R/E/P Community

R/E/P => R/E/P Archives => Bruno Putzeys (Designer) - Dave Hecht (Master Tech) => Topic started by: bushwick on January 06, 2011, 02:29:26 pm

Title: S/N in big consoles
Post by: bushwick on January 06, 2011, 02:29:26 pm
Do any of y'all have S/N specs on large format boards with all channels summed to the L/R buss. It came up in conversation today and I am very curious. Something north of 40 channels and along the lines of a 72 ch SSL, or 72 ch of 8078, or a legacy.

Thanks,
josh
Title: Re: S/N in big consoles
Post by: Dominick on January 06, 2011, 07:56:57 pm
Neve would include factory measured noise floor levels measured under those conditions in the service manual of every console built.
Commissioning Neve consoles in the mid 90's, I would measure, under those conditions, using an Audio Precision PortaOne, around -80 dBu (unweighted, 20hz-22kHz passband) on 60-72 input VR consoles in the NYC metro area with no mains filtering or exotic grounding schemes.
Title: Re: S/N in big consoles
Post by: John Roberts {JR} on January 07, 2011, 06:07:59 pm
bushwick wrote on Thu, 06 January 2011 13:29

Do any of y'all have S/N specs on large format boards with all channels summed to the L/R buss. It came up in conversation today and I am very curious. Something north of 40 channels and along the lines of a 72 ch SSL, or 72 ch of 8078, or a legacy.

Thanks,
josh


It's been a while so I don't have any numbers off the top of my head (the most I ever sent to a L/R bus in a console design was well over 100 (36 inputs, and 3 sends to L/R bus from each of the 24 monitor strips for another 72, + a handful of efx returns with assignment capability 4 or 8 more pairs).

The question suggests to me some discussion might be in order about the sundry noise sources in a console and how they combine.

In a console summing bus we are dealing with coherent and incoherent noise contributions. The coherent sources generally follow the number of inputs linearly  (actually n+1 for the popular virtual earth sum bus). Incoherent noise sources like the mic preamps and input strip build as the square root of the sum of the signals squared.

So pulling some hypothetical numbers out of my butt, a 72 input virtual earth bus with a good quality low noise bus amp (say -120 dbu ein) will contribute 73x that noise, or roughly -83 dBu.... But that is ignoring the rest of the console.

Lets say the input strips (ignoring the mic preamp) have a -100 dbu noise floor with fader at unity and pans centered. 72 of them squared and then square rooted will add another -82.5 dBu of noise.  Since these are incoherent with each other they will sum to about -80dBu... So that's your noise floor before one mic preamp is cracked open.

The theoretical ein for a low z mic preamp input is around -130dBu  so one perfect hypothetical mic preamp running 50dB of gain will come up to another -80dBu all by itself... and this is ASSuming the room noise is lower than the theoretical mic preamp noise.  Of course you will likely have more than one input turned up if 72 channels are punched up to the 2 mix. Of course not every input will get 50dB of mic gain, and not every fader will mix in to the L/R bus at unity gain fader.

I don't doubt there are differences between different console spec sheets, and there may even be slight audible differences listening to WFO bus noise floor with all inputs muted, but in practice most competent master bus designs should be well below the noise floor of the sundry input sources being combined.

While this noise analysis suggests to me there is no real noise concern between competent designs another consequence of the high noise gain virtual earth topology (73= 37dB noise gain) is that the pole caused by the compensation cap to stabilize the sum amp can lead to significant phase error in the top octave if total loop gain margin is not adequately high.  This should be well known to (big) console designers so managed using one of several strategies.  

Distortion will also rise with falling loop gain margin just like the phase shift if the design is marginal.

My numbers are crude, but my point is that noise is not the only or even the dominant problem in a large bus structure. In my judgement.

I hope this makes some sense. In my experience people read too much into bus noise floors and that number is not completely representative of final mix sound quality.

JR

PS: This is the short simplified answer and I rounded off some of my numbers for a neat result.  
Title: Re: S/N in big consoles
Post by: QUEEF BAG on January 10, 2011, 01:23:04 am
but to achieve that lowest available
noise floor from that large console,
your system has to have a very good
ground system, with nothing forgotten.

or a number of noises in addition to hum
may appear at low levels (or not so low)
Title: Re: S/N in big consoles
Post by: Jim Williams on January 10, 2011, 12:56:51 pm
It all depends on what your reference point is. If +4 dbu analog tape, not so great. If you use elevated levels as found from some DAC's, it improves quite a bit.

If your signal levels reside at +12 dbu or more, you can improve the S/N by the same amount, 8 db. Take 8+ db of buss noise off the mix and complaints go away.
Title: Re: S/N in big consoles
Post by: John Roberts {JR} on January 10, 2011, 02:30:50 pm
Jim Williams wrote on Mon, 10 January 2011 11:56

It all depends on what your reference point is. If +4 dbu analog tape, not so great. If you use elevated levels as found from some DAC's, it improves quite a bit.

If your signal levels reside at +12 dbu or more, you can improve the S/N by the same amount, 8 db. Take 8+ db of buss noise off the mix and complaints go away.


One thing I don't miss about selling consoles is the customers who would listen to consoles WFO with inputs muted and complain about the merest hint of output noise.

In the real world we are dealing with acoustic sources that have a dynamic range limited by nature. It is good practice to deliver a suitably lower channel noise to not compromise that natural dynamic range, but at some point we need to be practical. That is one important leg of the engineering stool.

Mic preamps and well designed bus structures have been quite good for decades. Mic preamps have bumped against theoretical limits for decades. With 70 odd inputs on a bus I find it hard to ignore the combined source noise.

Pushing the mix level to +12dBu is IMO rearranging the deck chairs on the Titanic. Any real source material will get it's noise floor raised 8 dB also. Unless you are mixing 72 channels of digital zero, i'd like to know what your friends are using as input sources that lead to complaints?  

JR

Title: Re: S/N in big consoles
Post by: Jim Williams on January 11, 2011, 10:38:28 am
I'm talking buss and console noise contributions, not source noise from the storage device. By elevating the signal levels throughout the console, the console's noise contribution is lowered. Signal levels throughout my analog console reside at around +14~16 dbu. It clips at +28 dbu. That still offers enough headroom to be able to boost EQ without overloading the system. The opamps I use also sound best at higher levels. The buss noise does degrade with common digital reverbs, exception being the Bricasti M7.

Noise problems were solved many years ago here. It's not something I ever hear anymore. With a console loaded with .9 nv/hz/sq noise opamps, it's below the 16 bit theoretical noise floor so noise is not encoded to CD's.
Title: Re: S/N in big consoles
Post by: John Roberts {JR} on January 11, 2011, 11:48:11 am
Jim Williams wrote on Tue, 11 January 2011 09:38

I'm talking buss and console noise contributions, not source noise from the storage device. By elevating the signal levels throughout the console, the console's noise contribution is lowered. Signal levels throughout my analog console reside at around +14~16 dbu. It clips at +28 dbu. That still offers enough headroom to be able to boost EQ without overloading the system. The opamps I use also sound best at higher levels. The buss noise does degrade with common digital reverbs, exception being the Bricasti M7.

Noise problems were solved many years ago here. It's not something I ever hear anymore. With a console loaded with .9 nv/hz/sq noise opamps, it's below the 16 bit theoretical noise floor so noise is not encoded to CD's.


Agreed, not noise, but phase shift and linearity still count.

FWIW, in the last "big" console I did, I included a gain trim on the master L/R bus that allowed for up to -10dB pad right at the bus amp.  This is pretty similar to running the bus 10dB hotter without the headroom compromise. (in fact better, because reducing the bus noise gain 10dB improves phase shift and linearity performance too).  Of course this gets complicated for full differential/balanced buses. I suspect a switched bus pad could accommodate that without compromising differential balance.

I wrote a column back in the '80s and in one issue I compared some practical differences between nominal +4dBu and -10dBV mixing platforms.  Even back in the '80s you could get decent performance from a (good) -10dBV system. Today with even quieter off the shelf electronics, I suspect the differences could be even smaller. (Yes there are practical benefits to passing hotter signals between chassis, but inside the box, not as much.)

I find it useful perspective to realize that most IC A/D converters run from 5V or 3.3V supplies. At the end of the day it's all about raising the bridge or lowering the water... to get a good ratio.

JR

PS: I'm not trying to be argumentative, just adding some perspective to the topic at hand.
Title: Re: S/N in big consoles
Post by: Geoff Emerick de Fake on January 11, 2011, 04:43:25 pm
Not trying to be argumentative either, but...
I think the unmoving nature of bus noise makes it appear more intrusive than it is in a complete mix.
Typically, one starts mixing with the master faders up and builds the mix by raising faders one after each other, so bus noise is there from the beginning.
I must say that I took a serious liking to digital mixers because of the lack of bus noise.
Title: Re: S/N in big consoles
Post by: Jim Williams on January 12, 2011, 12:33:42 am
Your buss noise should be pretty low. If you unassign all the mix buss resistors, it really goes down. That noise adds up from both the buss resistor's thermal and current noise plus the electronic hiss from the input channels. The resistors are fixed but the electronic hiss contribution can be lowered. There are some very low noise fet input opamps that do very well lowering noise in older designs that used larger value resistors.
Title: Re: S/N in big consoles
Post by: Geoff_T on January 12, 2011, 01:47:39 pm
Hi

Vintage Neve consoles had signal to noise measured in three stages.

1. The output noise with the 4T fader at minimum... usually up around high 90's

2. The output fader up and a single channel assigned to the 4T bus

3. A channel at 80dB gain assigned to the same path with 20 to 20KHz filter on all these measurements and the input terminated 200 ohms. That last test gave figures better than -45dB = EIN better than 125.

But note, they did not assign all the channels to a bus... just one channel.

All these tests are clearly listed at the rear of the technical handbooks along with a polaroid photo of the square wave response.

Plus all vintage Neve tests were referenced to 0dBu, not +4dBu and the noise was not subtracted from the maximum output like digital measurements these days and power amps.

Hence their wonderful figures!
Title: Re: S/N in big consoles
Post by: Jim Williams on January 13, 2011, 12:26:58 pm
John Roberts  {JR} wrote on Tue, 11 January 2011 08:48

Jim Williams wrote on Tue, 11 January 2011 09:38

I'm talking buss and console noise contributions, not source noise from the storage device. By elevating the signal levels throughout the console, the console's noise contribution is lowered. Signal levels throughout my analog console reside at around +14~16 dbu. It clips at +28 dbu. That still offers enough headroom to be able to boost EQ without overloading the system. The opamps I use also sound best at higher levels. The buss noise does degrade with common digital reverbs, exception being the Bricasti M7.

Noise problems were solved many years ago here. It's not something I ever hear anymore. With a console loaded with .9 nv/hz/sq noise opamps, it's below the 16 bit theoretical noise floor so noise is not encoded to CD's.


Agreed, not noise, but phase shift and linearity still count.
JR



Yes they do, at least around here. I have the analog console down 1 db at 200k hz. It's also flat to 2 hz. Those points were chosen because phase shift begins a decade above/below the target frequency. A phase plot on the Audio Precision shows a flat phase response 20~20k hz = no phase shift. The reason it's so low in the bass is the entire console is direct coupled, only 2 capacitors are used in the hybrid balanced sum amps, the rest is coupled with wires. I wish I could explain the quality of sound without capacitors or transformers, it's heavenly. It's as if a huge weight of crap has been lifted revealing the sounds, not the electronics.

Linearity is also important to me. THD+noise is .00045% at +20 dbu. That is the residual of the AP test rig. A CCIF IMD sweep shows .0002%. That's one input to mix outputs. Specs just qualify what I'm hearing, unadulterated music.
Title: Re: S/N in big consoles
Post by: bruno putzeys on January 14, 2011, 04:12:58 pm
The curious thing is that when one keeps perfecting the console and the converters in this manner, you start approaching the sound quality you can get by not leaving the digital domain at all.
Title: Re: S/N in big consoles
Post by: John Roberts {JR} on January 19, 2011, 11:15:33 am
Jim Williams wrote on Tue, 11 January 2011 23:33

Your buss noise should be pretty low. If you unassign all the mix buss resistors, it really goes down. That noise adds up from both the buss resistor's thermal and current noise plus the electronic hiss from the input channels. The resistors are fixed but the electronic hiss contribution can be lowered. There are some very low noise fet input opamps that do very well lowering noise in older designs that used larger value resistors.


I wish this really was the problem people say it is..  Back in 1980 I wrote an article on console design for RE/P where I described a way to bus numerous channels together without the N+1 noise gain of a virtual earth summing amp (replace the resistors with current sources).  

In practice the current sources have noise too, and the commercial executions I did using this approach were populated with TL07x in the current sources so no where near as good as modern opamps (I did this decades ago). Rather than target near infinite output impedance from my "real" current sources I was happy to attain 20dB reduction in the N+1 bus noise contribution.

Like I say, I wish this was a real problem because I have a solution, but like I say, at the end of the day one microphone is going to dominate the noise floor of any properly designed console, so this is mainly a mental exercise (for noise, not for the phase integrity and distortion that also degrade with N+1).

Then there is digital with near perfect summation...   Time to move on to other problems.

JR

PS: am I the only one here who thinks it is spelled "bus" with one 's'?
Title: Re: S/N in big consoles
Post by: Jim Williams on January 19, 2011, 11:49:35 am
bruno putzeys wrote on Fri, 14 January 2011 13:12

The curious thing is that when one keeps perfecting the console and the converters in this manner, you start approaching the sound quality you can get by not leaving the digital domain at all.


You have to leave it if you want music. It's an analog world. Digital audio is just encoding, nothing magical about that.
It's the tank and train, a storage and delivery service, it's not music, it's data.

Like 8 track was between 4 track cartridges and audio cassettes, digital is just a carrier medium to be decoded into an analog signal, eventually.

The encoding can never approach the resolution and sonic quality of the analog signal, it's an approximation at best. As such it will always be behind in quality to the analog signal. Digital audio doesn't start or end with data, it starts and ends with an analog waveform.
Title: Re: S/N in big consoles
Post by: bruno putzeys on January 19, 2011, 12:46:59 pm
Ultimately you have to leave it. What I'm talking about is leaving it to mix in the analogue domain and then turn it back into digital.
Title: Re: S/N in big consoles
Post by: John Roberts {JR} on January 19, 2011, 01:02:59 pm
A philosophy but not scientific thesis. One could argue analog is just a bunch of discrete quanta massed in large enough numbers to appear continuous.

At some point digital becomes another piece of wire (which isn't perfect either but like so what?).  Perhaps more correctly digital will be like wire with some small flat colorless delay which does need to be managed for OTB mixing of mixed analog and digital stems, especially if coherent signals are split across multiple stems, but again not a new concept in recording, and managing delay inside a digital final mix is not a huge issue with adequate clock resolution.

We are several decades into digital being a merchantable medium. The SOTA is quite good and getting more affordable. Perfection is a concept not a destination. It can't be attained in either medium.

There are still many details to work out as digital is less mature than analog, but past the wet behind the ears stage IMO, and I'm an old analog dog...pedaling as fast as I can to keep up.  

JR

Title: Re: S/N in big consoles
Post by: Bob Schwenkler on January 23, 2011, 02:51:18 pm
John Roberts  {JR} wrote on Wed, 19 January 2011 08:15

Back in 1980 I wrote an article on console design for RE/P where I described a way to bus numerous channels together without the N+1 noise gain of a virtual earth summing amp (replace the resistors with current sources).

Is there available or do you have a copy of this article?
Title: Re: S/N in big consoles
Post by: dcollins on January 23, 2011, 10:22:59 pm
Has anyone read this!?

http://collinsaudio.com/Muth_Summing/


DC
Title: Re: S/N in big consoles
Post by: John Roberts {JR} on January 24, 2011, 01:01:07 am
Bob Schwenkler wrote on Sun, 23 January 2011 13:51

John Roberts  {JR} wrote on Wed, 19 January 2011 08:15

Back in 1980 I wrote an article on console design for RE/P where I described a way to bus numerous channels together without the N+1 noise gain of a virtual earth summing amp (replace the resistors with current sources).

Is there available or do you have a copy of this article?


http://www.circularscience.com/des_art.pdf

The references to SOTA are a little dated since I wrote this over 30 years ago, but it holds up OK (I hope).. Only thing I am embarrassed about is I spelled bus wrong back then...   Surprised   But some still spell it wrong today.  Laughing

JR
Title: Re: S/N in big consoles
Post by: zmix on January 24, 2011, 10:51:20 am
dcollins wrote on Sun, 23 January 2011 22:22

Has anyone read this!?

http://collinsaudio.com/Muth_Summing/


DC



Hmmm.... his assertions from 2002 seem more dated than JR's from 1980 (and John stated that he can spell Bus correctly now so he goes to the head of the class!!)

I'd like to see the Geoff Tanner Article Chris referenced...

You out there Geoff?
Title: Re: S/N in big consoles
Post by: John Roberts {JR} on January 24, 2011, 12:17:30 pm
dcollins wrote on Sun, 23 January 2011 21:22

Has anyone read this!?

http://collinsaudio.com/Muth_Summing/


DC


I have never seen this before, and it starts out with one strike against it, since it spells bus wrong, but I was guilty of that for a time, before I figured out what the right word is in the context of combining signals, and what the wrong word really means (look it up). Laughing

OK, from a quick glance at the article I see several claims that jump out at me.. Since the document is copyrighted I will paraphrase.

#1 The more inputs to a system the more losses.   (pg 3)

That may be an awkward way of relating the increasing noise gain of a "lossless" (not my nomenclature) virtual earth summing system, but as I published a few decades ago, there are other ways to skin that cat without the linear increase in noise gain.

#2  twice as many inputs = twice the bus noise.. (pg 5-6)

The math is a little sloppy as incoherent noise sources do not add linearly, but my larger criticism of the comment is that source noise will still dominate console noise, even for large consoles (I have designed at least one console with 100+ feeds to L/R bus).

#3 digital combining cost 1 bit of (?) for every doubling of stems. (pg 7)

I'm basically an analog dog, but I do enough digital crunching to know that is nonsense.  Simply adding together multiple digital signals results in the exact same digital word resolution. Headroom or bus overload is exactly the same as analog, but digital can use a carry bit to buy another 6dB of dynamic range transiently.

If you combine with a multiply (digital fader/pan gain adjustments are preformed by a multiply) You actually gain resolution in an output digital product with even more bits! In fact the result will have the bit resolution of the input word, added to the bit resolution of the multiply word, or more bits than we can typically use or care about.

#4   Dither?

This seems to be a red herring, Dither is pretty well understood by now (I think).

#5  A mix of 64 sources needs to run -36dB   (pg 8 )

Again. sloppy or misinformed math. Perhaps for 64 identical sine waves but 64 incoherent sources only increase +18 dB, and we don't build mixes with 64 inputs all at full gain, all the time.


===================

Sorry, this paper reads like a marketing tout masquerading as technical information so caveat lector.  I really wish a fraction of those claims were true, because if there was a real benefit to mixing in the analog domain, I would have the technological edge to promote (my current source summing).

I have been searching for some actual critical flaw in digital combining for years and I haven't found it yet.

I would take this paper with a grain of salt (the size of a deer lick).   Rolling Eyes  As I have said several times, there are real performance consideration when combining a large number of analog stems, but noise is not the dominant one (distortion and phase shift are real consideration on the scale of that paper's comparison). Digital on paper at least, seems to have the edge.

JR

PS: For the record I have nothing on sale here. I have been out of the console/mixer business for more than a decade, and I see little reason to promote a technology that has been eclipsed by digital. I still have friends making and selling analog consoles but they don't resort to such merchandising.
Title: Re: S/N in big consoles
Post by: bruno putzeys on January 24, 2011, 01:43:18 pm
You just saved me quite a bit of time replying there...
Title: Re: S/N in big consoles
Post by: ssltech on January 24, 2011, 02:38:58 pm
Back when that paper was current, I voiced similar comments, and Chris Muth registered and posted some rebuttals... Several people whom I respect all dig Chris's work, and I've used some of his product before and it's certainly been very well made. -I wouldn't wish for a second to impugn his abilities or his product.

None the less, I'm glad that I'm not the only one with these impressions regarding the paper.

-Keith
Title: Re: S/N in big consoles
Post by: zmix on January 24, 2011, 02:59:58 pm
Yes, thanks John...!  (I edited my initial response to the "buss" paper 3 times to maximize politeness...  Rolling Eyes )
Title: Re: S/N in big consoles
Post by: Geoff Emerick de Fake on January 25, 2011, 07:30:45 am
Hey, what's wrong with a fishing boat?  Laughing
Title: Re: S/N in big consoles
Post by: dcollins on January 25, 2011, 06:44:35 pm
ssltech wrote on Mon, 24 January 2011 11:38

Back when that paper was current, I voiced similar comments, and Chris Muth registered and posted some rebuttals... Several people whom I respect all dig Chris's work, and I've used some of his product before and it's certainly been very well made. -I wouldn't wish for a second to impugn his abilities or his product.



That doesn't mean the gear won't sound great, but I was just surprised that both the analog and digital conclusions were off the mark.

Was the paper ever revised?


DC

Title: Re: S/N in big consoles
Post by: John Roberts {JR} on January 26, 2011, 02:56:11 pm
dcollins wrote on Tue, 25 January 2011 17:44

ssltech wrote on Mon, 24 January 2011 11:38

Back when that paper was current, I voiced similar comments, and Chris Muth registered and posted some rebuttals... Several people whom I respect all dig Chris's work, and I've used some of his product before and it's certainly been very well made. -I wouldn't wish for a second to impugn his abilities or his product.



That doesn't mean the gear won't sound great, but I was just surprised that both the analog and digital conclusions were off the mark.

Was the paper ever revised?


DC




I don't mean to cast aspersion on the product or person.

My comments apply to the claims made in the paper in your link.

I will leave it to others to speculate about any connection between that paper and product. I have met, and worked with my share of marketing types who were clueless about the minutiae of design philosophy.

I recall following one co-worker around at an AES show to attempt damage control on the spot. He had memorized a bunch of buzz words from listening to my spiel, but couldn't string them together into a intelligible sentence.   Laughing   Bless his heart he was a selling machine, but an embarrassment to me when I was within earshot.  


caveat lector.  

JR

Title: Re: S/N in big consoles
Post by: JGreenslade on January 27, 2011, 06:04:24 pm
Geoff_T wrote on Wed, 12 January 2011 18:47

Hi

Vintage Neve consoles had signal to noise measured in three stages.

1. The output noise with the 4T fader at minimum... usually up around high 90's

2. The output fader up and a single channel assigned to the 4T bus

3. A channel at 80dB gain assigned to the same path with 20 to 20KHz filter on all these measurements and the input terminated 200 ohms. That last test gave figures better than -45dB = EIN better than 125.

But note, they did not assign all the channels to a bus... just one channel.

All these tests are clearly listed at the rear of the technical handbooks along with a polaroid photo of the square wave response.

Plus all vintage Neve tests were referenced to 0dBu, not +4dBu and the noise was not subtracted from the maximum output like digital measurements these days and power amps.

Hence their wonderful figures!



This reminds me of a chat I had a while back with a certain pipe-smoker who looks after a gaggle of vintage Neves here in the UK. A client had remarked that, whilst their facility could live with the console's noise floor, they were interested in bringing it below that of the new-fangled hard disk recording devices. Said engineer decided to try an experiment, by equipping a selection of bus amp cards with uber-low noise Japanese epitaxial types (the type developed in the eighties for low noise MC cartridge preamps), putting these in place of the original cards that contained BC184s and seeing what difference this made to the broadband noise trace on the FFT...

...suffice to say, there was no difference at all. The transistors themselves are by far one of the least dominant noise sources - as I'm sure the Neve stalwarts here will be fully aware. I just thought it would make an interesting interjection to relay the tale. I guess he wished in his heart that he could walk away from the studio feeling smug that he'd taken 12dB off the console's floor... No cigar that day!

Justin
Title: Re: S/N in big consoles
Post by: John Roberts {JR} on January 27, 2011, 11:46:04 pm
JGreenslade wrote on Thu, 27 January 2011 17:04

Geoff_T wrote on Wed, 12 January 2011 18:47

Hi

Vintage Neve consoles had signal to noise measured in three stages.

1. The output noise with the 4T fader at minimum... usually up around high 90's

2. The output fader up and a single channel assigned to the 4T bus

3. A channel at 80dB gain assigned to the same path with 20 to 20KHz filter on all these measurements and the input terminated 200 ohms. That last test gave figures better than -45dB = EIN better than 125.

But note, they did not assign all the channels to a bus... just one channel.

All these tests are clearly listed at the rear of the technical handbooks along with a polaroid photo of the square wave response.

Plus all vintage Neve tests were referenced to 0dBu, not +4dBu and the noise was not subtracted from the maximum output like digital measurements these days and power amps.

Hence their wonderful figures!



This reminds me of a chat I had a while back with a certain pipe-smoker who looks after a gaggle of vintage Neves here in the UK. A client had remarked that, whilst their facility could live with the console's noise floor, they were interested in bringing it below that of the new-fangled hard disk recording devices. Said engineer decided to try an experiment, by equipping a selection of bus amp cards with uber-low noise Japanese epitaxial types (the type developed in the eighties for low noise MC cartridge preamps), putting these in place of the original cards that contained BC184s and seeing what difference this made to the broadband noise trace on the FFT...

...suffice to say, there was no difference at all. The transistors themselves are by far one of the least dominant noise sources - as I'm sure the Neve stalwarts here will be fully aware. I just thought it would make an interesting interjection to relay the tale. I guess he wished in his heart that he could walk away from the studio feeling smug that he'd taken 12dB off the console's floor... No cigar that day!

Justin


If those are the devices I am thinking of they date back to before the '80s. The 2SB737, and 2SD786, nice very low noise, down around .35 and  .4 nV/rt Hz. I actually used them in MC preamps, and mic preamps back in the day. They are obsolete now.

JR:
Title: Re: S/N in big consoles
Post by: JGreenslade on January 28, 2011, 08:05:55 am
There were a handful of transistors supposedly made specifically for MC carts - I believe up until the early eighties, but you might be right (the audiophile boom seemed to peak around the late '70s from what I've seen). Sanyo and Hitachi both offered devices that were claimed to be developed for this purpose. IIRC, Mackie used the Sanyo parts in one of their consoles. Well, say what you want about Mackie, but their consoles have never been noisy!

There's a Self-designed MC preamp that parallels the SB737 at the front end.

2SA872 - I seem to remember that off top of head.

Justin
Title: Re: S/N in big consoles
Post by: John Roberts {JR} on January 28, 2011, 10:38:15 am
JGreenslade wrote on Fri, 28 January 2011 07:05

There were a handful of transistors supposedly made specifically for MC carts - I believe up until the early eighties, but you might be right (the audiophile boom seemed to peak around the late '70s from what I've seen). Sanyo and Hitachi both offered devices that were claimed to be developed for this purpose. IIRC, Mackie used the Sanyo parts in one of their consoles. Well, say what you want about Mackie, but their consoles have never been noisy!

There's a Self-designed MC preamp that parallels the SB737 at the front end.

2SA872 - I seem to remember that off top of head.

Justin


I was not familiar with the 2SA872. That looks like a true low noise part for mic preamps or perhaps input stage of a DOA, but Rbb (as speculated from constant NF curves) looks larger than 737/786 pair that were optimal for a decade lower source impedance (tens of ohms, not hundreds).

As I recall the 737/786 parts were developed by a small Japanese company that was then bought by ROHM who distributed them in the US. I don't recall any other parts that were really close to them. I used a lot of these in mic preamps where they were probably lower Rbb than needed, and lower than modern replacements since they went obsolete, and IC based preamps that are now widely used.  

JR
Title: Re: S/N in big consoles
Post by: Jim Williams on January 28, 2011, 11:18:00 am
The 2SB and 2SD devices were the slow stuff, the 2SA and 2SC are the fast ones.
Besides those slower parts, there were/are faster varients of them.

The Hitachi 2SA1083 and 2SC2545 series were the quicker parts. Toshiba also makes similar parts, the 2SA1316 and 2SC3329.

Renesas also now makes the 2SA1084.
Title: Re: S/N in big consoles
Post by: Geoff_T on January 28, 2011, 11:20:36 am
JGreenslade wrote on Thu, 27 January 2011 15:04



This reminds me of a chat I had a while back with a certain pipe-smoker who looks after a gaggle of vintage Neves here in the UK. A client had remarked that, whilst their facility could live with the console's noise floor, they were interested in bringing it below that of the new-fangled hard disk recording devices. Said engineer decided to try an experiment, by equipping a selection of bus amp cards with uber-low noise Japanese epitaxial types (the type developed in the eighties for low noise MC cartridge preamps), putting these in place of the original cards that contained BC184s and seeing what difference this made to the broadband noise trace on the FFT...

...suffice to say, there was no difference at all. The transistors themselves are by far one of the least dominant noise sources - as I'm sure the Neve stalwarts here will be fully aware. I just thought it would make an interesting interjection to relay the tale. I guess he wished in his heart that he could walk away from the studio feeling smug that he'd taken 12dB off the console's floor... No cigar that day!

Justin


Hi

In the era of the 8038/8048 the production manager at the time (Niall MacDermot nick-named Flying Officer Kite) came up with a wonderful idea that the B- bus system should be a welded/brazed solid section of 1/4" copper, all the way back to the EP5 input connector. The solid bus bar structure was fitted in the metal shop at the time of console frame assembly and, once in, could never be removed.

The console was duly assembled and tested and the noise floor was no better (I recall it may have been a dB or so worse) and the idea was scrapped.

So there is one console out there somewhere with this wonderful welded bus bar...

Smile

Title: Re: S/N in big consoles
Post by: John Roberts {JR} on January 28, 2011, 11:53:50 am
Jim Williams wrote on Fri, 28 January 2011 10:18

The 2SB and 2SD devices were the slow stuff, the 2SA and 2SC are the fast ones.
Besides those slower parts, there were/are faster varients of them.

The Hitachi 2SA1083 and 2SC2545 series were the quicker parts. Toshiba also makes similar parts, the 2SA1316 and 2SC3329.

Renesas also now makes the 2SA1084.


Yup,  the larger base-emitter structure to get base spreading resistance down so low (typically 2 ohms in 737) has other consequences (higher capacitance? etc). I am inclined to say those others are not the same parts just faster, but instead tweaked differently with smaller geometry, so arguably different parts.

Since the very low noise parts are lower noise voltage than they need to be for most  current applications, RIP.  

While I haven't been keeping up with the new ribbon mics. Aren't they pretty low source impedance?

JR
Title: Re: S/N in big consoles
Post by: bruno putzeys on January 28, 2011, 01:23:51 pm
Yes, but even lower than that. I've not yet managed to make a transformerless pre for a ribbon with sufficiently low noise.
Title: Re: S/N in big consoles
Post by: John Roberts {JR} on January 28, 2011, 03:21:41 pm
bruno putzeys wrote on Fri, 28 January 2011 12:23

Yes, but even lower than that. I've not yet managed to make a transformerless pre for a ribbon with sufficiently low noise.


Lower than 2 ohms? Interesting, you'd need a transformer just to get it up to a typical mic impedance, and the transformer windings have to get significant down in single digit ohms..

Transformers were popular for MC preamps too. I got some favorable responses to my transformer less MC amp back in early '80s. .  

JR
Title: Re: S/N in big consoles
Post by: Geoff Emerick de Fake on January 28, 2011, 04:46:00 pm
A ribbon has a typical impedance of about 1/10th of an ohm. The necessary transformer uses a few turns of large-gauge wire at the primary.
Even with these precautions, most of the reflected impedance at the secondary is resistive.
Title: Re: S/N in big consoles
Post by: John Roberts {JR} on January 28, 2011, 05:57:51 pm
Geoff Emerick de Fake wrote on Fri, 28 January 2011 15:46

A ribbon has a typical impedance of about 1/10th of an ohm. The necessary transformer uses a few turns of large-gauge wire at the primary.
Even with these precautions, most of the reflected impedance at the secondary is resistive.


Thanks, that does sound like it wants to be a transformer input...  I guess you could parallel a silly number of these very low noise transistors, but they're obsolete now anyhow.

I've seen some circa 1 nV JFETs, I wonder if they are better candidates to parallel? While noise current into .1 ohm can't be that much of an issue.  

Perhaps 6 of one, half dozen the other, or  maybe six of one "and" a half dozen the other.   Laughing

JR

Title: Re: S/N in big consoles
Post by: Geoff Emerick de Fake on January 28, 2011, 06:09:01 pm
As Bruno mentioned, there have been some experiments. The essence of the problem is that, in order to achieve a good ratio of En to In, the input transistor compound would have to operate at several amps quiescent.
Title: Re: S/N in big consoles
Post by: John Roberts {JR} on January 28, 2011, 08:37:55 pm
Geoff Emerick de Fake wrote on Fri, 28 January 2011 17:09

As Bruno mentioned, there have been some experiments. The essence of the problem is that, in order to achieve a good ratio of En to In, the input transistor compound would have to operate at several amps quiescent.


The low noise parts already need a couple mA for MC amps.

That's why I mentioned the JFETs, while nominall starting at higher En, their In should be much lower than the bipolar, but still either way that In is only working into < 1 ohm so will generate modest noise voltage.

I'm more concerned about a suitable cascode/bootstrap topology to deal with all the device capacitance from a bunch o' parts in parallel.

But not worried enough to kill brain cells over this... unless ribbon mics get real popular.

JR
Title: Re: S/N in big consoles
Post by: Geoff Emerick de Fake on January 29, 2011, 06:22:07 am
With a <1ohm source, you need noise voltage <0.1nV/sqrootHz, which is theoretically achievable by running the input transistor compound at about 1 amp.
With FET's, you would need to parallel 100 of them to achieve that, with about the same quiescent current.
Title: Re: S/N in big consoles
Post by: Jim Williams on January 29, 2011, 12:14:24 pm
Actually less of them if you use the right parts. There are very low noise jfets available. Linear Systems makes the selected low noise 1 nv 2SK170. NXP (Phillips) makes the BF862 at .7 nv voltage noise and minimal current noise. That part does work well in mic preamp front ends without the noise trade-off of most other jfets. That also allows one to set the input impedance very high to minimize loading. It also allows you to reduce the value of the input blocking caps letting you select quality film caps in place of traditional electrolytic caps to block 48 volts.
Title: Re: S/N in big consoles
Post by: John Roberts {JR} on January 29, 2011, 01:26:03 pm
Jim Williams wrote on Sat, 29 January 2011 11:14

Actually less of them if you use the right parts. There are very low noise jfets available. Linear Systems makes the selected low noise 1 nv 2SK170. NXP (Phillips) makes the BF862 at .7 nv voltage noise and minimal current noise. That part does work well in mic preamp front ends without the noise trade-off of most other jfets. That also allows one to set the input impedance very high to minimize loading. It also allows you to reduce the value of the input blocking caps letting you select quality film caps in place of traditional electrolytic caps to block 48 volts.


Yup, my recollection is that some of these low noise JFET have Idss well into the tens of mA, so when biased up for lowest noise, you can get up into amps long before 100 parts. Of course you will need to manage package dissipation, etc.  

Somewhere I saved some data sheets for a silly low noise JFET but I am not sure those were audio bandwidth parts.  I used a 2sk117 back in the '80s in a phono preamp and that was a sweet part. No doubt they've improved on that in 30 years.


JR

PS: I also made a simple phantom powered DI for bass guitar (or whatever)  out of a pair of 2sK117, that sounded OK.
Title: Re: S/N in big consoles
Post by: Geoff Emerick de Fake on January 31, 2011, 01:58:33 pm
Jim Williams wrote on Sat, 29 January 2011 11:14

Actually less of them if you use the right parts. There are very low noise jfets available. Linear Systems makes the selected low noise 1 nv 2SK170.
That's exactly what I said; you need to parallel 100 of them to achieve 0.1nv.
Quote:

 NXP (Phillips) makes the BF862 at .7 nv voltage noise and minimal current noise.
Then you only need 50 of them.
Quote:

 That also allows one to set the input impedance very high to minimize loading.
The actual input impedance is more dependant on the topology and design choices than by the devices. The  typical Cohen preamp with LM394's has an input impedance >100k.
Title: Re: S/N in big consoles
Post by: Jim Williams on February 02, 2011, 10:54:51 am
Geoff Emerick de Fake wrote on Mon, 31 January 2011 10:58

Jim Williams wrote on Sat, 29 January 2011 11:14

Actually less of them if you use the right parts. There are very low noise jfets available. Linear Systems makes the selected low noise 1 nv 2SK170.
That's exactly what I said; you need to parallel 100 of them to achieve 0.1nv.
Quote:

 NXP (Phillips) makes the BF862 at .7 nv voltage noise and minimal current noise.
Then you only need 50 of them.
Quote:

 That also allows one to set the input impedance very high to minimize loading.
The actual input impedance is more dependant on the topology and design choices than by the devices. The  typical Cohen preamp with LM394's has an input impedance >100k.


Which is still not high impedance like you can do with the jfets, I'm talking meg ohms here.  100k bias resistors used alone with  a LM394 will cause a lot more resistor source noise than that part is designed for, a low input impedance to match it's noise and current noise specs. Once the 6.81k phantom resistors are fitted, that sets it low again anyway. With a phantom switching scheme that switches out the phantom resistors and input blocking caps you can raise that input impedance pretty high with the jfets without source impedance noise problems that occur with bipolar transistors.
Title: Re: S/N in big consoles
Post by: Geoff Emerick de Fake on February 02, 2011, 02:44:28 pm
Jim Williams wrote on Wed, 02 February 2011 09:54

 
Which is still not high impedance like you can do with the jfets, I'm talking meg ohms here.
What's the use?
Quote:

 100k bias resistors used alone with  a LM394 will cause a lot more resistor source noise than that part is designed for, a low input impedance to match it's noise and current noise specs.
This is irrelevant. Once the source impedance is connected, the impedance the input devices see is much lower. 200 ohms in parallels with 200kohms makes 200ohms.
Quote:

 Once the 6.81k phantom resistors are fitted, that sets it low again anyway. With a phantom switching scheme that switches out the phantom resistors and input blocking caps you can raise that input impedance pretty high with the jfets without source impedance noise problems that occur with bipolar transistors.
What are "source impedance noise problems"? What's the difference between 100k in parallels with 200r and 100Meg in paralles with 200r? The only effect on noise the input impedance has is the loading effect. 0.018dB difference. Nothing to write home about.
Title: Re: S/N in big consoles
Post by: John Roberts {JR} on February 02, 2011, 06:23:16 pm
Indeed.. I don't follow what Jim is concerned about.

I used to terminate all with 2k (bridging to 200 ohm) terminations, but have since read some mic designers designed for somewhat higher, but not 100k high. .

I suspect some preamp designers use high Z terminations, just so they'll sound different. "you can' t sound better if you sound the same as everybody else".   Rolling Eyes

---

Back to the ribbons, I am curious about very low noise JFETs mostly for academic reasons not for any imagined practical advantage, especially with ribbons.

JR

Title: Re: S/N in big consoles
Post by: Jim Williams on February 04, 2011, 11:30:47 am
No concern, it's just another way to skin a cat. Sonically an all jfet low noise high input impedance mic preamp design does sound different. As I mentioned before, the higher input impedance allows the designer to use smaller value, higher quality film caps to block 48 volts in place of common larger valued electrolytic capacitors. It does allow one to inject high impedance sources without a DI box, but that's not even needed.  Besides an all jfet design, I also have some with all transimpedance amplifiers, those are very fast and once again, different sounding. They run at over 100 mhz and sound very good.

Which ever topology I use, they all sound a lot better to me than the commercial stuff you can buy.
Title: Re: S/N in big consoles
Post by: John Roberts {JR} on February 04, 2011, 05:32:08 pm
Jim Williams wrote on Fri, 04 February 2011 10:30

No concern, it's just another way to skin a cat. Sonically an all jfet low noise high input impedance mic preamp design does sound different. As I mentioned before, the higher input impedance allows the designer to use smaller value, higher quality film caps to block 48 volts in place of common larger valued electrolytic capacitors. It does allow one to inject high impedance sources without a DI box, but that's not even needed.  Besides an all jfet design, I also have some with all transimpedance amplifiers, those are very fast and once again, different sounding. They run at over 100 mhz and sound very good.

Which ever topology I use, they all sound a lot better to me than the commercial stuff you can buy.


Speaking of other ways to skin the cat, I like the concept of just letting a DC coupled mic preamp float up to the phantom voltage and cap couple with higher impedance film caps after it's up at line level.  One could fly an A/D converter up there with the preamp and literally do a capacitor free audio path...  While I am still inclined to go with conventional (circa 2k) resistive terminations. Of course non-standard terminations could be an obscure feature for esoteric variants.

JR

PS: I think my designs sound good too...  Laughing  Doesn't everybody think they have pretty (or smart)  children?