Summing amp noise.

[iampoor]

Hi

I am currently working on a summing amp, and trying to figure out how to quiet it down. The power supply is a single ended 9v supply. It is well filtered with a 100ohm resistor and 220uf cap. I am powering it off my bench supply and from batteries. I cannot detect any 60hz hum, so thats a good thing! 
 
I have shot 2 videos showing the relative noise before and after the summing mixer. I hear some "white noise" and a low frequency rumble (Much easier to hear in the second video).

https://youtu.be/kHkypo87tcg
https://youtu.be/M49ljR7h_Tc

The PCB layout and schematic are attached. The ground plane on the bottom layer is solid. Opamp is a tl072.

I am assuming I should add a compensation cap across R17 and maybe move C15? What else could I improve in this design? The noise is quite bothersome. Maybe lowering the impedance and using 2.2k resistors (instead of 10k) in the summing amp would help?

[danadak]

I am very suspicious since you have no G in the system to speak of.

Do you get the noise when all inputs grounded ? What are you using to "listen" to the
noise, what G stage after the circuit you show ? If its input grounded is it "quiet" ?


Regards, Dana.

[iampoor]

I am very suspicious since you have no G in the system to speak of.

Do you get the noise when all inputs grounded ? What are you using to "listen" to the
noise, what G stage after the circuit you show ? If its input grounded is it "quiet" ?


Regards, Dana.

Hi Dana

The system is biased at half VCC (Thats the VB net). This noise is with the inputs grounded.

This is the output stage. I am running it into a guitar amplifier I use for testing. I am running it at a very high gain to exagerate thje isue so it is easier to hear. 

[danadak]

With your guitar amp input grounded at point your circuit connects to it,
eg, end of leads closest to OpAmp Circuit, do you still hear significant
noise ? If so its your Guitar Amp.

Also use a scope to look circuit points, if DSO put it on infinite persistence
or peak detect to get an idea at various point of the significant contribution
point.

Some ref material on noise.

[iampoor]

With your guitar amp input grounded at point your circuit connects to it,
eg, end of leads closest to OpAmp Circuit, do you still hear significant
noise ? If so its your Guitar Amp.

Also use a scope to look circuit points, if DSO put it on infinite persistence
or peak detect to get an idea at various point of the significant contribution
point.

Some ref material on noise.

Hi

I have noise from my lights (very nasty switching supply..), but this is clearly noise from the summing circuit. WIth the input grounded the noise doesnt exist, and at earlier (buffered) points in the circuit, there is no noise, but right when I get to the summing circuit,m we start to have noise issues. The second video illustrates it. It is a low rumbling noise that sounds like "wind" and a low frequency "beating" )Not sure how to explain that one..hehe). This is coming from the summing amp. It is not present earlier in the circuit. I think it might be an unstable op amp? 

Thanks for the reference material. I have been reading Douglas Self's "Small signal Audio Design" which has a fairly brief section on noise, but I rather like some of these app notes. Very good explanations.

I will report back with more findings from my DSO!

[iampoor]

Okay, so I have a few updates

I have started using an Ne5332 instead of a tl072, this reduces the white noise reasonably. I can tell that some of the noise is demodulated RF. (with certain lead lengths attached to the summing point, it makes quite a nice FM receiver  ) I also tried 2.2k resistors instead of 10k. This did not change the performance.

Here are my scope shots. The bottom shot is the noise in the system before the summing section. The top shot is after the summing amp. The noise performance is still not improved by using a battery, so I am ruling out the power supply being the issue for the time being. Can anyone think of anything besides my layout that might be causing this? It sounds like low frequency oscillation to me.

http://imgur.com/a/STTvU

[danadak]

Set the scope up for a match waveform, and use the filter function, LPF, cutoff of
5 Hz, and measure the period of the waveform. Looks like you get ~ 3 cycles in 200mS
per div, or 60 mS per cycle, or 16 Hz.


Any room environment stuff going on, like a flourescent light starter cycling trying to
start a light, just guessing.

Any RC time constants look like 60 mS ? Possible oscillator parasitics.

Last thought is it a feedback loop via Guitar amp driving speaker and induced power
draw affecting mains and the power to the OpAmp circuit.....?


Regards, Dana.

[Audioguru]

The datasheet lists the noise at a typical level of 18nV/root Hz which is better than a lousy old 741 opamp but not as good as an OPA2134 dual opamp with only 8nV/root Hz.
The datasheet has a graph showing the noise rising below 400Hz which causes the rumbling noise that happens in all opamps.
Get a better opamp or turn down the gain. 

[techricky]

Ok, that really sounds like it is oscillating at some very high frequency.

It's usual for a virtual earth mixing stage to have the + input connected directly to signal ground (in your case VB).
I can't see any decoupling capacitors on your VB bias line so its quite likely that this is where unwanted feedback is occurring..
Try adding some bypass caps to gnd as close to the bias resistors as you can.

[Audioguru]

Find out where the 16.7Hz is coming from.
Your schematic is missing a power supply and what produces Vb.

[Galaxyrise]

Nobody has called it out yet, so I'm guessing it's not a problem... why is it ok for those opamps to be directly driving such large capacitors?  I can't find mention of output capacitance in either datasheet.

[Audioguru]

Nobody has called it out yet, so I'm guessing it's not a problem... why is it ok for those opamps to be directly driving such large capacitors?  I can't find mention of output capacitance in either datasheet.

The opamps are driving 10k ohms and 1m (1M?) ohms which is fine, not 1uF to ground. 1uF into 10k produces a cutoff frequency of 16Hz and 1uF into 1M produces a cutoff frequency of 0.16Hz. I have never heard frequencies so low.

[iampoor]

Hi guys, realized I never updated this post.

I found the source of the problem, it was the VB supply line. I added some additional power filtering, and the noise is almost non existent. I duplicated when I have on the main VCC rail (100ohm resistor and 220uf cap to ground) on the VB rail and that cleaned it up nicely. I also added some local VB bypassing the new PCB revision. We will see if that helps.

Even though the issue is solved, I want to reply to a few comments with some additional thoughts, and lessons learned. 

Set the scope up for a match waveform, and use the filter function, LPF, cutoff of
5 Hz, and measure the period of the waveform. Looks like you get ~ 3 cycles in 200mS
per div, or 60 mS per cycle, or 16 Hz.


Any room environment stuff going on, like a flourescent light starter cycling trying to
start a light, just guessing.

Any RC time constants look like 60 mS ? Possible oscillator parasitics.

Last thought is it a feedback loop via Guitar amp driving speaker and induced power
draw affecting mains and the power to the OpAmp circuit.....?


Regards, Dana.

Hi Dana

I think a nice variable LPF is on the list of test equipment to buy, because my scope doesnt have one

Nope, all of the noises in the room electrical system have been pretty well isolated.
Thanks for the ideas, I will add them to the list of things to check for in opamp circuits.

The datasheet lists the noise at a typical level of 18nV/root Hz which is better than a lousy old 741 opamp but not as good as an OPA2134 dual opamp with only 8nV/root Hz.
The datasheet has a graph showing the noise rising below 400Hz which causes the rumbling noise that happens in all opamps.
Get a better opamp or turn down the gain. 

I replaced it with an ne5532. That helped significantly with the hiss. I have found that I really can only use the tl072 in very low gain or buffer style configurations, otherwise the noise is very audible. I really want to get a decent audio analyzer in the next 6 months so I can directly measure th noise performance of opamps.

Nobody has called it out yet, so I'm guessing it's not a problem... why is it ok for those opamps to be directly driving such large capacitors?  I can't find mention of output capacitance in either datasheet.

I am using them to block DC in surrounding stages. Guitar amps dont like 4.5v DC on the inputs.