Advertisements

Victoreen 710-104 Ionization Chamber: LMC6081 Electrometer Amp

The simplest possible electrometer amplifier that might work:

Electrometer amp - LMC6081 schematic

Electrometer amp – LMC6081 schematic

The general idea is that the op amp will drive the (essentially) open-circuited inverting input to match the 2 V offset at the noninverting terminal, so that the output will stabilize above the LMC6081’s minimum useful output voltage (of about 1 V) and the gamma-ray pulses will go downward from there (it’s an inverting amp). The rebiasing network downstream from the output cap doesn’t appear in the hardware.

The small cap across the feedback resistor that would compensate / roll off the high-frequency response isn’t possible, unless you have a stockpile of teeny glass vacuum capacitors: the leakage resistance must be far more than 100 GΩ and my collection lacks anything like that. A Teflon-insulated gimmick capacitor wouldn’t be stable enough and would probably still leak crazy current.

Normal electrometer amps operate at essentially DC and amplify an actual current from the ionization chamber. In this case, the radiation level is zero, there’s no chamber current, I’m looking for small pulses generated by gamma ray events, and the amplifier must have reasonable AC response. It’s not clear that circuit can work, but it’s a starting point.

It turned into a hairball:

Electrometer Amp - 10 Gohm Rf

Electrometer Amp – 10 Gohm Rf

The LMC6081IN is socketed, with the inverting input pin bent outward and soldered to the flying junction of the 100 kΩ input / OMG resistor (which prevents inadvertent shorts from the 24 V chamber bias) and the glass-body feedback resistor. The socket sits end-on in a puddle of epoxy, so I can swap op amps as needed.

The 100 GΩ feedback resistor didn’t work well at all:

LMC6081 100 G - out - noninv level - matched

LMC6081 100 G – out – noninv level – matched

The lower trace shows the 10.2 VDC offset at the noninverting input, which is what it took to drive the output from near 0 V to the 11.4 V in the upper trace, with no proportional change in between. I replaced the schematic’s 1 MΩ / 220 kΩ resistors with the 20 turn trimpot shown in the photo in order to find that bias condition. Obviously, the op amp acts as an open-loop comparator, not a linear amplifier, and no amount of waiting for it to stabilize would change that outcome.

It’s possible the resistor has failed open, but, frankly, the difference between “100 GΩ” and “open circuit” probably doesn’t amount to much. Note, however, that there’s absolutely no 60 / 120 Hz interference or noise, which continues to surprise me; removing the shield cap and teasing the twiddlepot slams the output with a 60 Hz trapezoid:

 

LMC6081 100 G - out - noninv level

LMC6081 100 G – out – noninv level

Swapping in a 10 GΩ resistor produced a smoothly changing output for biases between about 1 V and 8 V, so it’s behaving like an op amp should. Setting the non-inverting terminal to 8 V puts the output voltage at 6.3 V, which means the 10 GΩ resistor drops about 1.7 V to pull 170 pA from the inverting input (which is presumably at 8 V give-or-take a bit) and the chamber electrode. The LMC6081 spec says a maximum of 4 pA (for the I version, which is what I have), so:

  • My cleanliness isn’t up to par
  • The chamber delivers quite a bit more zero-radiation current than I expected
  • The op amp’s input got toasted despite my efforts toward a static-free installation

Hard to choose among those options, it is, indeed.

With the bias set to produce a 6 VDC output, the AC coupled signal doesn’t seem promising:

LMC6081 10 G Rf - out and Vplus in - AC 2 mV div

LMC6081 10 G Rf – out and Vplus in – AC 2 mV div

The lower trace is the bias voltage applied to the noninverting input, which looks reasonably clean. The sweep triggers from the power line; there’s still no 60 Hz interference.

All those flying components are, as you’d expect, microphonic beyond belief: jumping on the concrete basement floor produces a corresponding bounce in the trace that may be due to air currents or noise, for all I can tell. Even with the chamber sitting on a loose cloth pad, tapping the workbench produces 10 s of slowly decaying oscillation, admittedly at a much lower frequency than the noise in that trace.

A single gamma ray event producing an unreasonably high 10 fA chamber current will cause a downward (it’s an inverting amplifier) pulse that amounts to a mere 100 µV, a pulse that obviously isn’t visible against all that racket. You might convince yourself that the event at the center of the top trace comes from a gamma ray, but you’d probably be wrong.

In a normal electrometer amp, a stiff low-pass filter discards all the noise and isolates the DC signal corresponding to the steady-state chamber current from the ionizing radiation. Given that the pulses are on the order of 5 ms wide, there’s no obvious way to discard most of the noise without also tossing the signal.

Pending more thought, I’d say this was definitely fun while it lasted…

 

 

Advertisements
  1. #1 by madbodger on 2015-09-22 - 09:17

    As you’re no doubt aware, you’re pushing your luck with simple amplifier circuits here. I’m going to go with leakage (your cleanliness is probably a factor, as well as that socket). Another issue with transimpedance amplifiers like that is Johnson noise (it’s proportional to resistance so 10G is going to create a lot of equivalent noise). You could try a chopper-stabilized op-amp like a TLC2652A, but I’m not at all sure the op-amp is the problem, but the lack of drift could remove one factor. The reference to the “OMG” feedback resistor makes me laugh. I’m reading this avidly, as one of my projects will also require this sort of heroic amplification (very small currents through 500Ω in my case).

    • #2 by Ed on 2015-09-23 - 09:02

      pushing your luck with simple amplifier circuits

      That’s putting it mildly; I’m bulldozing my luck right over the edge…

      What makes it absurd is my childish belief that a trivially simple circuit can combine infinite input resistance with at least a few kilohertz of bandwidth to show minuscule pulses against a (reasonably) quiet background. I can get any two of those five goals …

  2. #3 by Red County Pete on 2015-09-22 - 09:51

    It looks like it’s pretty close to a dead end, but a couple of thoughts for reference. Cleaning the op-amp before assembly might(!) help. In a plastic package, I’d consider a surfactant in deionized water, or the best equivalent. In the stone ages, we used Alconox in DI, but Dawn dishwashing liquid has a good reputation as a degreaser, and steam distilled water should work if from a fresh bottle. I believe you have an ultrasonic cleaner; put the parts in the solution in a small beaker and run that mess in the ultrasonic for a while. We had a collection of 50 and 100ml beakers at work for small parts. (Water was the working fluid in the ultrasonic’s main chamber.) Rinse and air dry.

    (In the semiconductor lab at school — circa 1973, we used the following in order:
    Trichloroethylene [AKA TCE] — nonpolar solvent, good degreaser, good ozone layer depleter, great at aquifer destruction
    Acetone — in the middle between non and polar
    Alcohol — I think propanol, pretty polar
    DI water rinse. )
    [I suspect TCE replaced carbon tetrachloride. 1,1,1 trichloroethane (TCA) replaced TCE in the mid ’70s. I’ll skip the aquifer stories.]
    The first step removed the grease, while each succeeding step got rid of residue from the earlier one, in theory. Each solvent sort of mixed with the ones next to it.

    Putting the input in the air is good, but I’ve seen a few applications where the part was attached to the substrate in dead bug fashion and everything air wired. I’d use double stick tape. The TI datasheet also talks about the guard ring on a PC board, but sheesh.

    Not sure how one would check a 100G resistor; sounds like mating porcupines.

    • #4 by madbodger on 2015-09-22 - 09:56

      To check a 100G resistor, I’d put it in series with an ordinary 10M DVM set on the 200mV scale and put 1000V across the stack. The DVM should read 100mV. Essentially the same setup we used to use to measure capacitor leakage.

      • #5 by Red County Pete on 2015-09-22 - 10:53

        should read 100mV

        Er, I wouldn’t count on that, not until I knew the input impedance of the DVM. I’d use the 10M resistor and a low voltage supply to find that value. Set up the supply to put out a bit less than 200mV, then measure the voltage with the 10M in series. I wouldn’t count on the impedance being constant across DVM ranges.

        Once you have the input impedance, you’d use that value in parallel with the 10M. Being paranoid*, I’d start the 1000V stack measurement at the 2000V range and work my way down.

        The old analog VMs I grew up with had nominal 10M inputs at best, but that was decidedly approximate. A 1000V supply on a test bench gives me a strong case of the yikes. Still, porcupines have babies.

        [* National Lampoon’s Deteriorata “You are a fluke of the universe. You have no right to be here. Whether you know it or not, the universe is laughing behind your back.”]

      • #7 by Ed on 2015-09-23 - 09:30

        and put 1000V across the stack

        That’s easy for you to say!

        Once upon a time, we had an HP 3456A DVM in the lab: 10 GΩ input resistance, 100 nV / 100 µΩ resolution, six significant figures, true RMS, Kelvin inputs. These days, I’m happy with 100 mV resolution. [sigh]

        • #8 by steve on 2015-09-23 - 11:57

          Not to brag… but I’ve got 4 or 5 HP 3456As… 3 of which I tend to believe (all 3 read pretty close, so, good enough for the girls I hang out with). I doubt I paid more than $50 of any of them (not including shipping) and swapped parts around until they behaved. Oh, and taking apart the front panel to keep the keys from sticking.

          I didn’t realize they had 10 G Ohm input resistance; that parameter was never important to me (until now!)

          Steve

          • #9 by Ed on 2015-09-23 - 12:48

            swapped parts around until they behaved

            To judge from the anguished wails heard on the forums, you should suck the bits out of the known-good ROMs and tuck them away in a safe place: they’ll come in handy one of these days. By then, of course, you won’t be able to get old-school UV EPROMs to replace the bad ROMs…

    • #10 by Ed on 2015-09-23 - 09:19

      I’ll skip the aquifer stories

      IBM will be pumping and filtering the plumes under their (former) East Fishkill chip fab / substrate factory forever

      I’m definitely going to do a better job cleaning the chip, although I suspect it slides out of the shipping tube about as clean as I can possibly make it; the real problem lies elsewhere.

      • #11 by Red County Pete on 2015-09-23 - 11:04

        One of the many superfund sites in Palo Alto belonged to the local dry cleaners, though our old plant was the worst. And the powers that be insisted above ground tanks were ugly/fire hazard, so they got buried. And leaked.

        I have my doubts about residue from plastic IC tubes. The extrusion lubricants might not be present anymore, but it’s not the way I’d bet. Tape and reel should give a cleaner part. On the gripping hand, the crud should respond to a not-so-intense clean.

        • #12 by hexley ball on 2015-09-23 - 11:39

          “…they got buried. And leaked.”

          Not long after the powers-that-be realized their earlier mistake and changed the rules, I toured a new fab facility in Santa Clara (back when people were still building new fabs in Silicon Valley!), and the guide told us that all the piping now ran overhead inside the building hallways. His theory was that the rulemakers believed that if nasty stuff was dripping on the heads of the executives, they would tend to fix it right away.

          Same guide also said that the effluent from the plant had been processed so thoroughly that you could drink from the sewer line… But as RCP noted above, this was the heyday of 1,1,1 trichloroethane — and I’m betting that sewer line had its share. :-(