Noise at regular hz intervals

I've been working on a gainstage circuit for Ian's Poormans pultec. Its based off a API 2520 non-inverting line amp circuit.

So far its working very well, and I have a noise floor of -75db which I'm generally happy with, but I'm curious about the noise I'm getting  nonetheless.

When I analyzed the noise, I found the major peaks were at 50, 100, 200, 300hz etc and then also found in between that 150, 250, 350 etc in smaller peaks.

I live in Australia and we run on 50hz AC - so would this would explain what I'm seeing? Could this be from interference from the toroidal transformer I'm using?

If I wanted to look into reducing this noise, where would I begin?

Noise is probably coming from PS rectifier diodes charging the caps at mains frequency.

These current pulses generate energy at multiples of the mains frequency...

JR

First thing to do is check the noise the the input of the 2520 shorted to ground. If you still have then mains related spikes then the problem is either in the power supply itself or coupling from the mains toroid to your output transformer.

If the hum goes then it is getting in through the input or the EQ itself.

Cheers

ian

thanks for this info. I've spent a good bit of time looking into it.

I think the 50hz, 100hz etc peaks were a bit of a red herring and now after further testing I can conclude are not an issue and the fact they are present isn't a problem. The EQ section is fine, no issue there and grounding the input of the opamp didn't make a difference.

The issue seems to be the discrete opamp I was testing with - I think its just a noisier opamp than others. Its based off a UREI Mod1 opamp - a very simple design - http://sureshotstudio.blogspot.com.au/2011/09/urei-mod-1-opamp.html

So I switched it out for a GAR1731 discrete opamp - the noise went away. Also had the same success when I switched out with a OPA134 dip8 IC.

I may even be measuring my noise floor incorrectly. Should I be measuring the peak or RMS? RMS is -91.9 and peak is -70.1

Here is a sample of the noise - http://www.thraxeh.com/Music/eq_013.wav - you will need to boost it before you can see/hear the wav form.

really could just be a non-issue as its really quite a low level noise. What do you think?

Definitely looks like the original discrete op amp is the cause. Looking at its schematic it is very basic so you would not expect its performance to be great. There are lots of ways of measuring noise but peak is generally not one of them. Weighting curves are often use dand the latest IEC standard calls up a quasi peak rectifier. Your rms measurement is probably nearer the 'truth'. Noise below -90dBu is  a good result.

Cheers

Ian

thanks for that Ian.

My OCD got the best of me and I continued to work on this to improve the figure  :
I now have these figures with the MOD1 opamp in its final enclosure with all incidental components(EQ, VU, relay switching, transformers etc)
-100db RMS
-85db peak

This is now only 3-5db away from my converters noise floor. So I'm extremely happy with this.

However I want to make sense of a change I made to the transformer wiring which got me the extra 10db floor. Here is the output transformer I'm using.



My calculations measure the "N.F" winding to 525 ohms (side note, does anyone know what N.F means?). I had it connected as follows

1 - opamp output
2 and 3 tied together
4 - GND

5 - TRS Ring
6 and 7 tied together
8 - TRS Tip

However when I switched the opamp output and GND connections, I got a 10db better noise floor(output volume stayed the same).

Why is that? I would have though the only change would be to polarity?

10dB seems a lot. NF seems negative feedback, used across the transformer to correct the transformer non linearities and LF loss, there's some paper from lundahl and app notes on that. Pins 2 and 3 are not connected to anywhere else but themselves in your circuit?

The only difference swapping the direction of the entire winding would be the proximity of each end to the shieldings, if the NF is less shielded than the other half, having it in the lower impedance side will make the noise lower, but 10dB seems too much for such a small change.

JS

Pins 2 and 3 are only tied together, they are not connected anywhere else.

Should they be?

The 10db difference was quite a surprise.

Noise that is 100dB below the nominal signal level is very small and not easy to measure. You are talking about a few tens of microvolts. This level is probably well below the CMRR of your (electronically balanced) converters so it is not hard to see how swapping a winding polarity could alter this and change the noise reading.

Cheers

Ian

One thought: if the noise changed when you changed op amps, it probably means that the power supply has 50Hz and harmonics on it. The two amps will probably have different power supply rejection ratios, so they will each pass or reject noise on the supplies differently.

A simple fix would be to add a regulator to the circuit, or more effective filtering in the power supply. Try measuring the AC voltage on the power supply.

Also, noise is probably best measured as an RMS value. 'Peak' and noise are complex… it depends upon the time interval that you measure over and the nature of the noise. RMS gives you what you want with little guesswork.