Grounding and bussing in a high-Z mixer

[NewYorkDave]

I've been wanting to build a tube-based mixer for line-level signals. For the purposes of this discussion, let's ignore the question of whether or not this is a pointless thing to do and pretend it's 1955 :wink:

The classic approach of using lots of tubes and transformers and a low-Z, constant impedance mixing network is far from economical. So I've opted for a high-Z passive network feeding directly into the grids of tube booster/program amplifiers. The source Z looking back into any of the pots in a channel is 0-20K (depending on setting) and the mix resistors are 100K. The impedance of each mix bus--which depends on the # of channels connected and any other resistance shunted across the bus--will fall somewhere between 5K and 10K in my build.

The maxim of "ground follows signal" is easily implemented in a single-input, single-output system (e.g., preamp, guitar amp) but the best way to do it become less obvious in a multi-input, multi-output system like a mixer. The physical details of the bussing become particularly important in a high-impedance network. I've prepared three diagrams to show the options I'm considering. These are outlines, not detailed schematics, but they show enough to get the idea across.

PDF 1

PDF 2

PDF 3

Diagram 1 is, in my estimation, the best approach overall in terms of noise and crosstalk. Diagram 3 is much more convenient in terms of construction, and Diagram 2 is somewhere in-between.

In all cases, the thick bus lines represent heavy-gauge (#14) bare bus wire. Shielded, twisted pairs (diagram 1) are low-capacitance, "AES compatible" type. Coax (diagram 2) is small, low-loss video coax such as Belden 1855A. Connections from amplifiers to ground bus are made with #16 or larger hookup wire.

Comments invited :sam:

 

[emrr]

Cool; I'll think about these and come back.

I guess you've looked into the Western Electric 23 mixer which stayed Hi-Z at the busses? Not many channels, but larger physical layout. RCA BN-2A remote mixer used hi-Z bussing with multiple preamps. The tube Neve I worked on was Hi-Z also, and not very large buss wire.

 

[emrr]

Okay; I'll bite on this black art of grounding.

Diagram 1: I don't like that the input ground is connected directly to R buss ground, and no symetrical reference to L buss ground. Maybe this helps crosstalk? Maybe it's an error in the drawing? Maybe it's an error in my brain? I'd rather see it go to the main ground busswire, but then you get the L and echo busses further upstream from the inputs and the R buss.

I'm thinking even more complex could be good. I like the 14 gauge for the busses in diagram 3. What about going all out and having 14 gauge busses for 1) input ground 2) left buss ground 3) right buss ground 4)echo buss ground 5) summing amp / power / chassis ground 6) left buss 7) right buss 8)echo buss. With the first four ground busses hitting the summing amp / power / chassis buss in a star fashion.

What a pain; would it help anything? It's certainly not the easy choice.

 

[NewYorkDave]

I don't have any documents for the WE 23 mixer.

Diagram 1: I don't like that the input ground is connected directly to R buss ground, and no symetrical reference to L buss ground.

Look closer. The grounds from the input channels for L, R and Echo all connect, essentially, to the same point on the ground bus. There's a difference of maybe a couple of inches of heavy wire. (Note that the bus line bends upwards and terminates at the R booster amplifier).

Maybe I'm misunderstanding you, but using "car battery cable" for all connections would look impressive but it's unnecessary when you consider the tiny currents involved in the mixer circuit proper (not counting the amplifiers). Even #14 bus wire and #16 links to the bus from the amplifiers is overkill, but still within the realm of physically manageable. Multiple #14 or #16 runs going to each channel would just be gratuitous.

If you're talking about #14 busses (fed by all input channels) with separate ground busses for Input L, R and Echo... well, that's not a bad idea, but probably still on the side of overkill.

Using bus bars, as opposed to individual runs to each channel, is very attractive in terms of ease of construction and cabling density. But a ~2 foot-long piece of wire hanging off a high-impedance node, followed by 40 or 50dB of gain just seems too tempting an offering to the Gods of Noise :wink:.

The bus bar is not worse in terms of high-frequency rolloff due to parasitic capacitance, at least not compared to shielded cable, although locating the mix resistors at the amplifiers does isolate the cable capacitance from the amplifier inputs.

I'll say I haven't ruled out going with Diagram #3, but it does make me a little nervous.

Another option I considered was scaling down all resistances in the mix network by 10 and using 10K:600 input transformers. Mixing at low impedance would alleviate some of the potential issues with noise--but this is negated to some extent by having to provide another 12dB of gain at the back-end.

Step-up transformers at the inputs of the booster amps would really be needed for low noise if the source impedance of the mix busses was down in the ~600 ohm region. For the sake of fidelity, these couldn't be $10 M6-core jobs (because of the very low signal levels involved).

Thanks for the comments, keep 'em coming :sam:

 

[Jim Zuehsow]

NYD,

I just posted the WE 23 mixer at:

http://twin-x.com/groupdiy/displayimage.php?pos=-810

Not the best copy, but the only one I have with me right now.

JZ

 

[NewYorkDave]

Thanks, Jim!

It's a little hard to read, but better than nothin'! :thumb:

 

[emrr]

Probably as good as mine, which ain't scanned yet.

 

[emrr]

[quote author="NewYorkDave"]

Diagram 1: I don't like that the input ground is connected directly to R buss ground, and no symetrical reference to L buss ground.

Look closer. The grounds from the input channels for L, R and Echo all connect, essentially, to the same point on the ground bus. There's a difference of maybe a couple of inches of heavy wire. (Note that the bus line bends upwards and terminates at the R booster amplifier).[/quote]

I guess I'm seeing it as left and right pan grounds connecting to the left and right amps respectively, and the input ground following through the right buss ground to get to the #14 ground. May not matter at all that it's closer to the right than the left; still rather see the input ground go directly to the #14. Maybe just semantics with no basis in reality.

[quote author="NewYorkDave"]
If you're talking about #14 busses (fed by all input channels) with separate ground busses for Input L, R and Echo... well, that's not a bad idea, but probably still on the side of overkill.[/quote]

I like it! Again, probably total overkill. How many inputs is this thing? How physically large? Isn't that were these tweaks really become crucial? I recall hearing about the mid-90's rush to marry multiple Neve's into 'super consoles', and the main bussing and grounding was always a nightmare because of the physical scale. I swear I heard about 1" thick copper bussbar being used to join formerly seperate mix busses together and get the noise and crosstalk right. But, we're talking about 96 input 20 foot long consoles with passive low level mixing; low Z mixing also.


[quote author="NewYorkDave"] although locating the mix resistors at the amplifiers does isolate the cable capacitance from the amplifier inputs. [/quote]

Good point; but what's the real trade-off between that and resistors directly hitting #14 mixing buss that travels to the amps? Which way is the capacitance worse? You certainly can't shield the #14 so easily, but how much does it matter? What are the real world levels we're talking about? Hot enough to overcome how much interference?
[quote author="NewYorkDave"]
Another option I considered was scaling down all resistances in the mix network by 10 and using 10K:600 input transformers. Mixing at low impedance would alleviate some of the potential issues with noise--but this is negated to some extent by having to provide another 12dB of gain at the back-end.

Step-up transformers at the inputs of the booster amps would really be needed for low noise if the source impedance of the mix busses was down in the ~600 ohm region. For the sake of fidelity, these couldn't be $10 M6-core jobs (because of the very low signal levels involved).
[/quote]

If we're talking reasonably modern digital recording line level, then I doubt the 12db stepdown would be any real issue compared to the interference benefits. And 3 higher grade step-ups for the amps would be a small price to pay.


I'll stop. Hopefully this helps with some ideas while I'm talkin' out my ass here.

 

[Jim Zuehsow]

I built a small tube mixer back in the early 80 just for the heck of it. Nothing fancy, but very similar to a pair of Altec tube mixers. I did have a cathode follower in the output of each mic pre that drove the pot and pan circuit. It was reasonably quiet and worked well with the average analog tape recorder's noise floor.
By the way, most of the "giant" consoles did have huge ground busses, but they also sub-mixed groups of channels so as to not have the noise gain problem you get combining a large number of channels into one mixing circuit. Neotek, I think it was, claimed their mixer had "special" circuitry to eliminate the problem by using current mixing instead of voltage mixing, but it seems to me that if you're running into a resistor feeding the opamp, you have changed from voltage to current mixing! What's up with that? They did have a pair of transistor emitters being fed before the opamp, so I guess that was their secret weapon.

 

[gnd]

Dave.
If I understand your three diagrams properly, diagram 1 and 2 is actually taking hi level signal from each module through shielded cable, and summing resistors are placed at summing board? I like this idea, I thought about it for my mixer. (because placing summing resistor at each board lowers signal, then it is sent through buss wire, and amplified again with added noise.)

Going full level to master summing board surely seems better. Somehow I preffer grounding on diagram 2 with shielded cables carrying only signal. Thus you can ground all channel modules separatly with thick cables to one point of choice in mixer, because mix amp board may have space limitations to feed a junge of cables to it.

Will be plenty of cables? How many channels? If plenty channels, how about submixing channels to groups (4 or 8 channels) near channel boards with local summers, then sending submixes by screened cables to final mix board with master summing amp? I think it is called distributed (devolved?) mixing. It would reduce cable jungle, but then you need local mix amps. But local mix amps can have balanced outputs, and you can then go for balanced mixing, increasing SNR by 3dB for free.

 

[NewYorkDave]

RE signal levels, they're -16dB at the output of the pan and echo pots (with controls at "typical" settings). Mix bus attenuation is -24dB with sixteen channels, so that works out to an average level of -40dB at the mix bus. All these levels are relative to the channel input.

The -40dB level at the busses is why I was inclined, despite the inconvenience, to locate them right at the inputs of their respective amplifiers. The potential for S/N degradation and crosstalk would be reduced, although it's hard to say how much of an improvement this would be in actual practice. (It has a lot to do with other physical details, such as shielding of the mixer case, or how far bus wires would be from high-level signals, or each other, and so on).

Doug, to answer your question, I'm looking at 16 inputs to the main L/R bus, counting stereo returns. If bus bars were used, they'd be about two feet long at most, and I'd be pretty much compelled to use an all-metal enclosure. (I have a Ross 12x2 I picked up for cheap which I might use as an case, since it would save me a lot of metalwork WRT drilling blank panels).

Jim, how did you handle the physical details of the bussing in your mixer? Did you hold on to any documentation?

 

[Jim Zuehsow]

Dave,

I don't have any docs left on that mixer, which is long gone now. I did build the channels as plug-in pc boards. They plugged into vertically oriented 15 contact double readout connectors and there was balanced input mic with the shield telescoped to the card connector. Power ground was a seperate ground from the audio out ground. The card outputs were somewhat low impedance, since they came from a cathode follower. The level pots and pan pots on the front panel had a common ground which tied back to the star ground point. and the left and right busses from the pan pots were made of # 12 copper with build out resistors going to the pan pots. I originally tried using a grounded grid input stage for the mixer but found it didn't work as well as a conventional common cathode 2 stage amp with feedback to make it bahave. This drove the masters and the output was a white follower driving a 10K to 600 ohm transformer.
I didn't have any sends on this board, as all we were trying to do was drive an old Ampex 440.

 

[NewYorkDave]

Thanks, Jim.

Do you remember the impedance of your mix busses, approximately? Were your pots and buildout resistors in the range of tens or even hundreds of Kohms? Was the circuitry completely enclosed in metal?

What you built sounds similar to what I'm considering in Diagram #3, although I'm using a bus in place of your star ground. I'm also planning to feed the mix into the grid of a common-cathode amplifier (the MILA-1 circuit, most likely).

I understand the theoretical benefits and drawbacks of the different approaches--but some real-world anecdotal information from someone who's been there is always appreciated. Your mixer was quiet so obviously your approach worked.

 

[Guest]

In high-Z mixer grounding buzz is not a problem unless you've powering filaments by asymmetrical AC through parallel to ground wires. What is more significant, is capacitive coupling. Everything has to be carefully shielded.

 

[NewYorkDave]

It's true that ground path resistance isn't so critical when the currents are very small (as they are in a high impedance circuit). But loops should still be avoided. And yes, capacitive coupling and RF pickup are dangers to be reckoned with when a significant length of unshielded wire is hanging off a high-impedance node with lots of gain following it. That's why I was leaning toward shielded cables--although with high source impedance, that brings its own set of problems WRT high-frequency rolloff.

 

[Jim Zuehsow]

Since I was coming off the output of a cathode follower, I was able to use a bunch of 50K AB pots for channel level. I had 4 busses selectable with some pretty nice lighted Switchcraft switches that would term the non-selected buildout resistors to ground. Buildout resistors I seem to recall were about 47 K. The mixer cabinet was a metal slanted front Bud model, and the power supply was in it's own shielded box inside the cabinet. The power transformer came out of an old Heathkit amp that lost the output transformer. It had the copper strap and a pretty low radiation field. B+ and filament were busses between card connectors using #22 buss wire and the boards were spaced 3 inches apart and fit in a card cage made from aluminum stock bought at the local hardware store. No phantom power was provided since I had an AKG supply which handled 6 mics. All 8 of the input transformers were in the opposite end of the case from the power supply box. I never could measure any hum from power and only a very faint hiss when the unit was set at normal operating levels from input to output.
Input transformers were all Triads and they fed a EF86, which was a favorite way of doing things. That was coupled to a 12AU7 with half of it being the output cathode follower, so each card held 2 tubes and associated circuitry. The mix and output amps were built on a small chassie and not a card and all it's components were shielded in that box.
The circuitry was a ECC83 and a 6CG7 for the White follower feeding one of 4 output trannies. All balanced in and out. VU meters were bridged accross the output XLR connector, but I can be forgiven for that, since we didn't know back then that doing that could add distortion.
I wish I still had this unit, but it became part of the clutter and was eventually sold along with the Ampex 440-8. I have no idea where it is, or even if it is still working.

Jim

 

[NewYorkDave]

Hey, I think I have the same Bud cabinet... C-1893? I bought mine with a mixer project in mind, but ended up chickening out because of the 18-ga. steel. It would be a heck of a job to drill all those holes.

RE: the VU meter, it may not have added much distortion if the source impedance driving it was low enough. But if it was a real 600-ohm power-matched circuit, it could be as much as 0.3% and still meet VU spec.

 

[Jim Zuehsow]

Back a couple years ago when this was the hot topic over at the Ampex forum on recordist.com, I went and checked my 1/2" 2 track 440 which I use for mastering sometimes. The meters are hung right accross the output with the series buildout resistor, and listening to test tones with very high quality phones at a high level, I could hear a little fuzz on the signal with the meters connected. I really didn't want to build a little follower to isolate the meters, so I just put in a switch so I could disconnect them once levels had been set. Pretty low level, but there nontheless. Now, whenever I built something that has a non-linear component hanging off the audio, I always add an iso amp.

Jim

 

[NewYorkDave]

I think that also accounts for the popularity of VU meter "multipliers" in the old days. These were essentially 3900-ohm stepped T or Bridged-T attenuators to be inserted between the 3600-ohm buildout resistor and the meter. It was a simple passive way to isolate the line from the meter, as well as to provide range switching.

 

[Jim Zuehsow]

I was talking about the distortion with the 3.9K buildout that makes the VU look like about 7600 ohms hung accross the line. It seemed to happen more with the selenium bridge stacks they would use in meters. I think it may be less with a silly cone bridge. I still have 20 or 30 old +6 accross 500 ohm level meters I'll never use.[/i]