Basic Opamp Inverting Buffer HUM

[FearLabs]

I was wondering if anyone has any input on this:


I have attached a schematic.  It's a basic opamp buffer followed by an inverting buffer.   This is to create a phase splitter for an audio signal, the two outputs drive the next stage which apply quite a bit of gain, so keeping the noise floor very low is important.   

I attached FFT plots of both outputs.  The nonInverting buffer works perfectly, the noise floor on the output of this stage is so low, it's barely different than using a cable.   The Inverting stage has at least 25db more 60hz hum plus a ton of it's harmonics.  It's fairly low, but once I boost it in the following stage, it causes a limited dynamic range in the whole design. 

I tried a few different opamp IC's.   The decoupling caps are very close to the power pins.  I initially tried the design on a PCB with a ground plane, I tried re-laying it out on different PCB with star grounding, then on a breadboard, all with the same results.  I tried different resistor values, different decoupling caps, changing the orientation, all with no change.

Are inverting opamp stages more susceptible to hum somehow? 

Let me know what you all think, and thanks lots for the input!

[BradC]

You appear to have 2 non-inverting buffers in series. Have another look at the inverting buffer. Your +ve should be connected to gnd and the -ve input should be at the tap of a voltage divider between the output of amp 1 and the output of amp 2.

To put it another way, take the connection off the +ve output of the inverting amp and just switch it to the -ve input. Gnd the +ve input.

[FearLabs]

You appear to have 2 non-inverting buffers in series. Have another look at the inverting buffer. Your +ve should be connected to gnd and the -ve input should be at the tap of a voltage divider between the output of amp 1 and the output of amp 2.

To put it another way, take the connection off the +ve output of the inverting amp and just switch it to the -ve input. Gnd the +ve input.

You are right, however that is just a typo in the drawing!!!  Oops    I re-drew it from a bigger design.  The way you described is the way it is.   I corrected the schematic in the original post.   

Same problem as described tho.

[rob77]

in general (haven't had a deeper look at your schematic):
- add more decoupling - i can't see any electrolytes/tantalum there - use ceramic + electrolytic
- use a low noise op-amps
- use high quality "low noise" resistors for the inverting buffer - resistor might be a source of noise
- add shielding

[FearLabs]

in general (haven't had a deeper look at your schematic):
- add more decoupling - i can't see any electrolytes/tantalum there - use ceramic + electrolytic
- use a low noise op-amps
- use high quality "low noise" resistors for the inverting buffer - resistor might be a source of noise
- add shielding

- I've tried ceramic and poly caps up to 1uF.  Also paralleled them with some low esr 100uF electrolytics, as well as 22uF organic ones.  Decoupling doesn't seem to be an issue

- Started with a Burr-Brown opa2134 (very low noise opamps).  Also tried Tl074s and 5532 amongst others. 

- Using 1/4 and 1/2 watt metal film resistors of varying sizes.  The only time I was able to get measurable resistor noise was when I got to very high values.  The 60 hum isn't resistor noise.

- Tried enclosing the circuit in a grounded aluminum enclosure 1/8" thick.  No change.  I'm sure at some point with some expensive MU-Metal can I could get some NASA grade shielding, but it seems odd that the inverting buffer with NO gain is noisy compared to a noninverting buffer with NO gain.   Orientation of the circuit doesn't change anything either.

[rob77]

how good is the power supply ? give it a try running from batteries and measure the 60Hz hum. probably your supply has insufficient filtering.

[FearLabs]

how good is the power supply ? give it a try running from batteries and measure the 60Hz hum. probably your supply has insufficient filtering.

Good call, just popped a couple of 9 volts on there and BOOM, problem gone. 

Now to figure out how to add more filtering.  Any suggestions?  My first thought is to try some resistors in series with the power supplies, so those plus the decoupling caps make some RC filters.

Thanks, I was feeling like I was going crazy!

[rob77]

i would add more "raw" filtering (right after the rect. bridge) and add a RC filter to both supply rails for the op-amps - experiment with values,but i would start with something like 47R + 220uf for the RC filters. and of course keep the decoupling as close as you can to ther op-amps , do not use the decoupling caps as part of the low-pass RC filter.

[Galaxyrise]

but it seems odd that the inverting buffer with NO gain is noisy compared to a noninverting buffer with NO gain.   

Thanks, I was feeling like I was going crazy!

Well, you're half-way there: you know where the hum comes from (and why shielding didn't help.)  And since the inverting amp is taking the non-inverting amp's output, it's not surprising that it has more hum.  But I would expect twice the hum, not 10x...  I'm still curious where that comes from. 

As you can see from rob's suggestion, 60Hz is very hard to be rid of through decoupling.  It's just too low a frequency.

[BradC]

The inverting amp is directly ground referenced by virtue of the link between +vin & gnd. The non-inverting amp is ground referenced but via the 1M input & whatever impedance the input cap and source present, most of which will be filtered out because the output is going to sit at a virtual ground between +ve & -ve.
Try replacing the gnd connection (+ve in) on the second op amp with a resistive divider between the rails (decoupled). That'll take the earth noise out of the equation.

A schematic of the PSU would help.

[FearLabs]

The inverting amp is directly ground referenced by virtue of the link between +vin & gnd. The non-inverting amp is ground referenced but via the 1M input & whatever impedance the input cap and source present, most of which will be filtered out because the output is going to sit at a virtual ground between +ve & -ve.
Try replacing the gnd connection (+ve in) on the second op amp with a resistive divider between the rails (decoupled). That'll take the earth noise out of the equation.

A schematic of the PSU would help.

I just tried a voltage divider btw +15 and -15 with two 10k resistors.  I connected the mid point to the +ve on the inverting stage instead of to ground.   Same amount of hum, but now there was a subtle DC bias (a fraction of a volt really), because  +15 is 15.14v and and -15 is 14.94v. 

Attached is the schematic of the power supply.   I plan on adding reverse protection diodes to it, but they're not on there yet.

I've been experimenting with calculating and adding LC filters to it.  Did a lowpass filter with the -3db point at 1hz and no change (huge caps).  I dropped the ripple by at least 30db and no change at all to the operation of the circuit.     Ugh!  This is crazy!   

I'm starting to wonder if the wires or resistors are behaving as parasitic inductors or something.    I'm just surprised  I'm getting such an effect with gain=1.

[T3sl4co1l]

Why is your power supply generating so much 60Hz?  It should be 120Hz.  Is it grounded (and your signal source or load as well)?  Also, what's the PSRR on the op-amps?

I would be surprised if PSRR were the problem, sounds like ground loop.

Tim

[FearLabs]

Why is your power supply generating so much 60Hz?  It should be 120Hz.  Is it grounded (and your signal source or load as well)?  Also, what's the PSRR on the op-amps?

I would be surprised if PSRR were the problem, sounds like ground loop.

Tim

It's not.  When I look at the power rails on the oscilloscope, even on the smallest setting, there is no visible ripple.  I actually see the noise of the regulator way more more than any ripple.   

If you look at the FFT plot, 60hz does dominate, but there is lots of harmonic content too, 120,180,240,300, etc...   

Here's a datasheet to the opamp, shows 106db PSRR.  Most certainly not the problem.
https://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=2&cad=rja&uact=8&ved=0CC4QFjAB&url=http%3A%2F%2Fwww.ti.com%2Flit%2Fds%2Fsymlink%2Fopa134.pdf&ei=zCA-VNrxLoP2yQTF_4CoBQ&usg=AFQjCNHNHJLdDVmojLs1G7jCBRJB12rgAw&sig2=ln0Ru8yweQXp1WWX3DWOcw&bvm=bv.77412846,d.aWw 

Don't understand how a ground loop could occour if every ground connection is ONE POINT.   

Baffled.

[wiss]

Is the power-supply half-wave rectified?

Once I hacked up a +-12 V from a single secondary wall-wart, half-wave rectified into either rail and stabilized by 78/79-12`s.
Hummed badly...

[FearLabs]

Why is your power supply generating so much 60Hz?  It should be 120Hz.  Is it grounded (and your signal source or load as well)?  Also, what's the PSRR on the op-amps?

I would be surprised if PSRR were the problem, sounds like ground loop.

Tim

I should also mention that I've built this circuit with two different PCB layouts, plus about 3 times on a breadboard.   First PCB had a ground plane, the second a star ground.

I've tried 3 different "low noise" opamps plus several others.  Burrbrown, 5532, Tl074

I've tried two different power supplies.

I've tried ceramic, poly, and electrolytic decoupling caps.  Ranging from 100nF to 10000uF is various combinations.

My power rails have NO ripple (it's below the thermal noise floor of the regulators).

Everything is grounded to one point.

I've tried metal film as well as 1/2watt carbon comp resistors.

The output of the noninverting stage is PERFECT.  Same as a wire to my interface.

Tried 20ohm resistor with 10000uF caps on the power rails,  (lowpass filter at .8hz).  This makes the thermal noise on the power rails unmeasurable on my oscilloscope- perfect flat line).

Shielding inside a 1/8" aluminum enclosure does nothing.

Using batteries is the only way I've seen the hum go away (unfortunately not an option in this project).


So....  Maybe this is unsolvable?   Seems pretty crazy since it's such a basic textbook circuit.     

[rob77]

are you experimenting on a breadboard now ? or did you solder the RC filters to a PCB version of the circuit ?
just asking because some breadboards are excellent in picking up noise from the environment (especially the 50/60Hz in a noisy environment)

[IconicPCB]

How is Your source connected to the circuit?

How are their earths connected? if at all?

[FearLabs]

are you experimenting on a breadboard now ? or did you solder the RC filters to a PCB version of the circuit ?
just asking because some breadboards are excellent in picking up noise from the environment (especially the 50/60Hz in a noisy environment)

I'm on a breadboard now.  Honestly the ripple on the power supply is buried in the noise of the regulators when I really zoom in on it, and even that isn't much without the filters..  I'll give it a shot for sure.   I'll try in a bit and report back.

[FearLabs]

How is Your source connected to the circuit?

1/4" jack. 

How are their earths connected? if at all?

To the star ground point.

The signal is getting into the circuit just fine.  The first stage is working perfectly.  Something about the second inverting stage is adding about 25db to the level of the 60hz and it's harmonics.

[deephaven]

Try going back to your hum-less situation by powering off batteries. Then try touching just the 0V line of the power supply on the 0V of your circuit. Then turn your power supply on and off. This should narrow down where you need to look. Sounds like a ground loop problem.

[FearLabs]

Try going back to your hum-less situation by powering off batteries. Then try touching just the 0V line of the power supply on the 0V of your circuit. Then turn your power supply on and off. This should narrow down where you need to look. Sounds like a ground loop problem.

Right on!  Getting somewhere!   The hum cam back as soon as the ground on the power supply was connected, and not even powered on.   I'm assuming it's a loop between the power supply and the interface i'm using to get the FFT.  Using one of those dangerous ground-lifters on the power supply also did it.   

Hmmm... ok.   I think I can get on with my life! 

Thanks everyone!

[BradC]

An Isolation transformer is a handy dandy thing to have around. Before I scored my 2400W ex Hospital unit I used 2 24V transformers wired back to back for low power stuff.

[bobcat]

Have you looked at whatever is connected to the non-inverting output of your amp section. It is connected directly to the input of your inverting amp. Noise could be coming back in from the next section. Also, because it is DC connected, it could be affecting the bias point of the inverting amp.

[David Hess]

Why is your power supply generating so much 60Hz?  It should be 120Hz.  Is it grounded (and your signal source or load as well)?

It's not.  When I look at the power rails on the oscilloscope, even on the smallest setting, there is no visible ripple.  I actually see the noise of the regulator way more more than any ripple.   

If you look at the FFT plot, 60hz does dominate, but there is lots of harmonic content too, 120,180,240,300, etc...

If it was just lack of power supply and circuit ripple rejection, then the 60 Hz noise would not be there.  The full wave rectifier in the power supply produces 120 Hz ripple.

Don't understand how a ground loop could occour if every ground connection is ONE POINT.

Is the power supply ground connected to earth ground?  If so then when the oscilloscope probe ground is connected to the circuit ground, there is a ground loop through the oscilloscope and power supply and that will result in 60 Hz noise.  Even if the power supply output is completely isolated, capacitive coupling through the transformer may still cause this.