Selecting an interstage transformer for passive summing

[bjosephs]

Hi All,

I wasn’t too happy with my last summing build and want to go back to the drawing board for something with more color. This time I want to try passive into a transformer followed by a make up stage of some kind (probably tube). I’ve been scouring the many threads on this matter and I’m struggling to patch together the advice given on transformer attributes so that I can start to whittle down what’s out there on my own.

Passive mix networks seem to include a terminating resistor of a few hundred ohms. I believe there is a relationship between this termination and the transformer’s expected source impedance. What I didn’t see a pattern on was if a low impedance transformer was better than high (say a mic input transformer or a 600:600 vs a 10k). Also in some places a step down transformer was recommended or selected despite increasing the work done by the makeup gain. I’d have thought step up would be the default as long as the next stage had sufficient Z in. One thing that is clear is that signal levels are low and small transformers (perhaps even mic level) are adequate. I also see that for best low end response a high inductance and low source impedance are preferred.

I’m shooting for 12-16 stereo pairs, maintaining low end magnitude is important but I welcome a little THD across the whole spectrum. Make up may be tube (I’m thinking an SRPP DC coupled to a line driver) or maybe something op amp based because I really want to experiment with a controlled DC trickle into the output transformer like the Neve Silk circuit. First things first, however, how do I start to hunt down a transformer?

 

[ruffrecords]

A few things to remember about passive summing, especially with tube gain make up.

1. If you only have stereo pairs you don't really have to worry too much about crosstalk. If you have some mono inputs (red equally to left and right) you do need to consider it.
2, Crosstalk depends on the source impedance of the output driving the mix inputs, the mix bus resistor values and the slugging resistor values. You want very low source impedance, and a reasonably high ration of mix bus resistor to slugging resistor.
3. The ratio of the bus feed resistors to the slugging resistor also affects the interaction between channels - disconnecting on can increase the overall output level slightly. With 16 channels this should no be a problem.
4. With typical 2 x 5K input resistors and a 200 ohm slugging resistor, the bus loss will be 34dB. You will therefore need about 34dB of make up gain.
5. Tubes are noisy compared to most semiconductors. For a tube gain make up amp you therefore need as much virtually noiseless gain as possible before the tube amplifier itself. This usually means a step up transformer. For it to significantly improve the noise performance of the tube make up amplifier it needs a turns ratio of at least 1:5 (14dB gain).
6. The signal level at the transformer input willbe quite small. Usually too small for the transformer to have an audible effect. With a single input the level at the transformer primary will be -34dB.Even with 16 inputs blasting away it unlikely to be much more than 12dB higher (-22dBu).
7. Colour tends to occur at higher signal levels. So output tubes and output transformers are where they most commonly occur.
8. An SRPP stage is a good choice for a colourful gain make up stage.

Cheers

ian

 

[bjosephs]

Hi, Ian

Very cool, thanks. I have a couple JT-115k-e mic transformers that would probably be a good starting point. Jensen stuff is usually quite clean but if the interstage point isn’t the place to expect much color anyway then no reason not to start there. 20db of free gain as well.

What are the gotchas on mono inputs? Make sure to ground those connections if unused? I was planning to do most of this on DB25 connectors but I can put a bank of switching TRS in there specifically for mono signals.

Output impedance from the interface is 100 ohms. pretty good.

Thoughts OT brands/series for my needs? I think it’ll have to be a 10k:600 or similar. I’m guessing plain steel core with high inductance if I can find such a thing. Preferably one that fits in 1U.