Sub diode clamping voltage

[ElectronRob]

Hi All,

Im looking for a method to clamp the input voltage to an opamp cct. I'm using my circuit to amplify very small ac voltages, my signal of interest is in the order of 10uV to several hundred microvolts.

Due to the low input voltage I am running with fairly high gains in the order of 1k-10k. My input stage is fully differential. Due to the high gain i can easily saturate the amplifier, the ultimate purpose is to amplify the input signal upto full swing input for an ADC.

Currently i have some low leakage, fairly low capacitance PMEG2010EA back to back diodes across the input. At the tiny currents they begin to conduct in the 100mV range, naturally this means that an overload on the input will saturate most of the stages, recovery from saturation is slower than I would like (ie not instant).

Im am trying to think of a way to either use a low capacitance BJT as a diode, where I could bias the input to a point just before conduction, or perhaps bias a diode with a cc source to a point just before conduction to try and get the forward voltage down around 1mV....

Has anyone come across or done anything like this? Could perhaps pass over some ideas? Looking for something with a low capacitance and naturally very low noise.

All the best

Rob 

[Manul]

Well, I suggest to not use biased diodes, because most likely it will work very bad. Reason 1 - diode curve will not be sharp enough for this job, 2 - it may introduce considerable noise, leakage, etc. You should keep in mind both voltage and current noise. Which one is more important depends on signal source impedance. I don't know about transistor clamping, need a schematic, most likely very problematic as well.

One possible solution which comes to my mind is to use ideas similar to logarithmic amplifier. In essence - some kind of non-linear feedback, so it amplifies less if signal starts to be too strong. There are some simple examples with a diode in feedback loop and some more complicated examples.

[mawyatt]

Recall from long ago with a similar problem at low frequency we made the input amp inverting and placed the input protection back to back diodes from the inverting input of the op-amp to analog ground where the non-inverting op amp input was tied, also had back to back diodes in the feedback path to keep the op-amp from saturating. Since the shunt diodes "see" virtual ground they don't conduct any current. Later in the chain after much filtering an op-amp was used to "expand" and reverse the effects of the input amp feedback back to back diodes, so back to back diodes were placed on the input of this later in the chain op-amp.

Best,

[KT88]

For such low-level signals I would be reluctant to to put the limiter at the input. A two stage design with lower gain for the first stage would allow for clamping the output of the first amp. The second stage could be a less expensive amp as well. That way you won't need any exotic parts...
The influence of the offset of the second stage would be reduced by the gain of the first one, noise even more as it contributes only by a root sum square factor.

Cheers
Andreas

[ElectronRob]

Hi Manul,

Thank you for the reply. Sorry I should have said. The product has been in use for several years with back to back diodes, im at around 3nV sqrt Hz. I am also concerned about biasing diodes due to variations and also thermal drift etc.

A bjt used in place of a diode would have a sharp curve but given I am using a single supply I am struggling to come up with a nice way to bias anything, especially without hand selecting parts or post production trimming.

I like the idea of a logarithmic amplifier, I had overlooked them since I actively want a linear response however I suspect you could (as you point out) use a diode to create an alternative, log gain path.... hmmmmm. Thank you.

[ElectronRob]

Thank you,

I need to digest your reply a little with a pen and paper!

[ElectronRob]

Hi KT88,

Thank you, fortunately this is fully differential through out so offset hasn't been an issue. You are right, my first stage has a gain of only 10 and has been selected carefully so its input impedance matches the source well. I do also have protection between the first and second stage, this is much less of a problem due to the amplified signals.

The fight here is that the very small signal is created by quite a large transmit event, this is a similar problem to sonar however with a separate transmitter/receiver and over a much shorter distance. The transmitted signal is quite large and attenuated heavily, the received signal is small however the large transmit saturates the receiver amplifier.

So I might see 'crosstalk' of several volts on the input. I am able to clamp this to 100mv or so with back to back diodes however I suspect that also adds to a slow recovery time as current circulates around the diodes.

It is very likely that it is the second stage saturating rather than the first, I have a feeling that regular diodes on the input for esd protection combined with back to back diodes in the feedback path of the second stage might well be a winner here.....

It all gets a bit sticky because the output of my amp goes to a variable gain sampling head that I don't have much control over. This typically with have a gain of between 10 and 100 however that at least does seem to recover quickly.

Apologies, im really just rambling here.

Thanks all

[KT88]

It would be interesting to know the frequency (-range) and the source impedance. For single-ended signals an opamp implementation of a circulator exists. Another 'trick' would be available if you use a transmission line with controlled impedance. In this case you could use the principle of the so called hybrid that is used in 1000BASE-T phys. They use a voltage mode driver with the (split-) termination in series. A 180-degrees tap at the transmissin line cancelles the Tx from the Rx...

Cheers
Andreas

[jbb]

I suspect you want to keep the input signal clamping nice and simple. Anything you do with diodes or transistors will have tolerances and temperature drift.

Have you established which stage is taking the longest to come out of saturation? Maybe you only need to modify one of them for adequate results...

[Marco]

You're not going to get good clamping at those levels without quite a bit of amplification. How about some fast CMOS opamps? They have near instant recovery. It's the BJTs and capacitors which get ya.

[ElectronRob]

Hi thanks,

I need to check my notes for source impedance however from memory it ends up at around 100ohms, rising to several k at 200Khz.

I am working from 50khz to around 2Meg. Circulator sounds interesting. Early on in dev I did try various switching schemes, one which worked fairly well was a clamping fet driven by a high gain opamp, like a mute. However there maybe signal of interest during the transmit, unfortunately most tx/rx style switches just add too much noise.

Ultimately I am headed towards feeding in a small portion of the transmitted signal 180deg out of phase however this is quite a challenge to get right across a wide frequency range and with other varying factors.

All the best

[ElectronRob]

Yes thats a possibility, I should add that I am recovering within a 1us or two which isn't bad, but would prefer to get closer to a few hundred nS or ideally not at all!

I am using a bjt input opamp deliberately due to the source impedance, they are also easier to get as low noise devices. I am reluctant to go to discrete components because I know matching will be a pita during production.

[ElectronRob]

Looks like its the input stage saturating. I'm going to lower the first stage gain and add a little to the second stage to compensate, I've also increased rail voltage from 10v to 12v in the hope that buys me a little extra headroom. I have diode clamps between the first and second stage so probably won't make things worse.

Its all a bit of a trade off, these low noise opamps typically don't have fantastic gain bandwidth so I'm having to spread the gain across stages. All good fun 

[ElectronRob]

Had a think on your suggestion, seeing as I'm at home and the gear is in the office I've run a few TINA sims. I like it, adding back to back diodes within the feedback cct of the output (2nd) stage does a nice job of keeping the output within reason.

I'm having boards made next week so I'll add them (can always remove if they don't work as intended). Off to find some ultra low capacitance diodes me thinks!

I have back to back diodes between each stage, plan is to make a step down transformer so I can use a signal generator on the input of the amp, hopefully I can get a better picture of whats saturating etc, and by selectively removing diodes perhaps see if they are making life worse.

PS like the 993's(?) in your profile pic.

Thanks very much all.

[David Hess]

Check out how Tektronix implemented fast overload recovery on their 7A13 vertical amplifier.  It uses feedback from the output back to the input when the output exceeds a defined level which prevents the earlier stages from overloading.  Full documentation including circuit descriptions are available.

[PartialDischarge]

Check out how Tektronix implemented fast overload recovery on their 7A13 vertical amplifier.  It uses feedback from the output back to the input when the output exceeds a defined level which prevents the earlier stages from overloading.  Full documentation including circuit descriptions are available.

I think I've said before but I'll say it again: most people do not appreciate the excellent and amazing technical content that Tek and HP have developed and published over the years.

[ElectronRob]

Thank you chaps,

Funnily enough I spoke with my father about this topic yesterday - he is an old school EE with a ridiculous collection of Tek/HP gear from the 70s-90s and often says very similar things regarding the information available in service manuals alone from both. He claims to be digging out some manuals for a few HP spectrum analysers which have nice input protection/clamping circuits. I appreciate the 7A13 recommendation and will certainly be doing some research this weekend. 

All the best