Trident Series 65 grounding: theory and practice

[leigh]

Believe me, I would like to be able to put an accurate figure on the noise floor.

The HP 334 is listed as a Distortion Analyzer. Can it also measure noise floor, or are you thinking of a different unit?

In the meantime, in lieu of having specialized hardware on hand, there should be a way to accurately measure noise using a DAW, provided your AD converter is high quality enough.

Or, I could upload WAV files captured at the -10 dBV sensitivity, if you'd like to see the test recordings for yourself.

 

[JohnRoberts]

For characterizing extremely low noise levels just adding gain in later stages will often give results dominated by the first capture stage,

A handy box I used to use was a little battery powered gain stage with something like 40 dB fixed gain and "IHF A" frequency weighting. After that 40 dB bump the noise floor of the next stage was generally inconsequential. In general noise will be dominated by the first stage of the chain.

I do not question your measurements and progress, If you see deltas between measurement that suggests you are measuring real changes.

just saying...  caveat lector  8)

JR

 

[PRR]

> a Distortion Analyzer. Can it also measure noise floor

A classic distortion meter works by nulling the fundamental and reading EVERYTHING else, calling it distortion.

So yes, it reads total garbage.

Do not feed a signal through the unit under test, it reads Total Noise. (Minus a narrow gap around the expected fundamental; if garbage is truly broadband this should not matter; tuning to 120Hz will null power-ripple and a lower reading that way points to a problem.)

It will not tell you specifics. Like that rise around 1KHz. Or if 60Hz or 120Hz dominates. For that the Spec Anal tells you more. But IMHO its numbers gan be tricky or wrong.

> battery powered gain stage .... 40 dB fixed gain

We do not need $4 boxes now that we have computers.

 

[PRR]

> HP 334 is listed as a Distortion Analyzer. Can it also measure noise floor

http://www.hpl.hp.com/hpjournal/pdfs/IssuePDFs/1966-04.pdf

GOOD article on the 334, with block diagram, and uses. Yes, you can bypass the null filter and it IS a noise-meter.

BAMA has a full manual, which for old H-P is usually comprehensive.

 

[leigh]

A classic distortion meter works by nulling the fundamental and reading EVERYTHING else, calling it distortion.

So yes, it reads total garbage.

Do not feed a signal through the unit under test, it reads Total Noise.

Makes sense – if not fed a test signal, the distortion analyzer's meter still gives you an amplitude reading. However, in looking into the HP334, I'm seeing that its lowest range only goes to 300µV (here's a big pic of its faceplate), which by my reckoning doesn't give it the range it would need for this measurement.

To wit: in order to turn my summing bus noise measurement into a signal-to-noise ratio, I would measure the quiescent output voltage of the console, and then divide that by a reference voltage (1.23V if we are using +4 dBu as our reference). And then do the math to convert that result to decibels. Right?

So, let's say our console output measures exactly 300µV output at rest, which would swing the HP334's needle to the top of the scale, when set to its most sensitive mode.

SNR would be: 0.0003V / 1.23V = 0.00024. Convert that to db, about -72dB.

At a 0.1 on the HP's meter, that would indicate a tenth the voltage, or 0.000024, aka -92dB. Trying to read the physical meter any lower than that doesn't sound helpful. So we're looking at the HP being able to reasonably look at noise levels down to about -92dBu.

(If I've made any incorrect assumptions or done bad math, please stop me here.)

Point being, recording the noise sample into a DAW through a high-quality AD converter gives you at least another 30dB of measurement range past what the HP334 can offer.

I realize some may object that this particular measurement I'm doing is not a true signal-to-noise test – because I am not actually running signal through the board at any point! I am, instead, trying to measure summing bus and summing amp noise (with all channels assigned, but their signals muted), and measure progress in improving that. So that's another reason, radardoug, why these noise levels are so low!

 

[JohnRoberts]

You guys are making me jealous.. back in the day I couldn't just use a daw to read noise down 120+ dBu...

Have you benchmarked what your baseline noise floor is using the DAW?

Perhaps with a shorted or terminated input jack plugged into the DAW?

Then with the console connected but with console output faders full down?

========
[old man rant]
Another measurement trick with old school distortion analyzers like the 334 is to use the "product output".  There will usually be an output jack carrying the signal that the distortion analyzer meter is measuring. This is typically fed to an oscilloscope to look at the distortion, but you can feed this to another measurement device  to extend the measurement range.

My budget test bench combined a marginal distortion analyser with a truly antique spectrum analyzer (Singer). While nowadays you can get a spectrum analyzer plug in for just about free. Back in the '70s I paid $700 for a very used crude spectrum analyzer with all of 50 dB dynamic range.  So the 50 dB gain of the distortion analyzer with the extra 50 dB of spectrum analyzer gave me a decent measurement bench.  The reason I share this too long story is to observe that when you stack measurements in series like that, the noise and linearity is dominated by the first stage. My cobbled together super distortion analyzer was good enough that I could measure the distortion caused by the distortion analyzer.  I literally had to zero the distortion analyzer at -10 dB to keep from adding distortion to my source.

So using your DAW and boosting the result will never let you see below the noise floor of the DAW. That was the nice thing about my little gain box, being battery powered with IHF 'A" curve, I could extend the rage of my bench for noise measurements by tens of dB.

You kids don't know how lucky you have it...  {/old man rant]

JR

 

[radardoug]

First you need to realise that the noise level of your Trident is not at -120. It is probably around -70 to -80 below +4. If you had a conventional peice of test gear such as an HP334 this is what it would tell you.
The DAW method looks very pretty, but is quite misleading.
Because it gives you noise vs frequency, the individual noise levels do not show you the summed wideband noise.
While it is usefull in terms of showing specific frequencies where there is a noise peak, it is not representative of wideband noise.
Any of the HP 33x series instruments are cheap on Ebay, and were well designed workhorses.
They will allow you to measure noise accurately to -80. Reading the high quality meter to - 20 is very easy.
Analog equipment of the ilk of the Trident will only give you -80 or so on the outputs. Remember that a tape recorder in the 70's had a s/n ratio of about 60 dB, only improved by Dolby noise reduction. So the console was better than the recorder.
If you want to get really serious, buy an Audio Precision, also cheap (relative to original cost) on Ebay. The true Rolls Royce of audio test gear.

 

[leigh]

First you need to realise that the noise level of your Trident is not at -120. It is probably around -70 to -80 below +4. If you had a conventional peice of test gear such as an HP334 this is what it would tell you.
The DAW method looks very pretty, but is quite misleading.
Because it gives you noise vs frequency, the individual noise levels do not show you the summed wideband noise.

Right, a wideband noise measurement is a different thing than looking at a spectrum analyzer. Attached is a screenshot of a noise meter, without any in-the-box boosting from DAW trim plug-ins. Like the other noise samples I've taken, it was captured with the converter's input trim at -10 dBV sensitivity. The noise meter reads -95dB.

Because this noise meter references (digital) 0dBFS, we need to convert its reading to get a measurement with an (analog) +4 dBu reference. To make this translation: a test tone at a +4dBu level, input into the converter set for -10 dBV, produces a -6dB reading on this noise meter. If I'm keeping all my numbers straight, this would mean that the wideband noise reading would actually be -101dB below +4.


This sample was taken from the main mix output of the Trident console, with all channels and groups assigned to the master bus, but all with muted/faders down, and then the master fader at maximum. So, as I alluded to previously, it is not the true "signal to noise" measurement of the whole system, because all the self-noise from the 36 channels and groups themselves is being muted! What it's really measuring is the noise of the master bus (amplifying the differences between the VE summing amp's ground reference, and the "ground" potentials at 36 different points in the console), plus the self-noise of the summing amps and output drivers.

...And this is why I have not been concerned with proclaiming the noise floor in reference to an absolute level, but instead have been focusing on improvements at each step. In practice, obviously I'm never going to mix on the board with all channels muted! Instead, it has been a way to isolate the noise from the summing bus and summing amps, and improve on those specs... whether or not this goal, in isolation from all other noise sources, is quixotic or not is another question.

Aside from looking "very pretty", I'll say that the spectrum analyzer view has been much more helpful in this process of improvement than the wideband noise meter. I can see things like "oh, that change dropped the 60Hz peak, but the 180Hz peak actually went up".

An Audio Precision meter would be an improvement for certain measurements, I agree. Maybe someday.

 

[leigh]

Have you benchmarked what your baseline noise floor is using the DAW?

Perhaps with a shorted or terminated input jack plugged into the DAW?

Then with the console connected but with console output faders full down?

Oh, good ideas there, to benchmark the noise floor of the measuring system (AD converters and cables) itself. I haven't done that yet, will grab those next time it's all powered up.

 

[radardoug]

You will not be getting - 95 or -105 dB. You will typically get -70 to -80 below +4. That is it. That's the best the Trident could do.
This with faders at nominal levels. Do you mix with all your faders down? I don't think so.
Your DAW is putting you wrong.

 

[leigh]

Ah, radardoug, did you read all of my post? I believe I already acknowledged these issues you are bringing up.

We don't have to guess at the noise level with faders up, I can measure it. I don't typically mix with all faders up either, so let's say I take a measurement with 16 faders up, assigned to the mix bus, and unmuted. Will do that next time I'm in the studio.

However, you seem to be missing this point: I haven't been leaving faders down in order to artificially improve the SNR measurement. I'm not the manufacturer of the console, I don't care about bragging rights for its specs. All I care about is improving it from its stock configuration, and learning which changes actually work, and which changes don't work, despite popular mythology to the contrary.

I have been leaving faders down in order to isolate my noise measurements to the summing bus and summing amps. Eventually, in the real world, I have to take the noise of the whole system into account. But in order to focus solely on improving the summing stage, taking the channel noise out of the equation has been helpful.

 

[radardoug]

I still say you are getting mislead by your DAW measurements. But what would I know, I have only been doing this for 40 years.

When you pull the faders down, the summing bus amps are no longer a factor. All you have in circuit is the fader buffer.
And of course very good noise figures.

At some point you need to realise that there is only so much you can do with the configuration.  You have made mods as suggested to you by various people, and you have acheived some improvement. Perhaps now you should get on with making music, which is what the mixer is really for.

I see so many situations on the forum where people with little knowledge get very anal, and think they can do better than the original designers. The original designers were not dummies, but were working with constraints. And they produced products that were very acceptable at the time.  All engineering is a compromise, backed by absolute parameters, such as how much you can swing a power supply, and the theoretical acheivable noise level.

 

[JohnRoberts]

I still say you are getting mislead by your DAW measurements. But what would I know, I have only been doing this for 40 years.

I doubt you've been making DAW measurements for 40 years, but several posters to this thread have been melting solder that long and longer.

When you pull the faders down, the summing bus amps are no longer a factor. All you have in circuit is the fader buffer.
And of course very good noise figures.

Depends on which fader. While not representative of real world performance pulling down the channel faders will help isolate master sum bus noise.  Leigh has already been advised of the limitation of that, but it is his time and his console.

At some point you need to realise that there is only so much you can do with the configuration.  You have made mods as suggested to you by various people, and you have acheived some improvement. Perhaps now you should get on with making music, which is what the mixer is really for.

sounds like a personal decision to make.

I see so many situations on the forum where people with little knowledge get very anal, and think they can do better than the original designers. The original designers were not dummies, but were working with constraints. And they produced products that were very acceptable at the time.  All engineering is a compromise, backed by absolute parameters, such as how much you can swing a power supply, and the theoretical acheivable noise level.

As a former console designer I do not mind being second guessed and he has already found enough questionable stuff to dig a little deeper. He seems to be practical about the reality of too complex modifications. 

He will lose interest soon enough when he stops finding easy improvements.

JR

 

[leigh]

When you pull the faders down, the summing bus amps are no longer a factor. All you have in circuit is the fader buffer. And of course very good noise figures.

As I've said, that measurement was done:

...with all channels and groups assigned to the master bus, but all with muted/faders down, and then the master fader at maximum.

Channel faders down. Master fader up.

So summing bus noise (36 different versions of "ground") boosted up 37x by the virtual earth summing, as well as summing amp self-noise, are indeed a factor. Were it not a factor, there would be no difference in the "before" and "after" snapshots of installing the 1/2" copper bus bar.

I completely understand and respect that the original designers were working with constraints. But as you yourself said,

Remember that a tape recorder in the 70's had a s/n ratio of about 60 dB, only improved by Dolby noise reduction. So the console was better than the recorder.

So one of their design constraints was, get the console noise lower than the tape noise is going to be (by a decent margin), and call it good. Any further effort on that front would be wasted. Jump ahead 30 years, not so. There is room for improvement, no shame to the original designers.

 

[leigh]

We don't have to guess at the noise level with faders up, I can measure it. I don't typically mix with all faders up either, so let's say I take a measurement with 16 faders up, assigned to the mix bus, and unmuted. Will do that next time I'm in the studio.

Got to grab this measurement today. This is with 16 channels assigned to the master bus, all 16 unmuted and with faders and input line trims at unity. The master fader is at maximum, as before, for easier comparison.

The wideband measurement looks to be about -88 dBFS. Again, the noise sample was captured with the converters at -10dBV sensitivity, so as before, we would subtract 6dB to convert that to a +4dBu level reference, making it -94 below +4.

I'm happy to have anyone double check these numbers. Here is the noise sample as a 24-bit WAV, captured directly into the converter at -10dBV sensitivity:
https://dl.dropboxusercontent.com/u/19586192/Trident%20noise%20w%2016%20channels%20summing.wav

 

[radardoug]

Again, the DAW measurement is not particularly usefull, because there is no reference.
I put your file into Audacity, and apply gain to it. I get 70 dB of gain to take it from where it is to digital full scale. But without knowing what the reference zero was, I can't tell how noisy it really is. I can tell you it is 70 dB based on your file.
You need to record some tone at your VU meter zero into the front of the file, and then the noise. Then a simple gain exercise gives us the signal to noise.

 

[radardoug]

ASSuming the meter is accurate then, if it reads - 6 dB for a +4 input, then the noise level is -64 ref to +4! Oops! Getting worse!

 

[ricardo]

The wideband measurement looks to be about -88 dBFS. Again, the noise sample was captured with the converters at -10dBV sensitivity, so as before, we would subtract 6dB to convert that to a +4dBu level reference, making it -94 below +4.

The -10dBV sensitivity is for FS.

This is a common sensitivity spec for 'prosumer' A/Ds.

So the wideband noise is -10dBV - 88 = -98dBV at the output of the mixer under the stated conditions.

That's what I'd expect for a competent but not exceptional mixer.

 

[audiomixer]

While measuring with the DAW has some merit, go and get (lend) a Neutrik Minilyzer XL2. Autoranging input, true dBu, dBV, dB(A), HP400 filtered numbers, FFT and 1/3 oct view, very low self noise and no conversion errors.... real numbers from your desk. you might find that reducing 60Hz harmonics does surprisingly little to the overall noise, but still a very valuable improvement. an other issue is whether you terminate your inputs with say 200 ohms or short them out or leave them open .... a other case study in itself. usually you can hear the most disturbing noise component (hiss, white noise, hum....) what's your limiting factor?

can you measure your individual channel thru a direct out? that would be interesting in relation to the bus only and 16 channel routed measurements. to check if the numbers add up correctly.

the most important thing is that you are happy with your improvements, so rock on!

- Michael
PS - I'am not affiliated with Neutrik.....

 

[JohnRoberts]

I am not opposed to investing in test equipment (I used to sell test equipment TS-1), but you can learn a lot from simple relative measurements. Better is always better,  as long as you know  what you are looking at, thus my advice to bench mark the bench noise floor etc.

I could tell stories about how many times I was fooled by "too good" measurements because I was doing something wrong.

JR