Summing two Neve paths into one: Virtual Earth, Transformer, or?

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

atavacron

Well-known member
Joined
Jan 28, 2009
Messages
344
Have a look at these two options for summing two low-impedance unbalanced outputs into one high impedance input (schematic attached). It's a particularly Neve-ish problem, and I couldn't find any documentation on the web of 2-into-1 summing in an actual Neve console (or any less than 8-into-1 for that matter). If anyone's got a schematic of the ways the Mono amp got fed in the 53XX or 80XX desks, I'd love to see it.

There are some conditions here. The connections to and from the summing stage (at the dotted lines) have to be the same level. They may be shielded wires of up to several feet. They may go through some switching architecture to accommodate alternate routing. They have to stay in phase, thus you can't use the invert-the-balanced-output trick. And of course we're looking to maintain all the headroom we can, while minimizing noise gain.

Normally (on a Neve bus at least), you'd do a voltage mix into the next module's 10468 transformer and move on, with 6dB free gain as you head to the next stage. But in this case, the summing resistor values would be too low, the shunt too high, the noise gain unacceptable. Sub-5k feed resistors are a no-no, I understand, if you want to send signal more than a few inches afterwards. Also, we're not trying to add a whole 'nother module, we're just trying to compensate for the sum.

Seems like the solution is virtual earth summing, i.e., mixing into the inverting input of the op amp. In the case of the back end of a 3405, 3415, or 3416 module, that means returning into the Gain Boost pin, and inverting the balanced output. Or you could DIY something with a B340 / B440 and an LO1166 / LO2567, which might sound better, and you could use the same tactics. But we're trying to maintain phase before we hit the back end of the module or DIY thing. Ok, fine, so one would add a BA489/M or a BA384 in the middle, anything that'll hold two 438s.

Are 438s stable when inverted down around +3dB or +6dB makeup gain? How about unity?

What would happen if you used a 438 as the first inverting stage, and a 440 as the second? There's greatly increased line driving capability in the 440, but does that risk harm to the output stage 440? Why don't we see 440s used in the front half of modules that don't require 438-level gain?

Aren't we supposed to use a 406 for unity? How does one invert a 406? Were they even used on any of the dual-opamp daughter cards?


Due to the high cost of the vintage 10468, Joe Malone did some tests of alternate input transformers for mic preamps, and the best substitution seemed to be the 31267, wired in reverse, with only one winding connected on each side. (For reference, the normal series-series use is 10k:600, with 12dB loss). He talks about how the 31267 may just be two 10468s wrapped around each other, and how the two transformers seem to have similar headroom. They start to distort around +7dB, which I don't think is a level that a 31267 in this position would ever see. There is also discussion of repurposing the other windings as additional inputs, which got me started. So in this case, I'm using the two 1k2 windings as inputs, and one of the 4k8 windings as the output. The other 4k8 winding is terminated. The voltage is doubled and the current halved at the output.

What is required of the B440 when it sees the summed signal? Is there enough drive coming in? Does the 440 have the same oomph into the output transformer regardless?

There is also this, from Malone: "WARNING. Using only one primary and secondary coil on the 31267 disables the mechanical hum bucking due to there being two separate bobbins on the U laminates that sit mechanically on opposite sides and directions of the formed O core."

Does using both inputs take care of that concern?

Are there other ways of going about this in a way that would be approved by Neve engineers in the '70s?

And finally, feel free to correct my schematic, or perception of reality, or both.
 

Attachments

  • Transformer-vs-Virtual-Earth-Summing.jpg
    Transformer-vs-Virtual-Earth-Summing.jpg
    810.8 KB · Views: 116
Whoops, input level should be labeled as -30dB instead of -35dB.

Also, there is a Neve doc that details the use of a 10468 in reverse, series on the high side and parallel on the low side, for 10K:600.

Also also, I'd take any tips about values surrounding the two inverting 438s.
 
Cinemag has an unlisted trafo that may accomplish this with higher headroom and better transparency than the 31267, the mu-metal can, high nickel lam CM-17310. Schematic for this attached.

I’m not seeing any trifilar input trafos on the Jensen site, maybe I’m looking wrong.
 

Attachments

  • 892BFC5D-D281-4514-972E-D0536A499A45.jpeg
    892BFC5D-D281-4514-972E-D0536A499A45.jpeg
    583.7 KB · Views: 77
I’m not sure about this, and I’m not familiar with that particular transformer, but it seems to me feeding a transformer with a 1k+ source impedance is a recipe for distortion...and low end loss... has to be 10H primary just to be -3dB at 20Hz
 
dogears said:
I’m not sure about this, and I’m not familiar with that particular transformer, but it seems to me feeding a transformer with a 1k+ source impedance is a recipe for distortion...and low end loss... has to be 10H primary just to be -3dB at 20Hz

The 1k2 resistors in series with the primary would change the source impedance in that manner?
 
I think it is unnecessary to over complicate this. If you have two low impedance sources you want to mix into a high impedance inputs then all you need is a pair of equal valued resistors, one connected to each low Z outout and their other ends both connected to the high Z input.

Cheers

Ian
 
atavacron said:
The 1k2 resistors in series with the primary would change the source impedance in that manner?

Yes.  But whether it matters or not all depends on the transformer, and I don't know what that Cinemag is designed for. 
 
I've run the 10468 in reverse. Hairball audio uses a switching system in their Lola Mic preamp so one can have a line in for DI. They use an EA10468 which is Ed Anderson's remake with same wiring/pin out. They have it in the Lola build docs.
 
ruffrecords said:
I think it is unnecessary to over complicate this. If you have two low impedance sources you want to mix into a high impedance inputs then all you need is a pair of equal valued resistors, one connected to each low Z outout and their other ends both connected to the high Z input.

Cheers

Ian
+1
 
atavacron said:
Sub-5k feed resistors are a no-no, I understand, if you want to send signal more than a few inches afterwards.
I would say it's good to drive about 6 meters of shielded cable.

Also, we're not trying to add a whole 'nother module, we're just trying to compensate for the sum.
I see you consider mixing two sources results in 6dB attenuation. Most of the times, it's not true. Mixing two totally uncorrelated signals of same level results in -3dB. Passive mixing two highly correlated signals (such as the L & R of a stereo signal) results in very little signal loss.
 

Latest posts

Back
Top