Do all DACs sound more or less the same?

[Jud]

Hi Mike,

I am an engineer and DAC designer. I don't have any DACs of my design for sale right now (yet), one is coming up soon.

 

My philosophy as to what makes a really good DAC is very different than most others. For me the most important part to get right is the interpolation filters. Every one built into a DAC chip has some serious audible flaws. The mathematics of what needs to be done IS well known, but no chip maker on the planet actually implements that! Doing it right is very compute intensive and would make the DAC chips cost too much, (from the chip makers standpoint every penny makes a big difference). So ALL of them actually implement a complex filter that is supposed to do the same thing, but do it with a lot less horsepower. It is a cost driven compromise. Now the mathematicians that come up with these filters will tell you that you can't possibly hear the difference, but you CAN. For me, the single most important part of a DAC design is bypassing these filters and using a filter that at least doesn't have these complex filters. This can be done in the DAC with an FPGA, a DSP chip or can be done outside the DAC in software and feed the DAC with an already filtered high sample rate signal.

 

Of course there is more too it like keeping jitter very low, keeping noise from the digital parts from getting into the analog parts, very low noise power supplies to feed things etc. But if you don't change the filter, all these other things will not get you anywhere close to what can be had. Once you get the filter right you can really hear the differences these changes make.

 

There are other adjustments to the filter that can be made such as slope, the exact amount of pre-post ringing etc. BUT if you make those changes to one of these compromised filters, you won't hear much difference. But once you get to a simple filter and THEN try the tweaks they make a big difference.

 

There are a few companies that believe the same thing and bypass the chip filter, or use one of the few chips without a filter, but almost all of these are expensive DACs, (some VERY expensive). You just don't see this in inexpensive DACs.

 

Doing it right DOES take horsepower, and if you want to put that in the DAC it does add a fair amount of cost to the project.

 

John S.

 

John, the things you are concerned with in this comment (and others) remind me a great deal of what PeterSt writes about with regard to his software and DAC.

[pagan]

John, the things you are concerned with in this comment (and others) remind me a great deal of what PeterSt writes about with regard to his software and DAC.

 

 

An interesting thought, hardware compensated by software..

 

Makes you wonder, what is being compensated?

[Jud]

Hi pagan, it's actually better software in a PC or FPGA substituting for software programmed into the DAC chip (in the chip it's often called "firmware," but it's just programming code, not anything different in kind from the programs that run on computers).

[pagan]

Hi pagan, it's actually better software in a PC or FPGA substituting for software programmed into the DAC chip (in the chip it's often called "firmware," but it's just programming code, not anything different in kind from the programs that run on computers).

 

Hi Jud

That's the filter coding in the dac's you mean?

 

but not all Dac's have filter or other bypassing.

And by "Dac's" i do mean just the dac chip.

[4est]

What's your point?

Hi Jud

That's the filter coding in the dac's you mean?

 

but not all Dac's have filter or other bypassing.

And by "Dac's" i do mean just the dac chip.

[mayhem13]

 

 

 

And strangely enough the bass impact seems much increased, when there is no measurable change in the bass frequency response.

 

I have no clue as to exactly what aspects of the waveform are actually responsible for these changes in perception.

 

I'm sure that what is being perceived are changes in the pattern matching results due to small changes in the "tags" that the pattern matching system is latching onto.

 

Everything I have found out over the last 10 years or so seems to point to it being related to transient events, which are just what these changes in the filters primarily affect.

 

John S.

 

I commend you for your courage in this post.

 

Good luck with your DAC venture. In this recently extremely competitive market, a DAC that does something fundamentally different is what's needed IMO to stand apart from all others...........provided it sounds good! LOL.

 

With you being involved in the commercial side of the industry, although off topic do like to share with you a consumer's need....with me being selfishly portrayed as the consumer......

 

I'd like to see a stand alone network music player/DAC with cross platform controllability from a tablet. I'd like as many of the GUI features that can be found in JRiver and the like. I'd also like on board advanced DSP with the unit packaged with its own microphone and fully adjustable via the tablet as well as graphic room measurements so that the end user can see the changes to verify that manual adjustments were successful. I'd also like the unit to have on board storage capabilities. Do you think that such a unit could be competitive in the audiophile market?

[pagan]

What's your point?

 

Not all dac chips have the ability to turn off(defeat) the hardware filters in them.

 

So, If the software is still making a difference, (on a chip of this type), then it might be another issue yet unknown, or thought insignificant.

[Paul R]

I commend you for your courage in this post.

 

Good luck with your DAC venture. In this recently extremely competitive market, a DAC that does something fundamentally different is what's needed IMO to stand apart from all others...........provided it sounds good! LOL.

 

With you being involved in the commercial side of the industry, although off topic do like to share with you a consumer's need....with me being selfishly portrayed as the consumer......

 

I'd like to see a stand alone network music player/DAC with cross platform controllability from a tablet. I'd like as many of the GUI features that can be found in JRiver and the like. I'd also like on board advanced DSP with the unit packaged with its own microphone and fully adjustable via the tablet as well as graphic room measurements so that the end user can see the changes to verify that manual adjustments were successful. I'd also like the unit to have on board storage capabilities. Do you think that such a unit could be competitive in the audiophile market?

 

Sounds an awful lot like most AVRs these days. Have you looked into that market? Seriously, it has taken some leaps and bounds in the last five years or so. I am not entirely if all those travels have been forward though.

[Jud]

Not all dac chips have the ability to turn off(defeat) the hardware filters in them.

 

So, If the software is still making a difference, (on a chip of this type), then it might be another issue yet unknown, or thought insignificant.

 

Hi pagan - If the internal DAC chip's filtering isn't bypassed completely, then the software filtering will interact with the internal chip filter. I know the outcome of the interaction depends on filter characteristics, but other than that I couldn't tell you what characteristics determine whether we hear the effects of both, or primarily one or the other.

[Superdad]

Not all dac chips have the ability to turn off(defeat) the hardware filters in them.

 

So, If the software is still making a difference, (on a chip of this type), then it might be another issue yet unknown, or thought insignificant.

 

Generally, if you upsample in s/w to 352.8/384 (assuming the USB or S/PDIF input of your DAC will accept that), then you will essentially be bypassing/turning off the 8x oversampling/filter of most sigma-delta DAC chips. Even 176.4/192 really minimizes the chip's resourced compromised filter's effects. (Take a look at the filter graphs of some popular DAC chips at different input rates and you will see.)

 

BTW, John's post discussing recent audible filter parameter comparison/optimization was commentary on the weekend he spent at my house at the beginning of January. He brought pre-production prototypes of a number of DACs he is finalizing for other clients, one of which is the forthcoming Bottlehead DAC. Among other pursuits, we spent a bunch of time dialing in (yes, by ear, in very fine increments) the filter parameters for that DAC, which he then wrote down and will load into the FPGA of the final product. I wrote a long report about our session on the Bottlehead forum. That forum is down for maintenance at the moment, but when it comes back up I will copy and paste my text from that report back here. I think some will find it interesting.

 

--Alex C.

[Savage]

Generally, if you upsample in s/w to 352.8/384 (assuming the USB or S/PDIF input of your DAC will accept that), then you will essentially be bypassing/turning off the 8x oversampling/filter of most sigma-delta DAC chips. Even 176.4/192 really minimizes the chip's resourced compromised filter's effects. (Take a look at the filter graphs of some popular DAC chips at different input rates and you will see.)

 

BTW, John's post discussing recent audible filter parameter comparison/optimization was commentary on the weekend he spent at my house at the beginning of January. He brought pre-production prototypes of a number of DACs he is finalizing for other clients, one of which is the forthcoming Bottlehead DAC. Among other pursuits, we spent a bunch of time dialing in (yes, by ear, in very fine increments) the filter parameters for that DAC, which he then wrote down and will load into the FPGA of the final product. I wrote a long report about our session on the Bottlehead forum. That forum is down for maintenance at the moment, but when it comes back up I will copy and paste my text from that report back here. I think some will find it interesting.

 

--Alex C.

 

Look forward to it.

 

What is Bottlehead doing with their dac? There would have to be at least one set of tubes in there for consistency if nothing else.

[Superdad]

What is Bottlehead doing with their dac? There would have to be at least one set of tubes in there for consistency if nothing else.

 

That'a what I thought when John first started discussing that project with me about a year ago. But the darn thing has turned out so good that Doc (he's the Bottlehead in Chief) decided to skip the tubes. The main board is costly to produce anyhow. And I think Doc decide upon a battery supply.

[Jud]

Generally, if you upsample in s/w to 352.8/384 (assuming the USB or S/PDIF input of your DAC will accept that), then you will essentially be bypassing/turning off the 8x oversampling/filter of most sigma-delta DAC chips.

 

--Alex C.

 

The 8x oversampling filter, yes. The sigma-delta modulator, no. Would be interesting to know more about the specific effect of that part of the internal processing in most DACs (since most DAC chips have this modulator as part of the conversion process following the oversampling filter).

[Superdad]

The 8x oversampling filter, yes. The sigma-delta modulator, no. Would be interesting to know more about the specific effect of that part of the internal processing in most DACs (since most DAC chips have this modulator as part of the conversion process following the oversampling filter).

 

Well of course. The only way to get rid of the SDM is to use an R2R DAC chip. LOL!

 

But seriously--and I am generalizing here--once you are feeding high rates into the chip, the difference between most DAC chips is how they do the SDM. How many bits wide (6 bit typ. IIRC), how many levels, what order, what sort of loops, what other fancy tricks. Those are the things that make an ESS DAC chip sound different from a TI, a Wolfson from a Cirrus, an AKM from a Niigata etc.

 

So even if you manage to bypass the digital filters of all those chips (for some of those 384KHz is not hight enough to do so), there are still plenty of differences between them. And that does not even consider what if anything they have on their output side. Some include internal line driver opamps, some perform best into different analog output stage topologies, etc.

 

It's all enough to make a DAC designer want to skip off-the-shelf DAC chips altogether and do something unique. Funny, I happen to know a fellow capable of just that (wink). But we also know that people like Miska (HQPlayer) and Mallison (ESS) have a put a lot of sweat and expertise into their sigma-delta modulators. We just need to do more research and prototyping to determine how important the SDM is to SQ at very high rates when everything else in the DAC is done right. John's feeling has to-date been that getting the SRC and filters right is the largest part of the battle. Based on how insanely great we were able to make a cheap TI PCM5102/5142 sound last month in my system (as used in the Bottlehead DAC with very good clocking/isolation/PS, etc.), I have to agree with him. Still, my ambition spurs us to go much further…

(and no, I do NOT mean going back to R2R--even though my current reference is an NOS PCM1704K at 176.4KHz)

[gmgraves]

It was a joke, in reference to all the bogus advertising referred to in the cable thread.

 

[Anything that requires the Schrödinger equation (or the Dirac equation, etc) to describe accurately.]

 

 

Does this have anything to do with a dead cat?

[JohnSwenson]

I commend you for your courage in this post.

 

Good luck with your DAC venture. In this recently extremely competitive market, a DAC that does something fundamentally different is what's needed IMO to stand apart from all others...........provided it sounds good! LOL.

 

With you being involved in the commercial side of the industry, although off topic do like to share with you a consumer's need....with me being selfishly portrayed as the consumer......

 

I'd like to see a stand alone network music player/DAC with cross platform controllability from a tablet. I'd like as many of the GUI features that can be found in JRiver and the like. I'd also like on board advanced DSP with the unit packaged with its own microphone and fully adjustable via the tablet as well as graphic room measurements so that the end user can see the changes to verify that manual adjustments were successful. I'd also like the unit to have on board storage capabilities. Do you think that such a unit could be competitive in the audiophile market?

 

Well I am the hardware designer for the Community Squeeze project. We are taking over the reigns of the Squeezebox system. We will be offering a "player" which is based upon a dual or quad core iMX6 processor system, with my DAC design built in. It is just a box, but it does have an HDMI jack and lots of USB jacks. You can use any of the SqueezeBox controller apps that run on just about anything, to control it. You can use it on a network to talk to an external server, or run the server on the player with either local storage (USB drive or SATA drive) or from a NAS on the network.

 

The player CPU is plenty powerful to run quite a bit of DSP. Currently there is a bruteFIR plugin and SoX comes with it. There are several others that have been made to work with it. I'm currently very keen on getting Dirac to work with it.

 

There is no mic preamp in the box, but it does have a lot of USB jacks so a USB interface mic preamp/ADC will work.

 

It's not exactly what you want, but it is starting to get there.

 

John S.

[JohnSwenson]

Not all dac chips have the ability to turn off(defeat) the hardware filters in them.

 

So, If the software is still making a difference, (on a chip of this type), then it might be another issue yet unknown, or thought insignificant.

 

Many chips will use a simpler more benign filter for the higher sample rates. Thus doing an external filter CAN still help by doing the first 4x or 8X upsampling with a better sounding filter than what the chip would have done on it's own.

 

The problem here is that the filter parameters that sound the best when completely bypassing the internal filter are usually NOT the best when partially upsampling. For best results you need to tweak things for the combination you are using.

 

John S.

[mayhem13]

Well I am the hardware designer for the Community Squeeze project. We are taking over the reigns of the Squeezebox system. We will be offering a "player" which is based upon a dual or quad core iMX6 processor system, with my DAC design built in. It is just a box, but it does have an HDMI jack and lots of USB jacks. You can use any of the SqueezeBox controller apps that run on just about anything, to control it. You can use it on a network to talk to an external server, or run the server on the player with either local storage (USB drive or SATA drive) or from a NAS on the network.

 

The player CPU is plenty powerful to run quite a bit of DSP. Currently there is a bruteFIR plugin and SoX comes with it. There are several others that have been made to work with it. I'm currently very keen on getting Dirac to work with it.

 

There is no mic preamp in the box, but it does have a lot of USB jacks so a USB interface mic preamp/ADC will work.

 

It's not exactly what you want, but it is starting to get there.

 

John S.

 

Very encouraging for everything in the digital domain to be done in one chassis. Expand ability certainly makes things more attractive. Upgrade modules might be the answer as that gives customers options at different price points.

[gmgraves]

Hi Jud

That's the filter coding in the dac's you mean?

 

but not all Dac's have filter or other bypassing.

And by "Dac's" i do mean just the dac chip.

 

Can I suggest a posting convention here? In order to avoid confusion, how about we call the "box" with it's controls, and I/O a 'DAC' and call the chip a 'D/A'?

[mayhem13]

Sounds an awful lot like most AVRs these days. Have you looked into that market? Seriously, it has taken some leaps and bounds in the last five years or so. I am not entirely if all those travels have been forward though.

 

The area where AVRs is lacking for me is local storage capabilities and a nice jukebox GUI. DSP is a generic add on too.......they could do such a better job here but that might put the price out of reach for many considering the all-in-one market.

 

....but I suspect we'll see some of what I've mentioned soon enough.....computers/tablets and television are merging more and more each year.

[Superdad]

….I wrote a long report about our session on the Bottlehead forum. That forum is down for maintenance at the moment, but when it comes back up I will copy and paste my text from that report back here. I think some will find it interesting.

 

BH forum is back up, so as promised, and since Savage here said he was interested:

 

Posted on the Bottlehead forum on January 27th (Taking orders )

---------------------------

Hello everyone:

My first post here. My name is Alex Crespi, and I am the friend that John Swenson spoke of above. John and I have known each other for about 9 years. I first contacted him for an ahead-of-its-time digital project back in 2005--when Hovland Company, of which I was a co-founder, was still in operation. In fact the DAC which the Bottlehead prototype went up against on Janaury 4th is one of the NOS PCM1704K prototypes with discrete output stage that I inherited when Hovland closed its doors. It has had work done to it since: John put in a WaveIO async-USB>I2S board and better power supplies for the digital side. (If you want to know more about me or my system, I use the same Superdad handle over at ComputerAudiophile.com)

 

My custom room and system are pretty advanced, and I upsample Redbook to 176.4KHz with Audirvana Plus on a fine-tuned Mac mini booting a very slim OS off an SD card. I have spent a lot of time tweaking all the filter parameters of the iZotope SRC engine that A+ uses, so I have a good ear for what the parameters do. John's visit to my place this month (I'm about a 3 hour drive each way) was not his first, and we have spent time tuning a filter for a PCM5142 (the chip in the forthcoming BH DAC) before. It is amazing what a custom digital filter can do for an otherwise ordinary sounding DAC chip. The filters in most all other DACs are terribly compromised--especially if they are only the ones within the chip as the resources are terribly constrained. But the FPGA designed into this unit you have all been waiting for has plenty of room to load and run a really good filter.

 

At first we listened to the BH DAC fed by my Mac mini (both with iZotope upsampling and with that off to listen to the filter John already had in the BH FPGA). Power was provided by a prototype of a new PS that John and I will be producing in the next few months. The BH sounded pretty good.

 

John has techniques to output a list of filter coefficients from the excellent SoX sample rater converter--and to then load those into the FPGA. So what we did then was run SlimServer and SqueezePlayer on a Wandboard under COS (a customized Linux that the folks involved in the Community Squeeze Project have created--John is chief hardware designer for that non-profit project), and used the SoX plug-in to upsample to 352.8/384KHz (that rate turns off the PCM5142's own filter, and John turned off his FPGA filter). The filter already in the FPGA was based on some of the intermediate SoX parameters, but was not critically tuned to this DAC. The parameters are: cut-off, pre-ring/post-ring balance (that's the range of minimum-phase to linear-phase), filter length (that's number of virtual "taps"), steepness (for SoX this is controlled by frequency of cut-off and final cut, if I recall correctly).

 

Anyway, once we deciphered the numbers and ranges to enter the parameters (on an ugly command line for the SoX plug-in), we were able to spend about 90 minutes with the same four VERY revealing tracks (real instruments, real spaces, very challenging material top to bottom). We went one parameter at a time, bracketing wide, then narrow, until we were making the smallest possible adjustments. It is an iterative process (meaning we sometimes came back to one parameter after tuning another) and somewhat tedious, but John's ears, my system, and my 40 years as an audiophile/music lover (I started when I was 12) made it pretty easy.

 

In the end we were both grinning like fools. Really that DAC chip (and its built-in opamp output stage) had no right to sound as good as it did. Cymbals, piano, voices, strings, bass, drums--all top of the mark. I have heard a lot of DACs, and I am convinced this Bottlehead DAC will go toe-to-toe with some mega-buck units! Of course John quickly wrote down the numbers for the magic parameters we had settled upon and he's converting them for loading onto the FPGA.

 

So Doc, I think you'd best mark me down for a Bottlehead DAC too! (I don't mind coming after everyone else here who have waited so patiently.)

 

Anyway, I just though you might all enjoy a little insight into the final stage of what is going to be a GREAT and very musical DAC. It would be a steal even at twice the price of whatever Doc ends up asking for it. Based on what I heard with my own ears, I don't think ANYBODY will be disappointed.

 

Best regards,

 

ALEX

------------------------------

<<END OF BH POST>>

[Savage]

Thanks for posting Alex. I'll wander over to BH and poke around some more.

[mp9]

Thanks for posting Alex. I'll wander over to BH and poke around some more.

Long time since your post. As of 9/20/14 Doc is saying something should be posted in the next couple of weeks...

Can't help but wonder how the BH dac will compare to the similarly priced ES9018 USB DAC - Hi-Resolution System* HQ reference DAC ES9018S-USB PCM DAC version 2 group buy DAC-END R (ES9018) full assembled board - version 2 - group 2 - diyAudio