[hagtech]

Building up proto boards.  Got CHIME mostly built, HAGDAC still blank.  Here's a shot of the proto HAGUSB.



Plugs onto CHIME motherboard.  Generic enough to be put into existing dacs or maybe even a cd player.  Powered by USB, so you just connect the 75 ohm S/PDIF output.

Now to see if it actually does something...

jh

[orpheus]

Hi Jim-
     Wow, you work quickly.  Thank you for sharing your process here, it is very interesting.  I know that you have stated that your dac will upsample, and I think it would be interesting to compare the upsampled to the non upsampled output.  

I have had some difficulties understanding the real benefits of upsampling, and in my own limited experience with upsampling, I question it's transparency and benefits.  I have an admittedly low end player, a Phillips 963, which offers upsampling to 24/192 from 44.1/16.  The upsampling is switchable on the fly, which is great for comparison.  I've found that there is some difficult to describe quality that is lost when the upsampling is turned on, the easiest way to put it is that the timing is off with the music.  The notes don't seem to fall in the correct place, as if the attack has been smeared in some way.  The timbre changes slightly as well, in a way that I don't think is better.  There is more air in the sound, but I'm not convinced that the air in the sound isn't a euphonic distortion that actually cuts down on the clarity of the presentation.  It also could be making the best of a poor situation, a solution to the problems that 44.1/16 has had from it's beginning.


I know that all the major digital players are upsampling now, and that it is the defacto standard for digital reprodution.  At the stereophile show over the weekend, most of the players upsampled, often without mentioning that fact.  So, upsampling is so common place now that it is probably being used even if it isn't mentioned.  Upsampling interferes with the simpler is better axiom, although I suppose that digital reproduction in and of itself interferes with that axiom.

I'm wondering if you would be willing to listen to the dac with upsampling and without upsampling before you made a final decision on the design.  It is possible that the upsampling improves the sound, but it is also possible that it changes the sound, and better or worse is a subjective concern.  I realize that upsampling plays a role in the filter as well as jitter reduction, so it might be so built into the design that it can't be switched off, but I think that listening to it without might yield a surprising result, and if it sounds better with upsampling, then all the better.

Thank you for taking the time to share your progress with us.
-Aaron.

[BradJudy]

Building up proto boards.  Got CHIME mostly built, HAGDAC still blank.  Here's a shot of the proto HAGUSB.
...

Nice work Jim.  I received my boards for building something like this myself, but the pads for the IC really make me squint.  This one is going to be a definite challenge.  

[hagtech]

Brad, here's the secret:  Solder all of the pins at the same time!  Yup, glob it on.  Then heat up again and use a solder sucker (preferred) or wick to remove all of the excess.  What is left is a tiny amount of solder stuck between the bottom of the IC legs and the board pads.  It's amazing the chips you can solder this way.  QFN packages 5mm on a side with 32 pads, no problem.  Finally, touch up each pin again with a quick reheat and then swipe the iron away from IC.  This cleans up any residual pin shorts and leaves it looking real nice.  

Aaron, will have to wait until tomorrow to answer you...

jh

[hagtech]

orpheus,

There are some good reasons I choose not to use a non-OS architecture.  First off is the lack of filtering.  You can try the brick-wall approach, which we already know can be super irritating.  Or go sans filter.  But you aren't really unfiltered.  Instead, you send tons of ultrasonic garbage down the pipe to the rest of the system.  Now your cables, linestage, amplifiers, and speakers have to do this filtering for you.  Talk about system dependency!  All of this hash can really mess up some amps.  Personally, I would never count on an unknown piece of equipment to fix a problem my DAC generated.

The number one thing oversampling accomplished is the ability to simplify the filtering requirements.  Shifting the alias band farther out allows a nice phase linear analog filter.  I chose a 3-pole Bessel for perfect transient response, and the proper loading the converter's current output wants to see.  

The second benefit of OS is the extra bits you can calculate.  The filter has the ability to better approximate the sinc function, which is what is needed for ideal reconstruction.  It may not be obvious, but you can formulate a better guess with a filter than no guess at all.  I like to look at the -90dB 1kHz sinewave.  At 16 bits you get a choppy stairstep of three levels.  Upsampling to 24 bits gives you what actually appears as an almost sinewave.  Now, do we actually know what the true input was?  No.  Perhaps it was a choppy almost squarewave.  But I doubt it.  With music and real world analog signals, the sinewave guess is better.  Think of it this way, it is better to omit, than to add.  Less is more.  Should the harmonics be added when we are not sure they should be there?  I don't think so.  The difference is subtle, found in the micro-details and low-level information.  I am aiming for a very natural and relaxed presentation.  For that, all filtering is phase linear for optimum transient time domain response.  No added harmonics, no grit, no hash.

Doing valid A/B comparisons is not easy.  It is not really a simple change to switch out a digital filter.  Lots of things change.  Clocks, power supply loading, filtering, etc.  In fact, the effects of jitter might even be more prominent.  You may indeed hear more detail and more attack with a non-OS.  But I believe it is artificial.  Sort of like the sharpness control on your TV.  Some may prefer this.  

Then there is the simpler is better axiom.  I just don't agree with this.  Instead, I believe in elegance.  But simplest?  Then we should be driving one cylinder cars.  My phonostage should use only one tube.  There are many reasons I use three tube stages per channel in my Cornet phonostage.  Real reasons that make it way better than if I had used one tube (which I could have done).  These are all based on personal design choices, which makes your choice of designer important.  I guess that makes me a non-extremist.  Simplest is not best, yet I abhor complexity.  I despise the brute force methods of some (just throw money and/or hardware at it).  For me, it's all about elegance, synergy, and value.  

jh

[hagtech]

Oh yeah, I got HAGDAC partially up and running yesterday.  Got this 1kHz test squarewave from the Stereophile test CD3.



A good start.  PLL temporarily bypassed.

jh [/img]

[doug s.]

...Then there is the simpler is better axiom. I just don't agree with this. Instead, I believe in elegance. But simplest? Then we should be driving one cylinder cars. My phonostage should use only one tube. There are many reasons I use three tube stages per channel in my Cornet phonostage. Real reasons that make it way better than if I had used one tube (which I could have done). These are all based on personal design choices, which makes your choice of designer important. I guess that makes me a non-extremist. Simplest is not best, yet I abhor complexity. I despise the brute force methods of some (just throw money and/or hardware at it). For me, it's all about elegance, synergy, and value. ...

jim,

i like your philosophy re: "simple".  i like what al einstein said about this; seems yure in the same camp:

"make everything as simple as possible, but not simpler."

this is also analogous to colin chapman's philosophy for designing racecars.  his is a bit more extreme, tho!    

"...the car should fall in apart when passing the finish line. if it does earlier, it's to weak. and if it does after the finish line, it is to heavily built..."

regards,

doug s.

[orpheus]

Jim-
    Thank you for taking the time to answer my questions.  I appreciate your candor and your willingness to share your experience and philosophy of design.  I now better understand the role that up/oversampling plays in your design and why you have chosen to implement it over available alternatives.

I'm looking forward to hearing more updates as you get closer to your final design.

-Aaron.

[Julien43]

That is a very nice looking trace. Am I correct in assuming that it is a 16 bit trace?

[hagtech]

Yes, 16 bits redbook cd.  Got it cleaned up a bit since that photo.  The 20kHz sinewave is beautiful, too.

jh

[hagtech]

CHIME motherboard built.  Thought you might like to see it.  Sort of a cross between a Cornet2 and Clarinet, with the extra +/-8V regulated supplies added.



Front panel has two LEDs to indicate signal lock, plus a polarity switch.  And, of course, the input SELECT and VOLUME controls.

Progress is good, except on HAGUSB.  Can get Windows XP to recognize it.  Probably something real simple.  Just not obvious.

jh

[Todd Krieger]

I was compelled to "chime" in...

Hi Jim-
     Wow, you work quickly.  Thank you for sharing your process here, it is very interesting.  I know that you have stated that your dac will upsample, and I think it would be interesting to compare the upsampled to the non upsampled output.  

I was hoping the design would *not* have asynchronous sample-rate conversion...

In my opinion, asynchronous sample rate conversion (ASRC, aka upsampling) is the single worst trend I've seen in audio since the vinyl LP being phased out.  Aside from the processor running at 96 or 192 kHz, there is nothing, either sonically or in theory, which would improve the performance.  In fact, I think ASRC degrades performance.  For I've never heard a DAC or CD player with ASRC that I personally liked.  They all seem to introduce noise into the audio band and produce "synthesized" overtones on various instruments.  (Feel free to check out my opinions of asynchronous upsampling on Audio Asylum.)

I read in the Blog the note regarding the "slow rolloff" mode of the DF1704 used in conjunction with ASRC smears the time response.  I think the ideal design would be the DF1704 used in "slow rolloff" mode **without** ASRC.  Or in other words, I think removing the ASRC would be the best option, not reverting to "sharp rolloff" mode.

I have had some difficulties understanding the real benefits of upsampling, and in my own limited experience with upsampling, I question it's transparency and benefits.  I have an admittedly low end player, a Phillips 963, which offers upsampling to 24/192 ...

I realize the audio press has hyped what I would call a bogus technology, and there are a plethora of products out there that utilize it, with varying success.  The Benchmark DAC1 being the most-successful.  But for me, the ASRC kills it.  I had a DAC1 and sold it.

I did find a DAC that utilizes the DF1704 without ASRC, the Chinese-made Lite Audio 38.  While its sound confirms my thinking about the subject (I prefer it over the DAC1), it would be nice to see a good domestic tubed design.  The square wave response should be even better than the one posted earlier.  And I think such a DAC would have a real performance edge over the competition.

If the "slow rolloff" mode without ASRC still smears response, there must be something else wrong....  The "slow rolloff" mode should exhibit superior time response.

[hagtech]

Sorry for the confusion, HAGDAC won't be asynchronous.  My mistake for using the term "24/96" or similar.  Everything synchronous, the word clock is 352.8kHz, or exactly 8x the 44.1k input.  So it is upsampled, but not asynchronous.  I agree with Todd, anything else has unwanted side effects.  I suppose a harmonic upsample (non-integer) would be ok, but definately not two unrelated clocks.

My "reclocking" is nothing more than a complete regeneration of the exact same bit clock.  Same frequency and phase, just less jitter.  I am running the bit clock at 11.28MHz.  That's about 1/2 as fast as the PCM1704 converters can handle.  

Again, the main features of the (synchronous) upsampling are that it adds fewer error harmonics and makes the analog filter section simpler.

jh

[Todd Krieger]

Sorry for the confusion, HAGDAC won't be asynchronous.  My mistake for using the term "24/96" or similar.  Everything synchronous, the word clock is 352.8kHz, or exactly 8x the 44.1k input.  So it is upsampled, but not asynchronous.  I agree with Todd, anything else has unwanted side effects.  I suppose a harmonic upsample (non-integer) would be ok, but definately not two unrelated clocks.

My "reclocking" is nothing more than a complete regeneration of the exact same bit clock.  Same frequency and phase, ...

The clarification is appreciated!

That was how I thought the design was going to be...  If "slow rolloff" is at least an option (it does not necessarily have to be the default), I'd be *very* interested in trying it.

[doug s.]

Sorry for the confusion, HAGDAC won't be asynchronous.  My mistake for using the term "24/96" or similar.  Everything synchronous, the word clock is 352.8kHz, or exactly 8x the 44.1k input.  So it is upsampled, but not asynchronous.  I agree with Todd, anything else has unwanted side effects.  I suppose a harmonic upsample (non-integer) would be ok, but definately not two unrelated clocks.

My "reclocking" is nothing more than a complete regeneration of the exact same bit clock.  Same frequency and phase, ...

this sounds nice.  i am sure the reason i dint like the gw-labs dsp upsampling unit between my transport & my modded art di/o, was due to its asynchronous upsampling.  i am conwinced that if it had upsampled to 88.2 instead of 96, that it woulda been a *lot* better.

doug s.

[BobM]

Come on - make this product a reality already    I've been playing around with a non-oversampling DAC lately and don't necessarily like what I'm hearing as much as my modded DI/O.

I need an upsampling DI/O killer kit to get my grubby little rosin covered fingers on. It might not be yet, but I need a good project for the summer too.

Enjoy,
Bob

[doug s.]

Come on - make this product a reality already    I've been playing around with a non-oversampling DAC lately and don't necessarily like what I'm hearing as much as my modded DI/O.

I need an upsampling DI/O killer kit to get my grubby little rosin covered fingers on. It might not be yet, but I need a good project for the summer too.

Enjoy,
Bob

i also tried a non-os dac.  i am keeping my modded di/o.    

i don't think you will find a di/o "killer" at any price.  but, i would certainly like to find something *better*, at a reasonable price!  

doug s.

[doggie]

CHIME motherboard built.  

Progress is good, except on HAGUSB.  Can get Windows XP to recognize it.  Probably something real simple.  Just not obvious.

jh

Hi Jim,

I found out about your Dac-to-Be via a link to your blog. I was particularly interested in the HAGUSB. Would this be a separate product or part of the CHIME DAC? I am looking for a high quality USB to SPDIF converter to add to my own NOS battery powered DAC. Any chance you might sell this a board-only "kit".

Thanks a lot,

Paul

[hagtech]

Yeah, the HAGUSB can be sold separately.  Not sure if I sell it as a kit, since the surface mount parts are really small.   What do you think?  Anyone want to solder 0603 packages and the TSSOP?

jh

[doggie]

Yeah, the HAGUSB can be sold separately.  Not sure if I sell it as a kit, since the surface mount parts are really small.   What do you think?  Anyone want to solder 0603 packages and the TSSOP?

jh

Thanks for the reply. Nope, I am SMD challenged since I am not eight years old or a sparrow hawk

I guess by kit I meant just the populated board.

Best,

Paul