So, it could be the circuits in the TS24.ricardo said:
Where can I find this thread?
I was thinking of the next generation of engineers, those who designed the 900 series, the supergate and the 700 system.
balanced vs unbalanced mixer
- Thread starter audiomixer
- Start date
Help Support GroupDIY Audio Forum:
http://www.groupdiy.com/index.php?topic=47556.msg601313#msg601313abbey road d enfer said:
Ok, I've read it. The preamp in question is far from being the best piece of circuitry coming out of Borehamwood. It was designed within tight monetary constraints, for a line of cost-effective mixers. The inherent flaw in its implementation in the 200B is that the 75k current FB resistors limit the current available at less than 200uA, much less than the 1.5mA emitter current, so in this respect, it has the same limitations as the basic version using only one opamp, i.e. distortion at high level, common mode gain modulation and limited "input" headroom.ricardo said:
As to Doug Self mentioning that this is also the basis for a summing amp, I believe the real novelty is the use of an additional opamp to provide variable gain via an active FB loop (I believe it was used on Aux sends).
These circuits are a far cry from those that were created for the TS24/48, which were pure beauty IMO.
I don't have them in my computer, but I have good hope to have them tomorrow.
> according to doug self a simple unbalanced line input is much less noisier then a balanced
Are you looking at "A Balanced View, Part I" (April 1997), first sidebar?
"To balance...
Balancing offers the following advantages
* Discriminates against noise and crosstalk.
* .... "
_I_ would ban the word "noise".
While on one hand, ANYthing we don't want could be called "noise".
But as savvy users, we MUST be able to categorize a "noise" by its cause (and possible cures).
I tend to think of "noise" without qualification as "random noise" hisss.
But Doug Self is clearly talking about hum and buzz, nearby signals, also radio and welder interference.
All of which can be avoided. Run on batteries far out in a field.
Actually I've had excellent unbalanced systems 3 feet from a 208V 400A fuse-box; also run 100 feet above fluorescent and dimmered lamps.
Are you looking at "A Balanced View, Part I" (April 1997), first sidebar?
"To balance...
Balancing offers the following advantages
* Discriminates against noise and crosstalk.
* .... "
_I_ would ban the word "noise".
While on one hand, ANYthing we don't want could be called "noise".
But as savvy users, we MUST be able to categorize a "noise" by its cause (and possible cures).
I tend to think of "noise" without qualification as "random noise" hisss.
But Doug Self is clearly talking about hum and buzz, nearby signals, also radio and welder interference.
All of which can be avoided. Run on batteries far out in a field.
Actually I've had excellent unbalanced systems 3 feet from a 208V 400A fuse-box; also run 100 feet above fluorescent and dimmered lamps.
audiomixer
Well-known member
I was referring to 'small signal audio design' chapter 14 and self generated noise specifically. the basic balanced input is 14 dB noisier then a simple unbalanced input.
Of course the external noise must not add up of course, so careful filtering and short cables are a must. I am interested in the possibility to achieve a lower total noise floor, albeit only under very controlled circumstances regarding the external connections. I am not sure if there is a real advantage under these circumstances to have a balanced input. by the way many top hifi components run unbalanced and have a happy life....
I have started to simulate the different circuits in spice and will post some results. right now it seems that I am limited with pots and fader impedances as they seem to add most of the noise. the balance pots are at 50% in center detent position and this is much noisier then at full pan, measured and simulated. the channel fader is also not helpful and introduces most noise at the +2dB position.
my goal would be to match the OPA1632 I use at the outputs with a measured noise floor of -109dBu (with shorted input) when having at least four channels routed (faders @ 0db)... but this might be difficult to achieve...
- michael
Of course the external noise must not add up of course, so careful filtering and short cables are a must. I am interested in the possibility to achieve a lower total noise floor, albeit only under very controlled circumstances regarding the external connections. I am not sure if there is a real advantage under these circumstances to have a balanced input. by the way many top hifi components run unbalanced and have a happy life....
I have started to simulate the different circuits in spice and will post some results. right now it seems that I am limited with pots and fader impedances as they seem to add most of the noise. the balance pots are at 50% in center detent position and this is much noisier then at full pan, measured and simulated. the channel fader is also not helpful and introduces most noise at the +2dB position.
my goal would be to match the OPA1632 I use at the outputs with a measured noise floor of -109dBu (with shorted input) when having at least four channels routed (faders @ 0db)... but this might be difficult to achieve...
- michael
The problem is not balanced vs unbal .. but resistors. These have been known to resist the flow of electrons. For lowest noise, I recommend you get rid of all resistors. 8)audiomixer said:
eg a 10k 'hand carved from solid Unobtainium by virgins' Golden Pinnae resistor, has -112.6dBu noise over 20kHz. You are only allowed 2 of those before they nearly reach the -109dBu noise of your OPA1632.
> get rid of all resistors.
Quite right.
Resistance may be unavoidable at the microphone (coil/ribbon resistance, acoustic resistance and grid leakage in condensers) but after that you "can" do other things.
NFB network to set preamp gain can be transformer tap.
Volume/pan "pots" can be tapped transformers.
There's ways of mixing with stacked transformers.
True, at audio there's always some parasitic resistance but it can be 1/10th of the audio impedance.
(Stacking transformers tends to be problematic.)
> limited with pots and fader impedances as they seem to add most of the noise.
No.
It only looks that way because you are voltage-oriented.
A perfect (or good) resistor at reasonable temperature has a certain hiss POWER.
For a high resistance this is more hiss voltage and less hiss current. Contrariwise for a low resistance this is less hiss voltage and more hiss current.
You get (on paper) better S/N only by increasing signal POWER.
If you raise supply rails to 300V, 1Meg pots can be low hiss. (Stray capacitance will suck.)
Contrariwise if you dump lame 15V 1mA-10mA amps and use 200mA amps, you can drop pot impedances below 1K.
Since 40V transistors are cheap, and safe to touch, that's the more common route.
(Ultimately the techniques for very-high-power tend to trash-up your rails and "ground" with hash from other channels; in large consoles this tends to be a bigger problem than mere thermal hiss.)
> noise floor of -109dBu
Few domestic playback systems exceed 105dB SPL well. Here in the woods, acoustic background clutter may be 20dB SPL. It is hard to justify heroic electronics when consumers can't enjoy a lot more than 85dB dynamic range.
Quite right.
Resistance may be unavoidable at the microphone (coil/ribbon resistance, acoustic resistance and grid leakage in condensers) but after that you "can" do other things.
NFB network to set preamp gain can be transformer tap.
Volume/pan "pots" can be tapped transformers.
There's ways of mixing with stacked transformers.
True, at audio there's always some parasitic resistance but it can be 1/10th of the audio impedance.
(Stacking transformers tends to be problematic.)
> limited with pots and fader impedances as they seem to add most of the noise.
No.
It only looks that way because you are voltage-oriented.
A perfect (or good) resistor at reasonable temperature has a certain hiss POWER.
For a high resistance this is more hiss voltage and less hiss current. Contrariwise for a low resistance this is less hiss voltage and more hiss current.
You get (on paper) better S/N only by increasing signal POWER.
If you raise supply rails to 300V, 1Meg pots can be low hiss. (Stray capacitance will suck.)
Contrariwise if you dump lame 15V 1mA-10mA amps and use 200mA amps, you can drop pot impedances below 1K.
Since 40V transistors are cheap, and safe to touch, that's the more common route.
(Ultimately the techniques for very-high-power tend to trash-up your rails and "ground" with hash from other channels; in large consoles this tends to be a bigger problem than mere thermal hiss.)
> noise floor of -109dBu
Few domestic playback systems exceed 105dB SPL well. Here in the woods, acoustic background clutter may be 20dB SPL. It is hard to justify heroic electronics when consumers can't enjoy a lot more than 85dB dynamic range.
audiomixer
Well-known member
I had a go at the hardware jesterday evening
just for fun i tried a buffer LME49610 toghether with the summing amp (with the LME49610 in the feedback look of a OPA1612). very powerfull indeed, pushing over 150ma continous (I ran out of headroom and had only a 47 ohm load resistor available, almost whent up in smoke) to lower the mix resistor impeadances. but this is only conceptual, I think the whole thing was osciliating like mery go round - thd up in the mountains... maybe i will need to buy a scope after all. but the perspective of having a 2w capable summing amp is kind of cool (or warm!)
I looked at various mixer specifications from mida, studer and other and the faders @ unity to mix residual noise is not particularily impessive, true, but because I have a very limited requirement I thought I could get a better result (less noise per channel = less noise summed up, no?) implementation seem an other thing though, still trying my best.
the noise floor - as you mentioned is not overly critical in the end, maybe true, but hey, I could just start listening to mp3 compressed music at 96kbs...
- michael
just for fun i tried a buffer LME49610 toghether with the summing amp (with the LME49610 in the feedback look of a OPA1612). very powerfull indeed, pushing over 150ma continous (I ran out of headroom and had only a 47 ohm load resistor available, almost whent up in smoke) to lower the mix resistor impeadances. but this is only conceptual, I think the whole thing was osciliating like mery go round - thd up in the mountains... maybe i will need to buy a scope after all. but the perspective of having a 2w capable summing amp is kind of cool (or warm!)
I looked at various mixer specifications from mida, studer and other and the faders @ unity to mix residual noise is not particularily impessive, true, but because I have a very limited requirement I thought I could get a better result (less noise per channel = less noise summed up, no?) implementation seem an other thing though, still trying my best.
the noise floor - as you mentioned is not overly critical in the end, maybe true, but hey, I could just start listening to mp3 compressed music at 96kbs...
- michael
I first wrote about this in 1980 but it hasn't changed much, at least for analog console/mixers. The heavy lifting for S/N and fidelity is first the mic preamps, and secondly the sum bus. After that the "hierarchy of hard" tracks with noise gain, so EQ (15-20dB), post fader gain stages (+10dB), etc.
I/O circuitry doesn't even make that list since noise gain is generally modest. Even if there is a difference between unbalanced and balanced I/O the noise floor of either is well below other dominant noise floors so inconsequential.
It is good sport to make any path as good as possible, but in real use there are trade-offs associated with unbalanced I/O that could easily negate any slender noise benefit.
JR
I/O circuitry doesn't even make that list since noise gain is generally modest. Even if there is a difference between unbalanced and balanced I/O the noise floor of either is well below other dominant noise floors so inconsequential.
It is good sport to make any path as good as possible, but in real use there are trade-offs associated with unbalanced I/O that could easily negate any slender noise benefit.
JR
The 1612 has a very low noise voltage, but the noise current is much higher than a 5534, with 1.7pA/sqrtHz compared to 0.4. It makes the optimum source impedance quite low at about 600r. That means that if you want to take the benefit, you have to make sure that the inputs see no more than 600 ohms; it is far from impossible but it takes some thinking for getting away from the usual 10k.audiomixer said:
With the optimum source Z, the noise factor of the 1612 is about 1dB; for the 5534 it's 0.4dB.
These figures are to be taken with a spoonful of salt, because they are extremely difficult to achieve practically, the measurements are very critical, and lastly, the difference is so tiny it's unnoticeable in practice.
What is certain is that if you drop a 1612 or an lme49722 in place of a 5532, there is a fat chance the circuit will end up noisier.
audiomixer
Well-known member
I am still midst in the design phase and this will be an original design, so I am free to choose the source impedances. basically, everyone here says: man, just stick with NE5532 and 10k feedback resistors, that will be fine. but I wonder if there are ways to enhance the performance just a notch, just for the fun of it. so this input is very valuable I did not really think of current noise that much - still learning a lot. the OPA16xx series have some advice about the optimum source impedance so I will check that.
one thing I do not fully understand is why the noise goes up so much at intermediate pot positions. maybe I must try to install the whole thing in a real enclosure, and its just noise picked up but never the less I would not expect that kind of drop in performance. my spice simulation shows the same, but I can't really know if I can trust my spice models.
sure enough, my project is being delayed, but it's for the better.....
- michael
one thing I do not fully understand is why the noise goes up so much at intermediate pot positions. maybe I must try to install the whole thing in a real enclosure, and its just noise picked up but never the less I would not expect that kind of drop in performance. my spice simulation shows the same, but I can't really know if I can trust my spice models.
sure enough, my project is being delayed, but it's for the better.....
- michael
At both extreme positions, the wiper sees a very low impedance, zero ohm at bottom (CCW) and the impedance of the preceding stage at top (CW). In the intermediate positions, the stage to which the wiper is connected sees a variable resistance, which reaches a maximum when both "sides" of the wiper have the same value (halfway for a Lin pot, about 3/4 rotation for a Log). The maximum is 1:4 of the pot value, i.e. for a 100k pot, the stage sees 25k. The noise produced by the noise voltage is constant (it depends principally on the chosen transistor and the operating point), but the noise produced by the noise current, flowing in the pot, varies with the position of the pot.audiomixer said:
With bipolar transistors, noise voltage decreases vs. collector current but noise current increases. Most VLN opamp have significantly higher noise current because their input pair runs at relatively high current.
Note that FET's have a very small noise current, so in practice noise won't vary much when increasing the source impedance. However, FET opamps noise voltage is generally high, resulting in higher overall noise than bipolar opamps.
audiomixer
Well-known member
thanks abbey - finally a concise and eloquent explanation to what I was seeing in my tests. this validates your point about VLN opamps and current noise and also explains why we still need some fet opamps here and there. I tried to understand the phenomenon only with voltage noise (I knew of the variable issue of source impedance of the pot) but could not understand the 20dB shift in resulting noise. when testing different postfader opamps I tried jfet input opamps but then I just tested the extremes, where they do not excel....
will need to run some more tests with intermediate pot settings and a OPA1642 or similar. When I started designing the channel path I put a fet input opamp there, but I was under the impression this is only for the lower DC bias through the fader, not because of noise.
gracias,
- Michael
will need to run some more tests with intermediate pot settings and a OPA1642 or similar. When I started designing the channel path I put a fet input opamp there, but I was under the impression this is only for the lower DC bias through the fader, not because of noise.
gracias,
- Michael
Michael, what John is saying that getting the best S/N on even a simple mixer needs careful consideration of ALL stages. You pay close attention to the most sensitive circuits, the high gain mike preamps and summers to get everything to a level where 10k resistors do not limit performance.JohnRoberts said:
If you can find John's 1980 articles on selecting the operating levels in a mixer, they are a MUST read. Nothing has really changed for analogue mixing. The evils of unbalanced i/ps are FAR more likely to introduce crap than the odd 10k resistor.
What HAS changed today is you can go straight from mike into Digital and do all that nasty processing & mixing in your DAW with nice clean 24b (141.5dB dr [1]) digits. But even then you need to watch your evil resistors.
OPA1632 has potential noise of 1.3nV/rtHz. Fig1 in the datasheet shows it feeding ADC directly. But this has R1/2 = 2x1k on the input. They immediately add another 5.6nV/rtHz in the important 4kHz region. And this i/p has Zi = 2k which will make most of your sources unhappy.
What do you want to feed to your Line i/ps which need -109dBu noise?
[1] and if you believe that, I just saw Porky Pig fly past my window.
Abbey, please don't forget TS24/48.
audiomixer
Well-known member
I think I am in line with you guys except I do not have mic pre, equalizer, dynamics. so there remains the postfade amp (+10dB) and the summing. well that was the reason why I started to think of making the channel less noisy - summing 12 channels that are dead-quiet at -110dBu should be less noisy then summing 12 channels that are at -94dBu, no?
it is true that I do not need -110dBu noise floor in the end but I would like to go as far as I can take it. from your inputs this thread should maybe be called low noise design in a line only mixer and I should ask you guys for your reference summing amp topologies.
the line input section (balanced, with OPA1612) is built around two buffers at unity gain and then a diff amp with much lower feedback resistors. this is a nice design I have found in douglas selfs 'small signal audio design' and I found it obvious enough to give it a go, I have no restriction on opamp count or budget for opamps, as this will not go into commercial production - hehe. therefore I can have 1k resistors around the diff amp and maintain 22k or more input impedance. according to selfs measurements this is is already 5 dB better then the 10k traditional balanced input using NE5532, and using a more modern opamp in (correct implementation feedback res values) I expect it to be lower again.
the rest of my signal chain seems not to be up to spec, and I must work on that apparently. I did not realize that until we had a more in deep discussion about current noise in opamps - so I will need to redesign around an other postfader opamp. I have also gained some noise-floor by lowering the total fader resistance, any thoughts on how low I can go? right now I am considering 5k, anything against that besides more strain on the power supply?
- cheers,
Michael
it is true that I do not need -110dBu noise floor in the end but I would like to go as far as I can take it. from your inputs this thread should maybe be called low noise design in a line only mixer and I should ask you guys for your reference summing amp topologies.
the line input section (balanced, with OPA1612) is built around two buffers at unity gain and then a diff amp with much lower feedback resistors. this is a nice design I have found in douglas selfs 'small signal audio design' and I found it obvious enough to give it a go, I have no restriction on opamp count or budget for opamps, as this will not go into commercial production - hehe. therefore I can have 1k resistors around the diff amp and maintain 22k or more input impedance. according to selfs measurements this is is already 5 dB better then the 10k traditional balanced input using NE5532, and using a more modern opamp in (correct implementation feedback res values) I expect it to be lower again.
the rest of my signal chain seems not to be up to spec, and I must work on that apparently. I did not realize that until we had a more in deep discussion about current noise in opamps - so I will need to redesign around an other postfader opamp. I have also gained some noise-floor by lowering the total fader resistance, any thoughts on how low I can go? right now I am considering 5k, anything against that besides more strain on the power supply?
- cheers,
Michael
The main problem is that, as you reduce the impedances, you should reduce accordingly the ground resistance, both at PCB level and at interconnect level. The ground return current from the fader (or any other element that loads the signal) develops a voltage. For some reason (I believe Mr. Murphy should be blamed) these voltages find a way to sneak where they create the most problems. Most of the time you will find that your fader that is supposed to cut at -100dB will cut at only -70. Rail currents are a problem too, because they get reinjected in the ground via the decoupling caps. Having a separate garbage ground helps, but does not solve the entire problem.audiomixer said:
ricardo said:I know JR is very vocal about this and we had this discussion before. I guess John is not a studio SE; his POV is that of a designer. I know he can prove his point easily, just as I could, with just a few measurements or a plain and simple level diagram. But the AP's "needle" doesn't say all.JohnRoberts said:
I am both a designer, SE and musician. I know how to play the hat game. When I sit in front of a mixer, I don't think as a designer or a muso. And in this situation, the first thing that gets my attention is noise coming from the monitors with all faders down. Mix bus noise is abnormal (I know it is scientifically normal, but I mean unconsciously, I expect no noise when no fader is up, that's how stupid SE's are).
When I raise a fader, I expect, as anybody, technically trained or not, some noise to come up, so the "track noise" is something I will care later.
Although the hierarchy of priority is technically right, the mix bus requires more attention than the mic pre. It is relatively easy to approach theoretical limits of a mic pre by a dB or two; it will take a lot of work for improving - the law of diminishing return.
OTOH, mix bus are generally so far from the theoretical limits that they offers a lot of space for improvement. Between the standard VE mixer where noise grows linearly with the number of injection resistors and the theoretical "no-noise added" true-current mixing, through "multiple-partial" mixing, there are are several techniques that provide sensible improvement of summing noise. The fact that it is not very influential on the final mix does not justify deeming it as secondary. The comfort of the SE is also a significant parameter.
In that respect, I'm very happy with the absence of audible bus noise on my digital mixer.I don't.
The thing is, I finally got hold of the TS24 schemo, just to find that the mic circuit in there is the same as the 6000. My memory playing tricks with me. After talking with an ex-colleague, we came to the conclusion that the mixer I was thinking about was not the TS24/48, but the 3200. The schematic diagrams are about 1m x 2m, so I have to go to the ex-company and try to make copies, perhaps manual copies... The summing amps in the TS24 use a common-base bipolar PNP running at about 1mA, followed by half a TL0.
Ha, ha! I can almost see JR resisting the urge to comment at length since this was his area of expertise and he'd vowed never to visit it again. ;Dabbey road d enfer said:
I had a copy of one of his articles but can't find it.
The only things I could add to whatever JR may say to Michael are
- Draw up a Level Diagram. This is a chart that shows the Nominal Signal Level, Noise Floor & Gain for each stage. You may not be able to fill in all the numbers at present but trying is good. It will show you where you need extra care
- If you decide to use simple VE instead of the more sophisticated methods, consider switching OFF unused channels so they don't raise the noise gain of the summer. It's very rare you use more than a few channels.
- Consider doing the mixing in your DAW. As Abbey says .. no mix bus noise!
- Oh! And get a scope
Thanks Abbey. There's no urgency but the urge to re-visit old battlefields can be irresistable. I'm only after the Mike i/ps, the Balanced Outputs & Summing stage.
Similar threads
