[jb]

It appears, based on what I see in this and other audio forums, that most DAC tweaks are aimed at the audio side of the D-to-A process. I wonder why because there are benefits to be found tweaking the digital side. For example, I made this widget to replace the CS8412 in my Audio Note DAC.



It uses a CS8415A (mounted on the underside of the PCB) configured to operate like a CS8412 in mode 6 and allows the sample data to be right-shifted 0, 1, or 2 bits. Right shifting allows you to optimize the position of the 16-bit CD sample data in the larger field of an 18-, 20-, or 24-bit R2R DAC for best sound. Right-shifting the data reduces the maximum and RMS current output by the DAC chip and hence the maximum and RMS current through the I/V. It also reduces the maximum and RMS step size thereby reducing dI/dt and settling time, all of which can be beneficial.

When I first tried it using reference quality CDs I noticed only a small improvement in clarity and dynamics. I was disappointed but I left the widget in the DAC because it offered some benefit and didn’t do any harm. However, I was most surprised recently when listening to some ‘bottom-shelf’ CDs: you know, the ones with worthwhile music but poor sound quality – harsh, strident, and edgy. Most, but not all, were literally transformed into rich, lush music. I haven’t yet determined what recording characteristics are responsible for the change but I now have hundreds more ‘top-shelf’ CDs in my collection.

Here are the possible circuit configurations. Some status signals output by the CS8412 are not provided by CS8415A and others must be synthesized or buffered if needed by the DAC’s circuitry.

[whubbard]

I don't know too much about DACs but I do know a fair bit about computers. All I can say is that a digital signal is really just a digital signal and not much is going to change with better 'qualtily' components. All the digital side need to do is send the signal and its very hard, maybe improsible, to really improve on it.

-West

[jb]

I don't know too much about DACs

That's obvious.

All I can say is that a digital signal is really just a digital signal and not much is going to change with better 'qualtily' components.

I didn't say anything about better 'quality' components. What I described is by shifting the data on the digital side you can change the dynamics of the D/A conversion process and thereby improve the sound on the analog side. DACs are not computers.

[whubbard]

I don't know too much about DACs

That's obvious.

Yep.   

Would you mind explaining how shifting the data can affect the analog output. How exactly do you change the digital signal?

-West

[jb]

Would you mind explaining how shifting the data can affect the analog output. How exactly do you change the digital signal?

If you really do know a fair bit about computers you should know that right-shifting a binary number 1 bit is the same a dividing that number by two. Right-shifting it two bits divides it by four. Dividing each and every sample by two or four changes the operating characteristics of the DAC chip, as I described. In addition, the shifted sample utilizes a different set of current switches and a different portion of the R2R network and thus changes the overall linearity. There is no loss of dynamic range: It is already limited by the 16-bit input sample size. There is no loss of resolution because R2R DAC chips with 18 or more bits have good linearity to at least 18 bits and that’s the most I use. Whether or not right-shifting one or two bits is beneficial depends on the individual characteristics of the DAC chip, the I/V, and output stage. That’s why I provided for selecting 0, 1, or 2 bits.

[whubbard]

Yes, I understand what shifting a binary number will do, but when it comes to DACs I guess this is where the 'hole' in my information is. However, from your intial post I am able to see the benefits of what that does, so thank you. I must read more carefully next time. Now if I've understood this right, your design basically 'upsamples' the music and also makes the job the DAC chip has to do, easier.

Thank You,
West

[art]

It appears, based on what I see in this and other audio forums, that most DAC tweaks are aimed at the audio side of the D-to-A process. I wonder why because there are benefits to be found tweaking the digital side. For example, I made this widget to replace the CS8412 in my Audio Note DAC.

You are applying the tweak to the wrong side of the RX chip. All of the Crystal RX chips need serious work on the other side of the circuit.

IOW, a linear circuit between it and the input cable. Trust me.

Not to say what you are doing does nothing. It may very well. Just that you can get much more bang for your buck on the input side.

(Yes, I know the data sheet doesn't show any of that. Keep in mind that they also think SC transformers are good.)

Pat

[jb]

Now if I've understood this right, your design basically 'upsamples' the music and also makes the job the DAC chip has to do, easier.

NO! There is no upsampling of any kind. The only thing that changes is the absolute magnitude of each sample.

[jb]

You are applying the tweak to the wrong side of the RX chip. All of the Crystal RX chips need serious work on the other side of the circuit.

Tweaks on the analog side of the RX chip are well known and I have already tried most of them. I chose to tweak the digital side because, as far as I know, no one has explored that domain.

Not to say what you are doing does nothing. It may very well. Just that you can get much more bang for your buck on the input side.

If you haven't heard what I have done, what makes you think an analog tweak gives more bang for the buck? Regardless, no analog side tweak does what my circuit does. So, which gives more bang for the buck, apples or oranges?

[art]

Tweaks on the analog side of the RX chip are well known and I have already tried most of them. I chose to tweak the digital side because, as far as I know, no one has explored that domain.

Does not mean that any of the ones that you have tried are actually worth diddly. They may not be.

Those require some knowledge of RF. Yours does not, so may be more beneficial to the average DIYer.

If you haven't heard what I have done, what makes you think an analog tweak gives more bang for the buck? Regardless, no analog side tweak does what my circuit does. So, which gives more bang for the buck, apples or oranges?

Well, the only way to really fix it is to use a reclocking scheme. Anything short of that is a band-aid. "Analog" fixes don't require expensive parts, just knowledge and experience.

But, like anything else in audio, if you spend money on it, it has to be better. Ergo, lots of ugly looking "analog" fixes can be found on the 'Net. Therefore, they gain some degree of popularity, even when it is not warranted.

True, I have not tried what you are doing. Likewise, you have not tried what I have. So, to be truthful, neither of us are in a position to comment on the other's work.

Sorta.


BTW, what 24-bit DAC have you seen that is a true R-2R DAC? The B-B '170x series are only kinda sorta R-2R DACs. All of the later crap is not true ladder DACs. Point being that if you looked at how truly horrible their low-level performance really is, well, maybe more people would be jacking up the level in the digital domain.

SPDIF, notwithstanding. Which needs lots of help.

Pat

[geezer]

I don't know too much about DACs

That's obvious.

Lout.

[jb]

BTW, what 24-bit DAC have you seen that is a true R-2R DAC? The B-B '170x series are only kinda sorta R-2R DACs.

I use the term R2R to differentiate from sigma-delta DACs.

Point being that if you looked at how truly horrible their low-level performance really is, well, maybe more people would be jacking up the level in the digital domain.

Who cares about low-level performance? The input samples are 16 bits and I’m using no more than 18 MSBs regardless of the DAC width. Because the LSB of every sample is essentially noise, all I need is good linearity to 16 or 17 bits. That’s not too much to ask of any 18-, 20- or 24-bit R2R DAC chip.

[NewBuyer]

jb, this is really interesting. To your knowledge, is the application you have shared here, your own original idea? Is there even a single drawback you can imagine regarding its implementation, besides limiting to ladder DACs?

[jb]

jb, this is really interesting. To your knowledge, is the application you have shared here, your own original idea? Is there even a single drawback you can imagine regarding its implementation, besides limiting to ladder DACs?

As far as I know, the idea is original. I have presented my completed widget at the Audio Note Kit forum, where I got one response, and at HeadFi, where I got no response. I think the concept was over their heads. Also, I previously discussed the idea at diyAudio and diyHiFi and was jeered by the self-appointed 'experts' because the idea was outside the mainstream and didn't fit the message they preach, which is that jitter is the root of all-evil in the digital audio world.

There is no reason the circuit could not be used with a sigma-delta DAC chip although the audible effect may be different: I haven't tried it. I choose the route I did because it allowed a simple comparison: Unplug the CS8412 and plug in the widget. To try the circuit on any of the other eight DACS I have would have required major PCB surgery. I also enjoyed the challenge of implementing it within the confines of a 28-pin DIP footprint.

There is also no reason the circuit could not be adapted to work with any other DIR. I am presently working on a scratch-built DAC that incorporates the same idea but a very different implementation. The only drawback is that for each bit shifted you lose 6dB in output level and, in the extreme, could lose linearity and resolution. I wouldn't recommend this circuit for any 16-bit DAC.

[art]

As far as I know, the idea is original. I have presented my completed widget at the Audio Note Kit forum, where I got one response, and at HeadFi, where I got no response. I think the concept was over their heads.

Yes, I would tend to agree with you. But why were you surprised?

Also, I previously discussed the idea at diyAudio and diyHiFi and was jeered by the self-appointed 'experts' because the idea was outside the mainstream and didn't fit the message they preach, which is that jitter is the root of all-evil in the digital audio world.

interesting. Somehow, I sense this thread heading a direction that I won't have time to devote to, so back to the product(s) I am supposed to be working on.

Good luck, in any case.

Pat

[jules]

jb,

congratulations on a nice bit of lateral thinking and thanks for sharing it here on the lab ... should be more of it 

jules