schematic question - spectra sonics 110A
- Thread starter QUEEF BAG
- Start date
Help Support GroupDIY Audio Forum:
The MCI JH400 series, which used VE mixers, was designed in 1968. I don't know if they were first. When I started designing mixers in 1975, there were articles describing the advantages of the "new" technology and certainly the National application notes included it.
> virtual earth summing was probably predicted by analog computing.
Covered in Korn and Korn 1952.
Electronic analog computers: (d-c analog computers)
Granino Arthur Korn, Theresa M. Korn
McGraw-Hill, 1952 - Mathematics - 378 pages
Much of K&K is summary of existing practices.
How the idea snuck into audio, I can not say. However 1966-1976 was a burning time for putting opamps into audio (even when the fit was poor), and a few of those guys may actually have been aware of the 25 year traditions of analog computing.
There's also the fact that when you want high=Z inputs into a single transistor mix-recovery amp, it becomes quite natural to have mix resistors larger than the base impedance, then wrap what little NFB you can find from collector to base to stabilize gain. So it may have been done without knowing the concept.
Covered in Korn and Korn 1952.
Electronic analog computers: (d-c analog computers)
Granino Arthur Korn, Theresa M. Korn
McGraw-Hill, 1952 - Mathematics - 378 pages
Much of K&K is summary of existing practices.
How the idea snuck into audio, I can not say. However 1966-1976 was a burning time for putting opamps into audio (even when the fit was poor), and a few of those guys may actually have been aware of the 25 year traditions of analog computing.
There's also the fact that when you want high=Z inputs into a single transistor mix-recovery amp, it becomes quite natural to have mix resistors larger than the base impedance, then wrap what little NFB you can find from collector to base to stabilize gain. So it may have been done without knowing the concept.
Hello ... I thought I would weigh in on the original question and explain why the 101 and 110 are emitter inputs. First let me tell you of my background. I heard Bill Dilley give the original 101 paper at the Hollywood AES meeting. Shortly thereafter, the firm I worked for became his Hollywood distributor, and we were flown to the factory for training. We worked with and sold the products for a couple of years, then I left that firm and went to work for Spectra Sonics for four years in their newly opened Hollywood Office. I sold a lot of the product, designed a lot of systems / consoles, installed consoles all over Hollywood. During this period, I was personally tutored by Bill Dilley, and his design philosophy was drilled into me. After leaving, I have continued to use, design with, and build with the Spectra Sonics products. So that's where my background is ... let me address the question.
The reason the emitter input is used is several fold. First of all, it is inherently low-impedance. That means, when an input transformer is used, it can be a 1 to 1 ratio. A 1 to 1 ratio transformer will always outperform a step-up transformer. Secondly, the circuit is inherently quieter than the common emitter designs used in op-amps. For nearly all the op-amp designs, the lowest noise at the input will be with about a 10 k input impedance... so if operated from a microphone, without a matching transformer, you are not operating at the lowest point of the noise curve.
With the emitter input, whatever source impedance you have is in parallel with the emitter resistor, and therefore the input transistor, which determines the noise floor, sees an even lower impedance. The input impedance is 604 ohms, (or in special cases, 2.4k) and the typical dynamic 200 ohm nominal microphone is usually about 120 ohms, so the input transistor sees the parallel impedance of that 120 ohms across its 604 ohm resistor. This is important, because this input circuit gets quieter as the input impedance gets lower.
This configuration is noticeably quieter than most other inputs ... and the difference isn't subtle. There are several other things that contribute to the low noise floor, but that configuration was chosen specifically for its low noise characteristics. The original AES paper on the 101 actually spells out the derivation of this circuitry, and the whys and hows that it came to be.
Hope this helps answer your question.
The reason the emitter input is used is several fold. First of all, it is inherently low-impedance. That means, when an input transformer is used, it can be a 1 to 1 ratio. A 1 to 1 ratio transformer will always outperform a step-up transformer. Secondly, the circuit is inherently quieter than the common emitter designs used in op-amps. For nearly all the op-amp designs, the lowest noise at the input will be with about a 10 k input impedance... so if operated from a microphone, without a matching transformer, you are not operating at the lowest point of the noise curve.
With the emitter input, whatever source impedance you have is in parallel with the emitter resistor, and therefore the input transistor, which determines the noise floor, sees an even lower impedance. The input impedance is 604 ohms, (or in special cases, 2.4k) and the typical dynamic 200 ohm nominal microphone is usually about 120 ohms, so the input transistor sees the parallel impedance of that 120 ohms across its 604 ohm resistor. This is important, because this input circuit gets quieter as the input impedance gets lower.
This configuration is noticeably quieter than most other inputs ... and the difference isn't subtle. There are several other things that contribute to the low noise floor, but that configuration was chosen specifically for its low noise characteristics. The original AES paper on the 101 actually spells out the derivation of this circuitry, and the whys and hows that it came to be.
Hope this helps answer your question.
gridcurrent
Well-known member
- Joined
- Aug 28, 2009
- Messages
- 589
LdSpkrPro said:
Good to hear you chime in, Mr. G.
Not to argue any of your points, but lowering the input impedance loads the microphone, with purported negative effects.
Do hope you will discuss the design of the sound system at the Whiskey !
I would say that it HAS to be 1:1. The circuit's input impedance is governed by the emitter resistor. The transistor's Re (25/Ic), which is in parallels with it is in fact almost completely compensated by NFB, so the resulting impedance is probably about 2k. It is in accordance with the accepted standard of 10x the mic's impedance. Using a step-up xfmr would improperly load the microphone.LdSpkrPro said:
The actual source impedance seen by the transistor is dominated by the microphone, at about 200 ohms, and the 100r res to ground, so the transistor's operating point must be tailored for such a low impedance; that's why it runs at about 500uA.
Opamps optimised for lower source impedance exist though, having their input transistors operating at a few mA instead of microamps. I think they were rarer than hen's teeth in the 70's.
There is a law of diminishing returns there, because the loss introduced by the emitter resistor increases more rapidly than the noise decrease. When the input impedance is equal to the source impedance, the noise advantage is a tad above 3dB (3dB because of the Johnson formula, and a tad because the contribution of noise current is slightly smaller, but noise voltage is constant and dominant), but the signal loss is 6dB.
I must say I'm not convinced. My left brain says it cannot be better in terms of noise than a common-emitter stage. My right brain says if it was indeed better, why hasn't it become the de facto standard?
I haven't seen this paper; maybe I'm missing something, but I can't help thinking that the designer has found that a finely-tuned common-base stage performs better than an out-of-whack common-emitter.
Anyway, thanks for your valuable insight.
Grid Current ... you have me at a disadvantage ... you obviously know who I am ... but your are a mystery to me.
The 101's input impedance at 600 ohm was designed for dynamic microphones, and for most, is non-terminating. While marked 200 ohms, they are usually lower, in the range of 100 to 120 ohms.
The 110 was available in a higher input impedance version with the use of a 2.43 k input resistor. This was specifically requested for use with condenser microphones ... most of which at that time were transformer output.
The same input was used in the 104 which was a moving coil / riaa equalized preamp ... in that unit the input resistor was 50 ohms, an d it was designed to operate from 2 ohm output cartridges.
What do you want to know about the Whisky system ??? That discussion could go on for days!
The 101's input impedance at 600 ohm was designed for dynamic microphones, and for most, is non-terminating. While marked 200 ohms, they are usually lower, in the range of 100 to 120 ohms.
The 110 was available in a higher input impedance version with the use of a 2.43 k input resistor. This was specifically requested for use with condenser microphones ... most of which at that time were transformer output.
The same input was used in the 104 which was a moving coil / riaa equalized preamp ... in that unit the input resistor was 50 ohms, an d it was designed to operate from 2 ohm output cartridges.
What do you want to know about the Whisky system ??? That discussion could go on for days!
Abbey road d enfer ...
In circuit, measured, you'll find it to be just about 600 ohms or bit higher.
The fact that it has not become a de facto standard probably has something to do with the fact that the entire audio business is like a parade of circus elephants, every one of them holding onto the tail of the one in front of it, bitching about the view, but afraid to let go and do any independent thinking.
There is so little original thinking, instead, I see a lot of outright copying in the audio business. That and the fact that people seem content with Band-aids for their problems, instead of getting rid of them in the first place.
It has been my experience that people who want to argue about the input noise have usually had very little in-field experience measuring it. I used to install SS consoles, and the last part of the installation was to demonstrate the guaranteed performance to the purchaser ... and that included the 80 dB signal to noise ratio as shown on an HP 400GL. We were always able to demonstrate that ... while I know of other manufacturers who had to make cash rebates because consoles didn't meet their advertised specs.
In circuit, measured, you'll find it to be just about 600 ohms or bit higher.
The fact that it has not become a de facto standard probably has something to do with the fact that the entire audio business is like a parade of circus elephants, every one of them holding onto the tail of the one in front of it, bitching about the view, but afraid to let go and do any independent thinking.
There is so little original thinking, instead, I see a lot of outright copying in the audio business. That and the fact that people seem content with Band-aids for their problems, instead of getting rid of them in the first place.
It has been my experience that people who want to argue about the input noise have usually had very little in-field experience measuring it. I used to install SS consoles, and the last part of the installation was to demonstrate the guaranteed performance to the purchaser ... and that included the 80 dB signal to noise ratio as shown on an HP 400GL. We were always able to demonstrate that ... while I know of other manufacturers who had to make cash rebates because consoles didn't meet their advertised specs.
?? You paint an amusing picture, but I can't say I recognize that behavior in any peers, let alone an entire industry.LdSpkrPro said:
You need new friends, or at least different acquaintances. Your rose colored glasses are a little dark.
I recall some interesting disagreements (opinions?) about how exactly to specify ein, but I think Buff's article in RE/P back when there was an RE/P, was decent. Paul even sold a noise measurement rig for a while. I always preferred to think in terms of NF, for low noise design, since it is harder to game with BW or weighting curves.. This isn't rocket science but admittedly the rocket scientists were busy making rockets.
The low noise transistors I used (common-emitter) for mic preamps were actually tooled up by a small Japanese company to use in MC head amps, so more optimal at tens of ohms than hundreds. Now-a-days, off the shelf ICs can deliver a quite respectable mic preamp.
JR
Indeed, the schemo published by the OP indicates a 2.43k res; it may be a special version. This is the one I'm referring to. Who am I to know?LdSpkrPro said:
Not me , sir... I have designed mixers since 1976, achieving -128dBu EIN at 200 ohms/20k BW, which is 1.7dB NF with off-the-shelf transformers and commonly available discrete transistors. Once this not particularly difficult issue was addressed, I devoted most of my time solving the much more complex problem(s) of longitudinal noise.
I have also commissioned dozens of broadcast mixers, and I know they are not particularly demanding regarding EIN or NF (-127+/-1dB or ca 3dB NF is accepted), but they insist on output noise, x-talk and interference protection.
In fact, designing an input amp with -125dBu is nothing to write home about, but it would measure at 85dB S/N with a-40dBu input signal.
It is nevertheless very difficult to carry this performance to the outputs, when the signal has gone through all the stages, including the dreaded summing amps (I know that JR and me differ on the subject, but we agree on the mechanisms; our differences are on what is the most noxious).
It is very difficult to guarantee 80dB S/N ratio for 32ch. engaged, -40dBu input, al loaded 200r, +4dBu output, 20Hz-20kHz BW); that's the way the french broadcast tests large mixers. And they certainly don't expect 80dB S/N in these conditions, the theoretical limit being -70.7.
(-129.7dBu Johnson noise, 44dB gain, 15dB quadratic noise sum for 32ch)
For an on-air mixer with only a couple of mic ins, 80dB is doable, but doesn't imply an über-low-noise mic pre. The difficulty in that case is EMI/RFI and input overhead.
It seems that we are talking about apples and oranges ... or a least significantly different measurement conditions. When we demonstrated guaranteed performance after the installation of a Spectra Sonics audio mixing console, these were the measurement conditions: a - 50 signal was input to a microphone preamp, and the gain thru the console adjusted to produce +4 dBm at the output. All other faders down. The signal was removed and replaced by a 200 ohm resistor. Then the noise on the console output was measured over a 20 to 20 kHz bandwidth. It was guaranteed to be 80 dB below the + 4 dBm output. This was thru the entire console, installed.
In fact, one of the games Dilley used to do in selling consoles was to post a $10,000 bond with the purchaser's bank guaranteeing that the installed console would meet it's advertised specifications. It usually caught the attention of someone in studio management when other console manufacturer's wouldn't accept the challenge.
The other interesting spec was the harmonic distortion, again, thru the entire console. It was spec-ed and delivered at less than 0.01 %, 20 to 20 kHz, at output level. That was also demonstrated. And this was using 'off the shelf' transformers ... well.
But then, I think that was a different time. Today, people like to quote theoretical limits on noise as gospel... or excuses for mediocre performance.
In fact, one of the games Dilley used to do in selling consoles was to post a $10,000 bond with the purchaser's bank guaranteeing that the installed console would meet it's advertised specifications. It usually caught the attention of someone in studio management when other console manufacturer's wouldn't accept the challenge.
The other interesting spec was the harmonic distortion, again, thru the entire console. It was spec-ed and delivered at less than 0.01 %, 20 to 20 kHz, at output level. That was also demonstrated. And this was using 'off the shelf' transformers ... well.
But then, I think that was a different time. Today, people like to quote theoretical limits on noise as gospel... or excuses for mediocre performance.
gridcurrent
Well-known member
- Joined
- Aug 28, 2009
- Messages
- 589
JohnRoberts said:
Paul's measurements did not agree with anyone's. A theoritical limit of EIN of about -124 dBme, whatever that was supposed to be.
Certainly the latter.LdSpkrPro said:
This is only one measurement, not much related to the typical operation of a mixer. Who wants a mixer with just one input used?
Measurements done with all channels routed are much more challenging and more related to real life.
That is less than the Johnson noise of the 200r source. So these mixers had negative Noise Factor? I understand your desire to past the paint with rosy colours, but it must stand the test of scrutiny.
At what frequency? At 1 khz, most xfmrs have less than 10 ppm THD.
You seem to have a very pessimistic opinion of the state of the industry; it may be realistic for some of the cheaper brands, but there are a number of respectable manufacturers who devote themselves to the avancemnt of the audio art. I think Millenia, Forsell, RND, and many others do not deserve your despise.
gridcurrent said:
And why I avoided all the pissing contests, by using NF to design with personally, but I could never get customers and marketing types for let go of their attraction to three digit "ein" numbers, to declare they are so much better than the next guy when many such measurements were not even identically performed, so not comparable.
Yes, I consider much of this a paper chase, and gamesmanship since a few mics in use in a real room will surely dominate the practical noise floor. This doesn't mean that attention to noise in a design doesn't matter, only that we need to hold isolated specifications in perspective when a full audio path is involved.
I guess opinions vary.. I won't put words in anybody's mouth. I have known many competent designers with several hanging out here. This still isn't rocket science.
JR
Yes, that is only one measurement. However, you could pick any of the inputs and get the same measurement! And as to the point of you don't use only one input ... that is quite correct. If you start with a substantially lower noise floor on each individual input, then you wind up with a lower noise floor in the mix ... that's the whole point.
It's hard to call it painting the past with rosy colors, that's what I actually measured ... without any games, with HP 400GL. As I pointed out in an earlier post, the input circuit delivers better than expected noise floor because of the configuration. Everyone seems to miss that point. It delivers better than the theoretical.
But the real significant point here is, that was the measurement for a fully terminated console, with all 24 tracks and peripheral gear connected. And if you think that doesn't contribute to noise floor because of grounding considerations, they I would suggest you've not terminated very many large consoles!
I remember one console termination where we had to disconnect the clients ground rod which they had carefully, and with much difficulty driven 12 feet into the earth, right thru the trough beneath the console. It was a suberb RF antenna. Apparently, it didn't occur to them that the ground would be quite dry under the 100 year old building.
The harmonic distortion specification was for any frequency, 20 to 20 kHz.
I agree there is some interesting work being done ... but there's also a maniacal hanging on to variations of the TL0-72. And I seen copies of other manufacturer's printed circuit boards copied, without even removing their logos ... I think my characterization of the audio business as a parade of circus elephants is apt.
It's hard to call it painting the past with rosy colors, that's what I actually measured ... without any games, with HP 400GL. As I pointed out in an earlier post, the input circuit delivers better than expected noise floor because of the configuration. Everyone seems to miss that point. It delivers better than the theoretical.
But the real significant point here is, that was the measurement for a fully terminated console, with all 24 tracks and peripheral gear connected. And if you think that doesn't contribute to noise floor because of grounding considerations, they I would suggest you've not terminated very many large consoles!
I remember one console termination where we had to disconnect the clients ground rod which they had carefully, and with much difficulty driven 12 feet into the earth, right thru the trough beneath the console. It was a suberb RF antenna. Apparently, it didn't occur to them that the ground would be quite dry under the 100 year old building.
The harmonic distortion specification was for any frequency, 20 to 20 kHz.
I agree there is some interesting work being done ... but there's also a maniacal hanging on to variations of the TL0-72. And I seen copies of other manufacturer's printed circuit boards copied, without even removing their logos ... I think my characterization of the audio business as a parade of circus elephants is apt.
Maniacal? I recall when the TL07x was so new that the process wasn't very stable (I had to test them 100% for first few years), but it was a remarkable value for the modest price back in the '70s. Even today it is not a very weak link when properly applied. Of course we have a lot of better choices (for more money) today to use in more demanding sockets. I don't see much value in debating the hammers or nails without specific context of exactly how used, and not sure I want to go down that rabbit hole very far.LdSpkrPro said:
Part of the real work of design engineering is keeping performance vs. price in a yin/yang balance. The 072 generally won't hurt you for light lifting, while there are more modern choices.
JR
PS: I do not deny that there were elements of the audio business that prospered from near carbon copies of other company's successful products (Xerox engineering). It was not unusual to see mistakes copied too when they didn't have a solid grasp of the technology or product segment. But I repeat, this was not the majority of the industry, but a few isolated players. While always a certain amount of uncompetitive copying occurs in fringe markets. I've seen laughable knock-offs from So America and China over the years. That's why we have patents, but some of the deeper pocket players can get around even patents some of the time (don't ask me how I know that. :'( ).
This ends the discussion, doesn't it? You have the right to believe that the company you worked for had magic powers that allowed them to dispense with the laws of physics, you have the right to believe that you are so much above us mere mortals, I have the right to be uninterested in sterile divagations.LdSpkrPro said:
You don't know me, you don't know what my professional history is. I'm not going to start a pissing contest with you.
;D
I guess you think that referring to me as believing in magical powers somehow diminishes what I reported. The simple facts of the matter are the discussion was about the 'strange' input circuitry, and the whys it existed. I attempted to explain some of the whys of the circuit, and what the advantages were, and that the configuration allowed for a significant improvement noise-wise. I also reported what I measured, and the conditions under which it was measured. That you don't believe it possible, that is measures better than the so-called theoretical limits for a given impedance, only demonstrates that you've not personally measured it. I'm neither a fool nor a liar, nor do I believe in 'magic.' I do know what I measured, and I reported it accurately.
Lets keep this about things not people (or elephants).LdSpkrPro said:
Are you saying theory is wrong? if there is a significant deviation from theory, either the theory needs to be adjusted, or something is not what it appears with the measurement.
In noise voltage measurements one common error is related to bandwidth. Another is correlation between noise "power" and noise "voltage". Theory seems clear about specific noise power in dBm, while translating this theoretical power level into an in-circuit voltage involves accounting for true impedance and other possible errors. The input termination shares some of the common input noise power while this is much less of an issue in modern bridging terminations, etc.
Just thinking about it now decades later is giving me a headache. There were some knock down drag out fights back in the day between people convinced that each other was wrong. I tended to side with Buff who i knew personally and seemed very solid on this subject (and others).
I didn't waste a bunch of energy challenging other peoples voltage numbers, I preferred to just make my designs as good as I could and move on.
JR
PS: Just in case you haven't noticed some people participating in this thread consider themselves participants in the audio industry you are so unflattering about. We could have a more thoughtful discussion with less dismissive language and imagery.
PPS: Welcome to the forum, now what do you really think... 8)
