Negative Feedback Compressor

[abbey road d enfer]

That's exactly it!

 

[separateness]

I have a few additional questions about the circuit posted so generously above by @plexi , I hope it is not too improper for me to ask them here. I do not mean to second guess the engineers who devised this device but to understand why certain decisions were made.
1. Is it necessary that the compression portion of the BA5A be balanced? If so I imagine that it is to avoid thumping but I would not think that a little AC feedback applied gently to the cathode of a tube would make it thump. We aren't jerking the bias point around like in a varimu, just cancelling some G-K potential.
2. Why is the control voltage sampled from the input side of the circuit. Wouldn't taking it from the output make more sense?
3. Why the two nearly back-to-back transformers?

 

[abbey road d enfer]

1. Is it necessary that the compression portion of the BA5A be balanced?

Yes.

If so I imagine that it is to avoid thumping

Yes.

but I would not think that a little AC feedback applied gently to the cathode of a tube would make it thump.

More or less correct.

We aren't jerking the bias point around like in a varimu, just cancelling some G-K potential.

Instead of jerking the grid voltage with a few dozen volts, it jerks the cathode current in a way that produces similar thumps. Actually, the goal of the topology is to minimize thumps, goal achieved.
There is also DC flowing through.

2. Why is the control voltage sampled from the input side of the circuit. Wouldn't taking it from the output make more sense?

It's a feed-forward compressor with pre-delay. The control voltage is taken pre-delay.

3. Why the two nearly back-to-back transformers?

Because the designer thought it was the best way to insert delay and attenuation.

 

[plexi]

The basic circuit has to be balanced for all the reasons above. The original GE schematic is certainly strangely nuanced in spots....thank you for elaborating abbey road... but if we disregard the input amp/time delay LC/2 transformer pad, the line amp is really simple.

The GR happens with the pair of 6J5's in the feedback path. As their negative grid voltage comes up to around -24 they start to conduct and reduce gain.

They are biased somewhere at the beginning of this curve or slightly into it (slope control R90 3.3K pot).

Sidechain is negative powered and neg rectified pulsed, but inverted by the 6SN7 so it references negative but moves positive towards ground (conducting 6J5's and reducing gain)

I built the 12AU7 proto just to understand it's operation and see what it was capable of. Here is a schematic and gain/V bias curves with different tubes in the GR positions, you can see the AU has the most linear gain range. In the proto I have it biased at around -21V (a little into the knee)

The sidechain is really the trick to all of it, I made it like the original with mosfets subbd for the tubes. Gain/Phase split/current follower into negative rectifiers and timing circuit. I did use the final triode to buffer the signal.....here is the tricky part of the whole thing...the triode runs out of gas and cant pull all the way to 0V but it is what is required to stay out of the cutoff zone on the graph. So I did have to tweak the plate/cathode R relationships to get it right

It is AC and DC feedback into the cathodes, kinda strange and somewhat limited in operation. Too much and the DC component takes over and pushes the 1st stage into cutoff. But over 14 dB or so, it's good

 

[plexi]

Copy of GE spec sheet with GR curve

https://www.preservationsound.com/2011/05/pro-audio-hardware-of-the-early-1950s/
Read comments at end, good description of the HF phase time delay network and reasoning behind it

GE BA5A original thread from 2007 with 2017 restoration update from emrr

https://groupdiy.com/threads/ge-ba-5a-vari-mu-limiter-guided-tour-needed-any-info.20588/

 

[abbey road d enfer]

Copy of GE spec sheet with GR curve

https://www.preservationsound.com/2011/05/pro-audio-hardware-of-the-early-1950s/
Read comments at end, good description of the HF phase time delay network and reasoning behind it

Several misconceptions and misunderstandings in the various comments.
Those by Howard Compton are spot-on, though, particularly the one regarding the All Pass Filter.
APF's have been commonly used in broadast to modify the crest factor of spoken voice, by displacing harmonics.
In the BA 5A, it's also used as a program delay, while the side-chain is undelayed, allowing the unavoidable rectification delay to be compensated, resulting in pseudo instant attack time.
However, one of his comments says "This is a feed-back limiter.", which is ambiguous.
It's certainly using variable negative feedback for gain control, but in common parlance, a feedback compressor refers to one that takes the control voltage from after the gain reduction cell. Actually, the BA 5A takes its control voltage from the input stage, before the gain reduction cell, which makes it a variable feedback feed-forward limiter, or maybe a feed-forward variable feedback limiter...?

 

[plexi]

Yea some comments are not accurate to this unit. They got mixed up with a different weird GE limiter from that era... But Howard Compton's description yea thats how it all works.

abbey rd - NFB feed forward limiter LOL we struggle with the words... Seems like a strangely familiar but uncommon way of doing it

 

[Script]

On that magazine page they call it:

'inverse feedback gain reducing' circuit design

Maybe call it a
feedforward controlled variable subtractive feedback gain amplifier

?