Overshoot fix 8

[itsmoked]

I have a circuit here that essentially minimally amplifies a voltage coughed up by a DAC. It's a non-inverting amp that has a wee bit of gain because the DAC is only good for 4.096V and the output needs to reach 10V. Meanwhile I need to also PWM the output. Hence the FET that yanks the input to 0V without bothering the DAC.

It all works as I desire with the exception of the overshoot seen in my sketch. Does anyone have a simple solution to it?

Keith Cress
kcress -

http://www.flaminsystems.com

 

[itsmoked]

Mike: You need a bit thicker skin around here, don't sweat it.

If you want to insult me you'd need to be much, much more direct. Your inputs have all been welcome, don't turn into an introvert on us.




Keith Cress
kcress -

http://www.flaminsystems.com

 

[iop995]

Reduce gate resistor to .5k and add a resistor about 1k to gate and ground. Also as was posted, a cap of 1-2nF in parallel with 7.3k resistor may be a solution; what tell about rising time of PWM signal?

 

[ijl]

I built the circuit with some available components: Different transistor, constant voltage instead of the DAC. The result was an "overshoot" in the wrong place, at the trailing edge instead of the leading edge, see picture. The "overshoot" disappeared when I removed the resistor at the gate.

This makes me wonder about the reason of the original overshoot.

 

[LionelHutz]

I doubt it's the problem but you checked your scope probe compensation, right?

 

[Skogsgurra]

I have been bothered from start because of the "wrong" place for the voltage increase. ijl shows how it should be if the problem is charge injection from gate, namely at the trailing part. It is when positive charge is injected to turn the transistor on that some of that charge spills to drain and causes a temporarily increased voltage, just as ijl shows.
Question to Smoked: Did you observe the trace through a mirror?


Gunnar Englund

http://www.gke.org

--------------------------------------
Half full - Half empty? I don't mind. It's what in it that counts.

 

[mikekilroy]

Since overshoot is not at point that can be caused by injection voltage issue, I probably will be called a liar again for repeating myself, but it still appears you are exceeding the slew rating of your LM358.

You say it is changing at rate of about 1v/us, yet the LM358 is only rated at .3v/us, about 1/2 of even a 741. when I suggested a faster op amp, I assumed something with capability of around 100v/usec.....

the other scope pix shown has what looks like 8v/30usec so if he used same lm358 he prob would be ok and not see the leading rise overshoot in my opinion.

so i stand by my comment to add a cap around 7.3k resistor to slow down to match the BW availabe on your LM358 or change from 10 cent to 50 cent op amp.

 

[Skogsgurra]

No-one called you a liar before. But we are coming close to that situation now. A "reverse" action at the start of a transition can not be caused by any of the mechanisms you propose. Nor can it be cured by any of the changes you mention. Period.

Gunnar Englund

http://www.gke.org

--------------------------------------
Half full - Half empty? I don't mind. It's what in it that counts.

 

[itsmoked]

iop995; I didn't try any R in the fet's D or S yet.

ijl; Wow! Thanks for the effort. I too am using a constant voltage instead of a DAC in this study.
When I added the large gate resistor it didn't fix the leading edges and added the small overshoot you show!

LionelHutz; I haven't checked the comp and should have, but, I also, on my first sketch, failed to notice that I DID actually have the negative spike too. It was a bit buried by the cursor bar. However you'll see in the pics below that I can get rid of one of them so I don't think it's scope comp.

Skogs: Nope no looking glass.

Hi Mike; I see looking at the data sheet 1v/us. Why are you saying a faster one would cure this? Is it that the amp would respond faster and hence would allow it to correct the overshoot? I could use a 50 cent amp without a problem. As for the cap 'solution' see below.


All;
Here's a picture of the unchanged situation as not completely drawn correctly. Note the negative glitch too.
ASIS

Here's the large gate resistor as suggested. Note, no real improvement but now the trailing spike shows up.
LGR

I tried many caps across Rf. None did much until I got up into 10,000pF which is shown here where it wiped out the negative one but also threw a wiggle into the trailing edge.
CAP

Turned out I didn't have an 4066 laying around even though I turned my lab inside out looking for one. I did lay this all out in the product with provision for an Rf cap and using the 4066 so we'll see if that does the trick after I get the boards back. Seems charge injection has been argued out of contending anyway.



Keith Cress
kcress -

http://www.flaminsystems.com

 

[Skogsgurra]

OK. I should have known it. Smoked IS the liar.
The traces show that bad probe compensation is the problem.
Easy as that. Shame on someone. Do not sketch traces. Show them!

If you have an overcompensated probe and reduce the t_f while keeping the t_r more or less intact, you get exactly the traces shown in Smoked's last picture.


Gunnar Englund

http://www.gke.org

--------------------------------------
Half full - Half empty? I don't mind. It's what in it that counts.

 

[Skogsgurra]

This thread has generated lots of discussions. One could base a seminar on what has been said. I restrained myself to a short paper on probe compensation. I hope that it is OK with you, Smoked?

http://www.gke.org/presentationer/files/Smoked probe compensation and other topics.pdf

Gunnar Englund

http://www.gke.org

--------------------------------------
Half full - Half empty? I don't mind. It's what in it that counts.

 

[itsmoked]

Hi Skogs. No problemo. I'm still not convinced it's probe comp since I haven't been back to my lab yet to confirm it. I will say when I see things like this probe comp is my first thought. But since I didn't see the negative one buried in my on-screen cursor I discounted it. Maybe tomorrow I'll be able to make by my lab.

Nice paper. What you don't seem to describe really is why an overshoot. You show the RC curves expected with caps and resistors but not why that would cause a higher spike because of bad compensation.

Keith Cress
kcress -

http://www.flaminsystems.com

 

[Skogsgurra]

That's a good point. I think there will be a sequel. I'll add Bode diagrams for over- and undercompensated probes. And I see that I didn't include the smoking probes. Stand by - watching the New Year's consert from Vienna now.

Gunnar Englund

http://www.gke.org

--------------------------------------
Half full - Half empty? I don't mind. It's what in it that counts.

 

[itsmoked]

Juries in.. LionelHutz nailed it. It was the compensation. I on the other hand. Blew it.

New Years Resolution:
Check the compensation, especially when the horns are there.

I'm still glad I switched to a 4066.
Thanks for the help everyone and have a good new year.

Keith Cress
kcress -

http://www.flaminsystems.com

 

[Skogsgurra]

Why overshoot? Why calibration errors? Why do probes (perhaps) smoke?
The Sequel:

http://www.gke.org/presentationer/files/Smoked probe compensation and other topics the sequel.pdf

Gunnar Englund

http://www.gke.org

--------------------------------------
Half full - Half empty? I don't mind. It's what in it that counts.

 

[itsmoked]

Ahhhh
Much better. Thanks Gunner.

Keith Cress
kcress -

http://www.flaminsystems.com

 

[sibeen]

Skogsgurra, great posts, and great pickup. Takes me back about 25 years when I, as a junior, and a senior engineer spent about a week looking for a phantom, very similar to this case.

That wasn't as embarrassing as having the cro set to AC input, and wondering why my square wave wasn't acting correctly