PopularFX
Home Help Search Login Register
Welcome,Guest. Please login or register.
2021-10-21, 06:32:53
News: Registration with the OUR forum is by admin approval.

Pages: 1 2 3 4 5 [6] 7 8
Author Topic: "RF and molecular bond breaking Kanzius style"  (Read 102628 times)
Group: Guest
Yes, that was my thought as well. Got to get rid of that ringing on the Gate signal.

Itsu, is there a photo of the physical layout of the circuit that blows the mosfets?
   

Group: Elite Experimentalist
Hero Member
*****

Posts: 3189
Well, that 40Mhz signal on the gates was when the 4.5MHz module was in and that runs great, 500W out  at 82V drain voltage at 4.5MHz at 94% efficiency.
I am sure the gates signal with the 13.5MHz module that blows the MOSFETs is much cleaner  :D   But i have no picture of that, perhaps if i look back in this thread i find one.
**** found one back in this thread on page 2, the 13.5MHz gates signal looks like the yellow trace see picture below ****

See below a picture of the layout of the AMP with the 4.5MHz module installed (which as said runs great,)


I am uploading a video right now which shows the sweeping of the 13.5MHz output module as suggested by verpies, its a surprise to me how it looks like.


Regards Itsu
   

Group: Elite Experimentalist
Hero Member
*****

Posts: 3189
So either the VGS is exceeded or the 1000V VDS is.  Does Post Mortem ohmmeter measurement show the gate shorted or the drain shorted to source ?

Even the 1000V rating can be exceeded because of resonant rise in your LC network (due to bad tuning).  You should sine frequency sweep only the LC network to see its frequency characteristics - a lot can be gleaned from its amplitude vs. frequency plot.

Also, the voltage appearing on the drain influences your gate waveform through the Miller capacitance and/or the power supply line.  The distinguishing characteristic of RF MOSFETs is a very sensitive gate ( low VGS(TH) and VGS ).




Ok, did the frequency sweep of the 13.5MHz output module.
I added a 430pF capacitor simulating the both MOSFETs output capacitance (215pF each).
I also added the 24uH RFC with 100nF decoupling capacitor to ground.

input from the SG sweeping 5-15MHz monitored by the blue probe
output across a 50 Ohm resistor monitored by the yellow probe, see screenshot below.

Video here: https://www.youtube.com/watch?v=f_AtENQe6qA&feature=youtu.be

I am surprised to see such a flat response, as i expected to see a distinctive peak around 13.5MHz.


Regards Itsu
   

Group: Professor
Hero Member
*****

Posts: 2379
Ok, did the frequency sweep of the 13.5MHz output module.
I added a 430pF capacitor simulating the both MOSFETs output capacitance (215pF each).
That was thoughtful because the transistor's output capacitance is absorbed into the network.

I am surprised to see such a flat response, as i expected to see a distinctive peak around 13.5MHz.
That's because of the large toroidal inductor, which should be analyzed separately together with the MOSFET's output capacitance and the other cap to ground.
I was hoping to have only the LC network from the TP2 analyzed (frequency swept).  It should have the lowest impedance at 13.5MHz (measured from TP2 to output).
« Last Edit: 2015-07-13, 18:28:43 by verpies »
   

Group: Elite Experimentalist
Hero Member
*****

Posts: 3189

Hmmm,  thats the 4.5MHz output circuit, and i am missing the input capacitor of 450pF

For my 13.5MHz circuit i use these values, see circuit below in red taken from this website / pdf:

http://www.its.caltech.edu/~mmic/reshpubindex/papers/ClassE.pdf

The sweep of it is almost similar as the first one, see screenshot but not sure what you mean by:

Quote
(measured from TP2 to output).   
 
You mean scope probe tip to TP2, and its ground lead to the middle output plug (with or without the 50 Ohm resistor?).

I now have again the blue probe across the input (SG), the yellow across the 50 Ohm output resistor


Itsu

   

Group: Professor
Hero Member
*****

Posts: 2379
Hmmm, thats the 4.5MHz output circuit
I forgot to erase the frequency-specific values from the schematic. Done now.

, and i am missing the input capacitor of 450pF
Exactly. I was not interested in the capacitive reactance of that capacitor (450pF or 150pF) connecting TP2 to ground.

(measured from TP2 to output).
You mean scope probe tip to TP2, and its ground lead to the middle output plug (with or without the 50 Ohm resistor?).
No, SG-tip to TP2 and scope probe tip to the output (the right side of the schematic).  Grounds of SG and scope connected to the ground of the LCLC circuit.
The 50Ω load resistor present.
« Last Edit: 2015-07-13, 18:59:28 by verpies »
   

Group: Elite Experimentalist
Hero Member
*****

Posts: 3189

Ok,   well "that capacitor (450pF or 150pF) connecting TP2 to ground" is fixed in my module, so i have desoldered it, but without that cap, sweep looks similar as before
see screenshot yellow trace across the 50 Ohm resistor


Itsu
   

Group: Professor
Hero Member
*****

Posts: 2379
Ok,   well "that capacitor (450pF or 150pF) connecting TP2 to ground" is fixed in my module, so i have desoldered it, but without that cap, sweep looks similar as before
see screenshot yellow trace across the 50 Ohm resistor
So there is something wrong.
This LCLC circuit should have a different v(f) characteristic.

Let's measure each LC branch individually:
Apply the sine sweep from SG to point B and scope on the output socket with the load resistor present.  You may leave the LC branch between TP2 and point B, hanging in the air, if you do not want to desolder it.
You should see the classical notch LC frequency response of one LC branch.  Note the frequency of the lowest amplitude across the load resistor.

You can also apply the sine sweep from SG to TP2 and scope on the output socket with the load resistor present.  This time, desolder (interrupt) the series LC branch that grounds point B.
You should see the classical bandpass LC frequency response of the other LC branch. Note the frequency of the highest amplitude across the load resistor.

After these measurements you should have two frequency extrema - notch and bandpass, respectively.
« Last Edit: 2015-07-13, 21:16:18 by verpies »
   

Group: Elite Experimentalist
Hero Member
*****

Posts: 3189


something is wrong indeed,  no frequency extrema - notch and bandpass, respectively seen:

first screenshot is SG at point B
second screenshot is without the downward LC at point B

Itsu
   

Group: Professor
Hero Member
*****

Posts: 2379
something is wrong indeed,  no frequency extrema - notch and bandpass, respectively seen:
Could it be that the notch and bandpass are so narrow that you are not seeing it on such wide and fast frequency sweep ?
Alternatively, maybe the frequency extrema are outside of your sweep frequency range (i.e. not where you expect them to be).
Finally, maybe some RF capacitors have become shorted or opened.

The notch valley should be centered at the 2nd harmonic and the bandpass peak should be lower than the fundamental frequency.
Why lower and not equal? ...well because when the 450/150pf capacitor and the MOSFET's COSS capacitance is connected in parallel and when then this assembly of capacitances is connected in series (yes, in series) between the SG and TP2 in the second test (bandpass), then these additional capacitances will increase the center bandpass frequency, so it is very close to the fundamental frequency.
« Last Edit: 2015-07-13, 23:07:13 by verpies »
   

Group: Elite Experimentalist
Hero Member
*****

Posts: 3189

Zooming in on the frequency does not show a narrow peak/dip

I found the notch frequency (point B) at 20MHz see screenshot 1, no distinct peak seen on the bandpass, max amplitude is around 6.5MHz, but very broadbanded, see screenshot 2

Caps measure ok

Itsu
   

Group: Professor
Hero Member
*****

Posts: 2379
I found the notch frequency (point B) at 20MHz see screenshot 1,
Zooming in on the frequency does not show a narrow peak/dip
This notch center frequency should be at 2 * 13.5MHz = 27MHz.  So you can go ahead and correct it right away.

The notch is not narrow because the Q of this LC branch is low.  If you want to narrow it you can experiment with some of the methods listed under the graph below.
All while performing the frequency sweep to see the improvements in real time.

no distinct peak seen on the bandpass, max amplitude is around 6.5MHz, but very broadbanded, see screenshot 2
I'd say that LC branch is not working.  The Q would have to be incredibly low to exhibit such a wide bandpass.
It should have a much narrower bandpass with a center frequency that is much lower than the fundamental frequency (13.5MHz), when the other capacitors are absent (the 150pF + COSS).

You must troubleshoot this LC branch and get a narrow bandpass peak (high Q), the narrower - the better.  This LC branch is much more important than the other branch.



To increase the Q you may try:
- Decreasing the ESR of the capacitor (better cap or paralleling many small caps),
- Changing inductor winding turn-to-turn spacing to 1 wire diameter and trying to keep the coil's length/diameter ratio between 1:1 and 4:1 (see here).
- Shortening the intercomponent connections,
- Keeping the coil away from other conductors, e.g. the PCB copper laminate (taking it out of the box),
- Minimizing dielectric losses of coil wraps/carcasses/insulation.  Don't put insulation in the space between winding turns.
- Thickening the intercomponent connections (or Litz'ing them),
- More here.
« Last Edit: 2015-07-14, 06:02:53 by verpies »
   
Group: Guest
Maybe it's time to start using the FFT math function of the scope, and a slower sweep rate....
   

Group: Professor
Hero Member
*****

Posts: 2379
His frequency sweeps are already proven to yield good measurement results (see here )
Yes, longer sweeps are more accurate but he cannot make them too long, because at long time bases his scope goes into a roll mode ...which is annoying and begs for a firmware update.
   

Group: Elite Experimentalist
Hero Member
*****

Posts: 3189
The roll mode is active in auto triggering mode, in Normal triggering mode i can go down to any sweep time, presently i have it set to 1s on the Rigol, 100ms on the Tek which shows a nice updating picture.

By the way i am on the latest FW for this scope (2007 version  ;D ).

Thanks for the great info above, i have trimmed down the notch coil to just 1.1 turn (solid 2.5mm² copper wire) and its now dipping on 27MHz.
Its an air coil without isolation or a former.

My capacitors are all silver mica rated 1000V and show an ESR of 3 Ohm on the 1 to 2 nF ones going to 77 Ohm for the lower  (120) pF ones @ 100KHz.

Below:

screenshot 1  a 24 - 30 Mhz sweep.
screenshot 2  a broader (100KHz - 54 MHz) sweep

Itsu
« Last Edit: 2015-07-14, 11:57:31 by Itsu »
   

Group: Professor
Hero Member
*****

Posts: 2379
Thanks for the great info above, i have trimmed down the notch coil to just 1.1 turn
So the length/diameter ratio of this coil is not in the 1:1 - 4:1 range and Q suffers.  See the section Range of Inductor Form Factor in this article.

You can calculate the Q of your LC branch from the formula below:



My capacitors are all silver mica rated 1000V and show an ESR of 3 Ohm on the 1 to 2 nF ones going to 77 Ohm for the lower  (120) pF ones @ 100KHz.
That's a lot of Ohms !!

@TK
What capacitors do you use at these frequencies ?
« Last Edit: 2015-07-14, 14:17:41 by verpies »
   

Group: Moderator
Hero Member
*****

Posts: 2376


Buy me a beer
Hi Itsu

still in France but had a thought, I think you need to isolate your output LC when at 13.56MHz, I think there is some feed back through proximity with the driver at the higher frequency, also proximity to the ground plane creating capacitance which also at the higher frequency is detrimental to class E. At 13.56MHz maybe the C of the mosfet is sufficient as I have stated before, at 27MHz it would have to be reduced by a series cap for example.

I think that just with your setup at 13.56 you have been unlucky with your design setup (position) of components :-\

Regards

Mike 8)


---------------------------
"All truth passes through three stages. First, it is ridiculed, second it is violently opposed, and third, it is accepted as self-evident."
Arthur Schopenhauer, Philosopher, 1788-1860

As a general rule, the most successful person in life is the person that has the best information.
   

Group: Elite Experimentalist
Hero Member
*****

Posts: 3189


Mike,

you might be right, i could try to isolate the input from the output by means of a grounded shield or so.



verpies,

I modified the coils on the 13.56MHz module, see the picture.
The small coil is the notch filter on 2e harmonics (27MHz) and now is in the 1:1 range.
But the big coil (also 1:1) still produces a broad bandpass as can be seen in the screenshot (1 - 11MHz sweep) and shows
a mid resonance frequency of 6.5MHz which if you calculate the LC from here http://www.1728.org/resfreq.htm is right
(The coil measures 560nH with a Q of 8.5, the capacitor measures 1nF)

When looking at this design:  http://www.ixys.com/Documents/AppNotes/CO1.pdf and specific to Fig. 9 see below, then
i think that these designs are identical (they combine the Lt coil and the Lmatch coil in 1 inductor like i have, i only
have the extra notch coil in series with C0 to block the 2e harmonics).

They state that the tank coil should have a low Q like 2, so this means that also there they probably have a broad bandpass (or resonance tank)?

I tried severall different 1nF capacitors, but they all show the same bandpass and ESR (2.5 -3 Ohm).

With other words, i don't think i can get a better bandpass response.


Regards Itsu
   

Group: Elite Experimentalist
Hero Member
*****

Posts: 3189

Questions, questions questions,  sorry about that.

When sweeping the present series LC tank, see picture, i see a nice series resonance dip (expected / minimum impedance) around 7.5MHz, see the screenshot (100KHz - 15MHz sweep).
Adding or removing capacitors simulating a MOSFET does not influence this resonance point.

To me this means that the series tank circuit works as designed, right?
So then it seems to me that all i have to do is change the L or C or both so the series LC resonate at 13.56MHz, right?

During this measurement the notch leg (LC) was disconnected, but when connecting, there is no impact on this series LC resonance point.
It does influence the amplitude.


Regards Itsu
   

Group: Professor
Hero Member
*****

Posts: 2379
To me this means that the series tank circuit works as designed, right?
So then it seems to me that all i have to do is change the L or C or both so the series LC resonate at 13.56MHz, right?
I can't tell from the photo of the PCB where you are feeding in the SG and where you are probing, so I made the schematic below.  The red squares mean breaks/interruptions in the circuit.

If you feed SG to point A and scope across RL then you should get a bandpass response centered at 13.56MHz.
If you feed SG to point TP2 and scope across RL then you should get a bandpass response centered at much lower frequency (e.g. 6MHz - 8MHz)

Since the ESR of your C1 is high, then it might be prudent to increase RL to 500Ω to see the bandpass frequency response during sweeps.
Alternatively, short the output (RL=0) and scope the notch frequency response across the SG (across point A and Ground).  You may use a 50Ω - 500Ω resistor in series with the SG for its protection.

   

Group: Elite Experimentalist
Hero Member
*****

Posts: 3189
Thanks,

i will take a look at it again tomorrow.

My earlier photo of the PCB should show the red SG clip to your point TP2 and its black ground clip to your point B
Also the scope is on those same points, so probe tip to TP2, its ground clip to point B
I was only sweeping the L1C1 tank circuit.


This evening i was studying this design note:   http://users.skynet.be/BillsPage/ClassE030909.pdf
As always, things are more complicated as then first anticipated, and this design is without the notch coil.

I followed those design notes / calculations and not even need the circuit be designed for a specific frequency (13.56MHz),
but also for an intended drain voltage (24V / 40V etc.) and output (100W, 200W etc.).
  
So i calculated my amp. for 13.56MHz and for 24V and 40V  with resp. 100W and 200W output.
2 scenarios come out of that with fairly different values for L's and C's, see below diagram for these different value's.
Q of L1 was measured to be 8.5.

One question came in mind; does it matter if we have first L1 then C1 or vv (using verpies his diagram)?  (it seems that the L-match inductance is being combined with the resonance inductance,
so these 2 L's should be connected and not be separated by C1).


Thanks Itsu
« Last Edit: 2015-07-16, 12:05:17 by Itsu »
   

Group: Professor
Hero Member
*****

Posts: 2379
IMO nothing will change if you swap the position of L1 and C1
Of course the mutual inductance between L1 and L2 should be avoided by the usual tricks, such as: orthogonal placement, distance, shielding and so on...
   

Group: Elite Experimentalist
Hero Member
*****

Posts: 3189
That is not really what i mean, think about a design without the notch coil L2.
The L1 coil is made up of severall (2 or 3) different inductance's, like the main resonance inductance, the phase correction inductance and the L-match inductance.
With C1 in the righthand position, it is in between the first 2 inductance's and the last inductance.

See also my updated post above!

Itsu
  
   

Group: Professor
Hero Member
*****

Posts: 2379
From your signal path's point of view (from TP2 to Output) the L1 is a series inductor and L2 is a parallel inductor, so they are not the same.  Consequently L1C1C3 forms a fundamental bandpass filter and L2C2 forms a 2nd harmonic notch filter.  If you merge L1 and L2 then the node B disappears and the L2C2 notch filter with it.  The remaining C2 will just form a low pass filter, that will attenuate your fundamental at the output.

Without the L2C2 notch filter your output waveform will have a higher 2nd harmonic content and will be just less clean.
   

Group: Elite Experimentalist
Hero Member
*****

Posts: 3189
I understand, but you are talking about the design with the L2 notch coil.      

I am talking about the design without such a L2 notch (2e harmonics) coil.


Anyway, i will modify my output module to look like the above diagram (post #145) with the 24V / 100W values in green (as my L1 coil is close to the needed 519nH (now 560nH)),
so without the L2 notch coil to keep it simple (deal with the harmonics lateron).

Thanks,   Itsu

   
Pages: 1 2 3 4 5 [6] 7 8
« previous next »


 

Home Help Search Login Register
Theme © PopularFX | Based on PFX Ideas! | Scripts from iScript4u 2021-10-21, 06:32:53