signal to noise in the GSSL comps.

i built two of these a while back....i've been using them both (one is for a friend) testing them and such.

i just recently got around to doing a signal to noise test (hey, i've been busy)

gain at unity the box is at -65dB
gain at full the box is at -50dB

that's AWFUL specs. these things should be down around -100.

has anyone built one that has good noise specs, or is this noise floor just part of the design and PCB layouts?

any suggestions on how to get them down to something useable?

Hi Hall,

Are you sure you know how to measure S/N ratio correctly?

We have had this question before, I think. Try a search also.

Jakob E.

[quote author="gyraf"]Hi Hall,

Are you sure you know how to measure S/N ratio correctly?
[/quote]

HAHAHAHAHAHA

yes, i'm doing it correctly.

can some one with a working unit and some test equipment please tell me what your s/n ratio is at unity?

around -80@unity with unbalanced cables connected, -85@unity with nothing connected. are you running balanced or unbalanced? if you run unbalanced, ground the cold pin on the inputs, not the outputs mind you. this will dramatically decrease noise.

[quote author="Svart"]...ground the cold pin on the inputs...[/quote]Do you mean the "-" input, pin 3?

[quote author="Svart"]around -80@unity with unbalanced cables connected, -85@unity with nothing connected. are you running balanced or unbalanced? if you run unbalanced, ground the cold pin on the inputs, not the outputs mind you. this will dramatically decrease noise.[/quote]

ok....is there anything preventing me from modding the actual input/output of the curcuit to unbalanced?

i appreciate the quick response and the help!!!!

-80 still isn't great.

it should be around -100 at unity.

yes sorry, i do mean the - input.

to run unbalanced, just ground the - "cold" input and just hook up the + and gnd for the output(don't ground the - outputs!)

Do you measure it with shorted inputs?

I get more than 85dB - which is the limit of my test equipment..

Jakob E.

[quote author="j.hall"]
-80 still isn't great.

it should be around -100 at unity.[/quote]

No offense intended, but when you throw out these numbers, do you realize what they mean in terms of percentages?

-80dB below, say, a 1VRMS signal is 0.0001VRMS. That's pretty damn good in my book. I know I can't hear noise that's 80dB below my signal.

-100dB below 1VRMS is 0.00001VRMS. That's below the resolution limit of most test equipment... that's below the level of electronic noise that exists in most places on earth.

i agree with Gyraf and NYD, my meter shows -85 but i can't hear it at all. to hear it I have to turn up the output ALL the way, and turn up the input of the following equipment ALL the way and then i can barely make out white/pink noise.

what kind of power transformer did you use?
star grounding?


what is following the outputs of the compressor? are you inputting to a soundcard or something else?

i'm using a terrasonde audio tool box to do my s/n measurements

any piece of high quality pro audio gear should idealy have an s/n of -100dB

i agree that -80 is quiet enough....but we're building a known design of a piece of gear that has been tried and true for a long time. if reverse engineering it creates a noise floor this much higher then the original unit then something isn't right.

i'm not saying jakob's design is wrong, or bad, or anything like that.....obviously these two were put together by me, using components of my choice.

i just think -65 and even -80 are a bit high, compared to current standards for high end gear.

I'm not convinced you are doing the test right (don't get upset)
BUT
:roll:
even so it is typical that a compressor may not return the very high figures of 100dB
as there are designed to reduce dynamic range

compressor threshold and gain structure and headroom are all influences of a signal to noise test
too many variables to give a sensible opinion across a forum like this

How does the same test stack up on a typical well know commercial unit you may have ... give us something to compare to ??

Which Toolbox have you got ?
... or are they technically the same and just monochrome/colour stuff going on ?

[quote author="j.hall"]any piece of high quality pro audio gear should idealy have an s/n of -100dB[/quote]

I take it that you've never used a tape machine :wink:

Seriously, though, an UNWEIGHTED S/N of 100dB is very, very difficult to achieve. Don't go by manufacturer's specs, since they'll usually weight or otherwise bias the measurement somehow to make the figure as favorable to their sales marketing as possible.

And you are measuring with the input shorted, right?

-80 still isn't great.

Click to expand...

Just for your information this is the specs on the SSL version of the SSL stereo bus compressor:

Audio Performance -
Input XLR Electronically Balanced, Impedance 20 Kohm
Output Output XLR Electronically Balanced, Impedance 40 ohm
Frequency Response 20Hz to 20 kHz ±0.2dB
Dynamic Range >110dB
Noise <-85dBu (-90dBu typical)
RMS 20Hz to 22kHz at output, input terminated with 50 ohm resistor

Regards,
Jeff

Those specs Jeff just posted raise a good point.

JHall, what bandwidth are you using for your noise measurement? If you're using a wider bandwidth than the one specified by SSL, it's no surprise that you're ending up with a worse S/N figure.

> any piece of high quality pro audio gear should idealy have an s/n of -100dB

Why stop there? Ideally we can want more.

Because we have 96dB 16-bit stuff? Well, yesteryear we had 70dB stuff and tomorrow we'll have 120dB stuff, so "100dB" is so "to-day".

> a piece of gear that has been tried and true for a long time.

Since the days when 60dB tape was king. A lot of "great" stuff from that vintage isn't better than 60dB. Notably the LA2 type limiters: input (at grid) noise level around 5 microVolts, 1:50 gain after that. So re:1V we have 20mV/5uV= 72dB noise to reference level.

You know how you like to measure "S/N" but unless you specify conditions it is hard to compare results. In advertising, "S" often means "peak clipping level" and "N" means "heavy weighting filter" because that looks good in the ads. And peak clipping makes sense on digital media. In tape we used 3%THD but whanged the peaks up 10-15dB higher, so "60dB S/N" is really more like 75dB in a digital media. Inaudible power hum can add 20dB to a weighted measurement. When you read "S/N numbers", the conditions are often more important than the number.

And a limiter has a special problem. In varu-Mu (I have not studied VCA in depth) I can get an input range over 100dB. But the whole point of a limiter is to reduce dynamic range. If I expect to take up to 30dB of limiting, then the output dynamic range has to be 70dB. Similar conundrum in FET shunt limiters: an FET will distort over 30mV, for 30dB gain reduction we need 47K series resistance, noise level is at least 3uV, output dynamic range can't be over 80dB.

Noise levels in BJT VCAs are very funny. They've been trying to sort this out for 30 years. Many designs have excess noise around unity-gain, where both sides conduct, though they may be relatively quiet at high gain or low gain. Unfortunately that is where you idle a limiter.

I've been using 2 GSSL's that I built on some mixes , and I took my second one into a friend's design shop to set the distortion trims on the DBX VCA's ( the first one used THATS pre trimmed chips). We checked
S/N on his gear, and while I have forgotten the actual figure we got, I remember that we thought it was fine.

Personally, I couldn't give a rats ass anyway, because these units sound awesome, we had a Smart box in there for the session as well, and no-one could tell the diference sonically, (to take nothing away from Al of course, his box is fantastic and somewhat more sophisticated), so it is really an academic arguement - if you don't use the GSSL because you are worried about it's S/N, you are really missing out on a seriously useable compressor.

[quote author="Kev"]I'm not convinced you are doing the test right (don't get upset)
[/quote]

believe me....i'm doing it right.

How does the same test stack up on a typical well know commercial unit you may have ... give us something to compare to ??

Click to expand...

crane song STC-8

Which Toolbox have you got ?
... or are they technically the same and just monochrome/colour stuff going on ?

Click to expand...

it's the same test functions one has a color screen and few more test functions and features the the monochrome

my company has the monochrome version. i use it practically daily.

I take it that you've never used a tape machine

Click to expand...

i use plenty of tape....

i hardly consider the inherent noise that analog tape brings to the table a fair comparison

[quote author="j.hall"]crane song STC-8[/quote]

:roll:

I meant something a little more main stream and a little more analog from one end to the other
I can't remember the designers name but I think I did exchange an email or had a conversation on a forum ... these have a few tricks inside that put them just left of centre in the world of comps

As PRR has said S/N can be tricky
the conditions of the test say more than just some figures
and comps make it even more so.

I guess you could set up a max signal through the unit, on the verge of clipping and then short the input with the threashold at min and measure noise ....
:roll:
but then there is the make up gain section

comps just don't lend themselves to one on one S/N wars.
it is there performance in use that matters and I've never found this one to be noisey.

mine is noisier than I'd like it to be, I haven't done any sort of fancy measurement but I know that I can hear white noise with the makeup gain up much at all... I don't know if I'm being too picky or if there's something not quite right with mine.