schematic question - spectra sonics 110A

> Who wants a mixer with just one input used?

That's just argumentative.

Very few real mixes have all mikes at the same gain. (A choir may be an exception.) If one is more than a few dB hotter than the rest, it dominates the output hiss. Anyway it only assumes that the mix-network designer has done reasonably well. If mix-node impedance is similar to mike impedance, we can simplify and say that mike-amp gain must be greater than the number of inputs. An 8-in only needs 26dB mikeamp gain. Since the card in question goes down to 35dB which is 1:56, 16- and 24-in mixers should be dominated by mike-amp not mix-amp hiss.
_________________________

> 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.

So: input at "-50" (dbM or dBV or dBu?), total product adjusted for 54dB gain.

If "-50" is dBu or pseudo-dBm-in-600, that is 2.45mV in, a reasonable realistic level.

Now shunt 200 ohms, find output noise at -80 below +4dBm. -76dBm.

Referred to input, (-76)+(-54)= -130dBm.

Which is 0.245 micro-Volts.

A weeny 2N5087 at 0.3mA will do 0.280uV in 20KHz. The 0.28uV is datasheet typical; some will be worse, and some will do a bit better (without breaking the Law). For the time and budget, it is not hard to weed-out the hissy half and get a better average. A fatter device will do better. The very-spiffy LM194 seems to be under 0.2uV at 0.3mA.

0.245 apparently-measured, versus 0.28 or 0.2 for readily available devices, is tight but hardly worthy of scorn and doubt.

I do have quibs (not full quibbles). Is there a transformer? Is the "-50" input referenced to 600r, 200r, 150r, or what? What is the -50's source impedance? (If 50 ohms, then effective gain drops a small amount with the 200r resistor; if a 600r pad then gain rises on 200r source.)

I _am_ reminded of some other old designs with odd action. The first BiFet low-price consoles were a 1:1 tranny into an opamp INVERTER with 470 in and 0-500K gain-pot. Maxed, it had gain of 1:1,000; but if you DIS-connected the input it was unity-gain and utterly silent. Which is not proper measurement technique but actually a useful thing in live-sound where inputs did get pulled loose. (Easier on the audience, harder on me finding why the cowbell was gone.)

John, you are correct in pointing out the error is the measurements. A noise measurement without a specified bandwidth is ridiculous. And a measurement without any bandwidth limit is misleading. Some devices have a fairly narrow 40, 50 kHz bandwidth, while others have a megacycle or more. A wide-band comparative measurement on two such different systems will result in meaningless numbers.

I went thru a lot of these arguments in the late 60's and 70's, and even attended a meeting of the Hollywood AES Chapter meeting where Dilley was 'sandbagged' by some of the industry 'heavyweights' who wanted to take him to task. It was sorta like a 'roast,' or we're gonna take you to town for your claims. The result was embarrassing, and Dilley had a great time.

Here's the point that seems to get missed. Dilley knew the limitations. He also knew that the only way to do better, was to change the game. That peculiar input circuit is a game changer ... it literally improves on the theoretical. The input device noise floor is determined principally by that input resistor combined with the source impedance. For any impedance 600 ohms or lower, the input device will see a lower impedance than its input resistor, and the noise floor will be reflective of the combined lower impedance.  This is true whether the input impedance is 604 ohms, or the 2.43 k used in the 110A.  So in all cases, 600 ohms or lower, the amplifier gets quieter ... by as much as 3 to 4 dB ... and at those noise levels, that is one hell of a lot.

As far as measurements go ... I know that measuring other consoles, with exactly the same termination and measuring equipment, the readings were significantly poorer. Whether you agree with these specific measurements, the same technique and equipment on two different pieces of equipment, measured in exactly the same manner produced significantly different results. That means something is different.

I don't know Buff personally, but I do know other respected industry engineers who thought he was full of it ... on this subject.... frankly I don't remember much of what he said on the subject. I'd have to go dig out old copies of the magazine.

I'm sorry you took offense at my characterization ... I guess it comes from my too-dark rose colored glasses. My original intent was to contribute something of meaning to the conversation ... from an insider's viewpoint and experience. Not all that is new is better, and not all that is old is obsolete. There were serious engineering breakthroughs that are still worth examining.

PRR ... Good point. The really important thing is, if all the preamps are quieter, the resultant mix will be quieter...irrespective of what the relative 'gains' are between inputs.

My experience has been that in most consoles, the active combining network is often the noise limiting factor ... especially as the consoles got larger and larger. That configuration is NOT the quietest way to use an amplifier.

The -50 figure was set to dBm ... I remember because we had to use a 2.2 dB correction because the HP 400GL measured dBV ... reference 1 volt.

The input transformers were set for 600 to 600 (Triad A-67J) ... and as such, it slightly degraded the S/N because of the transformer's  insertion loss !

The gain of the input transistor changes with the lower input impedances.  Which makes it a monster when that circuit is used in the 104 (moving coil cartridge input) amplifier. Then transistors had to be selected for stability.  That input transistor that was employed had a 10 mHz small signal bandwidth.

There were some 1 to 3 or 1 to 5 Jensen transformers that Ed Reichenbach designed which actually improved the noise floor of a lot of those common-emitter amplifiers ... but by that time, everyone was being sold the idea that 'transformerless' was better.  And that's a subject I could go on about for days ....

One of the things I learned from Dilley was that the manner in which a device was employed was as important as the device itself. It's a pity that those who want to dismiss his accomplishments as 'magic' or 'impossible' have never actually tried them.

PRR said:

> Who wants a mixer with just one input used?
That's just argumentative.

Click to expand...

I beg to differ. Measuring a mixer with just one input active is relevant only if the rest of the chain is close to perfection, which is not always the case. I have found many cases where the noise contribution of the summing amps was significant. This worsens the one-input case, but becomes negligible with several channels engaged.
If I want to measure the mic amp's intrinsic noise (EIN or NF), I'll do that from the mic input to the next available node (insert send or direct output). Once I am satisfied that the performance I measured is accurate, I will measure bus noise. Gives me a better assessment of those two distinct noise sources. Measuring input-to-output is correct when doing frequency response or THD.

Very few real mixes have all mikes at the same gain. (A choir may be an exception.)

Click to expand...

There's more than one exception. That is as much argumentative as "> Who wants a mixer with just one input used?". We are talking about the protocols used in commissionig mixers, in regard of noise. I've never seen a broadcaster content with input-to-output noise assessment alone.

0.245 apparently-measured, versus 0.28 or 0.2 for readily available devices, is tight

Click to expand...

I don't know what protocol you are referring to, but in my book, EIN is inclusive of source resistance noise (that's the definition used by broadcasters in Europe), so your example would compute at 0.353 uV or -126.8 dBu.

but hardly worthy of scorn and doubt.

Click to expand...

Any claim of EIN lower than the Johnson noise of the source resistance is worthy of suspicion, so should be checked over and over until the flaw in measurement is found. I suspect the most probable cause is that the generator, being calibrated in dBm, delivers more than its expected voltage when loaded by the 600(?) ohms input impedance of the actual circuit. If the generator is set for 150r loaded, it delivers 4dB more when connected to a 600r load.(been there, done that, figured it out, learnt for life).

I do have quibs (not full quibbles). Is there a transformer? Is the "-50" input referenced to 600r, 200r, 150r, or what? What is the -50's source impedance? (If 50 ohms, then effective gain drops a small amount with the 200r resistor; if a 600r pad then gain rises on 200r source.)

Click to expand...

Exactly my idea. Makes more sense than believing in negative NF.

I _am_ reminded of some other old designs with odd action. The first BiFet low-price consoles were a 1:1 tranny into an opamp INVERTER with 470 in and 0-500K gain-pot. Maxed, it had gain of 1:1,000; but if you DIS-connected the input it was unity-gain and utterly silent. Which is not proper measurement technique but actually a useful thing in live-sound where inputs did get pulled loose. (Easier on the audience, harder on me finding why the cowbell was gone.)

Click to expand...

A very interesting twist on this concept was found on Audio Developments mixers in the early '70's. Patented by Peter Leavesley, it used a combination of voltage NFB and current NFB to define the input impedance. When the input was disconnected, current-FB took proeminence and reduced the stage gain and noise as well.

LdSpkrPro said:

John, you are correct in pointing out the error is the measurements. A noise measurement without a specified bandwidth is ridiculous. And a measurement without any bandwidth limit is misleading. Some devices have a fairly narrow 40, 50 kHz bandwidth, while others have a megacycle or more. A wide-band comparative measurement on two such different systems will result in meaningless numbers.

Click to expand...

This is why accurate measurements require precise bandpass filters.

Another more common error related to the noise of high gain stages is an apparent drop in measured noise caused by a roll off in HF  gain related to limited gain bandwidth of a given circuit. A soft top octave can make a few dB difference.

I am not suggesting that this is what happened, just sharing common types of mistakes. 

I went thru a lot of these arguments in the late 60's and 70's, and even attended a meeting of the Hollywood AES Chapter meeting where Dilley was 'sandbagged' by some of the industry 'heavyweights' who wanted to take him to task. It was sorta like a 'roast,' or we're gonna take you to town for your claims. The result was embarrassing, and Dilley had a great time.

Click to expand...

Group think can be present in any group. I found IEEE journals more professional. I have been trying for years to get the AES standards body to define Q of boost/cut EQ sections so the new generation of DSP can have consistent results when exchanging simple EQ parameter presets. Note: this all too common source of error  has been tolerated forever between sundry analog  1/3 octave graphic equalizers that despite the similarity in bandwidth nomenclature do not all boost/cut with similar bandwidth.

Here's the point that seems to get missed. Dilley knew the limitations. He also knew that the only way to do better, was to change the game. That peculiar input circuit is a game changer ... it literally improves on the theoretical. The input device noise floor is determined principally by that input resistor combined with the source impedance. For any impedance 600 ohms or lower, the input device will see a lower impedance than its input resistor, and the noise floor will be reflective of the combined lower impedance.  This is true whether the input impedance is 604 ohms, or the 2.43 k used in the 110A.  So in all cases, 600 ohms or lower, the amplifier gets quieter ... by as much as 3 to 4 dB ... and at those noise levels, that is one hell of a lot.

Click to expand...

We have a semantic problem here.. You can not improve upon the theoretical. Either the theory or the measurement is wrong. If he improved upon the theory, we need to reinspect the theory, which to my knowledge remains unchallenged.

As far as measurements go ... I know that measuring other consoles, with exactly the same termination and measuring equipment, the readings were significantly poorer. Whether you agree with these specific measurements, the same technique and equipment on two different pieces of equipment, measured in exactly the same manner produced significantly different results. That means something is different.

I don't know Buff personally, but I do know other respected industry engineers who thought he was full of it ... on this subject.... frankly I don't remember much of what he said on the subject. I'd have to go dig out old copies of the magazine.

Click to expand...

I will trust my personal judgement over your second hand here-say, and my personal experience with low noise design. There was controversy when his article published.. RE/P was not a technical journal and many people were, and still are confused about distinctions between noise power and noise voltage.

I'm sorry you took offense at my characterization ... I guess it comes from my too-dark rose colored glasses. My original intent was to contribute something of meaning to the conversation ... from an insider's viewpoint and experience. Not all that is new is better, and not all that is old is obsolete. There were serious engineering breakthroughs that are still worth examining.

Click to expand...

I am not easily offended, and am not now. I am just suggesting to have a thoughtful exchange of ideas with everybody here, lets hear more theory and facts, less ad hominum and bomb throwing.

JR

John: "We have a semantic problem here.. You can not improve upon the theoretical. Either the theory or the measurement is wrong. If he improved upon the theory, we need to reinspect the theory, which to my knowledge remains unchallenged. "

At the risk of whipping an already dead horse, let me try it again. I agree with you, I think the problem is semantic... and clearly it is my inability to find the right way to communicate what I'm trying to say.

Because of the unusual common base configuration, the input transistor sees a lower total impedance than the actual source impedance ... because its emitter load resistor is in parallel with whatever the source is.  Thus the input device sees that combined impedance, one that is lower than the source.  Since this input design gets quieter as source impedances get lower, the resultant noise floor is lower than would be expected from any given low impedance source.  Thus it produces a noise floor lower than the predicted theoretical for any given impedance source, because the input device is actually seeing a lower total source impedance.  Since this improvement is usually in the order of two to three dB, it really is a considerable improvement in an area where a dB improvement is hard fought to achieve.

I have a graph of the actual measured input noise vs source impedance, but unfortunately, I don't know how to post graphics on here.

By comparison,  the typical noise floor of a common emitter inputs, when plotting noise against impedance is a "U" shaped curve ... with the lowest noise point being considerably higher than the actual typical source impedance. So the this type of input circuit is typically operating up on the lower arm of this U-shaped curve ... and it's also why this type of circuit can become quieter with a small step-up transformer ... 1 to 3 or 1 to 5 ... which presents an more optimum impedance source to the devices.

I hope this better presents what I'm trying to communicate.

LdSpkrPro said:

John: "We have a semantic problem here.. You can not improve upon the theoretical. Either the theory or the measurement is wrong. If he improved upon the theory, we need to reinspect the theory, which to my knowledge remains unchallenged. "

At the risk of whipping an already dead horse, let me try it again. I agree with you, I think the problem is semantic... and clearly it is my inability to find the right way to communicate what I'm trying to say.

Because of the unusual common base configuration, the input transistor sees a lower total impedance than the actual source impedance ... because its emitter load resistor is in parallel with whatever the source is.  Thus the input device sees that combined impedance, one that is lower than the source.  Since this input design gets quieter as source impedances get lower, the resultant noise floor is lower than would be expected from any given low impedance source.  Thus it produces a noise floor lower than the predicted theoretical for any given impedance source, because the input device is actually seeing a lower total source impedance.  Since this improvement is usually in the order of two to three dB, it really is a considerable improvement in an area where a dB improvement is hard fought to achieve.

Click to expand...

Ok, but that topology does not present the expected bridging termination to the mic source impedance. 

The theory is still intact, the comparison is not apples to apples.

Microphone design engineers expect roughly 10x source impedance for termination impedance. A microphone mis terminated like this will behave differently with potential frequency response errors and more.  So the lower noise comes at a price. 

I have a graph of the actual measured input noise vs source impedance, but unfortunately, I don't know how to post graphics on here.

By comparison,  the typical noise floor of a common emitter inputs, when plotting noise against impedance is a "U" shaped curve ... with the lowest noise point being considerably higher than the actual typical source impedance. So the this type of input circuit is typically operating up on the lower arm of this U-shaped curve ... and it's also why this type of circuit can become quieter with a small step-up transformer ... 1 to 3 or 1 to 5 ... which presents an more optimum impedance source to the devices.

I hope this better presents what I'm trying to communicate.

Click to expand...

It may be a few dB quieter, and if it doesn't mess with the microphones too much, more power too you.

Since many active phantom powered microphones have a resistive output impedance, the lower input termination will harmlessly pad down level negating some or all of the noise benefit.

JR

John: "Ok, but that topology does not present the expected bridging termination to the mic source impedance.

The theory is still intact, the comparison is not apples to apples.

Microphone design engineers expect roughly 10x source impedance for termination impedance. A microphone mis terminated like this will behave differently with potential frequency response errors and more.  So the lower noise comes at a price.  "


That is quite true.  The 604 ohm input works ok with most dynamic mics.  The 2.43 K input works better with condensers, and it certainly provides a 10 X, or non terminating input ... while still offering the lower noise benefits of this particular configuration.