And presented in Dave's unique non-scripted overly enthusiastic style!
The secret world of Chopper (auto-zero) Amplifiers.
Very interesting, I never heard about these amplifiers, but if precision measurements come up I will know where to turn.
For those interested, Don Sauer pointed me to the history of his development of the National LM2001 chopper amp and their quest for the “perfect” chopper amp.
Some interesting reading:
Now I’m getting confused.
The circuitry posted in the link looks more like what I know as a “lock-in” amplifier, which are primarily used for measuring tiny signals buried in all sorts of noise. They’re quite useful for optical detectors (eliminates the background).
Sorry, didn’t mean to confuse. There are actually many different ways to produce a precision “chopper” amp. The one I described is the “offset servo” “compound amp” technique, probably the most common technique used today. Mention of this didn’t make the video cut I’m afraid.
Don’s article is about the design of the LMC2001 in particular which is a different again.
Some extra links:
I’ve actually just recently been working on a project where our goal is to get something stable down to the nanovolt level, measuring really low frequency signals (0-3Hz) based on the difference of two inputs. There seem to be quite a few trade-offs with the various chopper op-amps and chopper instrumentation amps out there, but the instrumentation amps don’t seem to match the discrete chopper amps on certain noise and offset considerations.
I’m wondering whether you might be familiar with considerations involved in using multiple chopper amps together (either to make some sort of custom instrumentation amp, or to chain some gain stages with low pass filters)? Does one need to get ones with a clock exposed to synchronize them all on the same clock even if the desired signals are quite low frequency?
Just curious if anyone has any thoughts or suggestions regarding slightly more complicated designs involving choppers.
There are on the market what are called nanovolt-preamplifier modules. These modules contain a chopper op-amp, offset adjustment means, and thermal insulation.
The bandwidth has been severely restricted, to about DC – 0.035Hz or so. This small bandwidth is necessary to reduce the rms equivalent input noise-voltage to about 0.4nV rms. This allows am A/D converter to have a 1/2 LSB resolution of 1nV.
JFets cannot be used for these functions, because of the 1/f noise of the JFets. For example, the integrated rms noise of an IF9030 JFet over the spam DC – 0.5Hz is about 80nV rms. This is way too high for 1nV measurement.
Try looking at the ADA4528 chopper op amp at very low bandwidth. Four of these in parallel will achieve an input voltage noise-density of 3nV/Rt-Hz from DC to 1KHz, compared to the gate noise-density of the IF9030 at 0.001Hz = 1.5uV/Rt-Hz.
After taking a look at some of the links above, I decided to look around at National Semi’s current crop of parts for chopper op-amps, and found their LMP2021/LMP2022 parts. While for a few specs they’re edged out by some of TI’s or Linear’s parts, these have some of the lowest input voltage noise I’ve seen out of any parts. Their spec quotes 11 nV/sqrt(Hz) at 1000x gain for low frequency signals, and they even provide some application notes on how to get 46nV RMS noise for 10Hz bandwidth situations. That, and the primary usage they give for these chips is in a bridge amplifier. I think I may have found a pretty good solution? (So long as what they’re quoting is realistic)
Still, if anyone is familiar with the design of things like nanovoltmeters or similar, I’d love any suggestions
What they quote in the datasheets is realistic, but it can be swamped by practical effects if you aren’t careful.
Thermal EMF’s caused by dissimilar connection materials will play a part down at these levels.
Keithley might have some good stuff on low voltage measurements:
Looks like they have the Low Level Measurements Handbook free if you register?
Agilent will likely have some app notes too, and Nationl Instruments perhaps.
For low noise systems at low frequency it is common to parallel op-amps to lower the 1/f noise of the amp. Not sure if this works with chopper amps though, never tried it!
Thanks very much for the reply Dave. I’ve read on a few datasheets of these amplifiers that being careful about thermal effects, minimizing the number of dissimilar junctions (sockets = bad), and trying to keep some parallelism in wiring. I’ll take a look through those resources you’re suggesting. I wasn’t familiar with keithley before, and I hadn’t thought to check Agilent. The thought about using parallel op-amps is interesting. I’ll have to give that a try and see what effect that has.
Always fun to have a new/interesting challenge
You’ve described the “chopper-stabilized amp”, quite a different animal from the original chopper amp. The latter (obsolete and mostly forgotten) uses its input switch to modulate a DC voltage into an AC waveform, which then goes through an AC amp without a care in the world for offset, and the result is synchronously demodulated back to DC. I’m playing with one right now, in a 1965-vintage HP Differential Voltmeter. It uses photocells for the switches. 10^10 ohms, 1000V common-mode range – take that, IC’s!
The lowest noise chopper was the mechanical chopper that was made by Airpax, Inc. in Fort Lauderdales Florida, the leading developer and manufacturer of these devices from in 1950s to 1970s. I was the Director of Product Development for Airpax in the 1960s. We made and tested a mechanical chopper circuit that worked down to the lower nanovolt signal level! Transistor choppers eventually captured the market, but they could not compete with the mechanical chopper at very low signal levels.
Great read, Ill come back
Interesting to still find some information on mechanical choppers.
Herb Cook was the president of Airpax, and the company was developing a relay for production after WWII. Hughes Aircraft Company bought some and discovered that they weren’t very good relays, but they were excellent choppers. It became the main product line and many of them were sold.
I was head of Product Development at Airpax, and we were forced into developing a transistor chopper when the silicon transistor came along and could compete with the mechanical chopper in most applications. However,no one ever built a chopper of any type that could measure as low a signal level device than one development model that could handle signals at levels below one nanovolt (very difficult to test at that level).
We developed other product lines at Airpax, including magnetic amplifiers, RPM sensors, motor controls, tape speed compensators, etc., all based on electromagnetics. We also designed a Doppler current meter for measuring ocean current velocities for the University of Miami.
I was interested by your comment. I worked at Airpax (Cambridege MD)from 1972 to 1979. They no longer made choppers at that time. However, I came to Airpax from another company in Cambridge MD called Cambridge Scientific, Inc. – President Donald Holdt, former VP of Airpax. CSI was still making choppers in the early 70s.
Join 799 other subscribers