Author Topic: Decoupling dilema  (Read 1460 times)

0 Members and 1 Guest are viewing this topic.

Offline paulca

  • Super Contributor
  • ***
  • Posts: 2116
  • Country: gb
Decoupling dilema
« on: April 19, 2018, 02:47:29 pm »
So I have a noisy power rail for an amplifier circuit which uses a virtual ground.

When I place a large decoupling cap on the power rail the noise is removed from the power rail but deposited on the ground.  The amplifier inputs are ground referenced and I still have noise output by the amplifier.

Similarly if I don't decouple the power rail the power rail noise is present on the virtual ground and is amplified by the amplifier.

What seems to work is to not decouple the power rail, but to only decouple the virtual ground.  But it sounds fishy to me.

How does one deal with this issue, or do I have some other issue?
"What could possibly go wrong?"
Current Open Projects:  3 Channel Audio mixer with DAC, BT, pre-amps and h/phone amp, WS281x LED controller Version 2 5V/5A w/Atmega328, FY6600 Power supply, 5A DC Load (Still!)
 

Online blueskull

  • Supporter
  • ****
  • Posts: 13214
  • Country: cn
  • Power Electronics Guy
Re: Decoupling dilema
« Reply #1 on: April 19, 2018, 02:53:09 pm »
1. Grounding design is an art. You need better grounding design.
2. Inductor helps.
 

Offline Audioguru

  • Super Contributor
  • ***
  • Posts: 1508
  • Country: ca
Re: Decoupling dilema
« Reply #2 on: April 19, 2018, 03:21:04 pm »
The virtual ground passes the noise. An amplifier is supposed to have good supply noise rejection.
The virtual ground should feed a very low current like only to the (+) input of an opamp. Then the "half the supply voltage reference" can be simply two 100k resistors in series from the supply and a reasonable filter capacitor value like 22uF.

Please post the schematic of the amplifier and how it is biased with a virtual ground.
 

Offline madires

  • Super Contributor
  • ***
  • Posts: 4957
  • Country: de
  • A qualified hobbyist ;)
Re: Decoupling dilema
« Reply #3 on: April 19, 2018, 03:32:22 pm »
In case you like to see the circuit and read the original thread: https://www.eevblog.com/forum/beginners/circuit-review/
 

Offline paulca

  • Super Contributor
  • ***
  • Posts: 2116
  • Country: gb
Re: Decoupling dilema
« Reply #4 on: April 19, 2018, 03:41:03 pm »
In case you like to see the circuit and read the original thread: https://www.eevblog.com/forum/beginners/circuit-review/

Thanks.

I recreated one channel of the circuit on a breadboard and was able to recreate the noise.

I tried lots of things to remove the noise being amplified off the power rail, but the thing that was almost perfectly successful was removing C3 completely.
"What could possibly go wrong?"
Current Open Projects:  3 Channel Audio mixer with DAC, BT, pre-amps and h/phone amp, WS281x LED controller Version 2 5V/5A w/Atmega328, FY6600 Power supply, 5A DC Load (Still!)
 

Offline paulca

  • Super Contributor
  • ***
  • Posts: 2116
  • Country: gb
Re: Decoupling dilema
« Reply #5 on: April 19, 2018, 03:43:28 pm »
The virtual ground passes the noise. An amplifier is supposed to have good supply noise rejection.
The virtual ground should feed a very low current like only to the (+) input of an opamp. Then the "half the supply voltage reference" can be simply two 100k resistors in series from the supply and a reasonable filter capacitor value like 22uF.

What about the feedback divider?  Ground or 1/2 Vcc VGround?
"What could possibly go wrong?"
Current Open Projects:  3 Channel Audio mixer with DAC, BT, pre-amps and h/phone amp, WS281x LED controller Version 2 5V/5A w/Atmega328, FY6600 Power supply, 5A DC Load (Still!)
 

Offline Cerebus

  • Super Contributor
  • ***
  • Posts: 4427
  • Country: gb
Re: Decoupling dilema
« Reply #6 on: April 19, 2018, 05:03:08 pm »
Here's what you've got, according to your circuit diagram:



Obviously I've just used a TL07x OPA because I don't have a model for the OPA551 handy. I've injected 0.1V of noise at 10 kHz into the power rail, and this is what you get as output with 100 mV input at 1kHz:



Here's the circuit reworked with a proper active virtual ground, and bypassing on the voltage divider that virtual ground is derived from:



There the second OPA could really be a TL07x or any other suitable OPA, it doesn't need to be a beast like the OPA551. And here's the output under the same conditions:



Noise gone! The issue was the impedance of the virtual ground, and adequate filtering on the source of that virtual ground.

« Last Edit: April 19, 2018, 05:05:38 pm by Cerebus »
Anybody got a syringe I can use to squeeze the magic smoke back into this?
 

Offline Kalvin

  • Super Contributor
  • ***
  • Posts: 1806
  • Country: fi
  • Embedded SW/HW.
Re: Decoupling dilema
« Reply #7 on: April 19, 2018, 05:12:06 pm »
If you connect Vin to virtual GNDA you will not have a problem with ground noise. For some circuits you can reference the input to virtual GNDA, but some circuits may be constructed so that you have to use the common ground and then you may have this kind of noise problems. Selecting your ground matters.

Edit: Fixed typo: s/AGND/GNDA/.
« Last Edit: April 19, 2018, 05:41:56 pm by Kalvin »
 

Offline paulca

  • Super Contributor
  • ***
  • Posts: 2116
  • Country: gb
Re: Decoupling dilema
« Reply #8 on: April 19, 2018, 06:12:03 pm »
Here's the circuit reworked with a proper active virtual ground, and bypassing on the voltage divider that virtual ground is derived from:

Thank you very much.  I did try an active virtual ground on breadboard last night, but I didn't consider giving the divider that much resistance. It didn't occur to me to split the divider into 4 and bypass between two of them like you did.

Quote
There the second OPA could really be a TL07x or any other suitable OPA, it doesn't need to be a beast like the OPA551. And here's the output under the same conditions:

When I read stuff about how much current the buffer needs to handle people say it needs to be as beefy as the amp, but when I look at it I can't understand why.  The current path for the main amp will mostly be via it's power rails and the feedback which has a 10K in it is the only route to the VGND.  So ... surely the virtual ground buffer only needs to supply the feedback loop with current when the amp is sinking it.

Surely an NE5532 would do for the buffer amp?

Quote
Noise gone! The issue was the impedance of the virtual ground, and adequate filtering on the source of that virtual ground.

I did try and filter the source with large caps, but each 470uF cap only reduced the noise by about 1/3, when I had 3 of them on the main power rail and I still had noise I knew something else was wrong.  LOL
"What could possibly go wrong?"
Current Open Projects:  3 Channel Audio mixer with DAC, BT, pre-amps and h/phone amp, WS281x LED controller Version 2 5V/5A w/Atmega328, FY6600 Power supply, 5A DC Load (Still!)
 

Offline Cerebus

  • Super Contributor
  • ***
  • Posts: 4427
  • Country: gb
Re: Decoupling dilema
« Reply #9 on: April 19, 2018, 07:02:37 pm »
Here's the circuit reworked with a proper active virtual ground, and bypassing on the voltage divider that virtual ground is derived from:

Thank you very much.  I did try an active virtual ground on breadboard last night, but I didn't consider giving the divider that much resistance. It didn't occur to me to split the divider into 4 and bypass between two of them like you did.

That ought to be three resistors, I just couldn't be bothered to coalesce the bottom two; out of curiosity I'd tried something else that needed the extra resistor first.
Quote

When I read stuff about how much current the buffer needs to handle people say it needs to be as beefy as the amp, but when I look at it I can't understand why.  The current path for the main amp will mostly be via it's power rails and the feedback which has a 10K in it is the only route to the VGND.  So ... surely the virtual ground buffer only needs to supply the feedback loop with current when the amp is sinking it.

Surely an NE5532 would do for the buffer amp?


That would be fine. The virtual ground needs a low impedance, it doesn't need to actually source or sink much current. The magic of negative feedback reduces the 10s of ohms of actual output resistance of the op amp into the milliohm region.

Quote
I did try and filter the source with large caps, but each 470uF cap only reduced the noise by about 1/3, when I had 3 of them on the main power rail and I still had noise I knew something else was wrong.  LOL

The trick is adding some significant resistance before the filter cap. That trick of splinting the top divider resistor and decoupling there can be seen in almost every discrete class AB power amplifier design ever made to get a quiet source for bias voltage from unregulated power rails.
Anybody got a syringe I can use to squeeze the magic smoke back into this?
 

Offline paulca

  • Super Contributor
  • ***
  • Posts: 2116
  • Country: gb
Re: Decoupling dilema
« Reply #10 on: April 20, 2018, 02:32:02 pm »
Spotted this...

https://www.tubecad.com/2018/02/blog0412.htm

Seems there are many designs for virtual ground splitting!
"What could possibly go wrong?"
Current Open Projects:  3 Channel Audio mixer with DAC, BT, pre-amps and h/phone amp, WS281x LED controller Version 2 5V/5A w/Atmega328, FY6600 Power supply, 5A DC Load (Still!)
 

Offline paulca

  • Super Contributor
  • ***
  • Posts: 2116
  • Country: gb
Re: Decoupling dilema
« Reply #11 on: April 30, 2018, 08:46:56 pm »
Thanks for the help fixing this up guys.

Here is the new version with it's new VGnd Opamp and decoupling caps.  Works a treat.

"What could possibly go wrong?"
Current Open Projects:  3 Channel Audio mixer with DAC, BT, pre-amps and h/phone amp, WS281x LED controller Version 2 5V/5A w/Atmega328, FY6600 Power supply, 5A DC Load (Still!)
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf