Trying to understand Vos in Op-Amps for a High bandwidth uCurrent

[BrunoGelli]

Hello, Guys!

I tend to write a lot, so I'd like to provide a short (tl;dr haha) version of my question first: In the uCurrent design, a big effort is put in finding a low V_os Op-Amp. Is it possible to use another Op-Amp (with higher bandwidth, for exemple) and just deal with the offset?

Now for the longer version!

First, here is some context: I work with particle physics and I'm often using Photomultiplier tubes and plastic scintillators. We measure the PMT signal using a decoupling capacitor, so the DC high voltage does not blow up our measuring devices.

But this capacitor deforms the waveform of our signals, which is a big pain in the butt when making time resolved measurements, e.g. measuring the temporal emission spectrum of a plastic scintillator.

This limitation forces us to caracterize the impulse function response (IRF) of our photodetectors with really short light pulses, which is expensive and not so easy to do. (Using a picosecond pulsed laser or a synchrotron light source, for example)

One solution I thought was to measure the current going into our device. I thought I could use a shunt and measure the voltage difference over it, just like the uCurrent idea. But for that, I would like to use a faster Op-Amp so I could resolve the fast signals.

Is it possible to design a uCurrent using a different Op-Amp with higher V_os and just deal with the offset? I mean, let suppose we have a 100 uV offset, which would be multiplied by 100, giving a final offset of 10mV. Lets suppose now that my current produces a 200uV drop in the shunt, which translates to a 20mV signal. It would be possible to see the signal clearly over the offset, even if the signal was smaller.

I was thinking about using the LMH6629 for its high bandwidth, and low noise. (and also because we have a reel of those laying around haha) Other option is the AD8009, although the V_os of these one is perhaps a bit too high.

Maybe I'm missing something really obvious, I'm just starting to dive into electronics.

Thanks a lot for reading it! Feel free to call any and every mistakes in my train of thought

P.S.: Some information about my system. It feeds 1600 V into the PMT at 250 uA base current (to drive the voltage divider of the PMT) and the signals should produce a 20 nA current at least, going easily up to 2uA.

[bdunham7]

Yes, and if you don't care about offset, you can just use an AC-coupled design.  Much of the difficulty with instrumentation amplifiers is the general requirement that they go to DC.  Take that off the table and lots of problems go away.

[magic]

It's a bipolar device so mind input bias current, which flows out of the input pins and has to go somewhere. That somewhere is likely your shunt resistor, together with your signal current.

Mind input current noise. It's specified at 2.6pA/rtHz which means 2.6nV RMS or ~15nVp-p over 1MHz bandwidth or 3 times that over 10MHz bandwidth and so on. The first 1MHz will be much worse due to flicker noise, but you can cut it out by AC coupling.

You may prefer a fast JFET or CMOS opamp perhaps?

BTW, if you don't care about low frequencies and DC, it's possible not to amplify DC offset of the amplifier by AC coupling the bottom resistor of the feedback divider. This is widely used in audio amps to cut their gain below a few Hz.

[Terry Bites]

The simple solution is to combine the speed of one opamp with the precision of another. https://www.google.co.uk/url?sa=t&rct=j&q=&esrc=s&source=web&cd=&ved=2ahUKEwjr6Ym2vPnxAhWMbsAKHS0bCOoQFjABegQIBRAD&url=https%3A%2F%2Fwww.analog.com%2Fmedia%2Fen%2Ftechnical-documentation%2Fapplication-notes%2Fan21f.pdf&usg=AOvVaw0LFWZi9ivKCFdMpmxUeSDm.

There is plenty of material on "composite amplifiers" out there. Have you looked at phototranistor/ diode amplifiers. High speed tansimpendance devcies for fast small current meaurements.
eg Texas Instruments OPA855/OPA855. It is common practice to measure the PMT output with a charge sensitive amplifier (CSA) rather than a voltage amplifier. This integrates the current pulse from the PMT and you set the dt over which you take your measurement.

If the the waveform distortion is systematic and you can charcaterise it, then you can compensate in hardware. Pulse shaping techniques are invaluable. Capacitors are not the all the same and choosing the right type and value may be crucial for your design. CR coupling circuits act as differentiators. Parasitc LCand Rs in your circuit and wiring and cables will cause  problems at high speeds if unaccounted for.

https://www.google.co.uk/url?sa=t&rct=j&q=&esrc=s&source=web&cd=&cad=rja&uact=8&ved=2ahUKEwjFheb1wPnxAhUOQUEAHRRPDn4QFjABegQIDBAD&url=https%3A%2F%2Fwww.eevblog.com%2Fforum%2Fprojects%2Fcharge-sensitive-amplifier-for-photomultiplier%2F&usg=AOvVaw2H_p4hP_JCqWeusMiVdRlA this may well be of interest. The coupling and linearisation of PMT outputs as discussed in detail. https://www.google.co.uk/url?sa=t&rct=j&q=&esrc=s&source=web&cd=&ved=2ahUKEwirmfipwPnxAhWsQkEAHYSJATsQFjAAegQIBxAD&url=https%3A%2F%2Fwww.hamamatsu.com%2Fresources%2Fpdf%2Fetd%2FPMT_handbook_v3aE-Chapter5.pdf&usg=AOvVaw0nqYUeBDMoRxuQRcpOrkWz

[David Hess]

Is it possible to use another Op-Amp (with higher bandwidth, for exemple) and just deal with the offset?

Sure you can change the operational amplifier, and then the offset can be either trimmed out or measured and subtracted from the measurement.

[BrunoGelli]

Hmm, interesting! That is what I was thinking about. I was only afraid this V_os could change overtime, like faster than one could trim it out.

I follow it with a newbie question: If it can be trimmed out, why the manufacture does not correct it?

Thanks a lot!

[BrunoGelli]

That is excellent! I really do not care about the DC part... and, to be fair, the DC level would only bring more trouble for the data analysis.

This DC decoupling is easy, right? Just put a cap to block the DC and I'm golden? This sounds too good to be true, but I guess we get lucky sometimes

Thanks, bro!

[magic]

For high frequency AC you really should use a configuration where DC offset is not even amplified in the first place.

Manufacturers do trim the DC offset by means varying from internal fuses or laser trimmed resistors (inaccessible after package molding) to internal fuses accessible externally after packaging (rare). Such parts are inevitably more expensive.

Trimming gets you only so far. Offset may still change due to package molding, slowly over device lifetime and with temperature variations too.

[BrunoGelli]

I have never heard about composite amplifiers before. That is a game changer! Thanks for the tip

We do use some real nice charge amplifiers from CAEN, but for this application we are interested in the full waveform, i.e. the temporal emission spectrum of our scintillators. They should emit light (from an instantaneous excitation) following an exponential decay law with a certain decay coefficient, but there are some evidences that this may not be true all the time.

We can calibrate and correct this effect with some effort while post-processing the data, but there are several downsides with that strategy.  For exemple, as the Impulse Response Function (IRF) distortion acts in the Fourier space, a deconvolution of the division (distorted signal(w) /  IRF(w)) gives back the undistorted signal. But having so may DFTs in a finite time acquisition window gives rise to an annoying ringing on top of the signal... Long history short: Perhaps I should be using a SiPM or an APD for this measurements 

[BrunoGelli]

Oh, I see! That is really interesting. I have access to a Focused ion beam microscope... It could be an interesting project to delid an Op-Amp and try to trim it (or, more likely destroy it while focusing the beam hahaha)

Thanks for sharing your knowledge!

[magic]

It probably voids the warranty

Only metal can and ceramic DIP packages can be opened easily, and those are somewhat rare these days.
As for plastic, very few strong acids can dissolve it without damaging the IC.

[David Hess]

Hmm, interesting! That is what I was thinking about. I was only afraid this V_os could change overtime, like faster than one could trim it out.

Precision parts also have low offset voltage drift with time and temperature.

I follow it with a newbie question: If it can be trimmed out, why the manufacture does not correct it?

Manufacturer's do, but it takes time and time costs money.  The parts can be graded into different categories of offset and dirft, or they may be trimmed to reduce their offset and drift, or some parts allow the user to trim the offset and drift via "offset null" terminals.