Quick circuit noise question...

[bluebird]

Pretend you had two noisy old graphic EQ's. Say you could set them exactly the same electronically. Gain through both circuits at all frequencies the same.
Pass a signal through one and the out of phase (differential) signal through the other. Then sum the output for 6db gain. Does the noise of the EQ circuit cancel or double? The noise all those op amps cascaded make, not RFI. The hisssss.
I'm guessing it doubles because its not exactly the same noise? Because noise is random?

 

[ruffrecords]

The signals should increase by 6dB because they are 100% correlated (identical) but the two noise signals are uncorrelated (completely independent) so their powers will add rather than their voltages and the overall increase in noise will be 3dB. This is also the reason why wider tape tracks are quieter.

Cheers

Ian

 

[bluebird]

Ahh thanks, so there is a -3db noise advantage, but having two identical EQ's (or double the circuitry in one unit) to achieve it is most likely not worth it.

Or could you, add both the out of phase signals, get just the noise and then add that to the addition of the in phase signals, but out of phase and cancel all the noise?

 

[JohnRoberts]

Ahh thanks, so there is a -3db noise advantage, but having two identical EQ's (or double the circuitry in one unit) to achieve it is most likely not worth it.

only 3dB S/N improvement for every doubling of paths (cost)... quickly diminishing returns.

Or could you, add both the out of phase signals, get just the noise and then add that to the addition of the in phase signals, but out of phase and cancel all the noise?

No, but wouldn't that be nice...

JR

 

[boji]

Or could you, add both the out of phase signals, get just the noise and then add that to the addition of the in phase signals, but out of phase and cancel all the noise?

;D
Foolish musing ahead:
Supercomputer would have to know where each electron is in space within an imperfect but mapped geometry of conduit that makes electrons move where, then calculate probability of moving from p1 to p2 during time t,  times the number electrons,  then the electrons bump up against each other... I'd have to guess it would quickly become unsolvable over even a fraction of a fraction of timespace.   

So... will we have computational power eventually that can predict both position and velocity?  Or will it always be veiled in secrecy ie quantum indeterminacy?

Edit: Until then, may the low noise be with you. 

 

[bluebird]

Supercomputer would have to know where each electron is in space within an imperfect but mapped geometry of conduit that...

Dude, mind blown right there ;D

I'm not sure if the doubling up of eq's was a hypothetical thought experiment, but,  doubling up (or quad for -6dB)  on the semiconductor that's dominating the noise is sometimes worthwhile and certainly less expensive than doubling up on the whole kit and caboodle.

Mostly hypothetical, but could have some real world applications if a lot of amplifiers were to be used in series to achieve some kind of manipulation. Nothing that couldn't be done better in the digital domain, but yeah was just thinking.

Your saying If I paralleled 4 NE5534's instead of just one it would give me -6dB of noise? Or would that not practically work?

That just reminded me of the  NE5532 "OpAmplifier" power amp that always peaked my interest.  32 NE5532's in parallel for  16 watts.

Here's a link to a PDF of the article and schematics

https://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=&ved=2ahUKEwj82OaTvJ7qAhVMrp4KHYItDRkQFjAKegQIBhAB&url=https%3A%2F%2Fwww.physicsforums.com%2Fattachments%2Fdouglas_self_elektor_amp5532_compressed-pdf.254881%2F&usg=AOvVaw0nC43peZf7Po8gK7OQGR3O

Ha! After reading that article again there's a lot of cool tricks in the circuit. A really neat balanced input circuit. Also a couple added transistors to the power supply that will shut down the opposite rail if the other rail fails.
Who is this Douglas Self guy anyways ;D

 

[JohnRoberts]

Dude, mind blown right there ;D

science fiction... 8)

Mostly hypothetical, but could have some real world applications if a lot of amplifiers were to be used in series to achieve some kind of manipulation. Nothing that couldn't be done better in the digital domain, but yeah was just thinking.

Your saying If I paralleled 4 NE5534's instead of just one it would give me -6dB of noise? Or would that not practically work?

EIN noise voltage is lowered while input noise current increases... no free lunch but can make sense in some applications.

This is well covered in the art... "low noise design" by Motchenbacher and Fitchen (C.1973) described cost/benefit of paralleling active devices. Of course 553x didn't exist yet, or modern uber low noise bipolars. His examples used 2n4403 transistors in low noise designs. 

That just reminded me of the  NE5532 "OpAmplifier" power amp that always peaked my interest.  32 NE5532's in parallel for  16 watts.

Sounds like a way to get rid of excess op amp inventory  ;D

FWIW I have paralleled op amps before if say I needed one section to drive lower Z... TL07x are good GP op amps but only rated to drive 2k loads... If you have a spare section unused in a quad, it only cost another resistor or two to double up, as needed selectively. 

JR

Here's a link to a PDF of the article and schematics

https://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=&ved=2ahUKEwj82OaTvJ7qAhVMrp4KHYItDRkQFjAKegQIBhAB&url=https%3A%2F%2Fwww.physicsforums.com%2Fattachments%2Fdouglas_self_elektor_amp5532_compressed-pdf.254881%2F&usg=AOvVaw0nC43peZf7Po8gK7OQGR3O

Ha! After reading that article again there's a lot of cool tricks in the circuit. A really neat balanced input circuit. Also a couple added transistors to the power supply that will shut down the opposite rail if the other rail fails.
Who is this Douglas Self guy anyways ;D