Channel Noise Floor

[synthjoe]

Hello everyone,

New member but long time musician/tech here. Just acquired a D8B in decent condition. Having worked with this desk back in 2002 or so, I believe that the analogue noise floor of my 'newly purchased' desk is much higher than what I was used to. I'd like to ask for your help confirming some noise level readouts on other desks.

Therefore I've made some measurements with the following setup:
- open inputs, nothing connected to either of the input connectors (analogue alt, inserts and others included)
- output signal taken from the master bus SPDIF out as 24 bit 48 kHz into a stereo channel of a similarly configured ProTools session
- relevant track armed record ready and fader to unity in ProTools
- on the d8b all faders down, except the master set to unity (0 dB)
- all inputs assigned to the main L/R bus
- all channel filters, dynamics, dither and other such stuff set to off, bypass or any other relevant setting to defeat their effect

With this configuration I get an initial -oo peak indicatin in ProTools, which is good.
Slightly and slowly raising the fader on any d8b input channel results in -138.5 dB peak readout, which again sounds good, confirming 24 bit resolution and proper working of the digital mixing engine.
Raising a single channel's fader to unity results in -87.3 dB peak indication and a noticeable amount of noise on the outputs (be it main or control). I doubt that this high a noise floor is normal - could anyone please confirm their peak readout in this configuration? The noise is quite uniform on all input channels, regardless of the input gain pot setting (except the last bit of their travel, whereby the noise can be raised to about -67 dB). Noise is slightly different on even and odd ALT inputs (L and R FX returns), and their level is -90 to -92 dB respectively.

Raisiing all input faders to unity (gain pots down to minimum) the noise for the 32 inputs (24+8) totals to about -66 dB on the main output, which sounds far too bad to me (but predictable from the -87.3 dB single channel noise floor). Also, the noise gets worse when switching to 44.1 kHz mode (increased noise floor and appearance of 'whining' in each of the analogue inputs). Did anyone make the same experience? By shorting channel inputs the noise floor does not change.

Bottom line of the tests, with master fader at unity:
1 channel at unity: -87.3 dB noise floor
1 ALT analogue input at unity: -90 to -92 dB noise floor
all inputs to unity: -66 dB noise floor

I assume that all dB readings are relative to 0 dBFS (rather than 0 dBu = -15 dBFS, making all figures 15 dB worse considering the headroom above 0 dBu), although I have not yet verified this. Anyone would be able to provide some assistance as to what these readouts are on other desks and whether the values I see fall within the band of normal operation?

[Petersueco]

I remember that OSv3 had some noise in the EQ. Switch off the equalizers and see if the noise goes a way.

Also check the channel filters. It 's better they are off.

Peter Holmquist.

[synthjoe]

Hi Petersueco,

Thanks for the swift reply! Channel filter is off, and so is EQ, as I wrote. I'll double-check EQ, however, just to be on the safe side. Indeed, I have not mentioned that I have V3 OS. On the other hand, some factual figures from other users would be interesting - maybe it's just my noise floor expectation going down too much in the last 9 years, or so...

[Crash]

What does it read with the channel filter set to on? I am curious. I tried the "off" trick with the d8b in a live setting and it was a noise spewing mess with the amount of wattage I was pushing.

[synthjoe]

I think it actually goes to -oo, but has 'spikes' (clicks, like a Geiger-counter ) when noise reaches the treshold. Mine was noiseless but did this 'clicking' when all faders were at unity with channel filter on. First I thought the PSU was faulty, suspected some static discharges until I noticed that the fewer faders were lifted, the less frequent the clicks became - with no clicks at all when all faders were down. Then I remembered the channel filter, which when set to 'off' got rid of the clicks, but I got static instead - as described above.

I share Crash's concern, in a live situation it must be nearly unuseable with the noise floor I experienced and the high powers put out by the FOH. However, just in these cases the channel filters might be OK - someting that produces quite an unacceptable 'chatter' in a studio situation. However, noise seems a little too much for studio too, for me...

I'll check tomorrow anyways and report on the dB reading with filter 'on', just to make sure.

[Petersueco]

That static comes from the equalizer when it is ON.

In version 3 the equalizers where noisy. That was one of the fixes that came with OSv5.

[synthjoe]

I assume that all dB readings are relative to 0 dBFS (rather than 0 dBu = -15 dBFS, making all figures 15 dB worse considering the headroom above 0 dBu), although I have not yet verified this.

Indeed, 0 dB is 0 dBFS for the above figures. However, from the manual I understand that 0 dBFS = +20 (or +22) dBu, which results in -15 dBFS equalling +5 dBu. I wrongly assumed this to be 0 dBu. However, it does not change the fact: noise on a single channel is about -72.3 dB referred to the +5 dBu nominal level. What is your reading/figure?

@Peter: I have checked EQ (using the 'on' button in fat channel), it had no effect on the noise (flat setting, of course, otherwise it cuts or boosts noise according to the EQ setting). Could it be that one cannot remove noise added by the EQ by simply switching them off?

@Crash: channel filter behaved as expected: reads -oo as long as noise does not exceed the treshold, then it jumps to the same -87.3 dB value and you can hear the 'click' in the speaker.

Any figures for comparison would be greatly appreciated!

[anyhorizon]

With what are you measuring these noise figures? My d8b is as quiet as a church mouse with line in selected and filters off. Just curious.

Peter

[synthjoe]

Hi anyhorizon,

As described in the opening post, I have attached a ProTools input to the SPDIF ouput (24b/48k) of the desk and registered readouts based on the channel peak indicator (in PT CTRL+click on the green number field below the fader). Have done two measurements using an M-Audio transit (with coax/optical converter) and an MBOX (direct coax) as the SPDIF input, results were precisely the same.

Indeed, I also feel that the noise of my desk is higher than what I was used to (PSU problem?), that's why I would appreciate others posting some figures or at least comparing these against their own reading.

Would you be able to quantify exactly how quiet yours is when all faders are at unity, gain pots down to minimum? Or maybe a single channel's noise floor at unity gain? That would be serious help! Thanks!

*EDIT* PS: to clarify, I'm measuring only analogue inputs (line/mic 1-24 and ALT 1-8 - I have an ANALOG I/O card in there), I have ADAT cards (digital input) on tape I/O ports and they are of course dead quiet.

[synthjoe]

Just a quick update. I suspected the power supply to be the culprit, as the line was too noisy to the codec board. So I created an extra low noise shunt regulator PSU from a PC PSU (idle noise was around 2-5 mV - compared to the 200-500! mV noise of the original power rail). However, connecting the new supply to the codec board resulted in the same 200-500 mV noise in the 10 MHz range, so I investigated further. As a test, I have recapped the analogue I/O card, using tantalum caps for the digital side of the A/D and D/A chips and low ESR electrolytes for the analogue side. Result: at least 3 dB better noise figures for a start. Encouraged by the results I have recapped the codec board the same way, resulting in 6 dB improvement of the noise figure (I'm measuring -93.4 dB when using the homebrew shunt regulated PSU, versus the former -87.3 dB) and I'm far from done. The +5V PSU for sure will need some overhauling as it generates some 50 mV 'hum' in the 100 kHz range, and I'll see the other supplies (particularly the +/-16V) for other possible improvements. Oh, after recapping the codec board, I'm now measuring a steady, near -95 dB noise floor on the I/O card, which is pretty close to the textbook theoretical limit. I'm satisfied - even though it is 'only' 80 dB (-15 dBFS) s/n or -85 dB (0 dBu), which is not the best on earth, but much better than what I've started with.

Since I've now learned to attach pictures (thanks, Peter!), here's a scary one of the shunt regulator in operation.

Do not try this at home! :)

16062011792a.JPG (Array KiB) Viewed 2735 times