Channel Noise Floor

[synthjoe]

A brief update on this.
I have installed EMI filters and additional capacitors on the CODEC board's power rail. Have cut the long power strip and supplied filtered power to each group of 3 (D/A and A/D) chips (will post pictures later).
Unfortunately no major improvement to report. Minor improvement is that all channels have a steady -93.4 dB peak noise reading using the original PSU, which is exactly a 1 bit noise (relative to 16 bits) in center panning. I think power is now as clean as it ever could be, noise on all channels are pretty uniform (regardless of sampling rate), so I'm okay with that. However, I'm not OK with the apparent 16 bits resolution and I still think that it should be possible to bring the noise further down.

I haven't investigated yet what the readout is on the digital tape out, but something tells me that input converters are working in 16 bits mode. Anyone any thoughts on this? I seem to remember having read somewhere that the main output sometimes gives 16 bits stream and that at unity gain tape outputs are truncated to 16 bits (without the OS patch installed, but I do have it installed), but now it seems to me that the actual channel input digital data is truncated to 16 bits. If I lower the channel fader, values do go into the 24 bit resolution range on the main output - so the mix engine and SPDIF output does seem to work fine at 24 bits.

[synthjoe]

OK, here I go with a couple of pictures as promised.
First, a close-up with the details.
1. cut the original power rail to form sections of 3 A/D or D/A chips
2. added EMI filters to each block of three and feeding from the main power rail - now replaced by the 'thick' black insulated wire
3. added 470u capacitors (and an additional 100n not visible on the picture) to further filter the power supply

Close-up of modifications

290920111208a.JPG (Array KiB) Viewed 1213 times

A picture of what the board looks like with all mods done

Ready to R&R

290920111207a.JPG (Array KiB) Viewed 1213 times

[synthjoe]

a few more pics:

Cutting the power rail trace

IMG_2712a.JPG (Array KiB) Viewed 1210 times

Soldering the EMI filter on the electrolytics' terminal

IMG_2725a.JPG (Array KiB) Viewed 1210 times

Adding the new 'power rail'

IMG_2736a.JPG (Array KiB) Viewed 1210 times

A view of the whole board before installing further electrolytics and the 100n ceramics

IMG_2732a.JPG (Array KiB) Viewed 1210 times

[High C Double G]

Well, first off very impressive. A drag there is not a sizable amount of improvement after the work is done. There are 16-bit pathways and they are well documented (or so I hear...). I could not quote them from memory anyway.

This desk is mid nineties and back then it was a different world altogether. I reference Nirvana --- Kinda get my drift? - I appreciate, but come on, just the outfits were trivial and lacking common sense for show business. I guess there was a vacuum and they filled the void. I love all music so I love them too.

But to the d8b, it is an amazing product. It is just as amazing today as it was back when, amazing. People are using this product. I think because it really still works and it still sounds good. With Marc G’s software I know it will be an added benefit.

I have found my way to use the mixer that works for me and the beauty is that it is very flexible and there is no wrong or right way to make it work for your set-up. Is it old? Yep, it is… Does it have flaws? Yep it does… Can you make a good recording on it? Yep, you can.

I use it as I do, and honestly someone here turned me on to a program that measures the bit length of an audio stream. I can’t locate it on this computer; anyway the bit-length is important but is it the end of the world? No it is not the end of the world. I am working on a recording now that has a 3.5mm (is that right? – I think that is the correct number) anyway… it is a headphone stereo (yea right!) out of a cell phone, to reference some audio. The source is whatever it is; I will send it through a 24-bit path to the computer or recorder, or both actually.

The whole bit thing is strange. To say you can’t get a good sound from a 16 or 20 bit recording is hogwash, equal is true that you always get a good sound from a 24 bit recording. All things being equal (with digital) the 24-bit will sound better and allow more headroom. The same is true of cassette tape; it does not inherently suck because it is a cassette. I know I am preaching to the choir. So now I will just fade out old school with the hand on faders. Ha ha!

I just wanted to acknowledge what work you did and them is some good photos too, but sorry it was not a marked improvement over what was already there. I am impressed that you tried something like this.

Michael

[synthjoe]

Thanks, Michael! Kind words, and I'm surely not depressed by the results - hope any of the information here will be of some use to someone else.

I would not even say that I expected something very spectacular - as far as I'm concerned, I heard/saw something I did not like and a steady 6 dB improvement of the noise floor on each channel after fixing it is quite a reward for me. If for nothing else, now the channel noise filter/gate starts to become usable, the chatter on quiet passages is much less noticeable.

However, I'm surprised that despite of using 24 bit converters not the full potential was exploited. On the other hand, the noise figure I have now is practically matching the worst case noise figure of the data sheet, which is not too bad - but the datasheet OTOH looks rather bad, wasting 8 bits out of the 24 just for the noise is not a good thing. Maybe that's why Mackie designers choose to use only 16 bits in the DSP out of the full 24 resolution of the A/D chip (not confirmed yet, but I'll investigate - any reference from anyone to the 16 bit signal path documentations would be very welcome).

Am I happy with the desk as it is? You bet! I love it, the most compact (72 channels, 48 with EQ, dynamics and flexible routing on less than 1 m suqare? - that's a good deal!) and usable piece of gear I've ever owned. I fully agree with all of your comments. However, if I can improve on it by simple means, you can also bet I'll do it. If the DSP input indeed works at 16 bit, then that is a real barrier I'll not attempt to overcome. But first I'll verify whether the A/D converters really do put out 24 bits resolution and whether noise figures are any better than the worst case figures in the datasheet (90 dB is only allowed for full scale signal, at -60 dB the noise floor should be -97 to -102 dB)

[High C Double G]

Thanks Joe,
I don't know what OTOH means, and yes 6db is 6db, if it happened to go the other direction it would be +12db from where you are now so that is good. (my guess… off the original handbook?)

It is my understanding that on the the AES/SPDIF card the digital input is 20-bit (continuous and not a way to switch back and forth between 20 and 24-bit) - it is 20-bit period; this is not the card it is in the software from what I understand. The topic of those extra “throw-away” 4 bits has been talked about ad nauseam. For me I go by the sound, if it sounds good I don’t care what the bit count is. There is much speculation on how close many manufactures get to a full 24-bit equipment piece. I talked with one designer about his 24, and he said “its close, it’s really close”

I have been toying with the idea of setting a AIO-8 card in the ALT slot, but if you have a digital card (OPT-8, DIO-8-with a 24-bit chip installed, or a PDI-8) in the ALT slot and you send signal from the Aux channel, the output of any of those digital cards will be 16-bit and there is not a way to switch back and forth between 16 and 24 bits, it is 16-bits period – again not the card, it is the software. – This will happen only in the ALTernate slot and only with the signal coming from the auxiliary send from any channel. The d8b is very flexible and there is a work around for all this if it is important to get the “full” 24-bit signal path.

I have mentioned before that I rarely use the 24-bit convertors (AD or DA) in my d8b anyway. As Crash might say… they don’t suck and I agree with this statement.

This and many other talked about issues has always raised the question, why doesn’t Mackie release the SDK? This would make great sense in this day and only be good for Mackie’s rep with their customers. My opinion is that they lost it, or it got destroyed, or someone has the rights to it. It makes no sense (at this hour) not to release it. Again a topic discussed ad nauseam. This post is already too long.

I am really impressed with your passion and knowledge. And that you take this stuff on is very cool!

Michael

Okay just got it, the lights came on... DOH! "on the other hand" right?