Help designing a low-noise TIA

Hi all,

I am posting to this forum at the suggestion of George Herold from Teachspi n, he says you are the people to help me out. I understand a decent amount about circuits, but not enough to design the low-noise TIA I would like to build.

We're trying to create a modern version of the Tolman-Stewart experiment ht tps://en.wikipedia.org/wiki/Stewart%E2%80%93Tolman_effect that was one of t he first proofs that electrons inside metals carry the current.

In this experiment, a coil of wire is spun to high speeds and then braked r apidly. The electrons keep moving and create a small pulse of current. Or iginally, Tolman and Stewart used a ballistic galvanometer to act as a char ge amplifier and integrate the current to find the total charge.

I?d like to use a TIA to convert the small current pulse into a vol tage, then record that voltage as a function of time. The problem is that the coil acts as a giant antenna and picks up all sorts of unwanted noise, so I?d like to get rid of that noise. In particular, it is really good at finding 60 Hz signals in the room.

Right now we?re using an OPA 140 with 1 GOhm and 10 pF as a feedbac k resistor and capacitor in parallel. We attach the coil (about 200 Ohm re sistance, 500 mH inductance) to one input and put the other input across 20

0 Ohms to ground. The large amplification leads to huge amplification of t he noise, and it is hard to see our signal. We expect the current pulse to be 1 nA of current, almost square wave in shape, and it should last the du ration of the braking, about 0.5 seconds.

Any suggestions appreciated!

--Matt Sullivan Ithaca College Physics

Reply to
ithacacollegephysics
Loading thread data ...

One approach would be to put another winding around it, electrostatically shielded inside and out using copper tape, and hang the TIA on that (inside the shield). You could get the coupling coefficient up to probably 0.5, and it's easily measured, so it wouldn't hurt the accuracy much.

The shield would look like a shorted turn , so you use two half-turn shields with a nice big overlap , insulated with kapton tape. Some capacitors going across the gap would help too.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs

On Tuesday, June 4, 2019 at 3:28:55 PM UTC-4, snipped-for-privacy@gmail.com wrote :

pin, he says you are the people to help me out. I understand a decent amou nt about circuits, but not enough to design the low-noise TIA I would like to build.

formatting link
that was one of the first proofs that electrons inside metals carry the current.

rapidly. The electrons keep moving and create a small pulse of current. Originally, Tolman and Stewart used a ballistic galvanometer to act as a ch arge amplifier and integrate the current to find the total charge.

oltage, then record that voltage as a function of time. The problem is tha t the coil acts as a giant antenna and picks up all sorts of unwanted noise , so I?d like to get rid of that noise. In particular, it is reall y good at finding 60 Hz signals in the room.

ack resistor and capacitor in parallel. We attach the coil (about 200 Ohm resistance, 500 mH inductance) to one input and put the other input across

200 Ohms to ground. The large amplification leads to huge amplification of the noise, and it is hard to see our signal. We expect the current pulse to be 1 nA of current, almost square wave in shape, and it should last the duration of the braking, about 0.5 seconds.

Hi Matt, Thanks for asking here... I guess I'm hoping that this will give you more ideas, than just you and I going back and forth on email.

The OPA140 looks fine. 1G ohm and 10 pF is a 10 ms time constant..

But your big problem is the 60 Hz pickup from your coil?

How big is your coil (diameter)? Do you see signals from the coil spinning in the earth's B-field? I was browsing this,

formatting link
(from your wiki link) and it looks to be a big coil. But hard to tell.

We do a trick in our Earth's field nmr where there are two coils in series (but wound in opposite directions.) A many turn inner coil that picks up the nmr signal. and a much bigger outer coil, with the same turns*area as the inner coil. And this picks up the same emf from the AC 60 Hz as the signal coil... and cancels it. I think JR will take a tur n or two off the inner coil to try and match the cancellation.

George H.

Reply to
George Herold

Great paper. It really gives you a feel for how physics was done back in the days before physics was "done."

I wonder what was responsible for the rapid variations in the local geomagnetic field that they had to compensate for. They say that they could detect the effect of cars driving by, but they also say that separating the compensating coil(s) from the experiment had little effect, suggesting that the fluctuations were coming from a significant distance.

I guess they could have been seeing the effects of solar activity, measuring the K index a couple of decades before the term caught on.

-- john, KE5FX

Reply to
John Miles, KE5FX

If the 60Hz getting in is E-field there is also the trick of winding with coax and grounding one side (and only one).

However, much of it will be magnetic and then Matt can only notch it out. If the detection is done in software notch filters are easy. If analog you'd almost have to use a switched-capacitor filter for that. Very likely several notch filters are required because when 60Hz is gone Matt will discover that there's also a lot of 180Hz and 300Hz. Also, make sure there is absolutely no 60Hz gear that turns on and off at randon, such as electric cooktops, because that's next to impossible to notch out.

--
Regards, Joerg 

http://www.analogconsultants.com/
Reply to
Joerg

The source impedance of the voltage pulse will be low, so you may not want a high-impedance TIA. The coil is more of a 200 ohm voltage source, about 200 nV.

If the braking lasts a half second, a lot of the noise could be filtered out. Or brake faster and get more signal.

I don't understand your circuit, where the 200 ohm things go. Can you post the schematic?

You'll have two sources of noise, electrostatic and magnetic. The electrostatic can be shielded.

Is it possible to split the coil into two sections, wired so that the voltages add for the deceleration but cancel for external fields?

--
John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  
 Click to see the full signature
Reply to
John Larkin

Huh.. I'm going to have to read it.

I've got optical pumping coils/ apparatus, that I take 'down home on the farm' to test.... In the living room I can see the fish tank pump spinning, it's a permanent magnet.

I test 'em in a bedroom; I can totally see a car pull into the nearby drive-way.

There's some daily variation in the B-field... I never tried to measure it. Heading back to read about B-fields in 1916, I can only think things have gotten worse since then.

George H.

Reply to
George Herold

spin, he says you are the people to help me out. I understand a decent amo unt about circuits, but not enough to design the low-noise TIA I would like to build.

formatting link
that was one o f the first proofs that electrons inside metals carry the current.

d rapidly. The electrons keep moving and create a small pulse of current. Originally, Tolman and Stewart used a ballistic galvanometer to act as a c harge amplifier and integrate the current to find the total charge.

voltage, then record that voltage as a function of time. The problem is th at the coil acts as a giant antenna and picks up all sorts of unwanted nois e, so I?d like to get rid of that noise. In particular, it is real ly good at finding 60 Hz signals in the room.

back resistor and capacitor in parallel. We attach the coil (about 200 Ohm resistance, 500 mH inductance) to one input and put the other input across 200 Ohms to ground. The large amplification leads to huge amplification o f the noise, and it is hard to see our signal. We expect the current pulse to be 1 nA of current, almost square wave in shape, and it should last the duration of the braking, about 0.5 seconds.

Yeah, Matt wants to measure current, but I'm not sure a TIA is the answer. (I think he has 2 x 200 ohms, but I'm not sure.)

I was thinking of a gated integrator... which I haven't used in years. You could also trigger a DSO and average. (which is like a gated integrator, with more time information.)

George H.

Reply to
George Herold

A ballistic galvonmeter measured current, but that doesn't mean we have to. I think the physics produces voltage.

--
John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  
 Click to see the full signature
Reply to
John Larkin

At school we only ever used them to measure charge. They integrated current over time of a current pulse, where the pulse's duration was a short fraction of the ballistic galvanometer's response time.

Never saw them being used to measure current; ordinary galvanometers were sufficient for that.

Reply to
Tom Gardner

WOrks.

--
John Larkin         Highland Technology, Inc 

lunatic fringe electronics
 Click to see the full signature
Reply to
John Larkin

Hi all,

Thanks for the replies! A lot to think about.

If you want to see our setup, here is a picture:

formatting link

Here is the simple circuit we are working with now:

formatting link
In the simulated circuit, we've added a current source to act as the electr on motion upon braking the coil.

Phil, I am not sure I follow your suggestions, sorry. We'd put another coi l outside the rotating coil to measure the signal from the rotating coil? Elaborations appreciated!

George, our coil is about 8" in diameter. I remember the Earth NMR from th e Food Truck when it came to Ithaca College. If you look at our setup, we have something very similar: we have a second, stationary, coil directly be low the rotating coil, and this coil is counter-wound. This definitely hel ps to reduce the 60 Hz noise (if there is other noise, it is drowned out by 60 Hz noise). At the time I thought our methods were essentially identica l. Do you think a larger stationary coil surrounding the rotating coil wou ld work better than a stationary coil of the same diameter beneath the rota ting coil?

Jeorg, right now we reduce the remaining 60 Hz noise by simply averaging ou r voltage signal over 17 ms, so we get one data point for every 60 Hz cycle . This does reduce our noise. I planned to try to remove the 60 Hz noise via FFT and then fiter out the high frequencies, but I wanted to try and fi nd an experimental solution first.

John Larkin, braking faster would be hard. The coil has about a pound of c opper wire 8 inches from the rotation axis, and we spin it up to 6000 or 70

00 rpm. It has a lot of angular momentum! Half a second is the best we've been able to do. And I am afraid we've worked pretty hard to get it spinn ing down even that quickly. As for the split coil, we have essentially don e that with the rotating coil and the stationary coil. They can't both be rotating, or the signal we are hoping for cancels out (the electrons in the two coils are counter-rotating, so you get two pulses of current in opposi te directions).

George, we did try a charge amplifier at first, which as far as I understan d is the electronic analog of the ballistic galvanometer. But it was hard to get the op-amps to be stable and still be able to integrate over a whole second. So I figured we have much faster electronics now, so to get the t otal charge I can just integrate the current, so that's why we are trying a TIA now.

Reply to
ithacacollegephysics

R2 just adds Johnson noise. That opamp is fairly noisy too.

The DC gain of that circuit is about 4e6, and the typical input offset of that opamp is 30 uV. Multiply those!

Is the amplifier and its batteries rotating on the coil? It would be a lot easier to export volts than nanovolts.

Optical coupling would be fun to get the signal off the spinning coil.

I still think it would be better to use a voltage amplifier rather than a TIA. Low noise diffamp.

--
John Larkin         Highland Technology, Inc 

lunatic fringe electronics
 Click to see the full signature
Reply to
John Larkin

We can remove R2.

Why is the gain only 4e6? Because we are using a current source, if we hav e a 1 nA pulse we should see 1V on the output, so a gain of 1e9 V/A.

What do you mean by multiply those?

The circuit sits on the table -- we pull off the raw signal from the rotati ng coil. We'd have to very carefully balance everything to get the amplify ing circuit and batteries rotating at 7000 rpm.

Pre- or post-amplification?

How would this work for such a low current? And if you put that current ac ross any resistance, it drops the current generated by the spinning coil. So best case scenario, it's across, say, 10 Ohms, and then you have a volta ge of 10 nV. A low noise diff amp can find that 10 nV signal and amplify i t and not generate noise? Happy to try it -- are there any schematics I ca n look at, or a section in Horowitz and Hill?

Reply to
ithacacollegephysics

You are not using a current source. You are using a coil of wire.

Can't simplify that. Multiply them.

--
John Larkin         Highland Technology, Inc 

lunatic fringe electronics
 Click to see the full signature
Reply to
John Larkin

Move the Apple monitor and the whole computer away from back there. They could be a source of line frequency noise. Also a whole lot of other noise (back light, long traces behind the screen). Nothing should be plugged into wall outlets there either. As a test turn off the ceiling lights in the room to see if that lowers the noise. Older style lighting has magnetic ballasts which leak some magnetic field.

A brute force option would be to pack the whole coil setup into a mu-metal box. If you bend that srt of metal you must anneal at the bends or it can lose much of its shielding properties. Of course, if this is a demo setup that would take away the "live" look and feel.

Like John suggested you could get rid of R2. 260ohms isn't going to make much of a difference though.

The gain is a bit high, factor of 1e9 current-to-voltage. Make sure noise does not rail the amplifier. I would try to get by with the minimum in gain and go digital as soon as possible. I such cases I often use the Behringer ECA-224 USB audio interface but not sure if it goes far enough towards DC for you. Probably it does and you could SW-compensate for the roll off.

Digital opens a lot of filtering opportunities and the use of pre-canned C-library code. However, you can't go into that from an opamp that is supplied +/-15V. Too much, and if you have an accidental big pulse the audio input gets fried. The supply voltages would have to be reduced and the gain as well.

I have seen such averaging. It sort of works but not very well. Too much residual line frequency noise. If this has to be analog you can use a switched-capacitor notch filter but if at all possible I'd do it on a PC. For quick experiments the best would be a laptop running only on its battery, no charger connected. This is because such chargers often introduce ground loop noise.

Listen to the ceiling when in that room. If you hear a muffled hum at times or all the time there may be a big air diverter motor operating. Those can spew 60Hz around. You could talk to maintenance if it's possible to turn that off for a while. Same below if this isn't on the ground floor and check what's on the other side of that white brick wall. Such walls do not shield 60Hz at all.

--
Regards, Joerg 

http://www.analogconsultants.com/
Reply to
Joerg

R1 and L1, in series with a current source, have no function in that circuit.

A coil is not a current source. If you short it with a low-impedance device, a galvo or a TIA, you can pretend that it is.

--
John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  
 Click to see the full signature
Reply to
John Larkin

On Wednesday, June 5, 2019 at 12:21:19 PM UTC-4, snipped-for-privacy@gmail.com wr ote:

ave a 1 nA pulse we should see 1V on the output, so a gain of 1e9 V/A.

So at DC (ignoring inductor and 10 pF cap) you've got 1 G ohm feedback R, and ~250 ohms to ground on the inverting input. A (voltage) gain of 4e6. You can think of the offset voltage (V_os = 30 uV typ) as being at the non-inverting input to gnd. With the amp just sitting there, what's the DC level of the output? Do you try and trim that?

ting coil. We'd have to very carefully balance everything to get the ampli fying circuit and batteries rotating at 7000 rpm.

Have you seen anything that looks like a signal yet? (or just noise so far?) Sometimes it's hard to know if something fundamental is wrong.. or if you are just making a bone headed circuit (or other) mistake. You might try putting a larger dummy signal through the coil and see if you can see anything. To test the circuit. (maybe you've done that already.)

George H.

across any resistance, it drops the current generated by the spinning coil. So best case scenario, it's across, say, 10 Ohms, and then you have a vol tage of 10 nV. A low noise diff amp can find that 10 nV signal and amplify it and not generate noise? Happy to try it -- are there any schematics I can look at, or a section in Horowitz and Hill?

Reply to
George Herold

AFAIU Ipd0 represents a current induced in the coil L1, for simulation or illustration purposes. It isn't a physical current source. R1 is probably the DC resistance of the coil because it is made from thin wire. At 446mH in air it has to be thin wire. At least at today's copper price :-)

A current pulse forms in L1 but it's wimpy, so Matt is trying to amplify the heck out of it by setting the TIA conversion ratio to 1e9. Which I think is a bit much. A mere 13-14nA of 60Hz noise from the coil could rail the opamp.

Ahm, but that's how pretty much all switch mode converters work. An inductor is charged up to a peak current and then dumps that same current into a capacitor, either until it has no more current or in CCM not all the way down to zero. The energy to create that current being stored in the core if it has one, else in air, sometimes in an air gap.

Active magnetic antennas work that way. The coil is connected to a very low (ideally zero ohms) input impedance amplifier that has a structure similar to a TIA. That sort of "B-field" antenna can be beneficial in situations with lots of man-made electrical noise.

--
Regards, Joerg 

http://www.analogconsultants.com/
Reply to
Joerg

On a sunny day (Tue, 4 Jun 2019 12:28:49 -0700 (PDT)) it happened snipped-for-privacy@gmail.com wrote in :

From wikipedia I just did read the paper:

formatting link

Fun!

So the experiment coil was loaded with about 40 Ohm.

It seems to me, if I was to tinker with that setup, to make a 'ballistic' meter I would do something this +9V +9V | | [ ] R1 o voltmeter + | |-------------------------o voltmeter - | | c | coil + --- b NPN === e Q1 | C1 | | /// /// +9V | [ ] R2 5k6 | coil - --------| +.7V | | | c -- b NPN e Q2 | //// How it works: A positive pulse on coil + results in base current in Q1 that is amplified and the collector current then discharges C1. C1 will slowly recharge via R1, The voltmeter indication is proportional to the current peak. R1 sets the gain in a way, te hhigher teh moresensitve, 5k6 is a nice value. Q2 creates a stable bias for Q1.

You are working with a very low impedance coil / source, so can use a low impedance sensing system to actually get some current, Using a high impedance sensor gives next to zero current. Transistors are basically current amplifiers...

In place of voltmeter perhap use a storage scope.... have a small one just right for this sort of thing:

formatting link

Or just digitize it at decent speed and then process it later, a hum filter may work, better is to find a place away from mains cables and traffic, how about a deserted island, now there you can have hula girls and bananas and sit in the sun zipping .. oh well, get a budget for the trip, who knows...

The original article goes into depth about screening, very nice job, canceling the H and V earth magnetic fields.

Hey. ..

Reply to
Jan Panteltje

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.