Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[mhartung]

I would like to know more about how some large consoles actually do the process of combining signals. ( i.e. passive, active, other??). I'm interested in API, SSL, and Neve topologies. Do any consoles use transformers on their L/R buss outputs?

Thanks

Mike

[Geoff_T]

Hi

Most "vintage" Neve consoles used transformers as part of the mix bus path with the exception of the PSM, 54 series, the three AIR Monserrat consoles and any console after the 8078... (81**, 51**, V series, etc)

The audio path generally went from the channel routing unit, to the mix bus via a mix resistor (usually 15Kohm), from the mix bus to the transformer (that lived inside the mix bus amp) and the t'other side of the transformer was taken back to the B- coppery bus and hence back to the send amplifiers in the routing unit who were connected to the main B+ and B- power busses.

The use of the chunky copper or aluminium power busses meant that the impedance from any channel, be it on the far left or far right of the console, was kept very low so that the circular path of send amp - mix bus - mix amp transformer - send amp was consistant throughout the console.

The 51 series was the first Neve console to use quasi-balanced mixing with vastly improved noise and crosstalk figures.

Up to the 8078, all Neve console inputs and outputs were transformer balanced and isolated. Later consoles could be fitted with transformers either as stock or as an option, depending on the console type and spec.

[mhartung]

Hey Geoff,

thanks exactly what I am looking for. I would also like to compare that with say an API topology. Anyone??

thanks.

Mike

[bobkatz]

If I recall correctly, API's summing buss uses the standard approach of a long, unbalanced low-impedance (virtual ground) summing buss. The main differences versus a 5532 would be

The API opamp is discrete in a potted module several times the size of a 5532. There are many types of discrete opamps using the 990 footprint, and the API opamp is actually a precursor of the 990, but it still sounds mighty good. Maybe it's the flavor of its distortion that attracts me?

The 5532 is a dual opamp and may even handle both channels of your stereo buss. The API opamp only handles one channel, which may give greater stereo separation. It is powered from significantly higher voltage rails, + and- 24 volts or higher, giving considerably more headroom than the 5532... The 5532 will be powered from + and- 15 volts, up to as much as 18, but that is pushing it and is dangerous.

I think API advertises clipping at +37 dBu, compared with +20 dBu in an unbalanced version of a console with 5532's, or +26 in a balanced version.


Does any of this make a sonic difference? It definitely sounds different and personally I prefer the sound of the API opamp. It sounds more "robust" and "open" to me, especially at high levels. But some of that may be due to the API transformers, which are always in the signal path--maybe I like the sound of the API transformer's distortion. Regardless, I'm impressed by the headroom and effortless sound quality of the API console. Call me a snob  

[John Klett]

Hi

I'll start on topic and then most likely veer off topic and generally babble...  I have to keep stopping to squeeze limes...  4th of July hosting makes it's demands.

All new large consoles made today, with few exceptions (depending on what you call large), are actively summed.  API and SSL consoles have always been actively summed.  Neve moved from passive summing to active summing, in music consoles, when they moved from the 80xx series consoles to the 8108, 8128 series desks.

I've covered this before but...

Active summing uses an operational amplifier to create a "virtual earth" buss.  The buss is set up similarly to what Geoff describes.  There are drop-on resistors (summing resistors) that feed signal to the buss, which is a circuit board trace or a wire that goes right to the inverting input of an opamp (maybe through a cap, which is not a good thing, but commonly done).   The output of that opamp is fed back to the buss (summing node) so that it always holds the buss to zero volts (or whatever the non-inverting input is referenced to).  That is why the bus is called a virtual earth.  It LOOKS like zero volts or earth.  You need an active amplifier element to make the summing function happen.  I sometimes call active summing busses "current summing" since the voltage on the buss is held to zero and what is actually moving is current.  

Passive summing busses need no amplifiers or active elements to make the summing action happen but because of the loss in signal level you need and amplifier to make up the lost level.  If you want to use a mic preamp for makeup gain you can...  though special purpose makeup gain amplifiers were used in most consoles...  1272's or 3415's in many of the vintage Neve desks.

AIR Monserat...   A few days ago I was talking with Robin Porter specifically about technical aspects of the three 4078 "AIR" consoles, of which the Monserat console was one.  These were consoles custom made in 1978/79 to satisfy the requirements and standards set by George Martin and Geoff Emerick.  I asked Robin specifically about the bussing and he says these were passive summing but most of the amplifiers were bipolar IC's (mainly TDA-1034's which are pretty much equivalent to NE5534's)  and all the transformers were toroidal instead of the flat pack "E core" types.  These consoles are unique because they blended a lot of the "new" electronics technologies that had been rolled out in the 54 series broadcast consoles with the well established console architecture of the 80 series music desks.  The toroids had much wider bandwidth as did the amplifiers so those consoles were flat out past 100KHz and were a little squirrelly as far as stability was concerned.  



Neve passive-loss summing consoles used a transformer-loaded buss.  The transformer provided some "free gain" before the actual gain makeup amplifiers.  The transformer is not required to make the summing action happen but it is part of the load and, because it's a transformer it will be wanting a reasonable impedance match between itself and the buss.

Other passive loss mixing consoles of the day...  like Trident A Range...  had passive busses loaded with a resistor (or resistors) and no transformer ahead of make-up gain amplifier.  Those consoles had typically had more bandwidth but the busses were very susceptible to noise pickup and the actual noise floor was higher simply because ALL gain off the buss was electronic.

All passive loss summing consoles had to have the drop on resistors terminated either to a low impedance source or to earth so that the buss impedance, and the buss loss figure, would remain constant.

Active summing consoles kept the tradition of terminating unassigned input resistors to but technically they did not have to do that.

API consoles used, mostly, 2520 opamps.  These were discrete transistor opamps in a blocky potted package that was standardized originally in instrumentation amplifiers made by companies like Analog Devices and Burr Brown.  RCA, AD and BB opamps sometimes turn up in API consoles along with Melcor...  which is the company API spun out of.  A bunch of people left Melcor to form API.  One of those was Saul Walker.  Saul designed the 2520 and about 20 revisions of it that followed over the years until Datatronix bought the company, and then eventually sold it to one of the design engineers who was working there, Paul Wolf.  So....  

oh yeah...  what? more limes?...

so...  uh...   2520 opamps operated on +/- 16 volt rails.  The output transformers would jam the maximum output level up to over +30dBu...  I'm not sure if it was +37dBu...  that is a LOT.   API did make a high voltage amplifier in the 2520 package.  That part was called the 2525 and it was whole other amplifier - not a 2520 at all.  I think it would run on rails up to +/- 32 volts.  It was very fast but the output stage was a little slower than the preceding input and second stages so it tended to oscillate and blow up...  but there are a few consoles around that are loaded with these and they do sound different.  They are much cleaner sounding.  Later on API introduced the 2510 amplifier - which is pin compatible with the GML opamp.  The 2510 is essentially a 2525 without the output stage.

more limes?  again?  

salt?  no salt?

cheese on that?

good? - okay

I have to wrap this up but I can come back in a couple days and try to untwist this...

oh...  so one last thing while I am on opamps...

The Jensen 990 opamp is actually NOT an all discrete transistor opamp.  It uses the LM-394 IC super matched transistor pair IC for the front end.  This is essentially 50 transistors made on one IC die that has fifty transistors interconnected in two parallel groups of 25 to make a matched "pair".

I had a couple thoughts that related back to summing busses and how various opamps work as summing amps...  I'll have to pick up on that some day soon when I am more coherent.

please forgive typos and lack of clarity or direction...

[mhartung]

Hi John,

Thanks for the info on the API. In an API console for example, I was wondering what a typical value of summing resistor would be used in a console of 24 inputs and still have the least amount of noise; taking into consideration the ground to be clean to begin with. Is there a golden rule for buss resistor value and feedback resistor based on the least amount of noise? I understand that the relationship of values will have to change depending on the circuit and where the gain comes from (opamp/tranny). How much gain does the API output transformer actually give?

Thanks to all you guys for the info, I'm learning a lot!

Mike

[bobkatz]

The thing about active (virtual ground) summing is it can stand a whole lot of inputs. The largest sources of noise will be the tape machine/line amps/console equalizers in FRONT of the summing resistors. You do want a VERY quiet opamp because effectively it can be modelled as a VERY high gain amp as the signal at the virtual ground bus is going to be in millivolts.

I've always used 10K ohm summing resistors by rote as they bridge the source and that's about it. Instinct tells me the noise of the resistor will be insignificant compared to all the sources being summed in front of the resistor array. But I could be wrong. Maybe 600 ohms would be quieter, but I've never seen a console with 600 ohm summing resistors.

[James Craft]

I am also studying summing buss applications and have spoke to John in the past about configuring the API 325 Summing amp card in a DAW situation (Hi John, this is J. Craft/Crosstown Recorders-Memphis) and he was very helpful with my particular investigation. My situation now is I have aquired two Auditronics 110A consoles circa 1983 which I will be making one up as a front end/monitoring console for my PT rig. The 110A is a stripped down 501 designed for broadcast/production, Jensen tranny/5532 pres, 3 band/2 freqs per band inductor based EQs, with some minor modifications this will fit my needs nicely. The frame/routing modules I have are 4-buss and stereo, but as a cost saving measure Auditronics made the summing amps and line output amplifiers do double duty via a switch/relay setup. You can have the four buss outs or the stereo out, but not both. The summing amps are all 5532 and the summing resistors are 10K as Bob suggested. My dilemma is cloning the summing amps in the console for the mix buss (they use an Auditronics A30 card for the line output of which I am in short supply of these modules, but may have a line on a box full) or do something new and different such as use the Forssell JFET 990 opamp via his provided schematic http://www.forsselltech.com/schematics/Summing%20Buss2.pdf. I am planning to populate the 24 input frame I have with 16 mono mic/line modules and 8 stereo line only modules to monitor/mix my PT rig, so I need both the 4 buss to mix mics to PT when needed (will use mostly pre-fader insert to patch direct to PT) and the 2 mix for the PT returns. I have 2 frames & power supplies, 40+ input modules (a mix of mono mic/line with EQ, mono mic line w/o EQ and stereo line w/o EQ), about 28 mono and stereo P&G 1100 series faders (thanx to John and all for the post on P&G faders)and two full sets of master modules. Any easily implemented, cost-effective solutions/recommendations for my 2-mix would be appreciated.

[ssltech]

The resistor value used should be tailored to match the optimal impedance of the opamp. -A couple of important design consideration thingies spring to mind, one of the more radical solutions is to 'bucket-sum'. This was the approach taken by Mr. Rupert Neve in the Focusrite consoles, and it gets very expensive when you do it for all the multitrack and aux buses as well, but for a master mix buss it's very achievable.

'Bucket summing' refers to keeping the individual buss lengths short, by limiting them to the width of each 'bucket' which in the case of Focusrite was 8 modules wide. Each bucket summed output then fed forward to the main summing modules. Think of it as branches feeding into a trunk, if it helps.

A virtual ground op-amp summing increases its gain as more and more inputs are connected to it... they reduce the source impedance, and the source impedance is the shunt leg of the negative feedback, so as the incoming signal becomes weaker, the gain increases apace, to keep signal levels consistent.

By virtue of the operating gain increasing, the gain margin used for negative feedback starts to decrease, and at higher frequencies -where the margin is already smaller- with more and more modules feeding the buss, distortion can start to creep up... it's not a nice distortion unless your op-amp has 'nice' distortion characteristics!

Keith

[bobkatz]

It is absolutely the best way to keep the virtual ground bus as short as possible and send high level signals as long as possible to the summing opamps. This means quite a dilemma for the console designer in keeping their wiring manageable, as you need multiple cables where you previously only needed one, and thus the "summing buckets" approach used in some consoles such as SSL (I believe). Look at the wiring of a classic API sometime... tremendous numbers of wires, but the sonic performance is excellent. If I recall correctly, API puts the summing resistors at the summing amps. This is the best place from the point of view of noise.

Virtual ground busses may seem wonderful, but they are prone to picking up RFI, ground noise, power supply nosie and other interference. I remember having to employ a trick in an Auditronics console in New York City that was definitely a compromise. They had a summing buss all along the length of the console that was picking up signals from the Empire State building. The only solution I found was to put a 1 K resistor between the summing bus and the input of the opamp. The negative feedback still goes to the summing bus, however. This let the input capacitance of the opamp do a natural rolloff of the RFI, which then became inaudible. But it also raised the noise floor of the console. A definite compromise.

[mhartung]

When it comes to muting signals on the channel strip itself using the MUTE button or pulling the fader down to zero, is it the same effect? Does using the MUTE button take more electronics out of the signal path than just pulling the fader down to zero (taking into consideration that the eq-dynamics are before the fader.

What I am also unsure about is if muting a channel is the same as disconnecting that channels source buss resistor, which would cause less gain from the summing amp, in addition to any noise from the electronics further upstream?

Thanks again!

Mike

[bobkatz]

mhartung wrote on Wed, 07 July 2004 10:37

What I am also unsure about is if muting a channel is the same as disconnecting that channels source buss resistor, which would cause less gain from the summing amp, in addition to any noise from the electronics further upstream? Mike

While it would be a great idea for the mute button in a channel strip to remove all sources of noise in front of the summing resistor, it is usually impractical to do this, and usually the mute button precedes a final active stage which is post the panpots. This stage is usually extremely quiet, though, so you don't notice the noise of many of them being summed.

In an active summer, you won't notice any change in the gain from the summing amp by lifting the source from the resistor. In a passive summer, the gains can change as well as the isolation, so if you mute the input of a passive summing matrix it is best to replace it with a load equal to the source impedance of the source circuitry that was there before you muted.

[ssltech]

Pressing the mute and pulling the fader down both turn the signal down, but don't forget that the fader isn't the last thing before the buss... that honour usually goes to the panning control(s).

Using the 'de-assign' buttons (which may be named something else) does indeed impreove the noise figure in a v-ground system. The gain essentially doubles as the source impedance halves... so halving the number of parallel resistances feeding the buss makes a significant difference... de-assigning a single module during a mix with 50 or so modules assigned won't make any sort of discernable or measurable difference. -Now if I'm mixing an 8-channel (stereo-pair-plus-spot-mics) classical gig on a 72-channel amek 9098i console (which has a total of 164 feeds to the stereo buss... two per module, plus stereo channels, plus FX returns etc...) de-assigning all the unused modules will have a very significant effect on the buss amplifier gain... reducing it by a figure approaching 26dB when compared to the "everything-assigned" setting.

Mutes are usually "soft" turn-downs, so that they don't click if you happen to press them during the high point of a low note on a bass waveform... hard disconnects like de-assign switches will occasionall click if you use them as mutes, it's unavoidable without complicating circuitry that re-introduces noise, distortion and nonlinearity at precisely the point where it's not needed!

My motto is to de-assign large numbers of unused channels, but not to become obsessive about it. -Always keep sight of the significance of what you're doing.

Keith

Keith

[John Klett]

bobkatz wrote on Wed, 07 July 2004 11:19

In an active summer, you won't notice any change in the gain from the summing amp by lifting the source from the resistor. In a passive summer, the gains can change as well as the isolation, so if you mute the input of a passive summing matrix it is best to replace it with a load equal to the source impedance of the source circuitry that was there before you muted.

actually the correct way of stating that is that the signal gain from one input to the buss output will not change a the number of drop-on resistors are added or subtracted but the NOISE gain does change.  Look at it this way.  In a standard inverting opamp configuration the voltage gain of the amplifier is equal to the feedback impedance (the feedback resistor) divided by the input resistance (input resistor).  

So with one input resistor equal to 10K and a 10K feedback resistor the voltage gain is unity (though the polarity is reversed).  Now if you have a console with 48 channels that are assignable to a buss and they all have 10K summing resistors the gain of any one input to the buss output is unity as before but the effective input "resistor" is 48 10K resistors in parallel, or 208 ohms.  The voltage noise gain is x48 or 34dB.  The input noise of the amplifier itself plus whatever noise appears on the bus or on the ground that the amplifier is referenced to is amplified by 34dB.

In an active summing buss - if you only connect to summing resistors that you want to assign and leave the ones you don't want to assign unterminated the noise gain is less.

Passive loss summing busses typically have a makeup gain of between 35 and 50dB but 12 or more of that can be in the input transformer of the buss amplifier and maybe another 6dB can be done in the output amplifier.  Transformer gain is called "free" gain because there is no amplifier noise involved.  A passive loss buss has to be designed around a fixed number of drop-on resistors because as they are added or subtracted the LOSS is the passive summing network changes and unless the gain amp somehow changes to compensate for this the buss output level will shift.  Passive loss summing busses can individually handle up to about 48 inputs before you run into design complications but summing smaller groups (bucket summing) can easily get around that problem...  i.e. the one passive loss mixer I know of that is in current production does summing in groups of 16 and up to six of those are summing together to make a 96 input mixer.... (Purple Audio)


In active summing busses the opamp itself is a key component.  It has to be fast, low noise and preferably have a beefy output.  Ideally you want to buffer the output of that amplifier so that it has to do nothing more than control it's virtual earth input.  In cheaper consoles you'll typically find an amplifier like a TLO-71 that really has no output gonads doing the summing and directly driving a buss insert...  this is too much to ask of such an amplifier.  On the other hand an API 2520 can drive a 75ohm load and CAN do the summing and drive an output transformer at the same time.

I generally prefer passive summing because all the components are, for the most part, resistors.  The trick is shielding the buss.  Voltages on passive loss busses are very low (figure the voltage at the top of a drop-on resistor will be a volt or less) and the buss level might be 40dB below that.  40dB below a Volt is 10mV and figuring you want a signal to noise figure in the 80dB range...  with signal set at -40dB you are looking at keeping your noise floor below -120dB...  that is below a MICRO-Volt.  This is why Neve consoles have those extruded aluminum boxes around their busses...  which is why it's not unusual to get signal/noise ratios of better than -86dB (that's better than 500 NANO-Volts.  The Purple mixers use a fairly zippy multilayer circuit board with interleaved ground planes to shield the busses.  The buss loss for 16 inputs still has to be 40 or more dB in order to keep crosstalk down to a reasonable level...  so shielding is a big deal no matter what.

Active summing is easier and certainly less costly in manufacture.

[mhartung]

What if... I wanted to summ 24 channels of balanced signal? Nothing else. I had a discussion with a friend who knows more than me and he recommnded that I should summ the balanced signals passively with resistors which in turn feed a diff op amp (all hi signals are summed and sent to the non-inverting input of an op amp and the lo signal sare summed and sent to the inverting input. Is there an advantage by doing this? Quieter???

mike