Function Generator protectionbox using EL2009's

[Itsu]

Here i will continue this topic which was started at http://www.overunityresearch.com/index.php?topic=272.msg58938;topicseen#msg58938, but was way off topic.

It handles a replication of a Function Generator protectionbox using EL2009's designed by verpies.

The box was build according to the info that can be found in that other thread.

There are some problems, so measurement where taken and presented in a video.

Below are some pictures / diagrams concerning this build.

Itsu

[Itsu]

If you scope the input signal at the input pin of the EL2009, what levels do you get with your various input terminations?
What output levels do you get when you apply a ±5V signal to the input pin of the EL2009, without any input termination (no pot/resistor to ground)?

Please put at least a 1kΩ load on the output of the EL2009 when you do the output measurements.

i did some measurements as requested by verpies.

some unexpected observations:

With the input resistors removed i have both channels working
without power on the box (±16V), when activating the FG outputs, there is a + and - 1V on the supply rails.
without input from the FG but with power on the box (±16V) there is +16V on the output pins.

Video here:  https://www.youtube.com/watch?v=L0jt-UM1cM4

Itsu

[Verpies]

With the input resistors removed i have both channels working

So that means that you are loading the inputs too much with them.
Do the narrow-pulse TDR with a long high-quality coax between the FG and the EL2009 Box so you can see the mismatchedZ reflections well separated in time. A short coax will not give you a good time separation (~10ns/m).

without power on the box (±16V), when activating the FG outputs, there is a + and - 1V on the supply rails.

I have that, too.  This is because the EL2009 rectifies the input just like a bridge rectifier when it is unpowered and its output transistors are reverse biased.
In the past, I had actually blown up my EL2009 this way - with an input signal levels exceeding the power rails.  After installing the protection Schottky diodes, it never happened again.

without input from the FG but with power on the box (±16V) there is +16V on the output pins.

My EL2009s do not do that!
However, I remember that my Rigol FG had a problem with  floating up the disabled outputs at some firmware version.
Test it with just a ½(V_CC +V_DD) resistive voltage divider at the input (1kΩ + 1kΩ should be fine)

Last, but not least, 1kΩ load on the output is fine but boring.  Why don't you connect some 12V automotive incandescent light bulbs as a load (try 15W bulbs or less).  10Ω noninductive resistors are decent loads, too ...but they will just heat up without giving you the joy of light

[Itsu]

I did the TDR measurements using a 5.3m long coax cable (RG58).
It turns out that both 100 Ohm pots need to be set around 50 Ohm (47.2 and 50.2) with the box powered on to absorb the pulse.

It also showed that 1 channel (EL2009) seems oscillating.

The working channel looks OK as the output follows the input signal very close, see video.

After a long look at the oscillating EL2009, i noticed a 100nF bypass cap to be installed kind of awkward (grounded at the input lug), see picture.
After changing it to be grounded at the output lug instead, the oscillations are gone.

Video of the TDR measurement and clean in- and output of 1 channel here: https://www.youtube.com/watch?v=U2XGycyHdtY

I will install the schottky diodes and do some further testing, thanks for now.

Itsu

[Verpies]

I just blew up a circuit that had a mixture of aluminum and SMD tantalum capacitors.  I got confused which type was which and reversed the polarity of one.
The tantalum capacitors have a bar on the anode (+) side while the aluminum capacitors have a bar at the cathode (-) side, which is sick!

I remembered making this mistake years ago, but the memory of that quirk came back only as I smelled the magic smoke... a little too late.  In human brain the olfactory cortex is closely connected to the long-time memory storage.

I hope you did not make the same mistake with your tantalums.

[Itsu]

Sorry to hear that, yes i know that situation when memories come back as one smells the magic smoke...again.
Electronics are unforgiving as too late is too late 

My tantalums (not smd) are ok, the bipolar 100nF bypass cap was the problem.

Itsu

[Verpies]

It turns out that both 100 Ohm pots need to be set around 50 Ohm (47.2 and 50.2) with the box powered on to absorb the pulse.

That 3Ω difference is not small.
This is probably caused by a bad BNC elbow/knee, etc... or a little damaged EL2009.

Anyway, since you are so close to 50Ω input termination, you can swap the pots for high quality fixed resistors.  Inductance can cause ringing and fixed resistors always have less inductance than pots.
BTW: I liked how your TDR measurement showed how the input impedance of the EL2009 changed when you powered them up.

It also showed that 1 channel (EL2009) seems oscillating.
After changing it to be grounded at the output lug instead, the oscillations are gone.

Do you have all bypass caps (of both channels) connected to the output lugs, now ?

I noticed, there is still some 150MHz ringing with the square wave.
Try to see if a cap between +V and -V pin of the EL2009 helps with that.  As an old ham, you'll know which caps are the best for low-loss RF conduction/shorting.

I will install the Schottky diodes and do some further testing, thanks for now.

Schottky diodes have such low capacitance, that they should not change anything when the I/O signals stay between the supply rails.

[Itsu]

That 3Ω difference is not small.
This is probably caused by a bad BNC elbow/knee, etc... or a little damaged EL2009.

Yes, could be, it turns out that my BNC connectors / knee's are not the best quality 
As the TDR measurement signal was overwelmed by a 104Mhz signal it is hard to position the pot exactly.

I tried to eliminate that 104Mhz signal by various caps on various +/- power positions, but non worked.
I suspect that a strong local 104Mhz FM radio station is infiltrating into the circuit via the 5m long coax cable input connection.
I will try to build a notch filter for that.  I don't think it is visible when direct connecting the box to the FG.

Do you have all bypass caps (of both channels) connected to the output lugs, now ?

Hmmm, no, the both channels have their + power connection bypassed to the input ground lug (closest lug) and their - power connection
bypassed to the output ground lug (closest lug).
Perhaps i need to create a new single ground point very close to the both power leads to rule out any RF difference between the power rails.

 

I noticed, there is still some 150MHz ringing with the square wave.
Try to see if a cap between +V and -V pin of the EL2009 helps with that.  As an old ham, you'll know which caps are the best for low-loss RF conduction/shorting.

right, i did see that, but kind of thought that would be normal at that high frequency, but i will recheck.


Thanks,  itsu

[CITFTA]

Hi Itsu,

Now that you have fixed some of the other problems do you still need the coax?  If you feel you still need it, then just change the length of the coax and see if that solves the 104 mhz problem.  104 mhz has a wavelength of about 2.88 meters so your coax is pretty close to being 2 wavelengths long.  Maybe a 6 meter or 4 meter length would be less susceptible to the 104 signal.

Carroll

[Verpies]

He just needed the coax for adjustment of the input termination to have the reflections separated in time.

[Verpies]

bypassed to the output ground lug (closest lug).
Perhaps i need to create a new single ground point very close to the both power leads to rule out any RF difference between the power rails.

Good idea

P.S.
In the future, if you ever need more output current (lower input impedance), you can stack multiple EL2009 on top of each other, but this requires forward thinking about the mechanical arrangement of the box now. Good heat sinking is important.

[Itsu]

Its not the local FM station (on 105Mhz), the signal is 103.3MHz with the 5.3m RG58 coax cable (vf=66%).
My spectrum analyzer sees both the local FM station 105Mhz and this 103.3Mhz signal and can demodulate (FM) only the FM station to sound, just a carrier on the 103.3 signal.

When attaching another (somewhat shorter) piece of coax (rg58 too) i see the peak now on 114MHz.

The 103.3Mhz signal on 5.3m RG58 coax does not make sense to me as it does not fall into any ¼, ½, 1 or 2 wavelength category.

I think the attached coax acts like a capacitor in combination with some inductance to oscillate on 103Mhz.

Screenshot shows the signal as seen on my scope at the output plug when loaded with 1K.
My Rigol spectrum analyzer does not want to get linked to my PC anymore (since latest firmware upgrade) using Ultra Sigma from Rigol, so no screenshots from that.


Itsu

[Itsu]

Good idea

P.S.
In the future, if you ever need more output current (lower input impedance), you can stack multiple EL2009 on top of each other, but this requires forward thinking about the mechanical arrangement of the box now. Good heat sinking is important.

Yes, i notice that the box gets quite warm when powering some 3W leds for some time.
I have the both EL2009's on ceramic heatsinks attached to the box, i will see if this is enough.


Itsu

[Itsu]

I was busy trying to improve the signals when both or 1 channel is not turned on at the FG.
There is a strong signal in the 140 / 150 Mhz range (2Vpp) in that situation, see screenshot
When both channels are on on the FG, then there is no such a signal.


I have installed fixed 50 Ohm resistors at the input,
i have removed the beads at the EL2009 power sides with slight improvement
i have created a fixed ground point just below the EL2009's and attached the ± rails bypass caps (100nF) there.
i have added 47pF (spectrum analyzer shows a steep bypass frequency around 150Mhz) to both ± rails to ground and in between the ± rails.

The screenshot shows the signal with both channels off on the FG.
The spectrum analyzer also sees those signals using a short antenna, peaks 1 and 2 are from 1 EL2009, 3 and 4 from the other.
I went over to USB connect to do this screensshot.

Not sure if this normal, but it forces me to set both channels to on to overcome that noise.


Using the schottky diodes messes up the signal (both on and off at the FG) big time, so i did not install them.


Itsu

[Itsu]

A new batch of 5 EL2009's was received and i selected 2 with minimum in/out voltage when using a diode measurement (9 and 10mV difference).

These 2 new ones cleared up the noisy oscillations i had around 150Mhz when the FG output (input for the box) was disabled.
So somehow the old ones where damaged or way out of specs.

Installing the schottky diodes to the outputs messes things up big time still, so i removed them again.

I will do some run tests with the box to see if it stays stable under different circumstances.

Itsu

[conradelektro]

A new batch of 5 EL2009's was received and i selected 2 with minimum in/out voltage when using a diode measurement (9 and 10mV difference).

These 2 new ones cleared up the noisy oscillations i had around 150Mhz when the FG output (input for the box) was disabled.
So somehow the old ones where damaged or way out of specs.

Installing the schottky diodes to the outputs messes things up big time still, so i removed them again.

I will do some run tests with the box to see if it stays stable under different circumstances.

Itsu

In the EL2009 data sheet it is recommended to read the application section of EL2008 data sheet.

http://pdf.dzsc.com/EL2/EL2008.pdf  (EL2008C data sheet)

Source Impedance

The EL2008 has good input to output isolation. Open loop,
capacitive and resistive sources up to 100kΩ present no
oscillation problem driving resistive loads as long as care is
used in board layout to minimize output to input coupling and
the supplies are properly bypassed. When driving capacitive
loads in the 100pF to 1000pF region source resistances
above 25Ω can cause peaking and oscillation. Such
problems can be eliminated by placing a capacitor from the
EL2008s input to ground. The value should be about 1/4 the
load capacitance. In a feedback loop there is a speed
penalty and a possibility of oscillation when the EL2008 is
driven with a source impedance of 200Ω or more. Significant
phase shift can occur due to the EL2008's 25pF input
capacitance. Inductive sources can cause oscillations. A
series resistor of a few hundred ohms to 1kΩ will usually
solve the problem.


And as far as I understand, the EL2009C can only do 90 MHz, no wounder you get strange oscillations at 150 MHz.


Greetings, Conrad

[Itsu]

Hi Conrad,


thanks for the info, i am aware the EL2009 can only do 90Mhz, which in 99.9% of the cases is enough for me.
I never tried it on 150Mhz (this FG cannot do that), so it was an internal EL2009 oscillation, probably due to some capacitive or inductive feedback.

Verpies also pointed me to the EL2008 specs, which has good data for this type of device.

But usually i do not take note on what load my FG is seeing (inductive, capacitive or resistive), so this is something i need to do in the future.
How to cope with it when using it in LC resonance circuits where the load can be either of these 3 depending on if the frequency is below, on or above
the resonance frequency i need to find out.


Regards Itsu

[Verpies]

When driving capacitive loads in the 100pF to 1000pF region source resistances above 25Ω can cause peaking and oscillation.
Such problems can be eliminated by placing a capacitor from the EL2008s input to ground.

Since Itsu's FG's output has impedance of 50Ω then according to this datasheet these oscillations indeed can occur with capacitive loads between 100pF and 1nF.
The FG's output is not an inductive signal source so the other aforementioned problem cannot occur.

It is possible to decrease the FG's output impedance below 25Ω with a unity gain buffer before the EL2009, in order to make a more robust and versatile device.
These op-amps are usually small SMD parts that can easily fit into the metal box together with the EL2009. 

Op-amps selected for this purpose should be able to operate with the ±16V power supplies and have at least 120MHz of unity-gain bandwidth, as well as sufficient slew rate to support the maximum signal amplitude at the maximum frequency, that the FG can output (in Itsu's case that's 2.5V @ 100MHz which calculates to 25 0V / μs), for example the LME49713 or LM6171 or THS3001*.   The additional bonus is that with the addition of two resistors this can be converted to a x2 (or x10) voltage amplifier for obtaining higher output voltages at higher frequencies (e.g. at 80MHz, Itsu's FG is limited to only ±1.25V ) and the EL2009 does not provide any voltage gain.

Alternatively, a 100pF cap from the EL2009's input to ground, as the datasheet suggests, ...but then there might be further loss of amplitude at higher frequencies.


* ^{I am using this op-amp because it is the fastest, which means that it can do high amplifications (x10) at high frequencies.}

[Verpies]

@Itsu

Do you have a 15W automotive light bulb to stress-load the EL2009, yet ?

[Itsu]

No, but i did use some 3W leds which worked stable
I will take a look in my automotive light bulb spare box in the car to see if there is such a bulb.

Itsu

[conradelektro]

@ Itsu and other owners of an EL2009C:

Where did you order your EL2009C ?

Greetings, Conrad

[Itsu]

Conrad, 

i got my from Ebay.com and more specific, here:

http://www.ebay.com/itm/5PCS-EL2009CT-Encapsulation-TO220-90-MHz-1-Amp-Buffer-Amplifier-/272162161605?hash=item3f5e2103c5:g:MQsAAOSw44BYW4Wy

Itsu

[conradelektro]

i got my from Ebay.com and more specific, here:

http://www.ebay.com/itm/5PCS-EL2009CT-Encapsulation-TO220-90-MHz-1-Amp-Buffer-Amplifier-/272162161605?hash=item3f5e2103c5:g:MQsAAOSw44BYW4Wy

Thank you, nice source.

China seems to be the coming source for most things.

Greetings, Conrad

[Itsu]

I have 3x  12V/5W   bulbs parallel and set the FG to max pp voltage (10Vpp over 50 Ohm) which gives an rms value of about 3.2V.
The 3 bulbs just glow (sine wave) and the box gets hand warm.


I will let it run some longer time and shoot a video lateron tonight.

Itsu

[Itsu]

Load testing with 3x 12V/5W bulbs, at 10Vpp = 3.2V rms meaning pulling about 2.5W when using square wave at 50% duty cycle.
The box gets fairly hot and reaches (using a black tape on the box) around 60°C.

Screenshot is with sine wave signal, video shows both channels one by one and different wave forms:

https://www.youtube.com/watch?v=JzrFctnzCkc

Itsu