Mono Summing - Active or Passive?

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

TomWaterman

Well-known member
Joined
Jun 4, 2004
Messages
1,151
Location
The Shire, UK
Hi folks,

I'm looking at the best way to implement mono summing in my monitor console design.

At this point I am undecided about needing a balanced or unbalanced summing network, still working on a few other things that determine this...anyway I was searching around and found the following:

http://www.rane.com/note109.html

While the Rane article is talking about crossovers for PA's the summing part is relevant.

While I am not so worried about cross-talk figure given as this is a stereo monitoring box and there is a reasonable amount of common-signal in stereo program material....I am worried about the following:

The low value input impedance of 1kOhm-ish.

I will probably be driving this network with some Forssell 992s and they don't have the drive capability so I'm worried about them crapping out so to speak.

The rane article mentions upping the values to improve crosstalk and input impedance which is cool and would solve that worry...however then its get sticky with regards to driving cable runs to the monitor amps etc. I don't want any HF droop due to large output Z.

n109fig1.gif


So, should I up the values to improve input impedance / x-talk and then buffer the output of the box with a line driver to eliminate cable capacitance affecting the Freq response and be done with it??

Or should I go for a 2 input active summing amp? This would improve x-talk, and isolate the whole system from long cable run problems etc, no? but would it be worth it? I also prefer the fact the I shouldn't lose any level with the active device...

If I was going to do the active version, would it be best to go for something like 10k summing resistors or lower?

Also if I want to sum L+R and have the mono signal appear at both speaker outputs can I just parallel the out of the passive or active networks?

Any thoughts, suggestions etc?
Much appreciated
Cheers Tom
 
> driving cable runs to the monitor amps etc.

Facts!

How long are these cables? How much droop will you tolerate?
 
Hi PRR,

well, umm the cable length is unknown at the moment, in my room it will likely be 3-5m, but in other larger rooms it could quite possibly be 10m too.

I'd say it would need to cover a range of 3-10m comfortably...

It would obviously be essential to have the controller flat to 20k but ideally have a much higher response >50kHz(?)

What say ye?
Thanks for your help again...

Oh and I just realised that the low input Z of the passive mono network is only a problem if I decide to use Forssell DLR-1's instead of 992s as line receivers. So I think its less of an issue if I stick with 992/3s. The DLR-1 is only happy into loads >3kohm.

Still wondering about an active version though, any obvious benefits?

Cheers Tom
 
The first rule is that your monitoring circuits must not load down, cause leakage to, or otherwise detract from your main signal path in any way. In reality, this is impossible to achieve absolutely, but we can try to come close. Here's a simple circuit which should do the job with good fidelity:

MonoSumForTW.gif


Net gain from input to output is -3dB. If you want some "gain in hand", increase R3 to 47K and R6 to 100K. R4 and R7 may need to be adjusted on test; use the values that give the lowest DC offset levels at the outputs of U1 and U2. I just chose approximate values. Use metal-film resistors throughout and good film caps for C1 and C2. Ceramic is acceptable for C3 and C4. You can use supply rails of lower voltage (with slightly decreased headroom) but do not increase the rails beyond +/-18V.
 
> 3-10m comfortably...

10m ~ 30 feet, ~30pFd/foot, 0-1,000pFd.

A 10K resistance driving 1,000pFd is down 3dB at 16KHz. So we need less than 10K source, but how much less?

> flat...

How flat is flat?

> ...to 20k but ideally have a much higher response >50kHz(?)

Roughly: a 3K source driving 1,000pFd will be -3dB at about 50KHz, not quite -1dB at 20KHz. 3K is a very light load for most good preamps. But the mix network will cause L-R crosstalk based on the summing resistors, the source resistance, and any load at the summing point. In this case we can't stand loading at the summing point. So assuming the "600Ω output" is more like 60Ω, and 2*3K is 6K, we have crosstalk on the order of 60Ω/6KΩ or 1:100 or -40dB. Some folks would call this "poor".

If the sources are actually 600Ω, crosstalk is like -20dB, which might bother me.

If we hang 60Ω on the summing point, crosstalk is over 70dB down, but we need about 40dB gain to get back up to line level.

Dave's plan uses 47K summing resistors and ~33Ω (effective, at top of audio band) load on the summing point. From Left to Right we have about (47K/33)*(47K/60) or over 90dB crosstalk. It also presents very low impedance to the pot wiring, and 75Ω to the output cable, shifting the -3dB point for 10m cable to 2Mhz.

Note that the Rane plan uses "balanced" connectors but will generally be unbalanced. Dave's plan is honest unbalanced.
 
[quote author="NewYorkDave"]The first rule is that your monitoring circuits must not load down, cause leakage to, or otherwise detract from your main signal path in any way. In reality, this is impossible to achieve absolutely, but we can try to come close. Here's a simple circuit which should do the job with good fidelity:

MonoSumForTW.gif
[/quote]

Thanks so much Dave, thats great. In my design the monitor path will be driven from a seperate DAC so the main process path of the console should not be affected by the monitor circuits.

I presume the second stage is purely an inverter to provide the appropriate output polarity? Is there a reason for the 5k pot other than level adjustment? could that be eliminated and replaced with a 5k resistor for -3dB in mono?

I am planning running the entire monitor path at +/-24V due to the small but useful increase in headroom and the Forssells.....I was thinking about using OPA2604s for the summing circuit as they are the only ICs I know of that run at the higher voltage. They will of course alter the 'transparency' of the console but for mono checking it may be less of a problem. I don't think they can drive as well as a 5532 but that shouldn't matter if the output is buffered again with a line driver right?

Maybe its better to go with a 5532 @ lower voltage and accept the loss of headroom when running mono?

Thanks again for the schemo.

PRR, many thanks for the explanation. It seems low capacitance cable is only a little lower than your assumed 30pF/ft, so the passive system certainly appears like a compromise of x-talk and freq response.

[Edit - something stoopid I wrote that I hope nobody saw...! :oops:]

Sources would actually be the o/p of a forssell 992 stage <600ohm or the output of my DAC which is 100ohms balanced. So I would expect x-talk in the region of 40dB. This is rather poor I agree.

Please could you explain how you calculate the approximate 33ohm summing load in daves design and the expected x-talk?

If I go for a balanced design, I can just double up on Daves arrangement, one for the +ve going and one for the -ve going half of the diff signal right?

Thanks for all of your help it is greatly appreciated.
-Cheers Tom
 
> It seems low capacitance cable is only a little lower than your assumed 30pF/ft

ALL cable is 30pFd/foot. OK, it is easy to find 25, even 22pFd/ft, and 35, 40pFd/ft. But you have to go to extreme dimensions and insulation to get far outside the 30pFd/ft zone. Unless you run spaced-pair on telegraph poles on glass insulators, 30pFd/ft is always the good-enough answer.

> explain how you calculate the approximate 33ohm summing load in daves design and the expected x-talk?

Dave's summing-amp is what we often call "zero impedance", "virtual ground/earth", etc. The summing node impedance is the (33K) feedback resistor, divided by the gain of the opamp. Opamps "have infinite gain", so the summing node is at zero impedance.

Of course that's too easy. A good audio opamp has gain-bandwidth around 15MHz, which implies that the gain is 1,000 at 15KHz, 10,000 at 1.5KHz, etc. So at 15KHz the summing node is really 33K/1,000 or around 33Ω, dropping to 3Ω at 1.5KHz, etc.

We could argue that this slanted line should be weighted according to the ear's sensitivity, or masking, or other factors. But life is too short.

Pretending 33Ω at the summing node, the Left input gets attenuated about 33Ω/47K which is better than 1:1000 at the summing node. Then it leaks to the Right input via another 47K and the ~60Ω output impedance of the Right mike amp. Counting on thumbs, this is about (1/1000)*(1/1000) or 1/1,000,000 or -120dB. And that at worst-case: the top of the audio band where the opamp is least perfect. It computes to -140dB midband; in reality you have ~-90dB leakage in grounds and wiring, you never get -120dB/-140dB unless you OBSESS over it. (Nor is it generally necessary to go nuts.)
 
[quote author="PRR"]ALL cable is 30pFd/foot. OK, it is easy to find 25, even 22pFd/ft, and 35, 40pFd/ft. But you have to go to extreme dimensions and insulation to get far outside the 30pFd/ft zone. Unless you run spaced-pair on telegraph poles on glass insulators, 30pFd/ft is always the good-enough answer.[/quote]

Yeah! Thank PRR - I was going thru some cable catalogues and couldn't for the life of me find any that was out of the 30pF/ft range, so I went looking for 'esoteric' low cap cable and couldn't find much less than about 22pF/ft....

Thanks very much for the explanation of calculating the x-talk, that was really helpful.

One last question, if I take a feed from the left and right bus and mix them into daves summing circuit, can I take two feeds from its output so that the mono signal goes to the left and right speaker outs simultaneously without too much hassle?

Or do I need to parallel a pair of summing circuits for L and R?

Cheers Tom
 
On another tangent... I just found this and though of this thread.

http://www.edcorusa.com/transformers/audio/pc/pc6410.htm

Edcor makes a transformer that's:

Stereo line level to mono unbalanced line level

They don't quote the impedance... but probably 10k to 10k I'd imagine. The frequency response is listed at 30Hz - 20kHz, so not too bad. Would this be usable in a monitor switcher??

Justin.
 
Hmmmm I dunno Justin, maybe...

It would work for sure but I dunno if it would be as clean as Daves 5532 circuit. The TX looks small, so I'm guessing it won't handle as much level and therefore produce some LF distortion...also Daves plan should have a much higher HF response too and for the price of the ICs, a few good caps, should be cheaper or cost the same as the TX.

-Tom
 
Any transformer will have the same type of crosstalk issues as other means of passive combining, as well as the usual transformer issues of loading and distortion at low frequencies, so I don't see advantages unless galvanic isolation etc. is desired.
 
Hi again,

Dave I have one more question concerning your circuit.....

The 4.7and 10uF caps are just for coupling and removing the 5532s DC offset right?

If I go for OPA604 instead (for the small headroom increase) can I omit these and use the OPAs DC offset trim circuit to minimise offset to a few mV?

I figure whatever monitor amp it feeds will have coupling caps or a servo.
The 75R on the output is for stability into capacitive loads but I can remove the 10uF and 47k resistor?

Thanks muchly again.
-Tom
 

Latest posts

Back
Top