noise in passive mixer

[solder_city]

hey- i want to make a simple passive mix to combine 8 to 12 line level sources, with level controls and maybe even pan pots. i need high (5k - 10k) input impedance and low (159 -600) output impedance. to minimize loss i figure i will need matching transformers on either inputs or outputs- stepdown input transformer and low impedance mix circuit, or high impedance mix circuit and stepdown output transformer. i m not sure of the best way to go. heres my thinking:

low z network means less resistor noise and less chance of noise pickup, but signal level will be lower due to stepdown input- so possibly no difference in s/n? also, more transformers needed (one for every input)

higher signal level with hi z network, but also higher resistor noise and possibility of noise pickup. also could be loss of bass response from high source z driving output matching transformer and or distortion from low signal level depending on core material. a very good trabsformer would have to be used here, advantage being only one needed per output. (inputs dont need transformers unless it has to be balanced).

your thoughts?

thanks

 

[JohnRoberts]

I don't see a specific question, but since you asked for thoughts....

Passive combining is lossy, but arbitrarily clean.

It is 6 of one, half dozen of another whether you passively combine X channels and then add make up gain to restore level, or actively combine X channels and suffer the noise gain of the summing amp.

I published a topology (two+ decades ago) using synthesized current sources in an active combining bus to knock down the bus noise gain some 20+ dB but like so what? In the real world any mix's noise floor will be dominated by one mic preamp at full gain, not to mention ambient noise, etc.

This strikes me as a new fashion, with some tail wind from digital mixers without enough clock ticks (or whatever) to get the digital job done. This too will pass as cheap digital gets better, or not.

JR

PS: I really wish there was some there there,,, I'd dust off my old analog mixer chops and pursue, but I don't see anything new or different.

 

[solder_city]

john i guess i could boild it down to this question: is one of the two *passive* approaches i describe 'better' than the other or is it six of one, half a dozen...

basically its a scenario of higher signal level with higher impedances or lower signal level with lower impedances.

i m not sure if thermal noise increases directly proportional to resistance, i ll have to look that up on google today. but i do know that a higher impedance node in a circuit is more apt to pick up noise (r.f.i. and so on).

the whole reason i want to do this is because i have a number of 'amplifier blocks' (preamps, line amps) i would like to put to work around my studio.

 

[JohnRoberts]

If 8x2 is the maximum configuration you can expect -18dB of loss from resistive combining of 8 sources. The simple one pot pan control with reasonable pan law has -10 db insertion loss, so were up to -28dB. Mix level faders with audio taper are typically set up with nominal zero at -10dB so in a fully passive topology we are down to -38 dB. Note: special taper (dual) pan pots, could reduce that 10dB insertion loss which is worth looking into, especially in this case.

Without active I/O, and knowledge of what sources look like 1:1 input transformers could help manage crosstalk between inputs and lack of differential for each input. Without transformers input grounds could be summed (averaged) and referenced forward to the output.

If the sources can drive 600 ohms cleanly, it seems like 1k faders, 5 or 10K pan pots, and say 15k summing resistors would work reasonbly well. The final source impedance of around 2k with almost 40dB makeup gain required doesn't seem like a sweet spot for typical mic preamps S/N gain stages.

It seems a larger configuration, say 24-36 channels might be more compatible with conventional mic preamps sweet spot.

An output step up transformer with maybe 3x turns ratio would get the source impedance friendlier for opamps and reduce the makeup gain needed some 10 dB, back down to 28 dB which is marginal high but probably OK from a modern high performance opamp. Note: I haven't studied what the optimal source impedance is for high performance opamps in decades so a modern low noise Bifet might like an even higher source impedance and so needing even less make up gain with a higher turns ratio transformer.

If you are willing to tweak, one of the common discrete mic preamp topologies could be dialed in for 2k source impedance and only 38dB of gain. (hint: run input devices at lower current density to get input noise current contribution lower.)

Caveat: This is off the top of my head, so not up to my $100/hr design integrity. In other words, I may have made mistakes in my back of the envelope analysis that would be revealed by actually drawing out and analysing circuits, building prototypes, etc.

Good luck. IMO it might be cheaper and easier to buy a decent old console and just upgrade to modern electronics. Some of those new generation opamps are quite good, but where's the fun in that?

JR

PS: AFAIK thermal noise in resistors is proportional with R.

 

[emrr]

I've got an interesting one that should really be a separate thread, but does relate: a 16x2 passive mixer using Langevin 600/600 ladder faders, followed by dual-10K pan pots, feeding mid-70's era RCA IC output section with discrete output transistor stages. It has a separate 12 input tracking section using the same faders, which I am hoping to combine for 28 channels. Luckily there is 1/4" square copper buss bar for the entire system common, which I understand can help a good bit.

Nice (I think) to have the ladder faders, which seem to solve many of the impedance related issues related to running a 600 ohm system, but the pan pots are a curve ball I don't find reference to in any notes I have. Unless referenced to 600 ohm ladder pan pots, which are truly rare unwieldy beasts. My assumption without doing any of the math (bad me) is that the pan pots bump the mixing system Z back up from the low point of the ladders to a higher point, but maybe not significant given the # of parallel channels and the relatively fixed low Z of the ladder output. I've not entirely reverse-engineered the path to the IC's, so there's a bunch of missing info still.

Aren't there plenty of 70's era Neve mixing circuits that would be good study examples? I've none on hand to look at.