Oh boy another active summing amp/bus discussion..

[Svart]

It's that time of day again, another summing amp discussion..

I'm itching to fix the only issue I have with my summing bus conversion.

As you folks know I modified my X2's master summing amp using FETblokes. It sounds great save for one problem. As I switch in more tracks I can hear the sound of the master output change. I know it's a bus impedance issue but I've been fairly happy thus far.

All that changed with the new studio build. New monitors combined with the new tracking and mixing rooms giving me sheer clarity that I never had before, I'm hearing the change much much more drastically now.

So now for the nitty-gritty.

Each channel has 3k summing resistors and there are 24 channels for a total bus impedance of 125ohms. This is driving a 220uf cap and a 2sk170 Jfet. The Blokes are NI with 15k feedbacks and 2k shunts. There are no shunts to ground from the audio path on the summing side.

It's similar to the output section of this but without the 50k pot:

http://www.forsselltech.com/downloads/schematics/8chsum_1.pdf

I start to hear the change at around 6 to 7 channels switched in. Below that I don't hear a change, more than that decreases the output of the master channel and changes the audio, mainly losing clarity on the top end.

I'm leaning toward working out a new value for the summing resistors, likely 10K-15Kohm, and installing a load resistor on the input of the summing amp. Since I start to hear a change around 6 channels switched in, that would give me a bus impedance of around 500ohms. Using this as my ideal bus impedance and reversing the math, that would put my summing resistors right at 12kohms.

So what are your opinions on this guess-timate?

Any novel and/or interesting ideas you want me to try? I'm game for being a guinea pig on this one for a while.

Otherwise, lets have some good discussion about this.

 

[Svart]

Oh yeah the output amps from each channel are (signetics/phillips)5532..

 

[JohnRoberts]

How about maybe measuring frequency response? That might help steer you toward a solution.

JR

 

[Svart]

That'll happen but currently(pun intended :green: ) I am looking for a good discussion on the topic..

EDIT: this started the discussion:

http://www.groupdiy.com/index.php?topic=22853&postdays=0&postorder=asc&highlight=summing+amp&start=30

And unfortunately at the time my resources limited my ability to hear the fault that I now describe. I was busy singing praises without being careful that nothing is gained for free..

Maybe I should get some ketchup for my foot to make it a little more tasty..

:green:

 

[Svart]

An interesting tidbit..

Installing 12k summing resistors on a couple channels caused those channel's audio to dip significantly when other channels were switched in. As little as one channel switched in could cause a change. It's fairly un-scientific but it shows that the theory is.. (wait for it..) sound.

 

[Samuel Groner]

It's fairly un-scientific.

You got a passive mix bus, so I'd say it is clear that using higher resistor on some channel will cause more attenuation for those.

Samuel

 

[Svart]

Sure, I'm just posting information so that others can benefit as well. I expected those channels to attentuate under the conditions that I created. (and yep I suppose that I should call it a passive bus with gain instead of truly active..)

Not being able to "borrow" equipment from work to do the testing I need to do on the console, I'm going to have to do a lot of unscientific inferring rather than scientific measurements.

 

[Svart]

since I'm talking to myself here, should I ask myself if I should just go ahead and change everything to a true active summing bus? :green:

The issue I see is changing the response of the active network as channels are added/subtracted. I can deal with attenuation as long as the sound/response stays the same.

EDIT: I have to also say that I figure that an active summing setup would also change the bandwidth of the gain stage as you add/subtract channels as well.

EDIT2: I see that D. Self mentions this:

The opamp therefore acts exactly as if it had a gain of 10, so input transistor noise is multiplied by 10, offset current is multiplied by 10, and bandwidth is reduced accordingly.

and

The inverting amplifier stage is actually noisier than a non-inverting stage with the same gain. For a non-inverting amplifier, the noise gain is equal to the voltage gain, but with an inverting stage, noise gain is equal to voltage gain + 1. When a large number of inputs is needed, a summing amplifier needs to have very low noise and wide bandwidth, or performance will not be as expected.

Any thoughts on this?

 

[JohnRoberts]

N vs N+1 for any appreciable number of inputs is insignificant.

Bus noise with any decent execution, will be below the noise floor of one amplified microphone at reasonable gain, so noise is not a paramount consideration (IMO).

The high noise gain of the summing amp also impacts distortion and phase shift (actually constrains the ability of NF to reduce open loop errors).

I wrote about all this in an article in 1980 which was posted here at least once, but I can't find it with a search so good luck.

The simple answer IMO, performance differences between inverting and non-inverting topology are insignificant. The practical utility favors inverting.

Regarding your current situation, I would be inclined to determine what is going on first as the flaw may not be related to the topology.

JR

 

[Svart]

To clarify the problem..

It all started with a bipolar transistor stage driving an opamp as the summing amp in the console. It sounded *bad*. Bypassing this stage sounded much better but of course I now needed gain. I simply replaced the stage with a DOA(fetbloke) and it sounded seriously improved, noise dropped as well. This gets back to my statement earlier, I wasn't able to notice the change in gain and in clarity while switching in/out tracks until now. A description might go something like this: Before: towel over speakers. After with only a few tracks switched in: Standing in the room. After with all tracks switched in: listening to a speaker(decent but not the same as before).

 

[Svart]

Anyway, one of the main reasons for this thread was to initiate a discussion that could evolve beyond the "do what's best for you" type of discussion that seems to happen often. I'd like to get more into the true theory of the different kinds of summing amps and why each one works or doesn't work.

 

[JohnRoberts]

Maybe if I find the time I will find and post my article again... I'm trying to get some work done now.

I'm not sure there's any new there there,,, pretty much design 101. If you run at high fixed noise gain, decompensate the opamp for best performance.

JR

 

[PRR]

> change at around 6 to 7 channels switched in.

The plan you linked has no switching.

It is scaled for 22K mix resistors; you say 3K.

> total bus impedance of 125ohms

When you have 6 channels at 3K it is 500R impedance.

Forssell's plan has the master gain pots and also a nice 604 amp...

> As you folks know I modified my X2's master summing amp using FETblokes

... don't know and not going to dig.

The lower impedance "should" not change the frequency response unless you have yards of cable or a high-C bus amp. But _I_ would measure it. Audio-sweep CD through mix to recorder's VU will quickly tell if something horrid is happening as you do whatever.

You could just be hearing noise-rise, or ground-garbage rise. Both can be the "mouse in the tweeter" effect.

> "I can hear the sound of the master output change." ...
> "It sounded *bad*."

"Change". "Bad". Bah. Describe the problem, accurately, in detail; you may discover the solution.

 

[Svart]

Hmm.

My bad, even though I said the design was only similar to the Forssell design, I should have been much more precise and said that only the NI opamp stage was similar.

I guess I wasn't clear in my description of what is going on either.

the 3k resistors are the STOCK resistors when using the BJT summing circuit.

I changed the summing circuit to JFET and now need to adjust the summing resistors for the change in load.

I calculated that 3k/24=125 ohms.

I hear the change when I get to around 6 channels worth of loading on the summing bus. This is equivalent to 3k/6=500ohms.(which was also stated in the previous postings..)

Figuring that around 500R of load on the buss is the point at which the load changes response and signal level, I use this in the equation X/24=500. X of course is 12kohm when solved. This gets me the 12k minimum summing resistor that I suspect would work.

The fetbloke is standard NI gain to make up for the bus loss. The master faders are after this in their own stage.

I'll describe the sound again.. A little differently..

less than 6 channels switched in: sounds like you are in the room with the instrument. clear, crisp and natural.

all channels switched in: sound like you sampled something at 22khz instead of 44.1. the clarity is gone. general sound is there but it's just not as clear. This sounds similar to how the summing sounded before changing the circuit out.

 

[PRR]

Pretend I think an "X2" is a 1950s airplane. So that link to Forssell was red-herring-ing me.

Pretend I'm a pre-MP3 man and all this talk of 22K vs 44K is greek to me. How does it sound? (Not what does it sound "like".)

Also it's not clear to me (I'm at work and I only bring half a brain) if "more than 6-7" means additional sounds or additional muted inputs.

I know you build and stupport electronics for a living. If your customer said "It's less clear", and said the hardware was like one thing but it is really something else..... because they pay you, you tear your hair, maybe make a field trip, trying to understand what they are really going-on about.

 

[Svart]

:wink:
gotcha.

When I mention "6" channels I mean 6 channels physically switched to the bus and the others are not physically connected at all. If I connect more than 6 channels to the bus I hear the change.

maybe I can describe exactly what I hear:

when you stand next to a vibrating cymbal, you can really hear it vibrating as well as hearing the note that it's vibrating at. When all of the channels are connected to the bus I cannot hear the vibration anymore, just the note.

OR I can switch to engineer speak:

It sounds bandwidth/slew limited.

Sorry, I tend to speak in layman's terms too much on the forum since I am aware of the general audience that is observing the threads.

 

[JohnRoberts]

Those parameters can be measured... it sure is easier to use test equipment than ears.

JR

 

[Svart]

I wish I could.. I don't think my boss would smile upon me dragging a couple 100k worth of gear home to test my console.. :green: And of course a 200Lb console isn't easy to drag into work and hide under my desk while I do *real* work..

I'm trying a few interesting things, purely subjective of course, but I notice a change so far.

I've been playing a looping clip of a cymbal and A/Bing the system to make sure I actually hear a difference.

 

[JohnRoberts]

It doesn't take $100k to beat human ears...

I suspect a decent computer sound card could get you some useful information.

JR

 

[PRR]

> dragging a couple 100k worth of gear home

I proposed a $0.25 starter test: sweep-tone CD to recorder meter.

There's many too-too-subtle things which could be wrong. But there's some real basic stuff. My preference is to alternate ears and eyes. I'll listen; but I also run basic "dumb" tests. Occasionally I've been mostly-satisfied with the sound, then discovered some major-boo-boo like a 10dB suck-out at 9KHz-13KHz. And other times my ears say "yuk" but the meter and quick 'scoping show clean.

It may seem argumentative, but that's just to trick you into finding the problem.