parallel DAC

MIDIbox FM
1

Hey guys,

I’ve been looking at the Midibox FM project lately and the dddac.de website came to my mind…

there, the dude takes several pieces of the same DAC chip and puts them all in parallel (he actually puts them one on the other!) to make the conversion more reliable. He uses 8 in parallel.

There’s some higher mathematics involved in this, but this does work. The same approach is used in highest-grade CD players and in some studio equipment.

So I’m wondering - would the same thing be possible with the Midibox FM?

If you get two chips, you probably have 2 DACs anyways :wink: And they’re pretty cheap to get - for example, cards with chips that are interesting to us can be bought for 1 zloty in poland (!) (that’s about 0.30 Euro)… and shipped for 5 zloty :stuck_out_tongue_winking_eye: So that makes, what, less than 2 euro for the whole thing? I think it may be worth trying it, don’t you? :slight_smile:

2

Have you already built your MIDIbox FM?

Mix all 4 outputs together, play the same sound, do you hear the difference? :wink:

Best Regards, Thorsten.

3

No TK, it’s not about hearing the difference! It’s actually about *not* hearing the difference! I believe those DACs are 12-bit, right? Due to different reasons, the actual output quality is probably somewhere in the range of 7-8 bits. This means 5 bits go lost, and we get more noise with less dynamics. They don’t entirely go lost, though - the quality of single samples, completely randomly, goes up and down between those 7 and the max 12 bits. Now a bit of probabilistics: if you repeat an experiment enough times, you’ll get more “correct” outcomes. This means: if you decode a single sample in 8 DACs, you’ll get the correct outcome (less noisy sample) with more probability. But (!) you never know which DAC it will come out of… so what you do is that you take all those signals together - some are only as good as 7 bits, some are better, and you mix all of them and get an average signal, which is still much better than those 7 bits. This way we could make the Midibox FM less noisy and more musical :slight_smile: Maybe the sound will be less “thin”. Just compare samples of a single DX7, and the same sample played back in Native Instruments FM7 - the synthesis is *completely* the same, the only thing that differs is the 12-bit DAC of the DX7 versus the 24-bit DAC of your sound card… :slight_smile:

4

You are writing this in a form where it could be assumed that you are an expert in this area - the YAC is not a common linear DAC, it’s provides a 10-bit mantissa and 7-step exponent (best for high dynamics), the analog output range is 16-bit.

Can you please determine the exact benefit with this type of DACs? what are the experimental results? Did you notice any disturbing parameters during your experiments which haven’t been mentioned at the website from where you copied this text?

Best Regards, Thorsten.

5

You are writing this in a form where it could be assumed that you are an expert in this area

I’m sorry - I never wanted to sound smart-ass. I just wanted to mark which possibilities could rise from such an application.

Can you please determine the exact benefit with this type of DACs?

Since you put this that way, I do not know what the benefits would be with those DACs. However, no DACs are perfect, and this makes them more “like they should be” - this can especially work with the YAC if it’s a simple circuit. I’d imagine the highest harmonics (thus, the ones with least energy - which means they’re “kept” in the less significant bits of the DAC than the “loudest” ones) of sounds would get less distorted in loud moments. I figure that’s one of the negative points of such DAC setups: good for dynamics, not as good for background detail. They’re much more suited for single instruments, as single sounds don’t have much background detail. I can’t tell - but I’m guessing such a DAC chip as the YAC used for completed music would result in mutilating it severely. You’d just have to try this, though. That’s just my guess.

Did you notice any disturbing parameters during your experiments which haven’t been mentioned at the website from where you copied this text?

I remember comparing my Yamaha DX11 with its copy, the FM7. The FM7 sounds better..

I did not copy this text from that website, I haven’t been there for at least half a year, until today to bring up some data. I’ve learnt this stuff from being an assistant in a recording studio, and from working in a high-end audio equipment shop. I still remember doing the math pen&paper style, but I don’t remember it all clearly anymore, this was like 2 years ago  :-X . I believe with a linear DAC, the bit quality rose as log_2 (sqrt(x)) where x is the amount of paralleled DACs. I remember the website has an error in the explanations, where they don’t include the logarithm in the formula. Oh well.

http://dddac.de/ma_dac21.htm#intro

Check it out yourself.

Besides, it doesn’t cost anything to test it, so if someone wants to, they can try. Not much work with soldering another chip on top of the old one, either. I’m just posting this for others to try - they (you too!) don’t have to, if they don’t want to.

Let’s not fight over this as if it was really that involving or hard to try out, or that important and crucial for the project :slight_smile:

cheers TK