Neve output/line driver ( transformer feedback winding over opamp)

[KrIVIUM2323]

Hi,
I've been studying Rupert Neve design of output stage for some times now and i can't get my head over them.

I spent a bit of time looking at Focusrite Isa range and Amek 9098 but thanks to Whoops who recently posted 33114 schematic i would like to start from it before complexifying it further with Isa and Amek stage. It make a nice historical approach to R.Neve design evolution too imho.

https://groupdiy.com/threads/neve.44877/page-4#post-1165516

Here is how i understand things on 33114: it's a non inverting (discrete) 'opamp' ( ba440) with feedback taken from tertiary winding of output transformer.

Reference Level before output stage is -15db, post ba440 unbalanced output ( pre transformer) is -8db so there is 7db gain in the stage.
As it is non inverting stage, gain is equal to 20log( 1+(r64/r62))= 20log(1+(1600/1200))= 7,33db.

So far so good, from this we can deduce the tertiary feedback winding is 1:1 ratio. From brochure we know psu is +24v and max output level is +26db so here again we can assume the balanced output winding have a 1:2 ratio ( 24v rail should be able to provide +20dbu, 6db gain provided by transformer).
So out transfo is 1:2/1:1. Because i searched a bit i know it's the same ratio used in Focusrite Isa 110/red range and in the 9098 desk. And it makes sense R. Neve used the same transformer over the different successive designs ( why change something which work?).

What i don't get is the value of cap used in the feedback loop (c64, 470uf) and it's role. I suppose it's here to limit low end freq of feedback loop but how to calculate it/what parameters it define?

Could someone explain further and if i'm on right track?
Thanks in advance!

 

[merlin]

Check out the BA440 app notes:
https://www.technicalaudio.com/neve/neve_pdf/0440_BA440_class_AB_line_amplifier_doc_4pages.pdf

"The inverting input at pin 2 presents a very low input impedance, but is not d.c isolated. Consequently an external isolating capacitor must be used. This input may be used for mixing different signal sources via suitable resistors in series with the input capacitor."

It doesn't say exactly what the input resistance of that pin is, and I haven't bothered to find out. I presume C64 was simply made arbitrarily large to ensure it has no effect on bandwidth down to subterranean depths. It looks to me like R64 (1k6) dominates everything anyway, so using that for the calculation I get 0.2Hz. Any further resistance simply pushes this figure even lower.

 

[KrIVIUM2323]

Thank for your answer Merlin.

I hadn't seen the ba440 circuit description only the schematic and i'm limited in my knowledge so unable to calculate input z.

So basically this line driver stage is a 'regular' non inverting opamp circuit with the transformer included into feedback loop, that's it?

 

[merlin]

Yeah you can basically think of the BA440 as an opamp with a 15k built-in feedback resistor.
Over time it got replaced by the B640 which used an actual opamp.

 

[KrIVIUM2323]

Oh Nice! A 5534 followed by a discrete buffer stage. As used in Focusrite and Amek.

If this card was used (ba640) in place of 440 would the capacitor still be needed?

Edit: let me formulate differently, is this capacitor required in both circuit because there is gain in the stage? If a unity gain stage was used would it still be needed?

 

[merlin]

If this card was used (ba640) in place of 440 would the capacitor still be needed?

Yes, there is 12V DC there. You gotta block that DC!

If a unity gain stage was used would it still be needed?

Gain makes no difference; you can't hook the transformer up to something with DC sitting there.

 

[KrIVIUM2323]

Ok, it's way clearer to me!
Yes we don't want dc on transformer primary as it would saturate core and induce distortion.

I will prepare next schemo i would like to take a look at and attach it to next message

 

[merlin]

Here's how Neve did it with the B640:

 

[KrIVIUM2323]

I see two new introduction wrt 33114 into the schem you posted Merlin:
First one is a secondary feedback path, taken from the 640 direct output ( not using the feedback winding on output transformers),
Second one is the introduction of c119/r133 and sibling c121/r134.

C121/r134 are on 640 direct output and c119/r133 is used for both feedback path.

As i understand this two cells they are low pass filter. They limit high freq of feedback loop that's it?
It seems to me that both feedback path have same driving level (1:1) ratio.

Am i right?

I don't get why Rupert decided to include the second feedback path.
I'm also puzzled at how to define gain of stage: R132 (4,7k) i can identify as the resistor to ground ( the gain boost pin is a clue as it offer a way to // a resistor and change gain of whole stage ) but i don't get where is rF or how to define it. The two feedback path confuse me even more in this...

 

[KrIVIUM2323]

It looks to me like R64 (1k6) dominates everything anyway, so using that for the calculation I get 0.2Hz. Any further resistance simply pushes this figure even lower.

The idea of such a low fc being to be a decade away from useful audio range in order to minimize possible phase/freq response anomaly from pole created. Not really different to what is done into a dc servo cell. That's it?

 

[merlin]

Yeah I don't know why the feedback network was done that way, either

 

[KrIVIUM2323]

I wonder if it's not because of high freq issues ( limited bandwith) within the transformer itself?

In the schemo you linked ( from which module does it come from? It's interesting from chronological order pov) there is different value of c in c121/r134 and c119/r133 and in my understanding the one from the direct output being half the value from the one coming from the transformer feedback winding, so the low pass fc should be 2x the one used in the transformer feedback path.

Devil is in the detail and i think it's one of this details.

Attached is the Isa110 ( console version) output stage. I suspect it to use the same ratio transformer (1:2/1:1) than used in the 33114 and schem you linked. It's nicknamed 'the pumpkin' as it's toroidal.
Afaik it's the same used in the Amek 9098.

The difference with the schem you linked is there is now only one C/R cell and the fact there is a resistor which in my understanding modify the ratio of direct output negative feedback/transformer winding feedback.

 

[KrIVIUM2323]

And as the last schematic ( in chronological order) the 9098.

Have you spotted the difference between the two? The one in upper is repeated constantly throughout the service manual but... whith the limitations of my understanding it can't work as is. The lower one have more chance imho. Spent litterally hours to finally find this possibly working one in the service manual.

Transfo use same ratio as pumpkin. Of course it's drawn differently. 5543 exchanged for 33078. Otherwise it looks like simplified version of Focusrite one imho.

 

[KrIVIUM2323]

And to close the historical pov, here is a reverse engineered schem ( not by me) from RND Portico 511, i recently discovered.

Ouput transformer is a 1+1/1+1+1+1 wired in 1:2/1:1 ratio, there is still 2 negative feedback path and i think there is a third positive feedback one which appeared (the 10n7 capacitor landing between the two 2k2 resitors).

The 'silk' function is interesting, it brings some dc to saturate transformer core in a controled way and it vary something in the rc cell within the transformer feedback loop.

Edit: taken from the Portico description page, silk:

'So how does it work? By reducing negative feedback across the output transformer and adjusting this feedback’s frequency response, the Silk / Texture circuit provides the sweet, musical saturation found in Rupert’s legendary vintage designs – but with complete & precise control over the amount of classic tone you want.'

After having read many things Focusrite related, i came to conclusion it's something related to the 'air' function implemented in Focusrite audio interface which contains a mic preamp. It's described as recreating a +4db bell boost at 24khz inherited from the ISA 110 circuit. I think it is some interaction with the L part from one of both transformers ( mic input - as snubber values are different from the one recommended by Lundahl for the 1538- or the ouput one as it seems there is some definitive 'flavour' in the Focusrite output stage) in ISA 110 schem.

I would'nt be surprised if the two silk modes ( blue/red) in RND gear is individual switching of the dc bias part/reducing feedback over transformer.