Summing and mix buss.. question on topology and design.

So i've been wondering lately about deficiencies in my console and others. I've been looking at a lot of designs and notice one thing that mine does a little differently..

each of the input points to my buss channels is implimented as you would a preamplifier, with the channel(s) being switched IN to the buss feeding a pair of 2n4403 which feed an opamp. there are no caps between the channels i switch IN to the buss either. the "preamp" is fixed gain. from there the signal goes to the insert jack then to a buffer opamp and then to the fader and ouput opamp stage.

I see a lot of summing areas that use resistors and an opamp and wonder what all the extra fuss in my channel is all about.

the master channels also have the same configuration.

so it leads me to ask, is there a specific reasoning behind someone doing this instead of something that might be simpler/cheaper/better?

:thumb:

damn i thought people would be all over this topic like a cheap suit.

:guinness:

Could you draw something up for us? I think it would be much easier to follow...

Samuel

[quote author="Svart"]damn i thought people would be all over this topic like a cheap suit. :guinness:[/quote]

:? Well very I'm interested but it's hard to figure out what you are talking about. Can you perhaps provide an illustration?

Are you talking about the summing amps or the way the channels are routed to the summing amps?

What console is it?

Lots of interest but no answers from me. :roll:

ok let me draw the circuit up for you guys or try to get this scanner to work..

:thumb:

ok got the scanner working.. here is the schematic of the master channel.. the busses are almost identical.

i'm really wondering if i could get away with removing the BJTs and using a simple resistor network and opamp in inverting mode.

http://www.groupdiy.twin-x.com/albums/userpics/10015/master.pdf

[quote author="Svart"]i'm really wondering if i could get away with removing the BJTs and using a simple resistor network and opamp in inverting mode.
[/quote]

Yes, but why?

Those BJT's are there for a reason. They improve the performance of the summig amps by lowering the noise factor. The two 2N4403's are the devices that determine the noise. Remove them (and their associated components) and you will most likely have significantly more noise.

This is tried and tested way of designing a summing amp with better than average noise performance. I have seen even more advanced versions with balanced feedback and DC servos, but the basic concept is the same.

I would not mess with it and I doubt that BJT and opamp change would give you a significant performance boost.

What size summing resistors is the console using? Lowering the value of the summing resistors is the single biggest improvement you can make to noise in summing amps but of course the channels need to be able to drive them!

thanks Cuelist. I had the notion of the BJTs being there for noise purposes but your post has cemented the idea. This was called into question by the numberous schematics of summing systems i have seen and more preference always seemed to be towards resistor w/ makeup opamp (usually discrete) versions.

what do you think of the ciruit as it stands? the opamps are now all JFET input and the capacitors are now gone. then what is the ideal summing circuit in your opinion?

I ask all this because my mixes get a bit mushy when many channels are summed.

:thumb:

[quote author="Svart"]...what do you think of the ciruit as it stands? the opamps are now all JFET input and the capacitors are now gone. then what is the ideal summing circuit in your opinion?

I ask all this because my mixes get a bit mushy when many channels are summed.[/quote]

Well did it sound mushy before you modified the circuit?

I would not be surprised if you have HF stability problems after changing from bipolar to JFET opamp and particularly removing the capacitors!

You'll need those capacitors in there and furthermore, a bipolar opamp is likely to work better in the impedance range this circuit is designed for. Before embarking on electronic surgery it always best to understand how the circuit works and what part does what.

I would suggest reverting to the original circuit, boring as that may be. :wink:

it was by far worse before. I changed to JFET inputs due to their DC offset resistance therefore allowing me to remove the caps. the resulting audio is MUCH clearer than before and further testing shows no dc offset change. I could get more into specifics later if needed.

however i think i might have made the situation sound a little worse than it is. it's 100% better than before but i'm still hearing the mix get "crowded", however i think it might be more of a problem with my mixes, not the gear, however my question was really a question of simplicity(resistors and opamp) vs. more complicated (BJTs and opamps) and which is "better".

sorry to confuse.

JFET input amps by themselves tend to have lower r.f. rectification susceptibility than bipolars, but one hopes there isn't much r.f. getting in at that point to begin with.

"Mushy" sounds like IM distortion, maybe slew rate limiting, but possibly some marginal stability is involved as well.

[quote author="bcarso"]JFET input amps by themselves tend to have lower r.f. rectification susceptibility than bipolars, but one hopes there isn't much r.f. getting in at that point to begin with.[/quote]

You'd hope but... long busses make excellent attenas!

[quote author="bcarso"]"Mushy" sounds like IM distortion, maybe slew rate limiting, but possibly some marginal stability is involved as well.[/quote]

And a good way to avoid that is to keep the 'out of bounds' HF garbage from entering the mix amps. Ferrite beads in series with the input as well as a small value resistor in paralell with a small inductor, again in series with the mix amp input, can certainly be of great help there.

Mix amps are at least as difficult to get right as mic pre's but very boring so few people spend much effort on it. It's usually the same 'ol opamp as an inverter...

not much RF in the area. no radio or tv or high powerlines outside, nothing running close to the board. the chassis has a THICK groundbuss bar and each channel has it's own connection. I've added ferrite to some channels without a marked reduction in noise as it is fairly low to begin with. the AC system has an extra deep ground rod sunk too.

[quote author="Svart"]I've added ferrite to some channels without a marked reduction in noise as it is fairly low to begin with.[/quote]

Just to be clear - the ferrite beads need to be at the input of the mixamps, not at the output of the channels.

And they won't make any difference to audible noise, their effect is way up in the MHz range...

A good broadband interference noise source is something with a good-sized carbon-brush motor, like one of those big red Mast*r heatguns. Also, cellphone radiation is difficult to keep out of things. If you don't see any effects from either of those you are probably all right.

Busbars and ground rods are low frequency references at best, and turn into good antennas at higher frequencies.

The original circuit is interesting; I don't see that the transistors would make a significant difference to noise.... In a conventional 'virtual earth' mixer stage, the noise is predominantly series resistor noise times the effective stage gain. It looks like the transistors are an attempt to 'match' the low resistor values to the 5532; academic and not really necessary. But by fitting Jfets you would be going the opposite way and making the situation much worse! (with self-noise)
Losing the transistors and using a good modern conventional silicon junction chip should be as good as anything. Output swing will be limited by the 1K5 feedback resistor, but again, not a serious problem, although with a 5532, there could be measurable distortion. :sad:

interesting ted, please explain more.

just to be clear, i changed the bjt input opamps with Jfet input opamps, not changing the discrete bjts to jfets.

the ferrite is at the inputs, mostly for testing/experimentation as there isn't much interference where it's located.

Although I initially thought that the BJTs were there for impedence/noise purposes but looking at it made me wonder about the reasons for doing this. thus the reason behind asking for opinions because it seems like overkill or at least trying to get around using high tolerance resistors maybe? either way, I am a MAJOR "less is more" person so when i see a number of parts doing what fewer parts can do i usually feel a compulsion to change it.

the 5532s have been replaced by a dual OPA627 on the master channel, OPA2604s on the buss channels. the TLO84s by OPA404s on all the channels.

I don't have any appliances near the console and cellphones are kept away from the console. I do hear a slight beeping when the cellphone updates to the tower if it is laying on the console itself.

> if i could get away with removing the BJTs and using a simple resistor network and opamp in inverting mode.

That's just the mix-amp. To study a mixer, we need to know the the size and number of mixing resistors and how they are switched/panned. These are typically on the channel card or assign switch.

With a glance and a guess: do you have ANY trouble with hiss noise? I suspect not. There is no gain between mixamp out and the fader, only 12dB gain after the fader. If you don't keep the fader very high, "normal" level at the main outputs could be very high level through the mixamp and the fader driver, which could be your "mushy". (Especially if the main faders are <10K and driven by TL084 chips.)

The mixamp feedback resistor is awful low value, 1K5. If we assume the channel mix resistors are also 1K5, resistor noise will be quite low. However channel-amp load will be 1K5 for one bus, 400Ω if the channel drives four buses (equally, without pan). A 5532 won't suck bad, but a '084 will be in great distress driving high channel levels in 400Ω.

Also the very low mix impedance means there will be great honking chunks of Class AB output current half-waves flooding the supply rails and thus the ground system. No special care is shown for the mixamp ground reference. The mix could be standing on a ton of garbage. This was pointed out years ago, by Doug Self and others: lowering the mix impedance lowers thermal noise but at some point (with multiple real inputs) ground-garbage dominates thermal noise. Solutions include transformers and differential mixing; MAJOR cost increase. Patches include GOOD local decoupling near every hard-worked chip, in the channels and in the mixamp. 0.1uFd per 5532 and nothing on the '084s is probably less than you want. I'd think more like a good 0.1uFd ceramic next to every chip, 10uFd within 2 inches, and 50-200uFd per board (either a lot of 10uFd or a bulk 100uFd in addition to local 10uFds).

Noise: 12 or 16 channels at 1K5 each is a 100Ω bus impedance. Thermal noise will be 0.2uV. Chip op-amp noise will be 0.5uV-2uV. So the transistors, working much richer than the 5532 inputs can afford to go for general use, will give a lower noise floor. At least with all channels un-muted but faded to zero. In any real mix, source noise should probably overwhelm mixamp noise.

So the BJTs are for test-spec. They may have little effect on noise in real use. Noise will be 114dB below +4dBm and 130dB below 0Dbfs. You need 21-bit recording before that is a problem. And 130dB mixes: there ain't no such thing. I've faced 120dB SPL peaks above 14dB SPL room noise, and that was an absurd situation (I made no attempt to capture the 106dB dynamic range, since I had 95dB recording and the judges would be in ~80dB S/N listening rooms.)

If the channel resistors are 1K5, then for 15 inputs the mixamp runs at noise gain of 15. If the mixamp were a low-spec naked '084, 3MHz GBW, gain is flat to 200KHz and we only have 20dB of NFB at 20KHz. For a stage that must handle high levels (because there is little gain to the main outputs), that's really not enough for a complex mix. The BJTs are biased/loaded for a gain of about 15, so they add to the 5532's gain. The combo gives gain of 15 at 6MHz, the naked 5532's unity-gain point. So we have 50dB NFB at 20KHz, which is enough to clean-up any decent audio chip. Note though that if we muted channels by floating their resistors, the 15:1 mix-loss would vanish and the opamp would have to be unity gain stable. This BJT-5532 combo is NOT unity gain stable, and it would take cleverness to make it so. (Actually there is some there already: the 100Ω + 0.03uFd termination. It might actually be stable with all channels floated.) The disadvantage of requiring all mix resistors connected (to channel or ground) at all times is that you always run at maximum noise; floating unused inputs reduces noise gain.

Before you go too crazy: try running your Masters at max and reduce all input trimmers to get back to proper output level. If it is less mushy, you have been overdriving the mixamp and fader driver.

Last night I was wondering about that 0.03uF cap... it's obvious now... it's there to improve low gain stability.... thanks PRR; I couldn't see the mechanism; it looked counterproductive, the last thing you want on a VE bus is a big cap to ground!

Yes, I agree totally with PRR, in real world use, those transistors are a waste of space, but then I'm sure both he and I would have done it all differently to start with!
My way would be to use a 'middle of the road' value of bus resistor, say 6K8, where if you needed to drive say 4 buses, the drive circuit would not fall over.... and I would have done the whole thing balanced, to overcome all those ground-path horrors.
With a decent post-fader buffer stage and balanced mixing, 128dB dynamic range is achievable; if anyone is loony enough to think that they need it. :shock:

Maybe that 'long-tailed-pair' approach could work as a balanced bus?? :?