Microvolts, how not to mess it up?

Morning everyone,

It is for another high-bandwidth current measurement attempt, in parallel to the Rogowski coil research. The bandwidth is assumed to be

2MHz; the shunt voltage is differential amplifier -> ADC driver -> high-resolution SAR ADC.

Ideally, the amplification should be ~40 times to cover the full ADC range. So:

  1. There are four-terminal shunts widely available, so there should be no issue. The connections may cause problems, e.g. due to the contact voltage. Welding instead of soldering and some digital offset cancellation might be necessary.

  1. Then there will be two cascaded amplification stages based on the OPA189 with a gain of 6.3 each. The specs are insane for that price:

400nV offset @ 14MHz BW. Unfortunately, I cannot find FDAs of similar performance, so the signal path by necessity would be differential input/single-ended output. I don't like that, but what can I do?

  1. The common-mode voltage will be well within the opamp's capabilities (it is rated up to 36V VCC), but not spoiling the accuracy would require insanely precise resistors. 0.1% would give me only ~60dB of CMRR. So floating the entire pipeline looks better, despite the added complexity. What do you think?

  2. What to do with that 1.25V signal to drive a differential ADC? Just buffer the +end with another OPA189 and apply the common-mode voltage to the -end or should I use a more fancy single-ended to differential converter? If so, what would be the secret sauce?

What else can go wrong?

Best regards, Piotr

Reply to
Piotr Wyderski
Loading thread data ...

On a sunny day (Sun, 6 Sep 2020 08:44:50 +0200) it happened Piotr Wyderski wrote in :

Have not tried it, but to measre uV DC maybe use a chopper system and AC amplifier, followed by rectification and an ADC? Cannot be that hard to periodically switch between ground and the signal from the shunt to get an AC square wave? CMOS switch?

Reply to
Jan Panteltje

Yes, this is the well-known approach, now integrated into that auto-zero opamps. But the BW is not even close to what is needed.

If a sub-microvolt offset part is available for less than 3 dollars, then a brute-force straightforward solution starts looking applicable.

The question then becomes "how do I not flush this available performance down the toilet?" :-)

Best regards, Piotr

Reply to
Piotr Wyderski

The chopper is there to get rid of any DC offsets - thermocouple voltages a nd the like. If you want lots of AC bandwidth, you can have it to - at leas t within each chopped segment. You might end up with two parallel signal pa ths with alternating positve and negative gain, which you'd fip between the ends of centre-tapped 1:1 ratio transformer (which happens to look very li ke a transmission-line transformer, since it is typically wound with twiste d pair, and should end up being 1:1 to about one part per billion, if you d o it right).

l

Always the tricky part of any design,

--
Bill Sloman, Sydney
Reply to
Bill Sloman

On a sunny day (Sun, 6 Sep 2020 09:20:53 +0200) it happened Piotr Wyderski wrote in :

10 turn trimpots are a great thing to have. If it is for lab testing only, where the temperature is sort of always the same, why not.
Reply to
Jan Panteltje

PS as I usually pick what I have used in the past, and possibly is in the junkbox, my first try would be a 74HC4053 for the switch, 10 turn trimpot for offset correction, switch at a few MHz, and then use one of those video 5 MHz wide AC amplifier chips.. Diode rectifier + RC filter or maybe some diode opamp thing, video FLASH ADC (32 MHz sampling).

But maybe I overlooked something :-)

Reply to
Jan Panteltje

I like to use a single-ended amplifier first, riding on one side of the shunt. That would probably need isolated power in your case. That gives a lot of gain first without any common-mode error. Get your signal from millivolts to volts before dealing with the common-mode.

A proper solder joint shouldn't have thermal offsets. But by all means measure the offset at known zero current, and subtract that out.

Gain is easier to get if there's no common-mode voltage to take out.

--
John Larkin         Highland Technology, Inc 

Science teaches us to doubt. 
 Click to see the full signature
Reply to
jlarkin

This is how I see it as well, but once you have isolated power and the signal in the volts range, floating the ADC too looks like a natural next step. Then the cheapest available digital isolator in SSOP to cross the whopping 15V barrier and it's done.

Yes; I just plan to do the calibration fully digitally to keep the analog path simple.

I see it the same way. An array of ultra-high-precision resistors to deal with the wide common mode voltage would cost more than the rest of the circuit. So floating it instead of going into excessive precision appears kind of logical.

Thank you for your contribution, it's very appreciated.

Best regards, Piotr

Reply to
Piotr Wyderski

If you attach two regular old noninverting amplifiers across the shunt, you can get a differential output with a CM gain of 1.0000 and a differential gain of whatever you like. (The input resistors to the noninverting stages go to the opposite end of the shunt.) One good thing about this measurement is that you don't care too much about high-Z inputs--it'll just perturb the effective value of the shunt resistance a little. This is a slight modification of the usual instrumentation amp front end.

A composite amp with something like an LM6171 inside the feedback loop of the OPA189 would allow you to get much higher GBW while maintaining the low input error. (I'm normally not a huge fan of composite amps because their long-time settling behaviour tends to be a bit strange.)

You can use a chopamp to force the offset voltage of your fave differential-output amp to zero. You'll need good resistors, but that's all happening after the main amplifier, so you don't need such good accuracy there.

Cheers

Phil Hobbs

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

I didn't know this composite opamp idea, but now, knowing the name, I was able to find some good papers on this subject. This is extremely interesting on its own, thank you!

How would that happen? Are we still considering the composite two opamp loop?

Best regards, Piotr

Reply to
Piotr Wyderski

Here's an isolated current shunt thing.

formatting link

formatting link

Probably too slow for you, about 160 KHz at 15 bits.

--
John Larkin         Highland Technology, Inc   trk 

The cork popped merrily, and Lord Peter rose to his feet.   
 Click to see the full signature
Reply to
John Larkin

It's easier with a single-ended circuit. For an inverting amp, you connect the chopamp as an (inverting) integrator and use its output to drive the noninverting input of the main amp. (It's often useful to drive the noninverting input via a voltage divider, so that the main amp's summing junction dynamics don't mess things up.)

For a fully-differential amp (with external resistor feedback), you can make the input error go to zero using false summing nodes. The FDA is going to servo itself to null out its idea of its differential input voltage, so you can't do much about that. However, if you put resistors between its inputs and the summing nodes, you can use the chopamp to dump some current into the FDA's input nodes such that the differential voltage between the two summing nodes gets zeroed out. (Full disclosure: I've never actually needed to do that.)

Cheers

Phil Hobbs

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

Piotr whats the low frequency limit? If it's fairly high, a capacitor would serve well & reduce offset errors.

Hul

Piotr Wyderski wrote:

Reply to
Hul Tytus

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.