noise in instrumentation amplifier

If you look at the datasheet, that op-amp has a strong peak of noise around 40kHz (130nV/rtHz) vs. less than 50nV/rtHz at 1kHz (still pretty noisy). You have a gain of what, 20, or something like that and don't forget noise is RMS, so P-P will be many times higher (maybe 6 or 8x)

Maybe try another op-amp. Your resistor values do seem a bit low for a

90uA op-amp.

--sp

--
Best regards,  
Spehro Pefhany 
Amazon link for AoE 3rd Edition:            http://tinyurl.com/ntrpwu8

Are you displaying the noise performance for the In-Amp itself or with it connected to the sensor ?

How "clean/stable" are the In-Amp power supplies ?

If the In-Amp is connected to the sensor, you need to consider the CMRR is affected greatly by the resistor tolerances of the diff amp.

50 nV/rtHz, Yuck! And they call it a low noise opamp. If it's a strain gauge or something you don't really need a low bias current opamp. You could even try an instrument amp. I use a lot of AD620's but the noise is not that good ~10nV/rtHz. There is a nice list of Int. amps on page 363 of AoE3.

So I agree with Spehro, try another IC.

George H.

C6 also hurts the CMRR at high frequencies, above roughly 100 KHz.

--

John Larkin         Highland Technology, Inc 

lunatic fringe electronics

Thank you for all the advices. Effectively it seems that the opamp is very noisy and I was not looking in the right direction. If I do the calculation I get about 1.5mV which is close to the measure I got.

In fact I was using the OP462 before and the noise was much lower, about

0.3 mV rms. I moved to the MAX 44245 because it has a very good EMI rejection Ratio (like 60 dB at 100 MHz). With the OP462 I saw DC rectification above 100 MHz. I need to find a low power op amp with decent noise and good EMI rejection.

:

th

the

heet

rong

ing

rrent

oise

y

on

t

OK, I've never had EMI getting into the front end. The "word on the street " (that I got from Joerg) is that only bjt opamps will have that problem. So ignore what I said before and get an FET or Cmos opamp. (They are typically a little noisier, but something less than

10 nV is not hard to find.)

George H.

Did y'all notice it's an auto-zero amplifier, to get its 7.5uV Vos max spec? That means it has input switches running, with charge injection, etc.

--
 Thanks, 
    - Win

Nope, but that perhaps explains the 30kHz noise peak.

To the OP; The few times I played around with making my own three opamp Instrument Amp. (IA) I found that getting the CMRR respectable was a matter of resistor/impedance tweaking... expensive except for a one of.

Better to buy someone's else's laser trimmed IA.

Do you have a copy of AoE3? There's a wonderful section on IA's. I wish I'd seen the CMRR graph vs frequency before a previous project. That alone is worth the price of admission, IMHO.

Why is there so much EMI? Can you shield it more? What 'cha doing? George H.

Didn't notice that- I guess that accounts for the noise hump. EMI seems to be an issue with some of the chopper amps, though this is one of the ones that claims to have filters:

"These devices also feature integrated EMI filters to reduce high-frequency signal demodulation on the output"

There are actually some out there which auto-zero ONLY on power-on. Once. Scary stuff.

--sp

--
Best regards,  
Spehro Pefhany 
Amazon link for AoE 3rd Edition:            http://tinyurl.com/ntrpwu8

I'm impressed by how much support you can get here. Thank you all !

I didn't knew that AoE3 was available now, I'm still on AoE2, but not for long ;)

What I'm trying to do is an Hall effect probe. Maybe I should try another IC like LMC660 (CMOS input) as suggested by George. I remember once it did the job in replacing a LT1014 for better EMI protection in a similar product.

Grin.. Well Welcome to SED. It's a pretty good bunch of guys (and a few ni ce gals.)

George H.

d

I thought RMS was 70.7 of peak. Converting RMS to peak is the inverse,

1.414 * RMS.

0.707? But that is for sine waves. With noise it's different. In theory if you wait long enough you can get an arbitrarily large peak voltage.

George H.

Yes, p-p is 2.8x RMS for a pure sine wave.

For Gaussian noise it should be less than 8x RMS p-p 98% of the time, but in theory it's unbounded.

--sp

--
Best regards,  
Spehro Pefhany 
Amazon link for AoE 3rd Edition:            http://tinyurl.com/ntrpwu8

Grin, It's not all that hard to see... Gain up a noise source to some reasonable ~1V rms value and watch how often it rails your

+/- 15V opamp.

George H.

d

The classic way of doing well with a Hall probe is to AC excite it and use phase sensitive detection to get your output.

There's a lot of literature on AC-excited bridges, lock-in amplifiers and t he like. The Art of Electronics mentions them in passing quite often.

You can generate your signal at a frequency where your amplifiers are relat ively quiet - above the 1/f noise corner - and amplify it as AC before demo dulation and low pass filtering.

You can band-pass filter the amplified signal before you demodulate it, but you do have to be careful to keep the phase shift in the bandpass filter a round the modulation frequency both predictable and frequency-insensitive - you will have to allow for it.

--
Bill Sloman, Sydney

** That ratio is only for steady amplitude sine waves.

Various ratios apply for other defined and steady waveforms, like triangle, half sine etc. But for a random noise signals, it must be measured to find the average rectified or RMS value that applies to the particular noise. T he measurement bandwidth of the test instrument will affect the result in m any cases too.

If the noise has low frequency components, the reading on a meter will vary so one takes a visual average.

.... Phil

There is a good video on conversion from rms noise to peak-to-peak values here :

** As if a novice like you knows what a "good" vid on the topic was.

That is not an informative vid, it is way misleading.

.... Phil