Cascaded OP-Amp

[jazzalex]

Hello again,

I am trying to cascade a number of OP-Amp circuits as shown in attachment 1 but only after two stages the signal indicates a significant amount of noise already (see yellow sine in attachment 2 - the green sine has no relevance). The input sine has a frequency of 10 kHz, while the oscillations have a frequency of approx. 250 kHz After three stages the signal becomes totally useless. Each amp is working fine in stand-alone operation.

I decoupled the stages with 100 nF capacities but the problem remains.

Can anyone help and tell me what the reason for the noise is and what I can do to get rid of it ?

Thanks in advance

Alex
 

[fcb]

So the 2nd opamp is driving a 50ohm source? (50 ohms is gruesome for a TL07x to drive).

[jazzalex]

Sorry, the sheet was made by qucs in order to S-parameter-simulate the circuit.

The actual load is 1 MOhm by the oscilloscope I am using.

[dannyf]

I am trying to cascade...

It actually will be a lot more helpful to you if you just pick an intro book on opamp. Like "Opamp for Everyone".

[jazzalex]

No, because the application itself is not related to "everyone". My interest lies in precisely this setup and not in a conventional setting. Pardon me for not stating why I need precisely this setup. It would lead to further discussions, which do not belong to this thread and/or forum.

[dannyf]

not related to "everyone"

By "Opamp for everyone", I actually meant "Opamp for Dummies": your issues are fairly simple to spot, even with rudimentary understanding of opamps.

[jazzalex]

Excuse me - I am not interested in bringing this to a personal level.

If this is a simple issue I am optimistic to get an answer in this forum because this is why I posted the question here.

[dannyf]

"Dummies" to signify the intro-nature of the book. Not stupidity of its readers.

[dannyf]

I went from "Opamp for Dummies" to "Opamp for everyone" thinking dummies may take offense to "Opamp for Dummies",

[T3sl4co1l]

It's dominated by noise because you made two differentiators (almost); in practice, they will probably be oscillators instead.  Capacitance to the -input of an op-amp is almost always best avoided.

What are you actually trying to do?

Tim

[katzohki]

Looks like you are trying to sample a signal and gain it up to look at it. You might try adding some impedance on the input of your opamps. That will help control the gain.

See this:
http://zebu.uoregon.edu/~rayfrey/431/notes9.pdf

Also, I normally think of "decoupling capacitors" as being connected across your power rail (Vcc to Gnd). Where they are now, might not be doing what you want.

You could give us more details, if you'd like, without giving away company secrets. Such as what voltage and frequency you are dealing with. And what you are trying to do.

[dannyf]

-Capacitance to the -input of an op-amp is almost always best avoided.-

so true. Rather than building on prior knowledge, people seem to be more interested in remaking the same mistakes.

[jazzalex]

You could give us more details, if you'd like, without giving away company secrets. Such as what voltage and frequency you are dealing with. And what you are trying to do.

Sure ! In context with a research project we are currently reproducing what had been done in the attached papers. They deal with negative group delays and how to correctly interpret them (of course not as "time machines"). In front of that background they cascade 4 OP-Amps in order to increase the negative group delay.

I just attached the 4th OP amp to the chain and the inserted sine (green) is advanced at the output (yellow) just like it is the case in the paper and as predicted by the simulation, however, the problem is the added noise obviously. 

[jazzalex]

I went from "Opamp for Dummies" to "Opamp for everyone" thinking dummies may take offense to "Opamp for Dummies",

OK - well understood. It's actually a great book. Do I get correctly that you are the author of it ?

[dannyf]

I wish I were that shameless,

[Zero999]

I went from "Opamp for Dummies" to "Opamp for everyone" thinking dummies may take offense to "Opamp for Dummies",

OK - well understood. It's actually a great book. Do I get correctly that you are the author of it ?

It's free to download.
https://www.google.co.uk/search?q=opamps+for+everyone&ie=utf-8&oe=utf-8&rls=org.mozilla:en-GB:official&client=firefox-a&gws_rd=cr&ei=KU7tUoeyCLSM7AbUuoHQCg

[jazzalex]

Yes, I know and have it already. Thanks anyways !

[kxenos]

Is it possible that the output you want is this? I think I haven't understood your circuit, but if you remove the capacitors of the following stages the noise goes away. Also, if you want to add 2 voltages you have to use resistors for both of them. 

[jazzalex]

Is it possible that the output you want is this? I think I haven't understood your circuit, but if you remove the capacitors of the following stages the noise goes away. Also, if you want to add 2 voltages you have to use resistors for both of them. 

Thanks a lot for the constructive feedback but I need to clarify a bit more what the goal of this experiment is:

1.) Although the design is supposed to work as a bandpass amplifier the amplification itself is not the field of interest.
2.) The negative group delay of this circuit is what matters is this context.
3.) My proposed circuit results in negative group delays for frequencies up to approx. 200 kHz.
4.) Again I attached the simulated group delay and transmission. Matter of interest is *only* the group delay below 200 kHz.
5.) With respect to my 10 kHz input test sine I expect a slightly phase shifted and slightly amplified 10 kHz sine at the output.
6.) However, the circuit seems to oscillate with the bandpass amplification frequency of approx. 500 kHz.
7.) Of course I'd love to see suggestions for alternative (more stable) circuits as long as they indicate negative group delays as described. above.
8.) The previously attached papers explain all this in deep detail but please let's not discuss the overall goal in this thread.

Thank you

Alex

[Ghydda]

I am trying to cascade a number of OP-Amp circuits as shown in attachment 1 but only after two stages the signal indicates a significant amount of noise already (see yellow sine in attachment 2 - the green sine has no relevance). The input sine has a frequency of 10 kHz, while the oscillations have a frequency of approx. 250 kHz After three stages the signal becomes totally useless. Each amp is working fine in stand-alone operation.

Well, you are kind of asking for it. Your feed back circuitry creates a zero at 80kHz. That means the signal gain increases with 20dB/dec above the corner frequency. Any signal on the input (including noise) will be amplified - and at 320kHz the gain is 12 dB.
@ 250kHz a bit less.
Adding stages on further multiplies the (unwanted) signal to a point were it is clearly visible on a scope and even exceeds in amplitude the fundamental signal.

I decoupled the stages with 100 nF capacities but the problem remains.

Decoupling is always recommended, however it cannot fix a non-error.

Can anyone help and tell me what the reason for the noise is and what I can do to get rid of it ?

Getting rid of it should be straight forward. Getting rid of it, without compromising the desired goal, that's another matter.
what is needed is at least one pole pr. zero in each feed back loop. But of course, in order to not interfere with your 10 kHz test signal the poles must be placed at a suitable frequencies.
In this case it's rather difficult due to your selection of the zero frequency compared to your 250 kHz noise. However, if you're willing to adjust a few parameters, it will be a lot easier to achieve a good noise rejection and still get what you want from the circuit: a negative phase at the test signal frequency.

Now if I may redirect your attention to the paper 'negative2.pdf' on page 4:

V. Experiment

A. Practical design

Figure 5(a) shows a practical circuit for negative delays [11].
The components C? are R? are added to the circuit of Fig. 1(b)
in order to suppress the gain for higher frequency.

The paper implements two poles to secure a stable working feed back loop, which rejects high frequency noise. I suggest you familiarize yourself with the chapter as it provides some good info to component selection.

The math behind ideal opamp gain blocks is very simple and easy to work out due to infinite gain, infinite bandwidth, zero offset, zero input bias current, completely noise free components and so on. In the real world it's a bit more complex, but knowing the shortcomings of the components makes all the difference and usually it is possible to get the math and the real world to show (almost) spot on results.

Happy hunting.

[Ghydda]

Hi Alex.

3.) My proposed circuit results in negative group delays for frequencies up to approx. 200 kHz.

Auch, in that case you need to be very very picky about your opamp. My first gut feeling tells me: choose something with a gain bandwidth product in excess of 10-20 MHz. And watch out for that slew rate figure as well.

5.) With respect to my 10 kHz input test sine I expect a slightly phase shifted and slightly amplified 10 kHz sine at the output.

That you shall have. Your existing schematic circuit surely supports that at this moment. Or does your measurements indicate otherwise?

6.) However, the circuit seems to oscillate with the bandpass amplification frequency of approx. 500 kHz.

What bandpass? The schematic shows no bandpass configuration in any of the feedback loops
I only see differentiator configurations.

8.) The previously attached papers explain all this in deep detail but please let's not discuss the overall goal in this thread.

Indeed they do, which make me think you haven't studied them in detail regarding the practical execution of the circuit. Combine that with a bit general knowledge of what a bug standard opamp is capable of, and you'll see you're asking too much of a TL07x series opamp.

Locate better parts, verify them against you requirements and implement. Then it will do as you ask.

A clarifying question: if your input signal source is shorted, do you still see the 250/500 kHz noise - and if so - what amplitude does it have?


Cheers

[jazzalex]

Hi Ghydda,

your answer does help very much !

>5.) With respect to my 10 kHz input test sine I expect a slightly phase shifted and slightly amplified 10 kHz sine at the output.
>That you shall have. Your existing schematic circuit surely supports that at this moment. Or does your measurements indicate otherwise?

Yes, this is what the circuit shows but with every additional stage the oscillations increase "dramatically". Nevertheless, the phase is looking allright.   

>>8.) The previously attached papers explain all this in deep detail but please let's not discuss the overall goal in this thread.
>>Indeed they do, which make me think you haven't studied them in detail regarding the practical execution of the circuit.

That is partly correct or better I neglected some parts of it. Actually in the meantime I installed a 15 kHz low-pass in front of each stage and the resulting sine is perfectly fine. However, this currently applied 100 Ohm/100 nF-low-pass generates a delay of 10 us, which is now the next problem to be solved.

>>What bandpass? The schematic shows no bandpass configuration in any of the feedback loops

With respect to the transmission characteristics (peak gain at 250 kHz) I called this bandpass amplifier but of course the circuit currently has no high or low pass filters involved. Sorry, if the term is confusing (or propbably wrong) in this context.

>>A clarifying question: if your input signal source is shorted, do you still see the 250/500 kHz noise - and if so - what amplitude does it have?

What do you mean by shorted (e.g. using a time limited pulse instead or simply a higher sine frequency) ?

Thank you,
best

Alex

P.S.: Just in case the now fine looking sine after the 4th stage - however, delayed by altogether 30 us due to the additional low-pass filters.

[Ghydda]

Hi Alex,

>>A clarifying question: if your input signal source is shorted, do you still see the 250/500 kHz noise - and if so - what amplitude does it have?

What do you mean by shorted (e.g. using a time limited pulse instead or simply a higher sine frequency) ?

I meant remove signal source and connect input to ground on your circuit  - in order to determine if the high frequency noise is coming from within the circuit or from the signal source.
But since your low pass filters seems to rid you of it, that suggests that the signal source might be the troublemaker.

P.S.: Just in case the now fine looking sine after the 4th stage - however, delayed by altogether 30 us due to the additional low-pass filters.

Well that's the trouble with low pass filters - you get a time delays.
Therefore you really should do as the paper suggest, and limit the upper frequency of the opamps via the feedback loops.
In fact, frequency limiting is very good design practice in any electronics circuitry. Reducing the bandwidth of any signal path to just what is needed is very good design practice, as it reduces problems with noise.

Cheers
/Ghydda

[jazzalex]

Hi Ghydda,

in the meantime I installed the missing components as suggested. This leads to the design as attached and works great for up to 5 cascades. After the 6th the same effect starts to appear, which allows to currently apply a max. of 8 cascades within my personal signal acceptance range. More stages lead to the same useless signal. However, the resulting negative shift of approx. 16 us is a nice reconstruction of what was done in the paper and currently I do not need to optimze further.

Thanks for your constructive feedback - I will get back with further news sooner or later,
best

Alex