L + R button panpot circuit

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.
EDIT: I see you use a weird panpot arrangement that is in shocking contrast with the level of dedication you put in this project.

Ack! What are you finding wonky?!?  I believe I faithfully followed the API circuit.  Those loose pins go to a dupont header.
 
Spin off of Ian's and 12ax's, but trying to keep selection as intuitive as possible:

DEHWDJl.jpg
 
According to Douglas Self in Small Signal Audio Design page 438, if you use the active pan pot in figure 17.11 then "it's good drive capability means that the value of the routing resistors can be much reduced, which reduces both capacitive crosstalk and noise in the summing system".

Right now it looks like you're using 47K when you could be using 10K which is a difference of almost 7 dB just in thermal noise.

Also the pan law is better.

Also, I assume you're switching out the mix resistors so that unused channels are not attached to buses.
 
Thanks Abby and Square!

According to Douglas Self in Small Signal Audio Design page 438
I think you have the first ed.  In case others read this: 2nd ed. info can be found on page 620, Fig. 22.11

Right now it looks like you're using 47K when you could be using 10K which is a difference of almost 7 dB just in thermal noise.
This and floating unused channel's bus resistors has been an ongoing point of concern for me.  I didn't get Self's book until after I followed the 70's API schema and started completing channels in the mixer, which do not include the things that people suggest would improve noise.  :-[ :-[

Why do you think API chose to run buses to ground for channels that are not in use? Wouldn't correcting for that  (leaving them floating when not used) mess with the gain of the ACA's as such?

However, I'm not opposed to improving things as I learn, and possibly upgrading the group cards with active with pan but as to it allowing a drop in bus resistors to 10k, I'm torn, as Paul Wolff mentioned: "10K is too low, and with only 16 channels it’s 625 ohms, so the 2520 will be noisier than if they were 20K, 16 channels @ 27K or 36K you won’t notice much bus noise to begin with."

 
I assume you're switching out the mix resistors
What a difference some jumpers would have made.  I suppose I could cut and lift the traces from the PCB's.  Does anyone think that would cause trouble, given I followed the 536 fairly accurately?
 
boji said:
Why do you think API chose to run buses to ground for channels that are not in use? 
To improve the swich cut-off. When not grounded, the leakage capacitance of the switch allows some signal to pass. But the bus injection resistors are not loading the bus, which is the right choice for noise. However, the signal leakage improves significantly when teh bus injection resistors are decreased. 10k would give 15dB less than 47k.

aul Wolff mentioned: "10K is too low, and with only 16 channels it’s 625 ohms, so the 2520 will be noisier than if they were 20K, 16 channels @ 27K or 36K you won’t notice much bus noise to begin with."
There must be a misunderstanding there: the lower the resistors the lower the noise. Indeed the OSI of the 2520 is about 10k, but with 47k resistors, a 16ch results in 2.7k bus impedance, which is already much lower, so the noise current contribution is swamped by the noise voltage. With 10k res, resulting in 588 ohms, not 625 (the FB resistor ups the count to 17), noise would be slightly lower, but almost imperceptibly.
Other DOA's, such as the 990, have a much lower noise voltage and a higher noise current, resulting in an OSI of about 500 ohms, which is much better matched to a low impedance bus, providing a real mesurable noise improvement.
Presenting a source impedance that is lower than the OSI does not increase noise.
I believe this Mr. Wolff mixed up noise and noise factor.
The only reason API used 47k is they wanted to minimize the number of active stages, because their DOA's were expensive and a significant part of the cost. That's why they didn't buffer the pan-pots and aux sends, which led them to use higher values.
Today, the cost of an active stage is negligeable, that's why almost everybody use 10k or less.
 
Other DOA's, such as the 990, have a much lower noise voltage and a higher noise current, resulting in an OSI of about 500 ohms, which is much better matched to a low impedance bus, providing a real measurable noise improvement.
Thanks for your time Abby, the above helps me better understand the choices that surround a 2520 vs 990. 

I know this now has little to do with active pan, but you raise a valuable point about 990's:
Do you see any problems with having mixed values of injection resistors on the program bus?

I ask because channel cards (already built) are two-stage 2520's.  If on the group cards I go with 990's can I keep 47k from channels to program, but say, use 15k's from group to program at the same time?  If not, I'll stick with 2520's for postfade boost.

 
boji said:
Do you see any problems with having mixed values of injection resistors on the program bus?
The injection resistor values depend entirely on the feedback resistor of what I am assuming is a virtual earth summing mixer. So yes it would be a problem because the gain would be different. But you could go back and change them all to 10K and then swap the 2520 with a 990.
 
So yes it would be a problem because the gain would be different. But you could go back and change them all to 10K and then swap the 2520 with a 990.
Ugh, I was afraid of that. I've built quite a few gar 2520's already, and have approximately 10 input cards up and running; it would be a shame to back out of the initial design at this point.  If noise becomes a real problem, perhaps that will force the issue.
As for the pan situation (In the group cards at least), if I have the space I'll make provisions for upgrading the passive setup with VCA's or Self's active design. 
 
boji said:
Ugh, I was afraid of that. I've built quite a few gar 2520's already, and have approximately 10 input cards up and running; it would be a shame to back out of the initial design at this point.
I'm probably not understanding your situation completely but isn't there only one DOA per bus? So use the 2520's for the mic pres (for which a 990 would not be appropriate anyway because of the lower OSI) of new input cards and use 990s for mixing at 10k.
 
use 990s for mixing at 10k.

If you think the second stage of a 536 as is, would support a drop-in 990 with adjusted nfb I could try it, but the postfade group and aux buses would need dropping as well then.
Also if I drop aux bus resistors, A1's prefade nfb would probably need to be tweaked. All the modification kinda intimidates me.

Edit: It would make a heck of a "super-1604" though I bet, if everything was dialed in.
 
I would think it’s about the sound/tone you are looking for.  The api console has a forward sound in the midrange with 2520’s and the output transformer create a round/loose bass with bandwidth limited sound.  I don’t use the program bus much. I use directs on the channels to send to tape/protools.  The mix bus is cleaner than program bus because it’s send is before the last channel out transformer.  So noise is not a problem from summing to mix with 24 Channel’s.  However noise from a reverb plate or actual older digital reverb can be heard even when driven correctly.  For me the api sound is 2 amps and 2 transformers.  And hence passive driven pans and sends are how that’s accomplished. 

I worked on a 40 channel amek with opamp buffered everything on channels.  The board required careful balancing for noise verse clipping which was a tighter window than my 24 channel api.  Also the cue output always sounded noisy if channels weren’t driven hard.  It was also easy to hear clipping in the eq’s . It had the RN channels.(rupert neve) version.  It was very clean sound once dialed in.   
I compare consoles to guitars.  Gibson acoustics have a different sound from a Martin.  I like both guitars . 

I would proceed with your plan before to much hybrid design.  Listen to the 990 verses the 2520 in summing and let your ears tell you before introducing op amps which sound clean but can be anemic in the chain.  To me the capi hybrid opamp 2520 has its place but sounds a little boring to a 2520 in a mic pre.
 
If you get really clever, feed the VCAs from a dynamics sidechain as well.

boji said:
I can't even imagine what that would do!  Are we talking dynamic pan/width ? Or just that it allows you to also have a dynamics processor included as part of the bargain?

OK, here's an overview of the whole thing.....


A pair of VCAs are fed from a common input.
The VCA outputs are left & right.
A common control voltage (from a physical fader or a DAC) varies the overall level up & down.
A pair of voltages from a physical panpot (or DACs) are used to offset the control voltages to the VCAs, giving you a pan function.
A mute switch slams, well, better if it rapidly ramps, the VCA control voltages to maximum attenuation, thus providing a mute.
Include a dynamics processor (compressor, limiter, expander, gate) in the signal path..... its control output would normally feed a VCA.... and we already have a VCA (actually a pair of VCAs) in the signal path.... so..... much use of control voltages, but only one pair of VCAs do all the work of level, pan, mute and dynamics. The devil is in the detail of how the control voltages are processed and presented to the VCAs.

The dynamics idea falls-down when we consider that the dynamics is effectively post fade. This can, however, be mitigated by feeding a third VCA from the input signal and controlling it only from the dynamics circuit, thereby providing a pre-fader signal under dynamics control.

There are various tricks that can be incorporated to improve the depth of cut provided by the VCAs and also to improve the distortion performance through the VCAs.

Width control would require phase inversions and cross-coupled blending of the left & right signals, so it is not impossible, but requires more circuitry downstream of the panpot. This is only really effective on a signal that is already stereo.
 
[A dynamics processor] output would normally feed a VCA.... and we already have a VCA (actually a pair of VCAs) in the signal path.... so..... much use of control voltages, but only one pair of VCAs do all the work of level, pan, mute and dynamics.

Ok yes, that is cool! It's workability sounds so obvious after your explanation. Thanks!

Width control would require phase inversions and cross-coupled blending of the left & right signals
It's exciting to know that would also work, as I had some vague intuition it would be possible with savvy control of VCA-controlled polarity.

There are various tricks that can be incorporated
A trick to you, would be sorcery to me. ;)
 
I would think it’s about the sound/tone you are looking for.
I'm mostly on the same page as you Fazer. Thanks for your contrast with with the Amek. I can hear those in favor of active circuits say IC's have much improved since it was made. I'm sure one day, perhaps in my lifetime, quantum pocket computers will faithfully reproduce the non-linearities of old transformer-driven circuits too.
But It is good to know you have the real thing (2404?), and can say that part of the sonic appeal and charm of API desks includes (and perhaps can be attributed to) the old methods of pushing signal around.
 
The VCA thing might be great but again the Amek had older VCAs. And mostly were in bypassed  as ITB tricks came into its own.  So I’m looking for tone and summing .  I believe you play the guitar you got instead of worrying about something you don’t have.  It can turn into a worm hole without boundaries.  Some music can only be made on a DAW.  The Neve API thing is a live thing.  It might work for synth plug ins but I want mics to be the main event.
 
Back
Top