a 'simpler' tube/valve mixer topology

[themagicmanmdt]

hey everyone - long time searcher and occasional information puller, feeling now well versed and have handled years of projects to open my mouth confidently instead of just posturing quizzically on the internet -

with my schematic research of boards of old as well as some of the hypothesized 'console' ideas on this forum and others, a much more stripped down line level tube mixer idea, i feel, hasn't been addressed.

why not simply have a line input passive mixer with a fader per input channel, aux sends, pan pot and then tube summing amp/makeup gain?

no mic preamps - i feel there is enough variation in sonic colors of both artists, recording environments and preamps that outside of a consistent small microphone classical recording setup that a small variety of mic preamps are used to get the desired tone/color/hi-fi signature depending on mic and source.

since the desk is line only, there are no inserts. anything desired outboard can simply be patched in between the recorder outputs and the mixer inputs via patchbay. impedance terminations shouldn't be bothersome if the line mixer is >10K input Z (probably 20k would be nice!)

noise performance of a line level summing network is better than using a high-impedance summing network (and the high value summing resistors that come with it), so a 12 channel line desk is no sweat and a 20-24ch desk completely feasible.

four aux sends should be enough to use for monitor mixes (if needed) and mixdown effects. each of these would have their own makeup gain amplifier, of course. the return from each can be simply routed via patchbay to a normal 'line in' channel not being used.



line input xfmr -> fader -> aux sends -> pan -> summing network -> tube makeup amp -> xfrmr line out.

the aux amps would possibly be wired with a 1/2 or 3/4 switch, simple pan pot and volume; no need for further complications.

i've researched and researched....nothing like this exists. the manley is close but has some SS in the signal path (as a transformerless balanced input?) - for reasons I'm still not sure why.

i like and flirt with the idea of a VU meter on the output buss, but also find it redundant because the mixdown deck usually has VU meters already, and the studio is calibrated throughout to have metering equal...and I watch the tape machine if I'm concerned, anyways.

ideas? thoughts?

it's basically a transformer coupled passive line mixer with 4 tube aux send amplifiers and 2 main L/R amps, all 6 of which can most likely be designed identical (sounds like the V72/V76 idea of a stock block amp used in various positions, just tailored to 10k line level in/outputs instead of 200 or 600 ohms)

of course, i may change my mind and want to build an inductor EQ on each channel, which would very much increase tube count and complexity. this would involve finding an EQ that I could use on everything. hmmm...those Pultecs sound great....it's possible!


thoughts from the crowd?
nice to meet you all, too.

i'll find a way to scan some sketches if necessary.

marshall

[kevinkr]

Interaction of mixed channels when passively summed can be a problem - I would at least sum the channels actively to reduce interactions between the channels. Passive summing can also add a bit of noise to the signal as well if the desire is to minimize interaction between the channels even when things are based on relatively low source impedances.

An active mixer stage could be a relatively simple unity gain inverting feedback amp with a white CF output or similar. Interaction at the summing junction would be relatively low compared to passive resistive summing.

I haven't even thought about the issue of balanced inputs and outputs, but given my proclivities I might use transformers.

[themagicmanmdt]

hey kevin - thanks for a reply.

i believe i mispoke when i said it was a 'passive mixer', as what i described was having a inverting makeup gain amplifier after the summing network - so i'm talking about what i think is an 'active' mixer after all.

as i've always been dealing with circuits where there is no summing - mic preamps, tape machines, power amps, etc - summing has a few considerations and definitions i haven't dealt with yet, although i'm well versed in tube amplifier circuit design/implementation in a variety of ways...just not mixing...yet!

so!

from what i've deciphered from several places, an 'active' mixer is one that is buffered from the input source (let's say, via an input transformer terminated appropriately) and that features an inverting amplifier after the summing network. is this correct, or does the mass information lead the terminology of 'active' mixing astray?

secondly,

i have thought of what merits exist if an active buffer stage for each channel is placed somewhere before the summing network and after the pan pot...

balanced inputs and outputs would definitely be via transformer. i may experiment with an unbalanced tranformerless CF output for the master L/R outputs if I felt it sounded preferrable.

i mean it, basic schematic soon - just need to figure out a scanner issue...

[ruffrecords]

There is really no such thing as 'active mixing'. The two main types of mixing are passive and virtual earth. A passive mixer simply feeds each source (after its fader) via a series resistor. Provided the value of the fader is relatively low compared to the value of the mix resistor then each fader sees a load equal to its series resistor in series with the parallel combination of the remaining series resistors and there is very little interaction between the controls. The loss in such a passive mixing system is simply 1/n where n is the number of channels. So a 6 channel passive mixer has a mix loss of 1/6 = 16dB. A 16dB gain make up amplifier is then needed to restore the signal to its normal operating level.

A virtual earth mixer uses an amplifier with shunt derived shunt applied NFB which effectively creates an input impedance equal to the feedback resistor divided by the open loop gain. Once again each channel is fed from its fader via a series resistor to the virtual earth. People believe this reduces interaction between controls because the virtual earth is very small. Unfortunately this is only true in op amps where the open loop gain is very large and the NFB loop is closed at dc. In tube versions of virtual earth amplifiers this is never the case because the pen loop gain is much lower than that of an op amp and also because tube amplifiers tend to become unstable at high values of NFB (which is what you need for a good virtual earth) because they have no dc path in their NFB networks. In practice virtual earth mixing has no real benefit in tube mixers.

Cheers

Ian

[ruffrecords]

Originally Posted by themagicmanmdt

why not simply have a line input passive mixer with a fader per input channel, aux sends, pan pot and then tube summing amp/makeup gain?

Probably because it is not quite as simple as you might at first think. Once you have a fader, a couple of AUX send pots and a pan pot sitting across the transformer secondary it becomes a non-trivial task to ensure those pots present a reasonably constant load to the transformer no matter what their positions. Also, if you have 10dB in hand in the fader, and with a simple pan control another 4dB loss there too, then your main stereo bus is working at 14dB below the nominal input level.

A possible answer is to have 0dB in hand on the fader and a more complicated pan pot with zero insertion loss but you have already made a compromise and increased the cost.

noise performance of a line level summing network is better than using a high-impedance summing network (and the high value summing resistors that come with it), so a 12 channel line desk is no sweat and a 20-24ch desk completely feasible.

Noise performance is more likely to be determined by the noise performance of the gain make up amplifier (whether you use passive or virtual earth mixing does not alter this).

Suppose you use 100K mixing resistors (pretty high). The noise in a 100K resistor is just -105dB, but that is not the bus noise. If you sum 10 such channels the bus resistance is 100K/10 = 10K and the bus resistance noise is only -115dBV. For a passive mix bus the bus attenuation will be 1/10 or 20dB so with a perfectly noiseless gain make up amp the noise out from it will at best be -95dBV. In practice you will find it very difficult to design a simple 20dB tube amplifier with an output noise of better than -86dBV when fed from a 10K source which is over 10dB more than the bus resistance noise.

You might think this does not apply to virtual earth mixing but the noise gain of a unity gain virtual earth mixer fed from 10 sources is 20dB and so it is just as noisy as the passive gain make up amplifier.

the aux amps would possibly be wired with a 1/2 or 3/4 switch, simple pan pot and volume; no need for further complications.

Good idea.

i've researched and researched....nothing like this exists. the manley is close but has some SS in the signal path (as a transformerless balanced input?) - for reasons I'm still not sure why.

I agree. I have seen it talked about in several places but no one has built one as far as I know.

New York Dave did a basic one without AUX sends at prodigy pro

of course, i may change my mind and want to build an inductor EQ on each channel, which would very much increase tube count and complexity. this would involve finding an EQ that I could use on everything. hmmm...those Pultecs sound great....it's possible!

You might be interested in my Poor Man's EQP1A project here:

Poor Man's Pultec EQP1-A


Cheers

Ian

[martyh]

Speaking of mixers… I need to mix the left and right input to a power amp for a summed sub output. In the attached schematic the L & R inputs are dc coupled to the input grids of the L & R channels. Will this work better than a simple resistor network?
Thanks,
Marty

[Wavebourn]

Originally Posted by martyh

Will this work better than a simple resistor network?

No. You will get significant intermodulation such a way.

[themagicmanmdt]

Originally Posted by ruffrecords

Probably because it is not quite as simple as you might at first think. Once you have a fader, a couple of AUX send pots and a pan pot sitting across the transformer secondary it becomes a non-trivial task to ensure those pots present a reasonably constant load to the transformer no matter what their positions. Also, if you have 10dB in hand in the fader, and with a simple pan control another 4dB loss there too, then your main stereo bus is working at 14dB below the nominal input level.

A possible answer is to have 0dB in hand on the fader and a more complicated pan pot with zero insertion loss but you have already made a compromise and increased the cost.

I can live with the mixer losses and slight drifts in impedance that I've seen with the single center tapped pan pot; however, I've seen that adding the aux sends sends it into orbit a bit too much for me.

I'm thinking of incorporating buffers for the aux sends (raises the tube count one per channel, each side of a dual triode can be fine for one stereo aux, 1/2 and 3/4, I'm thinking, with a bootstrapped CF topology). Some of the older 'triple triodes' might be fine if I'm feeling frisky to incorporate the other CF buffer for either a 5/6 aux send or buffering the pan pot - however, that's a few more caps and a CF in the signal path...echhh. Maybe just for 5/6.

Suppose you use 100K mixing resistors (pretty high).

Sounds about right for 10x voltage loading for ~10k input + pan pot.

The noise in a 100K resistor is just -105dB, but that is not the bus noise. If you sum 10 such channels the bus resistance is 100K/10 = 10K and the bus resistance noise is only -115dBV.

How did you derive this? Both of these, actually, especially the resistor noise equation.

For a passive mix bus the bus attenuation will be 1/10 or 20dB so with a perfectly noiseless gain make up amp the noise out from it will at best be -95dBV. In practice you will find it very difficult to design a simple 20dB tube amplifier with an output noise of better than -86dBV when fed from a 10K source which is over 10dB more than the bus resistance noise.

I may fiddle with the 'Aikido' elaborate noise reduction scheme to lower that figure. Although, as an analog tape (and LP) enthusiast, I willingly and gleefully sacrifice a bit of noise floor for a much more euphonic and emotionally charging listening/recording experience. Then again, noise floor is actually something that has never really bothered me.

So, if and since this won't meet technical figures of other consoles of the SS days in terms of noise and maybe max output headroom, probably only around +22dbu currently with transformer outputs, that's fine - I don't hit tape that hard to begin with Easy meters means hi-fi.

An artist chooses his brush, so to speak.

You might think this does not apply to virtual earth mixing but the noise gain of a unity gain virtual earth mixer fed from 10 sources is 20dB and so it is just as noisy as the passive gain make up amplifier.

It only makes a difference when the number of inputs is very high then?

I'm aiming for either 12ch or 20ch to coincide with the best track width tape formats - 8/1" ; 16/2" track.. .

I agree. I have seen it talked about in several places but no one has built one as far as I know.

Hey, time for that to change, and I might need some services from members here, since these forums are one of the ones I keep coming back to...

You might be interested in my Poor Man's EQP1A project here:

Poor Man's Pultec EQP1-A


Ian

Trust me, I am. I've been thinking of trying to make a 'super Pultec', 3 band...just one stereo pair for hassling around things for some specific creativity. Biting off a lot. That's down the road, though.

Finishing the 'sum' amp drafts now, will post when I can.

MT

[ruffrecords]

Originally Posted by themagicmanmdt

I'm thinking of incorporating buffers for the aux sends (raises the tube count one per channel, each side of a dual triode can be fine for one stereo aux, 1/2 and 3/4, I'm thinking, with a bootstrapped CF topology). Some of the older 'triple triodes' might be fine if I'm feeling frisky to incorporate the other CF buffer for either a 5/6 aux send or buffering the pan pot - however, that's a few more caps and a CF in the signal path...echhh. Maybe just for 5/6.

You probably need just a pair of CFs, one pre and one post fader (assuming you might want to switch an AUX from pre to post fader).

Does anybody actually use stereo AUXes?

Sounds about right for 10x voltage loading for ~10k input + pan pot.

How did you derive this? Both of these, actually, especially the resistor noise equation.

There is a standard equation for the Johnson noise in a resistor at room temperature in a 20KHz bandwidth. It basically simplifies to:

Vn = (R/3096)^0.5 microvolts where Vn is the rms noise in the resistor R

First divide the resistor value in ohms by 3096, in our case this is 100K/3096 which is 32.3 then take the square root which gives 5.68 uV.

To turn this into dBV take 20log(5.68) which is 15.09 and subtract 120dB to give -105dBV (you subtract 120dB because one microvolt is -120dB).

I may fiddle with the 'Aikido' elaborate noise reduction scheme to lower that figure. Although, as an analog tape (and LP) enthusiast, I willingly and gleefully sacrifice a bit of noise floor for a much more euphonic and emotionally charging listening/recording experience. Then again, noise floor is actually something that has never really bothered me.

The only noise the Aikido reduces is power supply noise; in all other respects it is as noisy as any other standard tube circuit and power supply noise should not be your limiting factor.

So, if and since this won't meet technical figures of other consoles of the SS days in terms of noise and maybe max output headroom, probably only around +22dbu currently with transformer outputs, that's fine - I don't hit tape that hard to begin with Easy meters means hi-fi.

I have never quite understood why today people still want to design tube circuits to drive 600 ohm loads (unless they are headphones). Almost everything it will be asked to drive these days will be 10K bridging or higher so why not design for that. A simple mu follower through a 600:600 transformer will drive nearly +34dBu into a 10K load which is a healthy 30dB headroom above a normal +4dBu operating level.

It only makes a difference when the number of inputs is very high then?

Under all circumstances VE and passive mixing have almost identical noise performances.

Trust me, I am. I've been thinking of trying to make a 'super Pultec', 3 band...just one stereo pair for hassling around things for some specific creativity. Biting off a lot. That's down the road, though.

I have just such a beast on test with a client right now. It also has a tube version of the Helios Type 69 EQ.

Cheers

Ian

[themagicmanmdt]

You probably need just a pair of CFs, one pre and one post fader (assuming you might want to switch an AUX from pre to post fader).

Does anybody actually use stereo AUXes?

Can one just simply use one CF and have the pre/post switch before the input to the CF?

I suppose it's more of a 'buss' than an AUX. Headphone monitor mix when needed (although I use mono, some people get itchy on stereo), perhaps parallel compression, stereo reverb send mix, and even useful as a drum submix while tracking. I find I'll just sneak the snare mic, if used, in with the overheads, along with any other 'spot' mics or possible room mics if the track count is high, especially since I'm currently still working on 8 track (Scully 280).

Good to know about noise -

I have never quite understood why today people still want to design tube circuits to drive 600 ohm loads (unless they are headphones). Almost everything it will be asked to drive these days will be 10K bridging or higher so why not design for that. A simple mu follower through a 600:600 transformer will drive nearly +34dBu into a 10K load which is a healthy 30dB headroom above a normal +4dBu operating level.

The only crutch is the interfacing with old gear that is 600 in and no going around it. 1176LN, for one, older gear, etc. I'm totally a fan of considering we move on and move to 10k as a universal standard even for new tube gear line inputs, or recreating old classics. Going down to 600 and back is a lot of iron squeeze....

I'm going to try to avoid followers in the main audio path, as yes they are super neutral (having liked a 12AU7 the best for a follower), but I enjoy the stage sound without...

I have just such a beast on test with a client right now.

ooh ooh ooh me me me!