How to amplify 0.9nV sine wave(really small signal)?

Discussion in 'General Electronics Chat' started by richiechen, Nov 26, 2012.

  1. richiechen

    Thread Starter Member

    Jan 1, 2012
    93
    0
    Hi all

    I am trying to amplify a really small signal at 0,9nV amplitude.
    The signal was obtained from a transimepance amplifier. The signal should go for a second or even third stage amplification.

    Here comes my question:
    How to amplify a such small signal?

    I have tried with OpAmps, but OpAmps seem not be able to amplify such small signals.. With 1k and 1M amplify resistance, the input current will be 0.9pA pp, which is too small for any OpAmps...

    Any idea will be appreciated.

    Thank you very much.

    Richie
     
  2. Dodgydave

    Distinguished Member

    Jun 22, 2012
    5,000
    745
  3. bertus

    Administrator

    Apr 5, 2008
    15,648
    2,348
    Hello,

    Where does this very small voltage come from?
    If we know that, we can give a better advice.

    Bertus
     
  4. WBahn

    Moderator

    Mar 31, 2012
    17,777
    4,804
    Measuring nV scale signals is not trivial. Just the thermal noise voltage produced by the thermocouples made by the electrical joints between your signal source and your first stage amplifier can easily swamp such a signal.
     
  5. richiechen

    Thread Starter Member

    Jan 1, 2012
    93
    0
    The signal was firstly from Photodiode. The output current of the photodiode has two parts: very large DC and small amplitude 1kHz sine wave.
    (DC Voltage)/(amplitude of ac)=1.1*10^9;

    Both the DC and ac part of the signal are needed. The DC will be measured through an ADC. In order not to saturate the ADC, the DC will be amplified to 5V only.

    Therefore, the ac will become 0.9nV. The signal will then go through a High pass filter. That means the output of the high-pass filter will only be 0.9nV.

    I want to further amplify the signal so the signal is large enough to be measured by a Lock-in amplifier.

    My question now is how to amplify the 0.9nV?

    Thank you very much.
     
  6. richiechen

    Thread Starter Member

    Jan 1, 2012
    93
    0
    Thanks WBahn

    The signal will be actually measure through a lock-in Amp.
    Is the noise still a main concern?

    Richie
     
  7. electron_prince

    Member

    Sep 19, 2012
    93
    3
    instrumentation amplifier. instrumentation amplifier is made from op-amps.

    [​IMG]
     
  8. richiechen

    Thread Starter Member

    Jan 1, 2012
    93
    0
    Could you please explain a little bit more?

    Or any material will be appreciated.

    Regards
    Richie
     
  9. electron_prince

    Member

    Sep 19, 2012
    93
    3
  10. takao21203

    Distinguished Member

    Apr 28, 2012
    3,577
    463
    I wonder about the replies.

    Dealing with 0.9nV is practically impossible.

    There is something fundamentally wrong with your design and/or the photodiode you use.

    In a VCR you get some 10s microvolts from the drum. And the circuits occupy large PCBs.

    I have never studied the detailed makings of such PCBs.

    But it gives me the knowledge that 0.9nV is basically nonsense.
     
  11. WBahn

    Moderator

    Mar 31, 2012
    17,777
    4,804
    Not necessarily nonsense - depends on where the signal is coming from. When I was working at NBS/NIST doing superconductor measurements, we needed to measure nV scale signals all the time. Was not easy and was not cheap.

    One chip I designed use a photodiode array to image a pump-probe laser in which the probe signal at 100kHz was eight to ten orders of magnitude smaller than the background signal of the laser that the chip was staring into. We did that by putting a lockin amplifier in each pixel.

    But the notion that the guys is gonna through some passive filters and a couple active circuits at the thing isn't going to fly.

    A 50 ohm resistor at room temperature has 1nV of thermal noise at a bandwidth of just 1Hz. At 1kHz, it would have a noise voltage of nearly 30nV.

    Then there's the noise voltage of the capacitors.

    Not to mention the input-referred noise of the opam circuit.
     
    richiechen and takao21203 like this.
  12. takao21203

    Distinguished Member

    Apr 28, 2012
    3,577
    463
    At least we have one expert here.

    Would it be usual practice for the scientists and engineers doing superconductor experiments to ask on public forums?

    And have the whole setup outlined in a thread.

    Yes you get the thermal noise, and not to speak of EMI.

    As I said, even 10uV would be difficult, for a circuit, and to describe it accurately on a forum.
     
  13. richiechen

    Thread Starter Member

    Jan 1, 2012
    93
    0
  14. richiechen

    Thread Starter Member

    Jan 1, 2012
    93
    0
    Thanks for your reply!

    The 0.9nV is at the output of the first stage TIA.
    The reason of the filters/second Amps is that I want to measure the 0.9nV signal. I want to amplify the signal first and then go to Lockin Amps.

    The signal needs to go through Band Pass filter to get rid of the DC inherited from the photodiode.

    Could you please give me some advice?

    Thank you very much.

    Richie
     
  15. richiechen

    Thread Starter Member

    Jan 1, 2012
    93
    0
    Can the 0.9nV signal be detected through a phase lock loop?
    Or what is the minimum 1kHz sine wave magnitude(after 1st stage) that can be detected through a PLL?
     
  16. WBahn

    Moderator

    Mar 31, 2012
    17,777
    4,804
    I wouldn't go that far. I learned a lot and know some of the tricks, but I can't claim to be an expert. Some of the guys there most certainly can.

    Back when I was there, the World Wide Web was very much in its infancy and all of our computers still used Sneaker Net. But I can easily imagine them asking questions on public forums. When doing applied research, you are always needing to do something that you have never done before. Some researchers take the attitude that they can do anything and don't need help, but most realize that the cost of gathering information is cheap compared to the cost of reinventing something that is likely to be quite inferior to something that is common place elsewhere. When I was working for a company designing ASICs and when I was doing research at the USAF Academy, I was always prepared to ask questions on newsgroups and forums -- sometimes it paid off handsomely and sometimes it was a waste of time, but usually it was somewhere in between with me learning something that help guide my continutes search.

    I don't think I've seen anything close to that, here.

    Would I be too surprised to see a whole setup outlined on a public forum. Hmmm. No, I don't think so. There would be lots of instances in which I would expect things to be held closer to the vest, but in many, many instances the "secret" that has to be guarded (after all, most researchers only win when they publish results before everyone else) is conceptual. So the researcher has a hypothesis about how something behaves and they are trying to take measurements to explore the validity of that hypothesis. They can often be completely open with what they are trying to measure and how they are trying to measure it without giving away much information about what their hypothesis is and how the data is going to help explore it. In many cases, they have already published their methods in previous papers and are now trying to improve or extend their measurement capabilities, so they have little cause to be secretive about their techniques.

    I remember one time my project leader and a couple of other people were trying to take some measurements and the chart recorder kept responding to some spurious signal that was causing it to flail up and down several centimeters at an erratic rate of around a couple hertz. I was standing about twelve feet away with a rubber vacuum hose in my hand waiting to ask my project leader what he wanted done with some stuff. While I was standing there I was just idly bending the vacuum hose up and down and my ears locked on to the fact that the sound of the chart recorder pin was synched to the motion of the hose. So I stopped the hose and the chart recorder stopped. I flicked the hose up and down quickly and the chart recorder responded in kind. So I go, "Ah, Jack, I think I found your noise source." The tiny bit of static charge on the hose was enough to completely mess up the measurement if it was moving, even a dozen feet away.
     
  17. takao21203

    Distinguished Member

    Apr 28, 2012
    3,577
    463
    So it's kind of "switch your iPhone off before the experiment" :)

    Actually older IR remote control receiver assemblies used to be embedded into a metal shielding. And these modules are still prone to produce spurious signals. Because their amplifiers are sensible.

    I don't think there is much hope to build such an amplifier which can deal with 0.9nV

    And if it is possible, then it would be expensive. You would need shielding as well.
     
  18. WBahn

    Moderator

    Mar 31, 2012
    17,777
    4,804
    I wouldn't want to try it unless everything (all the stuff below the microvolt level, anyway) was being done on an IC.

    We used an analog Keithley 147 and those things were like gold in the low-voltage measurement community. Keithley had stopped making them because one of the key parts was no longer available, so people would spend serious bucks to snatch up a broken one to use for parts. The digital K148 was available (for ~$4000), but it didn't have the performance we needed.

    Then, one day years later I was looking on E-Bay and discovered several K147s that were available. My first thought was -- gee, these guys don't know what they have! I called my old Group at NIST to turn them on to them and discovered that the K148s (or its descendents) were now good enough that the K147s had all been ditched or used as doorstops.

    Part of the lesson here is that we never even tried to design a circuit to work with nV scale measurements. Our job (and it was a bit of challenge) was to get that nV scale signal up out of the liquid helium and into the front end of the K147 without corrupting it. The K147 came with a special low-thermal probe that had alligator clips at the end. Our voltage probes coming out of the dewer were tightly wound twisted pairs of #32 copper wire that were soldered to pads on a PCB using indium solder. The alligator clips absolutely had to be connected so that the non-hinged jaw made contact with the pad. Then (and this was crucial, too), the whole connector block had to be packed in lead shot bags in order to establish a thermally stable environment at the connections.
     
    richiechen likes this.
Loading...