Question about preamp for measuring noise.

[matiasro]

How are you?
I am writing from Argentina.
My oscilloscope is a Rigol 1052 and HP3478A multimeter.
I am a DIYer and audio enthusiast, although I have done many successful projects in the field instrumentation i am a newbie.
I make this query as I begin to measure noise levels at regulators and power supplies.
The noise level of my Rigol is very high for this and I know I should build a preamplifier to amplify the signal to achieve and exceed the measured noise level of the instrumental.
I believe that for this task I need to use an amplifier with a factor equal to or greater than 1000x and low noise.
I was looking for information on this and there is so much material with words associated with other issues hampering my search.
On the other hand I saw few people carried out projects for this purpose.
Someone could tell me if you use any preamplifier with your multimeter or oscilloscope?
It would also be important if you know of a working circuit to achieve my goal.
I send a hug from Argentina and from already thank you very much.
Matias

[tekfan]

Measuring noise with a digital oscilloscope is very hard. Not only is there analog noise in the front end but there is also digital noise that is made by the AD converters.

Even if you do build a sensitive preamp to amplify the noise it will still be hard to measure on the Rigol.

This is what analog scopes do best. If you will do these measurments often you may want to find a used analog scope.
If measuring noise below several mV then you will have to be very careful about ground loops and exposed conductors.
If you use unshielded wire you will get meaningless readings.

Your best bet is to measure the noise with your HP3478A set to AC volts. That will give you a true RMS reading of what the noise is from 20Hz to 100KHz. If you can build a preamplifier so that you can use the 30VAC range on your multimeter you will get 20Hz to 300KHz frequency range. If all you're doing is audio then the direct measurment without the preamplifier should be enough. The good thing about using a multimeter is that it is floating and doesnt introduce ground loops to the measurment. If the device you are trying to measure has a low impedance output you won't even need shielded wires.

[Tube_Dude]

The good thing about using a multimeter is that it is floating and doesnt introduce ground loops to the measurment.

Thats the great superiority of handheld multimeters in workshop, because working with batteries, they are  inherently free of mains induced ground loops and noise.

100% agree with the use of a analog scope for low noise level inspection, and even for Audio work in general. 

Cheers

[jahonen]

I'm not sure if the preamp is a bad idea, I once build one with 80 dB gain and was able to get very meaningful results with it. I didn't even have any external shielding although I first thought that I absolutely need one (but incidentally, PCB was a 6 layer one, because I put it on a free space on DUT PCB panel which would have been discarded otherwise ). I used it to characterize the output noise of a low-noise voltage regulator. I also had a lowpass filter on preamp output to get meaningful and defined noise bandwidth.

Preamp is built around ADI microphone preamp SSM2019 (40 dB gain) and a post-amplifier with AD797 opamp (also 40 dB gain). Worked like a charm with our DSO at work. Even slight bending of the DUT PCB was clearly visible due to minuscule reference voltage change.

I can post the schematic if anybody is interested.

Regards,
Janne

[matiasro]

Thank you all for the answers.
They have been very useful to me. For now I'll settle for being able to measure the noise level of a simple circuit with a LM317 for example.
I know it's hard to find people who have done some preamp to measure these signals, so any input is very helpful for me.
Janne, I want to see that circuit did not understand if the PCB was 6 layers or would have been ideal if they had 6 layers.
My English is not very good and sometimes translators do not do a good job.
Could you clarify that point?.
Of course thanks and waiting for new contributions.
A hug.
Matias.

[jahonen]

Thank you all for the answers.
They have been very useful to me. For now I'll settle for being able to measure the noise level of a simple circuit with a LM317 for example.
I know it's hard to find people who have done some preamp to measure these signals, so any input is very helpful for me.
Janne, I want to see that circuit did not understand if the PCB was 6 layers or would have been ideal if they had 6 layers.
My English is not very good and sometimes translators do not do a good job.
Could you clarify that point?.
Of course thanks and waiting for new contributions.
A hug.
Matias.

"Host" PCB was 6 layers so I had to make it so to fit the preamp parasitically into same panel (no incremental cost). It can be certainly done with less layers (all internal layers were just voltage or ground planes), but of course having a stiff ground around is always good for low noise and it helps enormously with noise immunity, which is crucial for this kind of circuit. Here is the schematic attached. It is pretty much identical to Walt Jung's preamp which he used with Audio Precision audio analyzer when characterizing his super regulator. You can find that here. Look at his 1995 article series, part 2 A has a schematic for a preamp. Since scope is not so sensitive as an audio analyzer, I added a 100x post-amplifier. Thus total gain is around 80 dB. Raw bandwidth is around 1 MHz, so you might want to limit that to get less intrinsic noise and to make more sense in audio circuits.

Here  is a picture of assembled PCB:



As you can see, the capacitors didn't quite fit in so I had to improvise somewhat.

I'll post some results later, to get an idea of what kind of performance one might get with this and a DSO. But for quick stats, intrinsic noise seems to be around 20 mV RMS and 178 mVp-p (as seen by the scope), so considering 80 dB (10000x) gain, equivalent input noise voltages are 2 µV RMS and 17.8 µVp-p respectively.

Regards,
Janne

[jahonen]

And here are some measurements. Note that voltage levels shown should be divided by 10000, because of the preamplifier gain. First, a scope screenshot with inputs shorted (just like in my above post of the PCB):



Then, 10 Vpp 1 kHz signal via 1 M/1R resistor divider (attenuation of 120 dB). Signal level is therefore 10 µVpp:



Picture of the attenuator :



And same attenuator and signal levels, but 100 kHz frequency:



Some averaging helps digging the signal out of noise:



And finally, to put these voltage levels into some perspective, the output noise of 7805 regulator (output current is around 25 mA). Peak to peak noise seems to be quite high.



Regulator circuit:



Regards,
Janne

[Mechatrommer]

can someone teach/lead me on how to calc the opamp noise at this staggering 80dB gain? i know there's long formula in the datasheet, but cant we have anything simpler? (dylexia? ) or do i have to just live with it? i thought that (or some) of the noise (above pictures) came out of the opamp itself instead of the input signal?

@Janne: i've managed to study 1/1000 attenuation and re-amplified 1000x (dumb thing) without much noise (as far as the scope can see) thank you on the other thread
but as always said, opamp Voffset becoming clearer/important on the scope at this level of gain onward i suspect. it was just uA741

[jahonen]

can someone teach/lead me on how to calc the opamp noise at this staggering 80dB gain? i know there's long formula in the datasheet, but cant we have anything simpler? (dylexia? ) or do i have to just live with it? i thought that (or some) of the noise (above pictures) came out of the opamp itself instead of the input signal?

@Janne: i've managed to study 1/1000 attenuation and re-amplified 1000x (dumb thing) without much noise (as far as the scope can see) thank you on the other thread
but as always said, opamp Voffset becoming clearer/important on the scope at this level of gain onward i suspect. it was just uA741

Yes, the noise at those 1 kHz/100 kHz pictures is mostly from the preamplifier amplifier itself. Purpose was just to show that even such low level could be measured with this quite crude setup. As I now notice, it would be better to use the standard deviation measurement on the scope, because DC offset adds to normal RMS measurement. Standard deviation is equivalent to AC RMS.

One thing which is important for multi-stage amplifiers is to optimize the first stage for lowest possible noise (haven't actually done that ), because all noise of the first amplifier will be amplified in every subsequent stage. So in this case, the output noise of the SSM2019 will be amplified by 100x (40 dB) by the AD797, in addition of the noise it itself produces. I therefore think that 2nd stage intrinsic noise plays therefore minor role for the total output noise voltage.

I think that the maths is a bit hairy for noise calculations, haven't found out a shortcut to success there Have you read chapter 10 of Ron Mancini's Opamps for everyone? It explains the noise calculations in more detail. Unfortunately it seems to ignore the current noise contribution, but that matters more when source resistance is high.

Regards,
Janne

[branadic]

I've build a noise amplifier last year that was designed by Jim Williams in AN83. In one version I build the genuine circuit with 60dB for measurements at the scope. In a second version I changed the gain in both stages to 40dB (full gain of 80dB) and spend a voltage divider (Vcc - 200R - ADC input - 200R - GND) that inputs to the 12bit ADC of a STM32F103. The high gain ensures that noise floor of the amplifier is abouve noise floor of the adc. Some firmware and software makes it a nice and modern noise test site.
There are some critical points in the circuit. One is the peaking of the LT1028, the other is the right choise for the input cap.
Jim used selected Oscon 330µF/6.3V electrolytic cap. This parts are very expensive today and need some selection to low leakage current what makes this part less interesting. I replaced this part and used a Vishay  TR3E337M010C0060 330µF/10V tantalum instead. For measurements on higher voltages I use AVX FFB54D0117KJC, 110µF/75V three in parallel.

There are some very interesting threads an µC.net

http://www.mikrocontroller.net/topic/241224#2452507

http://www.mikrocontroller.net/topic/250656#2573926

[Kurski]

I really had some desperate need for that kind of circuit. I would like to bring some headphone amp breadboarding data on the big screen through Owon 1022i oscilloscope. Now the limit is 40-50mv. I would like to work with signals under typical audio interface noisefloor. Lets say - 120-130db. Accidentally just yesterday I got those chips from mouser.

[Kurski]

Sorry. I just noticed you posted the info already! Thanks.