a little the circuit to have a better response, see the attached asc file. I reduced the signal generator output voltage to 10mV (at f=3868 kHz), I chose the
series DC resistance of L1 to be 0.1 Ohm what you measured with the Ohm meter, I chose the ESR of the 100 pF capacitor also to 0.1 Ohm (a good quality capacitor should have
an ESR < 0.1 Ohm).
Note that with 500 mV generator input the resonant frequency of L1C2 also changes, this change can be considered normal I think, driving voltage levels change
semiconductor parameters.
Note also: if we choose 100 mV input for the generator, the resonant frequency changes to 3878 kHz to get maximum resonant voltage across either L1 or C2
as per the simulator.
Regarding where to measure the generator voltage: I think across the 10 Ohm emitter resistor and the scope be in AC coupling to get rid of the DC component.
OR try to measure directly across L1 with 2 probes in differential mode (probes alligator clips floating).
Here is the simulator display of the voltage across the 10 Ohm, (1st pic), the AC wave rides about on 10.312 VDC, if I use the cursor for this V(n003) AC voltage and I
substract the bottom peak value from the top peak value, I get 10.314413 - 10.311302 = 3.111 mV, (this seems to remain from the 10 mV generator input). I blew up the
display for the V(n003) AC emitter voltage vertically for clarity, cursor shows the bottom peak value (riding on the DC value) which is 10.311302.
The 2nd picture (blown up horizontally) shows the direct generator input voltage (20 mVpp in red color) and the voltage across the L1 coil (3.79 Vpp in green color).
So the Q of the L1 coil would be 3790 mV / 20 mV = 189.5 and considering the same ESR of 0.1 Ohm I used in the simulation for the 100 pF, we would need to halve this,
getting about Q = 94.7 for the LC circuit WHEN we use the direct generator input voltage.
Using the AC emitter voltage which is only 3.111 mVpp in this simulation, the calculated Q would be higher, i.e. 3790 mV / 3.111 mV = 1218 and halving this would give
Q = 609
Considering the inductive reactance of L1 which is around XL1 = 406 Ohm at 3868 kHz and considering its 0.1 Ohm DC resistance, the calculated Q from these data would be
Q = 406 / 0.1 = 4060 in theory. The core loss and the virtual emitter resistance surely reduces the real Q and it should be (much) higher than 7 to 11 or so values.
Hopefully the measurements would reveal this.
Gyula
Can the 2N3055 handle 3868kHz in this configuration ?
Does LTSpice have its precise model or is it just using a generic or ideal BJT model ?
Surely the 3.8 MHz is little high for the data sheet's 2.5 MHz fT where the small signal current gain (hfe) of the transistor has a magnitude of 1. I suspected
this can be a drawback in the tests and the next step would be to replace this type with a better one.
LTSpice has transistor models given by manufacturers, for each type the component library includes. The 2N3055 model came from ST Microelectronics. Probably they did their best to fit their model to a real 2N3055 as close as possible as most manufacturers did.
Itsu, if you have for instance BD135, BD137 or BD139 for instance, these have 40 -50 MHz or higher transition frequencies, depending on manufacturers.
Or if you have some other power types in junk box, consult their data sheet. 1.5 Amper max collector current is already enough with the fT of 40-50 MHz.
EDIT it is possible that no Darlington config is needed with the above types.
Gyula
Quote from: gyula on 2026.04.16, 17:33:46
Hi Itsu,
I corrected a little the circuit to have a better response, see the attached asc file. I reduced the signal generator output voltage to 10mV (at f=3868 kHz), I chose the
series DC resistance of L1 to be 0.1 Ohm what you measured with the Ohm meter, I chose the ESR of the 100 pF capacitor also to 0.1 Ohm (a good quality capacitor should have
an ESR < 0.1 Ohm).
Note that with 500 mV generator input the resonant frequency of L1C2 also changes, this change can be considered normal I think, driving voltage levels change
semiconductor parameters.
Note also: if we choose 100 mV input for the generator, the resonant frequency changes to 3878 kHz to get maximum resonant voltage across either L1 or C2
as per the simulator.
Regarding where to measure the generator voltage: I think across the 10 Ohm emitter resistor and the scope be in AC coupling to get rid of the DC component.
OR try to measure directly across L1 with 2 probes in differential mode (probes alligator clips floating).
Here is the simulator display of the voltage across the 10 Ohm, (1st pic), the AC wave rides about on 10.312 VDC, if I use the cursor for this V(n003) AC voltage and I
substract the bottom peak value from the top peak value, I get 10.314413 - 10.311302 = 3.111 mV, (this seems to remain from the 10 mV generator input). I blew up the
display for the V(n003) AC emitter voltage vertically for clarity, cursor shows the bottom peak value (riding on the DC value) which is 10.311302.
The 2nd picture (blown up horizontally) shows the direct generator input voltage (20 mVpp in red color) and the voltage across the L1 coil (3.79 Vpp in green color).
So the Q of the L1 coil would be 3790 mV / 20 mV = 189.5 and considering the same ESR of 0.1 Ohm I used in the simulation for the 100 pF, we would need to halve this,
getting about Q = 94.7 for the LC circuit WHEN we use the direct generator input voltage.
Using the AC emitter voltage which is only 3.111 mVpp in this simulation, the calculated Q would be higher, i.e. 3790 mV / 3.111 mV = 1218 and halving this would give
Q = 609
Considering the inductive reactance of L1 which is around XL1 = 406 Ohm at 3868 kHz and considering its 0.1 Ohm DC resistance, the calculated Q from these data would be
Q = 406 / 0.1 = 4060 in theory. The core loss and the virtual emitter resistance surely reduces the real Q and it should be (much) higher than 7 to 11 or so values.
Hopefully the measurements would reveal this.
Gyula
Gyula,
Thanks for the info.
What strikes my from the start is that you are able to produce 3 different Q values on this one circuit depending on how / where you measure: Q = 94.7, Q = 609 and Q = 4060.
This sound very odd to me.
Anyway, i can reproduce your values in the sim, but when i try it on my real circuit, things are very different.
Firstly, the 10mV (20mVpp) input from the FG is so low, my scope has problems showing it, see the blue trace (23.7mVpp) in the screenshot below.
But also this low input voltage seems problematic for the rest of the circuit as the voltage across the 100pF capacitor is still very low (274mVpp, see yellow trace).
When using my differential probe across the coil L1, it not only influences the resonance frequency (3620kHz), but it also shows a low value as 285mVpp, see purple trace:
(https://www.overunityresearch.com/index.php?action=dlattach;topic=3691.0;attach=55808;image)
In this situation with 20V input, there is 1.12A running through my 10.4 Ohm emitter resistor which get hot quick.
I have to increase the FG input to at least 150mVpp to get some readable signals, but still the Q will be around the familiar level of 11:
(https://www.overunityresearch.com/index.php?action=dlattach;topic=3691.0;attach=55810;image)
Further increase of the input (500mVpp) lowers again the Q, see here:
(https://www.overunityresearch.com/index.php?action=dlattach;topic=3691.0;attach=55812;image)
I indeed think the 2N3055 has problems with the used frequency, and also the probes seem to have again an impact on the Q as well.
I will see what i have in the junk box, would 2SC5200 do? I have some.
Itsu
Ok Itsu, and thanks for your efforts. Sorry that I did not consider the frequency specs of the 2N3055.
IF you use the Darlington config with another power transistor, use a 220 kOhm trimmer or normal potmeter for R2, this would help find better bias points.
In the BD 135 etc family the types are marked with A, B and C in the suffix (if you happen find such) meaning the types are sorted for hFE, C meaning the highest (over 150 - 200).
Always fine tune the generator frequency for resonance whenever you change something, as so far you surely did so. 8)
Gyula
Gyula, i have some 2SC5200, with a hFE of 55 and an fT of 30Mhz, i will try one tomorrow.
Itsu
You mean 2SC5200 rather than 2SD5200 ? Is it made by Toshiba?
I cannot find 2SD5200 data sheet.
Quote from: Itsu on 2026.04.16, 20:03:47
In this situation with 20V input, there is 1.12A running through my 10.4 Ohm emitter resistor which get hot quick.
Are your EL2009 buffers stable at such currents ?
Gyula,
sorry, yes the 2SC5200 from Toshiba
verpies,
i am using my FY8300 FG for such experimental measurements, its cheaper to replace :)
An initial test with the 2SC5200 looks better, but i just soldered the 2SC5200 in place of the 2N3055 without changing something else.
I have 214mVpp in from the FG and 5.5Vpp out across the 100pF cap pointing to a Q of 25 (differential probe across L1 removed).
I will use a 220K trimmer for R2 and look for a sweet spot later on today.
Itsu
Gyula,
i installed a 220K trimmer pot as R2 (2N2222 base to ground) and played around with the input voltage, input signal from FG and trimmer setting while keeping peak resonance.
It boils down to a max. reachable Q of 25.9.
The input voltage is 20V, the current through the 10.4 Ohm emitter resistor is 1.38A, the trimmer pot is set to 213KOhm and the input signal from the FG is 280mVpp (blue trace).
This results in a resonance voltage across the 100pF capacitor of 7.2V (yellow trace), with a slightly distorted input signal, see screenshot:
(https://www.overunityresearch.com/index.php?action=dlattach;topic=3691.0;attach=55818;image)
Itsu
Hi Itsu,
Many thanks.
Because the AC voltage is surely less across the 10 Ohm than the AC input from the generator, this means the Q is higher than 25.9 (the 7.2V then is divided by a less than 280 mV value).
(Small note if I may: Probably you may have connected the 220 k trimmer between the base of the 2N2222 and the positive rail.)
The LTSpice includes transistor 2SC5200 with a suffix of letter O, so the hFE ranges between 80 to 160, https://www.farnell.com/datasheets/316951.pdf (https://www.farnell.com/datasheets/316951.pdf) I will use it in simulation without a Darlington config first.
Good night,
Gyula
Gyula,
i can make an AC measurement across the 10 Ohm resistor later today.
Not sure what you mean by your "Small note...", as i had mentioned that i have connected the 220K trimmer from base (2N2222) to ground, so do you mean i can try to put it between base (2N2222) and the positive rail?
Itsu
Hi Itsu,
I referred to this you wrote above: i installed a 220K trimmer pot as R2 (2N2222 base to ground) In the simulation circuit drawing R2 is connected between the base of the 2N2222 and the positive rail, see Reply #55 what I mean:
https://www.overunityresearch.com/index.php?topic=4927.msg118871#msg118871 (https://www.overunityresearch.com/index.php?topic=4927.msg118871#msg118871) This is what I wanted to clarify: I had thought R2 labeled as in the schematic, that is all.
EDIT 1 You may have meant R3 for using trimpot, not R2 ? If yes, just let R3 remain a normal 100 k resistor and indeed connect the trimpot to replace R2 33 k resistor. This may give a wider bias point adjustment possibility.
EDIT 2 Would you use another 100 pF capacitor instead of the present one? Just for closing out unwanted surprises.
Gyula
Gyula,
All my references to the resistors R1, R2 etc. are to the original schematic you presented in your post #30 (even when using the darlington setup):
(https://www.overunityresearch.com/index.php?action=dlattach;topic=3691.0;attach=55828;image)
So my 220K trimmer pot is now R2, meaning in place of the original 33KOhm resistor.
To make it even more complicated, my R1 is now 33K due to a swap you made in your simulator which i had followed :D
QuoteBecause the AC voltage is surely less across the 10 Ohm than the AC input from the generator, this means the Q is higher than 25.9 (the 7.2V then is divided by a less than 280 mV value).
I made a new screenshot, with the FG output signal (now 310mVpp after cold startup), in white, and in blue the AC voltage across the 10.4 Ohm emitter resistor, while in yellow we still have the voltage in resonance across the 100pF capacitor:
(https://www.overunityresearch.com/index.php?action=dlattach;topic=3691.0;attach=55830;image)
I used another 100pF which measures the same as the old one.
When using the white trace (FG) as input, the Q = 26.8, when i use the blue trace (across Re), the Q = 9.9
By the way, my real 2SC5200 reads: 2SC5200-O
Itsu
Hi Itsu,
I see. Well, from the single 2N2222 schematic we changed to the Darlington configuration and I thought we are on the same page in component labels since then. :D No problem.
In the screenshot, I do not understand how the AC voltage (836 mVpp) can be way higher across the (10 Ohm) emitter resistor versus the direct generator input of 310 mVpp?
A single emitter follower amplifies say 0.9 and a double emitter follower also, together 0.9 x 0.9 = 0.81 so a generator input of 310 mVpp should be around 0.81 x 310 = 251 mVpp.
Gyula
Itsu,
I simulated the series LC circuit alone, driven from a voltage generator via a 14.7 Ohm resistor, see the attached circuit. The generator has 0.1 Ohm internal resistance,
the coil and the capacitor are the same 16.7 uH, 0.1 Ohm and 100 pF ESR 0.1 Ohm.
With these component values the simulator gives the same screen shot you showed in your Reply #65 https://www.overunityresearch.com/index.php?topic=4927.msg118886#msg118886 (https://www.overunityresearch.com/index.php?topic=4927.msg118886#msg118886)
Generator voltage in the simulator 140 mVp (which gives the 280 mVpp you had) and the AC voltage across the 100 pF is 7.24 Vpp, resonant frequency is 3894.6 kHz.
So the question now is what introduces the loss in the Darlington circuit you tested and took the referenced screen shot from? This loss is equivalent to a 14.7 Ohm resistor
in series with the LC circuit, I arrived at this resistor value by trial and error to get the 7.2 Vpp voltage across the 100 pF.
This loss IMHO cannot come from the Darlington circuit because the 10 Ohm emitter resistor is shunted by the virtual emitter resistor
which is rE = 26 mV / IE so even an emitter current of say 10 mA causes rE to be 2.6 Ohm and this would shunt the actual emitter resistor (which is 10 Ohm here).
A possible loss factor might be the ferrite ring core, this is a guess of course. Maybe you have an Amidon soft powder ring core labeled as T56- or higher OD, you could wind a 16-17 uH coil and see
how it behaves in the circuit.
Gyula
Gyula,
i have rewired the part from the 2SC5200 in with solid wire and soldering connections as there were clip leads there, but it did not make a change.
I still have the bigger AC voltage across the 10 Ohm emitter resistor compared to the FG input voltage so i cannot find any losses in that last path.
So it could be the ferrite.
Anyway, i was looking for an Amidon T56 toroid, but they seem to be non-existing, are you sure they exist?
I do have some others like (from small to big): T106-2, T200-2, T225-6-1 and T520-2
Itsu
Itsu, the Amidon T50-6 core type was in my mind but I 'managed' to write T56- sorry.
Regarding the higher AC amplitude across the 10 Ohm, maybe the Darlington circuit regenerates, oscillates, I do not know. Try to reduce the 20 V supply voltage, this maybe a remedy.
May I refer to your single transistor test with the 2N2222 in your Reply #46 https://www.overunityresearch.com/index.php?topic=4927.msg118821#msg118821 (https://www.overunityresearch.com/index.php?topic=4927.msg118821#msg118821)
where the generator input was 526 mVpp and across the emitter resistor the amplitude was 518 mVpp what you measured, this is the normal operation for an emitter follower.
Regarding your available Amidon cores, choose perhaps the T200-2 type, see Pages 3 and 15 in this data sheet https://alexsradioshop.de/wp-content/uploads/2023/01/AmidonAMI.pdf (https://alexsradioshop.de/wp-content/uploads/2023/01/AmidonAMI.pdf) The AL is 120 (uH/100 turns)
so to have 16.7 uH on this core the number of turns would be around 37. Perhaps use the same kind of Litz wire for this coil but a normal enamelled copper wire would also do it (and easier to solder), say use around OD = 0.3 mm wire.
Gyula
Gyula,
QuoteMay I refer to your single transistor test with the 2N2222 in your Reply #46 https://www.overunityresearch.com/index.php?topic=4927.msg118821#msg118821
where the generator input was 526 mVpp and across the emitter resistor the amplitude was 518 mVpp what you measured, this is the normal operation for an emitter follower.
Yes, i see what you mean.
QuoteRegarding your available Amidon cores, choose perhaps the T200-2 type, see Pages 3 and 15 in this data sheet https://alexsradioshop.de/wp-content/uploads/2023/01/AmidonAMI.pdf The AL is 120 (uH/100 turns)
so to have 16.7 uH on this core the number of turns would be around 37. Perhaps use the same kind of Litz wire for this coil but a normal enamelled copper wire would also do it (and easier to solder), say use around OD = 0.3 mm wire.
I had one T200-2 with 40 turns of 1mm diameter magnet wire lying around, so i removed some turns and at 37 turns i got an inductance of 16.8uH, so i used that.
This was the result without any other changes since yesterday, so still at 20V with 1A through the 10.4 Ohm emitter resistor:
(https://www.overunityresearch.com/index.php?action=dlattach;topic=3691.0;attach=55839;image)
Yellow is the output across the 100pF capacitor
Blue is the input from the FG (set at 280mVpp)
White is the AC voltage across the 10.4 Ohm emitter resistor.
After reducing the input voltage to 12V with 500mA through the 10.4 Ohm emitter resistor i got this result:
(https://www.overunityresearch.com/index.php?action=dlattach;topic=3691.0;attach=55841;image)
Yellow is the output across the 100pF capacitor
Blue is the input from the FG (set at 280mVpp)
White is the AC voltage across the 10.4 Ohm emitter resistor.
So in both cases also with this new toroid, the AC voltage across the 10.4 emitter resistor is higher than the FG input AC voltage.
Perhaps the darlington setup is not appropriate for this situation?
Itsu
Hi Itsu,
Thanks for your efforts. Whether the Darlington config is appropiate here or not, I do not know the problem, basically it should be.
Perhaps it would be worth doing a test with the 2SC5200-O alone, remove the 2N2222 and use the 2SC as a single transistor emitter follower. The top bias resistor (between the base and the positive rail) would be the 220 kOhm trimpot, the bottom bias resistor
(between the base the negative rail) could be 100 kOhm, emitter resistor remains 10 Ohm. Start with 20 V supply voltage.
Gyula
Thanks Gyula, i will try that lateron to see if the darlington setup is the problem.
Meanwhile, i was looking for a way to use the nanoVNA to measure / calculate the Q and i found this website: https://coppermountaintech.com/determining-resonator-q-factor-from-return-loss-measurement-alone/
It shows how to do it and how to circumvent the low 50 Ohm impedance of the nanoVNA by loosely coupling (inductive or capacitive) the nanoVNA to the DUT.
I choose to use 10pF capacitors to couple the nanoVNA to the DUT, so my measurement circuit looks like this:
(https://www.overunityresearch.com/index.php?action=dlattach;topic=3691.0;attach=55845;image)
I did a calibrate (open, short, load, isolate and thru) by using at thru the 2 10pF caps in series connected (series measurement configuration) and lateron connect each of the 10pF caps to the DUT.
Using the return loss or S21 Log Mag (dB) Graph to plot the resonance trace and from that take Fres. and the both -3dB points to calculate the Q, see here:
(https://www.overunityresearch.com/index.php?action=dlattach;topic=3691.0;attach=55843;image)
It once again turns out to be a Q of 13.1
Itsu
Okay Itsu, that is a good find. You could try to reduce the 10 pF coupling caps to even lower values to make the nanoVNA loading effect less and less, perhaps try to go down to say 3.3 pF, or even 1 pF. Of course the driving level and the sensitivity of the instrument sets a limit for using too low value coupling caps.
Did you use the bucking coils or the new coil on the T200-2 core?
Gyula
My lowest value capacitors are these 10pF ones, but i could try to put several in series, or build some out of parallel wires.
I guess this measurement is similar as with the scope probe measurement, which also has a capacitance in the 8 to 3.9pF range with similar Q results (13).
I was using the original bucking coils.
Gyula,
Still with the T200-2 core.
Running with the 2SC5200 alone, 220K trimmer between positive rail and base (set to 3.6K), 100K between base and ground.
20V input at 460mA, FG set to 5Vpp input, but the scope shows only 500mVpp so the base is loading down the FG signal, so we can not use it as real input reference i think.
Input measured across the 10.4 Ohm emitter resistor shows 2.01Vpp and the output across the 100pF capacitor is 28.4Vpp pointing to a Q = 14.1
Screenshot here:
(https://www.overunityresearch.com/index.php?action=dlattach;topic=3691.0;attach=55853;image)
Yellow output across 100pF capacitor
White input from FG
Blue input AC voltage across 10.4 Ohm emitter resistor.
Itsu
Thanks Itsu.
With the latest 2SC5200 test alone, please clarify where the probe of the white channel is placed? You wrote:
"FG set to 5Vpp input, but the scope shows only 500mVpp so the base is loading down the FG signal, so we can not use it as real input reference i think."
And your wrote this under the Screenshot:
"Blue input from FG
White input AC voltage across 10.4 Ohm emitter resistor."
IF the white trace shows the AC voltage (500 mVpp) across the 10.4 Ohm resistor, then why did you divide the cap voltage of
28.4 Vpp by 2.01 Vpp? (The latter is supposed to be the loaded FG input voltage from the 5 Vpp, right?)
So if we divide the cap voltage 28.4 Vpp by the 500 mVpp AC emitter voltage, we get Q = 28400 / 500 = 56.8
This is a better value already but came from using the T200-2 core.
In this test you had 460 mA, this current established a virtual rE emitter resistance of 26mV/460mA= 0.056 Ohm.
IF you would use say 1 kOhm or 1.2 kOhm fix resistor now to replace the 220 k trimpot, the emitter current would become about 1 Amper, further reducing rE.
Note that the Amidon T200-2 core is capable of producing Q > 300 or 350 when the number of turns is between 25 and 40 (i.e. L is between 8uH and 20 uH) in the frequency range between 3 MHz and 4 MHz.
See the graph in Page 15 here again https://alexsradioshop.de/wp-content/uploads/2023/01/AmidonAMI.pdf (https://alexsradioshop.de/wp-content/uploads/2023/01/AmidonAMI.pdf)
So the emitter follower with the 460 mA emitter current ruins the expectable Q > 350 to nearly Q = 57 value IF the white trace indeed shows the AC emitter voltage of 500 mVpp.
Thanks,
Gyula
Hi Gyula,
i swapped the blue and white traces :D sorry about that. I have corrected it now in my earlier post.
As can be seen in the screenshot, the FG (set to 5Vpp) signal (white trace) when connected to the 100nF cap gets distorted and pulled down for some reason, so i do not think it is the correct value (500mVpp) to use as the input.
The blue trace is the AC voltage across the 10.4 Ohm resistor and reads 2.01Vpp, so the real FG input signal should be somewhat higher according to you, so say 2.2Vpp, but somehow it is not.
If i install a 1K fixed resistor for the 220K trimmer, i indeed get an emitter current of 1A.
Without further changes, the output signal across the 100pF cap is now 25Vpp and the AC voltage across the 10.4 Ohm emitter resistor is 520mVpp this a Q = 47
Thanks for the clarification.
Itsu, in the meantime I found this paper on experiments with coils and Q measurements https://hrsasa.asn.au/downloads/files/coilq.pdf (https://hrsasa.asn.au/downloads/files/coilq.pdf)
In Page 2, the Q measuring principle of the HP4342A Q meter is shown, I attached a screenshot.
It insures as low as 0.001 Ohm driving impedance to a coil plus a cap in series with it. The transformer primary has 50 turns, the secondary 1 turn, the primary is terminated
with an 50 Ohm resistor, this is driven by a signal generator having also 50 Ohm output impedance, meaning the transformer steps down from 25 Ohm to 0.001 Ohm
BUT
there should be a typo : the stepped down impedance should be 0.01 Ohm (and not 0.001) with the given turns ratio and 25 Ohm primary impedance.
Nevertheless, the 0.01 Ohm would already be the lowest to be attained so far. I do not understand why the paper says one milliOhm output impedance. Turns ratio 50, the square of this is 2500 and the 25 Ohm primary impedance appears as 25/2500= 0.01 Ohm.
IF you agree, this method would also be worth testing as the last one in this Q measuring journey which so far has yielded low Q values.
Of course, it is possible the step down transformer measuring method may still yield a Q of around 13 - 15 with the bucking coils.
If this would be the case, it would be good to check the Q of the T200-2 type core with this method, the Amidon data sheet indicates Q > 350 with turns between 25 to 40 (between 8uH - 20uH) in the 3-4 MHz frequency range as I wrote in my previous post.
I just noticed your additional post above, so the Q now is 47 with the T200-2 core. IT would be good to attain at least a Q of 300 for this core in a test circuit and the method shown in the paper I refer to may insure this. Basically this is a refined version of what Smudge
suggested earlier, using a coupling coil, now with a single turn secondary.
The FY8300 FG has a 24 Vpp max sine wave up to 5 MHz if I am correct, so after the 50 times voltage division there remains > 400 mVpp across the 1 turn output to feed the series LC circuit. If you agree with this, use the T520-2 core, easier to wind the 50 turns. ;)
What do you think?
EDIT: the 24 Vpp output of the your FG will be halved when the 50 Ohm resistor across the primary coil of the matching transformer terminates its output but the remaining > 200 mVpp is still enough to see it on the scope.
Gyula
Gyula,
that is an interesting paper with good details on how it works.
I will have to read it several times to fully grasp it, but the principle is clear.
One thing i noticed when comparing the data from my real circuit and the simulation circuit (or datasheets) is that when adding a 4K resistor across the 100pF capacitor (mimicking a scope probe), the Q is drastically lowered as could be expected.
I am not sure if the simulator is including this "load" when placing a probe, but i expect its not
Anyway, i am halfway to adding 27pF smd capacitors onto each turn (42 total :D ) of the 2 delay line coils on my ferrite toroid.
When done i can make the measurement to compare my Q data with and without these delay lines attached to see if Smudge was right.
It would be interesting to use this new Q measure method later on to see if it is doing what it supposes to do.
Thanks, Itsu
Hi Itsu,
No any load is imposed on a circuit in any circuit simulator when you use any probe in it.
In the real circuit, when a probe is hooked up in parallel with the 100 pF in the present LC circuit, the probe's self capacitance is added in parallel to the 100 pF, and as you correctly expected, it is not like at all when you deliberately shunt the 100 pF with 4.7 kOhm resistor to mimic the capacitive reactance.
IT is always good to have several probes and when you are ready to test this 'new' Q measuring method, would you use them (you wrote about in post #36)?
Basically, all of their self capacitances should be 'absorbed' as a small capacitance added to the 100 pF but the probes' 10 MOhm internal resistance is surely transformed into the LC circuit as a certain loss.
However, the equivalent parallel resistance of a decent 100 pF capacitor may be as low as some MegaOhm due to frequency dependent dielectric losses, unfortunately.
As an alternative capacitor check, would you consider paralelling 4 SMD caps (27 pF each) to replace the 100 pF cap you have used so far? The extra 8 pF should cause only a few kHz change in resonant frequency.
This is in case if the T200-2 core would perform way under the Q=300-350 range, we can trust in the Amidon cores Q wise too.
(In the HP4342A Q meter they used a good quality variable capacitor in series with the coil to be measured and a built-in RF voltmeter measured the resonant voltage across the variable cap to know the Q of the coil.)
Thanks also for your kind efforts,
Gyula
This results post is with reference to my post #36 quoted here below where i made some baseline Q measurements without the delay lines attached:
Quote from: Itsu on 2026.04.13, 18:46:55
I used 3 probes to measure / calculate the Q of my "parallel LCR" (16.5uH, 100pF and 0.1 Ohm) as shown earlier, by using the "frequency / bandwidth" method.
Scope TDS-3054B:
probe P6139B (passive) 10x (10MOhm / 8pF) Q=13
probe P6202 (active) 10x (10Mohm / 2pF) Q=11
Scope MDO-3054:
probe TPP0500B (passive) 10x (10MOhm / 3.9pF) Q=11
Surprising outcome as the probe with the highest load (P6139B) gives the highest Q (13)
Gyula his bipolar transistor probe schematic was also tested on the above device, but in "series LCR" configuration (had to run the scope channel in AC coupling)
Bipolar transistor 2n2222 (active) 11.2V input, 3.5V DC across emitter resistor (220 Ohm), 55mVpp input from FG, 565mVpp across 100pF cap. Q=10.7
So i think this is enough to have a baseline Q (13) to see if when adding the delays it will increase.
Itsu
The results with the delay line attached (2x 21 turn coils with each turn a 27pF smd capacitor attached, see picture below) can be found here:
Scope TDS-3054B:
probe P6139B (passive) 10x (10MOhm / 8pF) was Q=13, now with delay line: Q=14.1
probe P6202 (active) 10x (10Mohm / 2pF) was Q=11, now with delay line: Q=13.4
Scope MDO-3054:
probe TPP0500B (passive) 10x (10MOhm / 3.9pF) was Q=11, now with delay line: Q=14.2
Lateron in post #76 i also used my nanoVNA to make an initial Q measurement without delay lines resulting in Q=13.1
(https://www.overunityresearch.com/index.php?action=dlattach;topic=3691.0;attach=55843;image)
The same nanoVNA measurement with delay lines results in Q=14.6
(https://www.overunityresearch.com/index.php?action=dlattach;topic=3691.0;attach=55860;image)
The initial test with Gyula his bipolar transistor probe schematic which resulted in Q=10.7 was also repeated, but this made little difference as Q now with delay lines was measured (Fres/bandwidth) to be 10.1.
We tried many different configurations and may be we can present some better results lateron.
All in all i think that the Q with delay lines attached does increase compared to the Q without them
(https://www.overunityresearch.com/index.php?action=dlattach;topic=3691.0;attach=55862;image)
Itsu
Itsu,
That is encouraging. Is it possible to connect the two coils not bucking and see whether the delay line then reduces the Q. And a measurement with the coils as primary and secondary of a transformer to see what the delay time is? That will all help with the math.
Smudge
Smudge, I guess this would be no problem, I will see what I can do.
Itsu
Well, there is a problem.
The problem i have is that the both delay lines are active now, so i cannot make a comparison between the 2 litz wire coils in aiding mode "without delay lines active" and "with delay lines active".
I don't think it is valid to do a Q measurement in the "delay lines active" situation and compare that with the bucking coils "without delay lines active" situation.
To remove or disable the delay lines i have to desolder the 42 smd capacitors return wire.
If still doing the measurement, this is the result:
Putting the 2 litz wire coils in aiding gives an inductance of 194uH (was 16.5 in bucking).
The parallel capacitor stays at 100pF, so the resonance frequency lowered from 3770kHz to 595kHz.
The low -3dB point becomes 486kHz and the high -3dB point becomes 716kHz which is a difference of 232kHz.
Q = 595kHz / 232kHz = 2.56.
So considerable lower, but is this a valid situation to compare? I can remove the SMD caps return wire from them if needed, should i do that?
Itsu
Well, could you use a 8-9 pF capacitor instead of the 100 pF, this would bring the resonant frequency back to 3.8-3.9 MHz range.
Gyula
But i still would not have a comparison between "with delay lines" and without delay lines" in the new aiding mode situation.
Itsu,
I see your problem. The data you have given me I can work with to further our understanding so please carry on with the original experiment to see whether inceasing the delay increases the Q.
Smudge
Smudge,
By increasing the delay i have to change out the capacitors, now 42x 27pF smd caps, for say 100pF caps which i have also 42 available.
So if i need to remove the present 27pF caps, i can also do your earlier request of measuring the Q in aiding mode and the time delay between primary and secondary.
As i already have the Q when in aiding mode with the 27pF caps delay lines active i now need to do the time delay measurement between primary and secondary.
So i tried to measure the time delay between primary and secondary when still having the 27pF delay lines active, but its hard to do so as there is quite some distortion at the secondary when using a square wave / pulse signal.
Also the time delay is different at different frequencies, so i tried to measure the time delay at 1Mhz (10Vpp) with a sine wave signal with max. amplitude selected (some reasonable vertical traces to compare the time difference), i get a time delay between primary input and secondary output signals of the transformer of 1.84ns.
If there is another better way to do this measurement, please let me know.
If not, i can now remove the 27pF caps delay line and do the same delay time measurement as above and later on the aiding coils Q measurement to compare with the one made yesterday.
Finally then i can install the 100pF caps delay lines and do further Q and time delay measurements.
Itsu
No response from Smudge, so i guess the time delay measurement method was OK,
I removed the 27pF delay lines and did the time delay measurement once again which turns out that there now is a time delay between primary and secondary of 1.46ns (was 1.84ns with delay lines active).
Then i did the aiding coils Q measurement without the 27pF delay lines active with as results:
Putting the 2 litz wire coils in aiding gives an inductance of 191uH (was 194uH with delay lines active and 16.5 in bucking mode).
The parallel capacitor stays at 100pF, so now the resonance frequency becomes 960kHz (was 595kHz with 27pF delay lines active!!!).
The low -3dB point becomes 695kHz (was 486kHz with delay lines) and the high -3dB point becomes 1300kHz (was 716kHz with delay lines active) which is a difference of 605kHz.
Q = 960kHz / 605kHz = 1.58 (was Q = 595kHz / 232kHz = 2.56 with delay lines).
So the Q WITH delay lines active is higher (2.56) than without delay lines active (1.58).
Strange i think is the difference in resonance frequency WITH delay lines (595kHz) compared to WITHOUT delay lines (960kHz).
I will now install the 42 100pF smd caps delay lines and redo the Q measurement and time delay measurement......
Itsu
Quote from: Itsu on 2026.04.22, 20:26:27
Strange i think is the difference in resonance frequency WITH delay lines (595kHz) compared to WITHOUT delay lines (960kHz).
Yeah, the resonance frequency affects the Q calculation and other things.
Quote from: Itsu on 2026.04.22, 20:26:27
No response from Smudge, so i guess the time delay measurement method was OK,
Sorry, I have been busy trying to get a wheelchair adapted vehicle suitable for my needs and one I can afford. I assumed that a simple phase measurement between sine wave input and output voltages would suffice where the output is into a resistive load, not a reactive load, and at the frequency of interest. I admit I am out of touch with the complexities arising in bench measurements, it is years since I last used an oscilloscope. I am seriously looking at converting a corner of my room in the extra care facility where I now live into a small laboratory where I can react to these problems.
Smudge
Quote from: Verpies on 2026.04.22, 22:05:50
Yeah, the resonance frequency affects the Q calculation and other things.
Yes, the delay lines do influence the resonance frequency more when the 2 coils are in aiding mode (595kHz with versus 960kHz without) than when in bucking mode (3535kHz with versus 3642 without), of course the inductance thus resonance frequency is different, but still.
I can imagine that the different resonance frequencies have their influence on the Q.
Quote from: Smudge on 2026.04.23, 06:59:36
Sorry, I have been busy trying to get a wheelchair adapted vehicle suitable for my needs and one I can afford. I assumed that a simple phase measurement between sine wave input and output voltages would suffice where the output is into a resistive load, not a reactive load, and at the frequency of interest. I admit I am out of touch with the complexities arising in bench measurements, it is years since I last used an oscilloscope. I am seriously looking at converting a corner of my room in the extra care facility where I now live into a small laboratory where I can react to these problems.
Smudge
No problem Smudge, the situation on the bench (and not every bench will be the same) can throw things at you, you don't expect, therefor i try to be so specific as possible to transfer the data i see onto this thread.
Itsu
I installed the 100pF capacitors delay line and made the 3 measurements (bucking coils Q, aiding coils Q and the time delay) using the "Fres / bandwidth (-3dB)" method on the first 2.
I was using the MDO-3054 scope with TPP0500B probes (10MOhm / 3.9pF) with the following results:
Bucking coils:
Base Q 11
27pF DL Q 14.2
100pF DL Q 18.9
Aiding coils:
Base Q 1.58
27pF DL Q 2.56
100pF DL Q 4.3
Prim / sec time delay:
Base 1.46ns
27pF DL 1.84ns
100pF DL 322ns!!!
This last (100pF DL) time delay measurement seems wrong, but i tried it several times with the same outcome (using a 1MHz sine wave signal of 10Vpp at the primary all the time)
If i reverse the secondary probe leads i get an even longer delay of 796ns.
Itsu
Quote from: Itsu on 2026.04.26, 15:57:40
I installed the 100pF capacitors delay line and made the 3 measurements (bucking coils Q, aiding coils Q and the time delay) using the "Fres / bandwidth (-3dB)" method on the first 2.
I was using the MDO-3054 scope with TPP0500B probes (10MOhm / 3.9pF) with the following results:
Bucking coils:
Base Q 11
27pF DL Q 14.2
100pF DL Q 18.9
Aiding coils:
Base Q 1.58
27pF DL Q 2.56
100pF DL Q 4.3
Prim / sec time delay:
Base 1.46ns
27pF DL 1.84ns
100pF DL 322ns!!!
This last (100pF DL) time delay measurement seems wrong, but i tried it several times with the same outcome (using a 1MHz sine wave signal of 10Vpp at the primary all the time)
If i reverse the secondary probe leads i get an even longer delay of 796ns.
Itsu
Maybe phase is not the best way to get delay time in this instance, sorry about that. Any chance of having a fast leading edge pulse and seeing the delay from that?
I appreciiate all you are doing on Q measurements and hope that work improves results from this bench.
Smudge
Quote from: Smudge on 2026.05.02, 15:56:14
Maybe phase is not the best way to get delay time in this instance, sorry about that. Any chance of having a fast leading edge pulse and seeing the delay from that?
I appreciiate all you are doing on Q measurements and hope that work improves results from this bench.
Smudge
Smudge,
i tried it earlier with fast pulses, but there are no valid responses as far as I can see.
Here i use a fast leading edge pulse of about 600ps (does not registrate on my data for some reason) in yellow, which is put into the primary (no delay lines active) and in purple the response on the secondary:
(https://www.overunityresearch.com/index.php?action=dlattach;topic=3691.0;attach=56058;image)
So i do not see a valid secondary pulse on which i could measure the difference with.
Itsu
Quote from: Itsu on 2026.05.02, 20:52:12
Here i use a fast leading edge pulse of about 600ps (does not registrate on my data for some reason) in yellow, which is put into the primary (no delay lines active) and in purple the response on the secondary:
The purple trace seems to manifest a disturbance before the first rising edge of the yellow trace ...as if the effect preceded the cause.
Is you probing deskewed ?
Quote from: Verpies on 2026.05.02, 23:51:24
The purple trace seems to manifest a disturbance before the first rising edge of the yellow trace ...as if the effect preceded the cause.
Is you probing deskewed ?
I have seen this type of disturbance apparently occurring earlier than the instigating pulse on oscilloscopes in the past, and put it down to the length of the coax 'scope leads where you can get a travelling wave both inside the coax and one travelling along the outside of the coax. The outside wave travels faster then the inner one, so if the outside wave can influence what the scope is seeing you get this apparent early effect. I would be inclined to try a ferrite ring around the probe lead for the secondary and see if this eliminates the early effect.
Quote from: Verpies on 2026.05.02, 23:51:24
The purple trace seems to manifest a disturbance before the first rising edge of the yellow trace ...as if the effect preceded the cause.
Is you probing deskewed ?
It was not O0 i normally have my differential probe at channel 3 (purple), but changed it to the normal voltage probe yesterday, but probably the skewing characteristics of the different probes are not reset when reattaching a probe, so i guess there was still an old skew factor active.
I have reset the skewing and now the screenshot looks better:
(https://www.overunityresearch.com/index.php?action=dlattach;topic=3691.0;attach=56072;image)
There still is some (premature?) pulsing visible which disturbs the scope time delay measurement function (it looks for the first rising pulse between the 2 signals), so i had to do it by cursors as best as i can, which seems to show a 560ps delay between the 2 traces.
Mind you this is WITHOUT delay lines active, so a base measurement.
There however are still the OPEN delay line coils around each half of the core, so perhaps they interfere with the measurement (causing these premature purple pulses).
I can remove those open delay line coils, but then its much harder to make further delay line active measurement.
(https://www.overunityresearch.com/index.php?action=dlattach;topic=3691.0;attach=56066;image)
Itsu
Instead of the OPEN delay lines, i now reattached the (42) 100pF smd caps to the return line, so we now have a 100pF delay line active.
Doing the same measurement as above, we now see the time delay increase from 560ps to 620ps between the yellow peak and the purple peak:
(https://www.overunityresearch.com/index.php?action=dlattach;topic=3691.0;attach=56068;image)
Itsu
Quote from: Itsu on 2026.05.03, 11:19:41
Instead of the OPEN delay lines, i now reattached the (42) 100pF smd caps to the return line, so we now have a 100pF delay line active.
Doing the same measurement as above, we now see the time delay increase from 560ps to 620ps between the yellow peak and the purple peak:
(https://www.overunityresearch.com/index.php?action=dlattach;topic=3691.0;attach=56068;image)
Itsu
There is something I don't understand here. The 560pS delay along the open core I can accept, but only a 60pS increase with the delay line active doesn't seem right. You previously measured 52uH for a 6 turn coil so a single turn coil would be 1.444uH. A delay line of 20 sections where each has L=1.44uH and C=100pF gives a time delay of 12nS per section (t=sqrt(L*C), a total time of 0.24uS for 20 sections. Can you use a rectangular pulse with your fast rise time and look to see whether the output is then a sort of rectangular pulse but with that 0.24uS delay? Perhaps use 1uS pulse width and say 1KHz prf.
Smudge
Quote from: Smudge on 2026.05.05, 14:39:41
Can you use a rectangular pulse with your fast rise time and look to see whether the output is then a sort of rectangular pulse but with that 0.24uS delay? Perhaps use 1uS pulse width and say 1KHz prf.
Well, you asked for a fast rise time and he has delivered a pulse that rises in picoseconds. I think this is as good as it gets.
Getting a rectangular pulse with such edge characteristics is unrealistic.
Smudge,
i used your suggestion and dialed 1us long 5V DC rectangular pulse with a PRF of 1kHz on on my FG.
This pulse is the best my FG can do, and its edge is way slower than the pico-pulse i used earlier which is not from a FG, but from a homemade gadget which only produces the shown yellow pulse in that earlier post.
This is the result with NO coil attached, just the FG output on the scope:
(https://www.overunityresearch.com/index.php?action=dlattach;topic=3691.0;attach=56100;image)
This is the same pulse, but now connected to the primary:
(https://www.overunityresearch.com/index.php?action=dlattach;topic=3691.0;attach=56102;image)
We see some disturbance coming up.
Now see what happens when i add a 2nd probe to the secondary:
(https://www.overunityresearch.com/index.php?action=dlattach;topic=3691.0;attach=56104;image)
Where or what should i use for a starting point for measure the time delay between the yellow (primary) and blue (secondary) signals?
Remember, this is without any delay line active (coils are there, but open).
Itsu
Are these voltage waveforms ACROSS INDUCTOR ?
Screenshots 2 and 3, yes.
Screenshot 2 is yellow across the primary, screenshot 3 is yellow across primary, blue across the secondary.
Screenshot 1 is just yellow across the FG only.
I can see something like 40+ nS delay which is far more than the 560 pS from your earlier measurement. I take the time between halfway up the leading edge of the applied pulse (1V) and the 1V point of the leading edge of the output pulse. I have access to earlier work conducted by Graham Gunderson while employed by Chava LLC where I find the measured delay was 34.8nS. This earlier work was not looking at bucking coils, but had the secondary loaded with a capacitor where theory suggested the input at the primary might exhibit negative resistance. My job was analysing all the results, so I have a wealth of information that could benefit these present explorations. I will now publish this work in a separate thread, but here I put one of my papers that shows the similarity with Itsu's set up. I also include a paper that shows a measurement I did on a smaller toroid.
Smudge
Smudge,
QuoteI can see something like 40+ nS delay which is far more than the 560 pS from your earlier measurement. I take the time between halfway up the leading edge of the applied pulse (1V) and the 1V point of the leading edge of the output pulse.
I "zoomed in" on that specific area and this is the result (just the FG and yellow probe at the primary and the blue probe at the secondary):
(https://www.overunityresearch.com/index.php?action=dlattach;topic=3691.0;attach=56108;image)
So the time delay (without active delay line) is 43.5ns measured at that specific point.
Why the pico-pulser measured the 560ps i don't know, perhaps the ferrite is reacting different at faster pulses?
Itsu
When i use the same setup and scope settings, but instead of the FG driving the primary coil i use the pico-pulser, i get this result:
(https://www.overunityresearch.com/index.php?action=dlattach;topic=3691.0;attach=56110;image)
This clearly shows a different reaction to the pico pulse and the purple vertical cursors measure a difference at the same 1V points of 1ns, which probably by zooming in further decrease more to the earlier measured 560ps.
Itsu
It strikes me that your homemade gadget producing the picosecond pulse is sending something via a ground loop. Can you tell us what this gadget is and how it is connected? The signals from your FG are giving respectable results, and it will be interesting to see what happens when the delay line capacitors are connected.
Smudge
QuoteIt strikes me that your homemade gadget producing the picosecond pulse is sending something via a ground loop
Perhaps the open delay line coils?
QuoteCan you tell us what this gadget is and how it is connected?
this pico pulser is build around the LT1073 and the specific circuit can be found on page 14 of this datasheet: https://www.rlocman.ru/i/File/2023/04/14/lt1073.pdf
(https://www.overunityresearch.com/index.php?action=dlattach;topic=3691.0;attach=56112;image)
QuoteThe signals from your FG are giving respectable results, and it will be interesting to see what happens when the delay line capacitors are connected.
Ok, i will use the FG as source (1us pulse at 1kHz rpf) and put the 100pF smd capacitors delay line on.
The "lumped constant delay.PDF" shows that the secondary was terminated with 47 Ohm load, is that needed or wanted?
Itsu
100pF delay line active, same setup and scope setting as in the last screenshot of my post #107 (1us pulse 1kHz PRF):
(https://www.overunityresearch.com/index.php?action=dlattach;topic=3691.0;attach=56114;image)
Again here the problem where to measure the delay.
Itsu
Here the difference between WITHOUT and WITH the 100pF delay line active, but now both with a 50 Ohm load at the secondary:
(https://www.overunityresearch.com/index.php?action=dlattach;topic=3691.0;attach=56116;image)
(https://www.overunityresearch.com/index.php?action=dlattach;topic=3691.0;attach=56120;image)
The difference at the 1V point is about 136ns (86ns WITHOUT D.L., 222ns WITH D.L.).
Itsu
In that last measurement the mid point of the rising edge of the blue trace is some 200nS from the input yellow trace, and that is of the order that I calaculated in a previous post. I am OK with that and would expect a sinusoidal input to show a phase delay commensurate with that order of time delay.
Having looked back I see that I started a topic (https://www.overunityresearch.com/index.php?topic=3742.msg73171#msg73171) about using magnetic delay on 6th March 2019 that didn't get anywhere. I then started another topic (https://www.overunityresearch.com/index.php?topic=3847.msg78018#msg78018) on 23rd September 2019 where Itsu did much work similar to what he has done here. There I posted much more information and I had forgotten about that. That work ceased on 20th August 2023. I would point out that the work does show that we can create magnetic delay, and it seems theoretically possible to obtain excess energy using that delay by various means of which the bucking coils is but one possibility. (Incidentally bucking coils were looked at in that previous Itsu work, and I had forgotten that). Regarding where the excess energy comes from I have long held the belief that the electron spins that create the magnetic fields in permeable material can be a source of energy if we could get somewhere in the system the Lenz H field or mmf from the load current in the secondary to NOT be 90 degree shifted from the B field or flux. Electrical delay lines with appropriate terminations have the ability to create differing phase between voltage and current along the line, hence a magnetic one should also have this ability with respect to mmf and flux.
Smudge
@Smudge
Please guide Itsu how to integrate these waveforms so the input and output energy can be compared.
Me thinks, he will need current waveforms, too.
For what its worth, i switched to the TDS 3054B scope which has 2 P6302 current probes attached besides the 2 voltage probes.
It was powered on for some hours and a fresh calibration was done.
I still have the 100pF delay lines active.
I hooked up the yellow voltage probe and the green current probe to the primary.
I hooked up the blue voltage probe and the purple current probe to the secondary which has the 50 Ohm load resistor.
The FG was still set at 1us pulse with a 1kHz prf and 5V DC.
As this scope can make 1 math calculation at a time, i took 2 screenshots with the math function (red trace) set to first calculate input power yellow x green over 1 cycle (screen) followed by the math function (red trace) calculating the output power blue x purple also over 1 cycle (screen).
Results can be seen here:
(https://www.overunityresearch.com/index.php?action=dlattach;topic=3691.0;attach=56122;image)
Input power 1 cycle
(https://www.overunityresearch.com/index.php?action=dlattach;topic=3691.0;attach=56124;image)
Output power 1 cycle
Then i increased the shown cycles and set the FG to prf 150kHz and toke the same measurements over multiple cycles:
(https://www.overunityresearch.com/index.php?action=dlattach;topic=3691.0;attach=56130;image)
Input power multiple cycles
(https://www.overunityresearch.com/index.php?action=dlattach;topic=3691.0;attach=56128;image)
Output power multiple cycles
I rechecked by swapping the current probes and redo the above measurements which showed very similar results.
I will redo these measurements tomorrow again.
Itsu
Quote from: Verpies on 2026.05.06, 18:24:47
Please guide Itsu how to integrate these waveforms so the input and output energy can be compared.
Me thinks, he will need current waveforms, too.
I am not looking for OU in this pulse measurement as it is not the bucking coil arrangement. However Itsu's new power measurements showing COPs of 1.07 and 1.13 are interesting and worthy of consideration. If he repeated this measurement over reducing pulse widths and plotted COP against pulse width it might tell us something. I note that in the 2019 work Itsu got some COP>1 results but when repeated the COP disappeared, so this latest result might follow that pattern. If COPs>1 can be consistantly demonstrated to not be a measurement artifact this non-bucking approach is worthy of continued development.
Smudge
I agree, differences are very marginal, so further tests (swapping probes, using other methods like measuring primary and secondary separately to avoid ground loops, using different scope etc.) need to be done first.
Itsu
Today i was doing some extensive testing on the primary and secondary coils with regard to input and output power measurements.
There was a marginal difference between input and output in favor of output power using the older TDS-3054B scope using two P6139B voltage probes and 2 P6302 current probes.
Using one set of voltage and current probes (measuring the primary, then the secondary) and / or swapping the current probes over did not change this.
When i used the younger and more sophisticated MDO3054 scope with one voltage probe and one current probe measuring the primary, then the secondary showed something different.
There was hardly any difference in input and output power and the difference (0.3mW) was in favor of the input as one would expect in this circuit.
When on the TDS-3054B instead of the current probes using two 1 Ohm 1% induction free current sensor resistors at the primary input and secondary output, it was confirmed that there is slightly more power going in then coming out.
The problem i think is that the TDS-3054B has a limited range of manual de-skewing (+10 to -10ns) a probe while the P6302 current probe is specified as having a skew of 30ns.
So setting the voltage probe to +10 and the current probe to -10 will leave a difference of 10ns.
Probably either the input or output is more prone to be influenced by this remaining 10ns skew which shows up in the positive result for the output power.
The MDO3054 has a preprogrammed list of probes and their delays and can be automatically set to compensate them, so they all are equal.
So, nothing unusual here IMO.
Itsu
Thanks Itsu. Back to the bucking coils test which is looking for some induced negative resistance when power is supplied to the connected coils. This should show up when using sinewave input at a frequency associated with the tranmission line delay time. I had suggested looking for a peak Q value at the appropriate frequency but you could instead look for a reduction in input power. A plot of input power against frequency might show a dip at the right point. Perhaps include a small load resistor in series with the coils so that you are measuring a serious quantity of power and not just circuit losses.
Smudge
Ok, i can do that, but you need to be more specific about "a small load resistor", like 50 Ohm?
Could this "small load resistor in series with the coils" be in between the 2 bucking coils?
There is NO capacitor (series or parallel) involved in the experiment, right?
Itsu
QuoteA plot of input power against frequency might show a dip at the right point. Perhaps include a small load resistor in series with the coils so that you are measuring a serious quantity of power and not just circuit losses.
I put a small 51 Ohm SMD resistor in between the 2 bucking coils.
I feed in a 5Vpp sine wave sweeping from 1Hz to 10MHz during 1s, and monitoring this with the yellow probe, and this is the result:
(https://www.overunityresearch.com/index.php?action=dlattach;topic=3691.0;attach=56146;image)
So we have a 10MHz sweep from left to right, so each division is 1MHz
We see a dip around 5.7Mhz (and some further up, not visible).
Is that what you had in mind?
Itsu
Quote from: Itsu on 2026.05.08, 15:27:29
I put a small 51 Ohm SMD resistor in between the 2 bucking coils.
I feed in a 5Vpp sine wave sweeping from 1Hz to 10MHz during 1s, and monitoring this with the yellow probe, and this is the result:
(https://www.overunityresearch.com/index.php?action=dlattach;topic=3691.0;attach=56146;image)
So we have a 10MHz sweep from left to right, so each division is 1MHz
We see a dip around 5.7Mhz (and some further up, not visible).
Is that what you had in mind?
Itsu
Not quite, that is just input voltage. (Interesting that the dip occurs at a frequency near where the delay time is one full cycle.) Measuring current and voltage then doing the math to get input power at each frequency is what I suggested. That voltage dip may not be a power dip. I realize the power measurement can't be done using the frequency sweeping, it requires selecting whole cycles for the math. I was expecting a dip in input power at a frequency where the delay line gives a 90 degree phase delay, like 1.25MHz for a 200nS delay. So I would concentrate on looking around that frequency to find the dip (if it is real and not a figment of my imagination).
Edit. Forget what I said!! Of course minimum voltage across a resistor is also minimum current. I had a senior moment there. Is that dip at 5.7MHz there without the 50 Ohms, i.e. the coils feeding an open circuit (just scope probe). Then try monitoring current into a short circuit looking for a dip.
Smudge
Ok, you asked for input power, sorry about that.
I am not sure why you say "power measurement can't be done using the frequency sweeping", does that mean that the following math function is not valid (voltage sweep x current sweep = power sweep?).
Anyway, put it up anyway, and got this result:
(https://www.overunityresearch.com/index.php?action=dlattach;topic=3691.0;attach=56148;image)
Sweep was from 1Hz to 20MHz as i did see another dip further on (19MHz).
Yellow: voltage across bucking coils
Green: current through bucking coils
Red: power into bucking coils (yellow x green).
I measured the shown power between the red vertical cursors (6.4MHz), and it was the biggest negative value (dip) besides the one further on.
Looking around 1.25MHz does not show any peak or dip.
Itsu
At 1s/div I doubt that the voltage and current are sampled frequently enough to approximate a series of products that are representative of continuous power.
OK, but it could be an indication of it, right?
I am manually sampling a frequency range around the mentioned 1.25MHz (1Mhz to 1.5MHz) in 10kHz steps (no sweep) to see if there is something there and plot it in a graph, see here:
(https://www.overunityresearch.com/index.php?action=dlattach;topic=3691.0;attach=56150;image)
It is just a small part of the above shown sweep output, but the slope runs in a similar way as the same part of that sweep (2nd part of the first division).
Itsu
To display the slow trend of instantaneous power vs. time, use you scope's histogram feature to display the average value of the math trace that is the product of current & voltage sampled at high rate.
The i*v multiplication must occur at high sample rate but the averaging of the resulting products can be slow.
Itsu,
I have looked at your sweep results and it seems to me you are measuring the input voltage to the series coils plus resistor. You are not measuring the voltage across the resistor. On that basis the results are quite revealing.
I disagree with Verpies, I think all the results are continuous values including the math. What is not known with the voltage and currents values that all sit about the zero line is whether they create real or reactive power, and the math samples tell you this. It is the peak values that we can easily discern and estimate on each waveform. Staring at the 1Hz point we see by eye the voltage samples at 1.2V pk and the current samples at 28mA pk. That ratio is 42.9 Ohms which is close to the actual 50 Ohm load, and at this frequency that is what the input sees. The math tells us it is real power (all positive samples) reaching 35mW pk. We would expect a 50 Ohm load to draw 28.8mW from a 1.2V input so the numbers from the waveforms are all in the right order. Looking at higher frequencies where the inductive reactance takes over we see that the voltage reaches 2.4V pk so that seems to be the output from the unloaded 50 Ohm signal generator, hence the 1.2V figure is right for having a 50 Ohm load.
Moving along the waveform we see current reducing and voltage increasing as expected as the inductance of the coils take effect. We also see the math moving towards having the same + and - peak values as we get into reactive power. Minimum peak current is at about 4.5MHz where it is about 2mA pk with the voltage at about 2.4V pk, so the input is seeing about 120 Ohms impedance there. The math tells us the power is again real at a value of 5mW peak. At the 5.7MHz dip in voltage (now 1.4V pk) the current has increased to a maximum value of 20mA pk and the math is showing a real value of 30mW peak. That is showing a 70 Ohm input resistance.
Beyond this 5.7MHz dip the negative power highlighted by Itsu could be a real effect so that needs investigating thoroughly. A non-sweep test using a fixed frequency sine wave input should show more detail and allow better values to be determined.
Thanks for looking around the 1.2MHz region and not finding anything startling. It seems something happens at higher frequencies. That second dip at 19MHz is at a frequency 3.3333 times the first dip. Are all those threes a coincidence?
Smudge
Quote from: Verpies on 2026.05.09, 09:58:12
To display the slow trend of instantaneous power vs. time, use you scope's histogram feature to display the average value of the math trace that is the product of current & voltage sampled at high rate.
The i*v multiplication must occur at high sample rate but the averaging of the resulting products can be slow.
I see what you mean with this "histogram feature", i have to play around with it some more.
Itsu
Quote from: Smudge on 2026.05.09, 13:23:51
Itsu,
I have looked at your sweep results and it seems to me you are measuring the input voltage to the series coils plus resistor. You are not measuring the voltage across the resistor. On that basis the results are quite revealing.
I disagree with Verpies, I think all the results are continuous values including the math. What is not known with the voltage and currents values that all sit about the zero line is whether they create real or reactive power, and the math samples tell you this. It is the peak values that we can easily discern and estimate on each waveform. Staring at the 1Hz point we see by eye the voltage samples at 1.2V pk and the current samples at 28mA pk. That ratio is 42.9 Ohms which is close to the actual 50 Ohm load, and at this frequency that is what the input sees. The math tells us it is real power (all positive samples) reaching 35mW pk. We would expect a 50 Ohm load to draw 28.8mW from a 1.2V input so the numbers from the waveforms are all in the right order. Looking at higher frequencies where the inductive reactance takes over we see that the voltage reaches 2.4V pk so that seems to be the output from the unloaded 50 Ohm signal generator, hence the 1.2V figure is right for having a 50 Ohm load.
Moving along the waveform we see current reducing and voltage increasing as expected as the inductance of the coils take effect. We also see the math moving towards having the same + and - peak values as we get into reactive power. Minimum peak current is at about 4.5MHz where it is about 2mA pk with the voltage at about 2.4V pk, so the input is seeing about 120 Ohms impedance there. The math tells us the power is again real at a value of 5mW peak. At the 5.7MHz dip in voltage (now 1.4V pk) the current has increased to a maximum value of 20mA pk and the math is showing a real value of 30mW peak. That is showing a 70 Ohm input resistance.
Beyond this 5.7MHz dip the negative power highlighted by Itsu could be a real effect so that needs investigating thoroughly. A non-sweep test using a fixed frequency sine wave input should show more detail and allow better values to be determined.
Thanks for looking around the 1.2MHz region and not finding anything startling. It seems something happens at higher frequencies. That second dip at 19MHz is at a frequency 3.3333 times the first dip. Are all those threes a coincidence?
Smudge
Thanks Smudge,
Nice info, to be clear about my present setup, this is how i have it right now:
(https://www.overunityresearch.com/index.php?action=dlattach;topic=3691.0;attach=56152;image)
If i need to make some measurements across the 51 Ohm resistor, please say so.
QuoteBeyond this 5.7MHz dip the negative power highlighted by Itsu could be a real effect so that needs investigating thoroughly. A non-sweep test using a fixed frequency sine wave input should show more detail and allow better values to be determined.
OK, i will do that on that highlighted part.
QuoteThat second dip at 19MHz is at a frequency 3.3333 times the first dip. Are all those threes a coincidence?
Well that 5.7MHz i mentioned for the 1st dip was an estimate as it is hard to determine the exact midpoint from that screenshot, same for the 2nd dip at 19MHz, so it very probably is a coincidence.
Itsu
Quote from: Itsu on 2026.05.09, 15:40:10
Nice info, to be clear about my present setup, this is how i have it right now:
(https://www.overunityresearch.com/index.php?action=dlattach;topic=3691.0;attach=56152;image)
Do you realize that your setup is equivalent to the following with the toroidal turns and delay line added ?
(https://www.overunityresearch.com/index.php?action=dlattach;topic=4525.0;attach=56154)
...and there is a classic transformer induction taking place between these two parallel half-overlapping current loops ?
You can find out how much this induction affects your measurements by making these current loops perpendicular to each other.
Quote from: Verpies on 2026.05.09, 16:34:04ÿ
Do you realize that your setup is equivalent to the following with the toroidal turns and delay line added ?
(https://www.overunityresearch.com/index.php?action=dlattach;topic=4525.0;attach=56154)
...and there is a classic transformer induction taking place between these two parallel current loops ?
You can find out how much this induction affects your measurements by making these current loops parallel to each other.
That is not the case when feeding the coils in bucking mode since they drive currents in opposite directions around the toroidal loop.
If Itsu did his sweep while using the voltage across the load his math channel should read positive mean power values. If they also show negative values at the higher frequencies we know tbey are false. This bench set out to look for negative resistance which shows up as negative power, but I did not expect it to show up so soon. I was expecting a reduction in positive power. It would be good if the negative values are genuine.
Quote from: Verpies on 2026.05.09, 16:34:04
Do you realize that your setup is equivalent to the following with the toroidal turns and delay line added ?
(https://www.overunityresearch.com/index.php?action=dlattach;topic=4525.0;attach=56154)
...and there is a classic transformer induction taking place between these two parallel half-overlapping current loops ?
You can find out how much this induction affects your measurements by making these current loops perpendicular to each other.
I did not realize that, but is this also the case when the delay line is not almost one full turn, but as it is really is, consists of two halves?
(https://www.overunityresearch.com/index.php?action=dlattach;topic=3691.0;attach=56156;image)
Itsu
Quote from: Smudge on 2026.05.09, 17:20:00
That is not the case when feeding the coils in bucking mode since they drive currents in opposite directions around the toroidal loop.
Yeah, "around" in which direction ?
If the currents are opposite and equal then their sum over the loop should be zero.
Quote from: Itsu on 2026.05.09, 18:41:58
I did not realize that, but is this also the case when the delay line is not almost one full turn, but as it is really is, consists of two halves?
(https://www.overunityresearch.com/index.php?action=dlattach;topic=3691.0;attach=56156;image)
Every current element will generate an inductive magnetic flux. Even in loops that are not complete. Currents must flow to charge and discharge these capacitors of the delay line.
If the magnetic flux generated by these currents integrates to zero over the area spanned by the other loop, then and only then the mutual axial induction will be zero.
Since you are an experimentalist and magnetic flux is a directional quantity, you can observe the influence of this unexpected induction on your measurements by gradually making one loop's plane perpendicular to the other loop's plane.
Ok, but the problem is that my core with the coils and delay lines are sort of poured in concrete (hot glue), so it will be hard to "gradually making one loop's plane perpendicular to the other loop's plane".
Quote from: Smudge on 2026.05.09, 17:38:19
If Itsu did his sweep while using the voltage across the load his math channel should read positive mean power values. If they also show negative values at the higher frequencies we know tbey are false. This bench set out to look for negative resistance which shows up as negative power, but I did not expect it to show up so soon. I was expecting a reduction in positive power. It would be good if the negative values are genuine.
Well, i zoomed in on that 6 to 7MHz range which showed that -9mW Power by manually again measuring every 10kHz the voltage and current and let the math function calculate the power, with this as result:
(https://www.overunityresearch.com/index.php?action=dlattach;topic=3691.0;attach=56158;image)
So no negative power there as you expected, thus an artifact of the scope using this sweep method.
I do see that the power is continue to decrease and has a minimum of a few hundreds of uW around 12MHz.
I will make a similar 10kHz step power calculation there the next days or so.
Itsu
Quote from: Itsu on 2026.05.09, 20:41:04
Ok, but the problem is that my core with the coils and delay lines are sort of poured in concrete (hot glue), so it will be hard to "gradually making one loop's plane perpendicular to the other loop's plane".
So don't move the core and its delay lines.
Move the plane of the other loop.
Well, the "other loop" are the 2x 6 turns litz wire bucking coils which also are tightly secured with hot glue and even if i could move them, i have to manipulate them around or under or over the delay lines capacitors / return wire:
(https://www.overunityresearch.com/index.php?action=dlattach;topic=3691.0;attach=56152;image)
Perhaps when done with making the present measurements.
I found an area in the sweep where the scope seems to measure negative power.
I had to switch over from the TDS3054B to the MDO3054 as with the increasing frequency (~18MHz) the unable to compensate 10ns deskew of the current probe poses increasing problems.
But the MDO3054 scope his current probe TCP0020 is less sensitive (10mA versus 1mA) compared to the P6302 current probes used on the TDS3054B, and we need to measure in uA's to make some sensible calculations.
So i settled for using the P6302 current probe on the MDO3054, which was properly deskewed, and seems happy to show negative average power (the TCP0020 never shows negative power when used on the MDO3054, but seems lost in dealing with uA's, which could be the reason).
Anyway, for what its worth, here is a frequency range from 16.5MHz to 18.5MHz where i manually measured in 100kHz steps the voltage and current and let the math calculate the mean power.
Note that in the orange area, i had to switch both the current and Pmean vertical settings to avoid clipping of the signals, and this influenced the calculations somehow.
(https://www.overunityresearch.com/index.php?action=dlattach;topic=3691.0;attach=56164;image)
So there is at least 1 area in the sweep where my equipment has a hard time dealing with and displays a negative power.
Itsu
At all, especially Barry (hope you are well).
What you should be looking at is not "a" coil fed from either end, but 2 coils (bifilar wound) with one fed from one end the the other from the other end.
In my case it is a trifilar coil where the two outers are fed one way and the center fed the other in a sequence of currents, and so magnetic fields.
The BUCKING is of the magnetic field and NOT as a standing wave within the wire.
Of course the magnetic fields must not meet head on, in the same place, you will not like it.
Mike