[hagtech]

Is there an advantage of using the PCM2902 over the PCM2704?

jh

[BradJudy]

Is there an advantage of using the PCM2902 over the PCM2704?

jh

I don't know.  I haven't even looked at the spec sheet for either one, and even if I did, I'm not sure I could answer the question.  

[hagtech]

Ok, I think I have this circuit worked out.  The PLL flywheel concept looks good, and should work quite well.  In fact, I discovered pretty much the same architecture at TentLabs, while browsing for chip info.  I do the PLL a little differently, and use different parts for the rest of the circuit, but it is comforting to know that the flywheel reclocking scheme is valid.  

Only issue is that all of the parts come in surface mount.  And I don't think that would make a good 1/2-kit.  Small outline packages and 0603 resistors are not suitable for the typical DIYer.  Therefore, I am thinking of doing the product in modules.  The HAGDAC would be a small 4-layer surface mount daughter card that would plug onto the main CHIME motherboard.  All prebuilt and tested.  Dang, these parts are not cheap!  It's like $50 just for the DACs.  The digital filter is $15.  The cost of parts adds up pretty quickly for such a small board.  My guess is that lowest retail price would be $495.

The CHIME motherboard would then be a standard $95 1/2-kit item.  You buy it and the HAGDAC to build a CLARINET-style chassis.  Will have two RCA inputs with optional SELECT switch, and optional VOLUME control.  Front panel also has optional analog meter showing what the PLL is doing (frequency tracking).  A third input would be an optional USB daughter card.  That too, is a surface mount circuit, so would be a pre-built and tested board for $95.  It would plug onto CHIME.

So CHIME contains the input and output circuitry and power supplies.  Outwardly it would look almost like a CLARINET, with same tube configuration (although I may opt for a 6922 current source loaded gain and output buffer stages).  You then buy and install pre-made HAGDAC and the optional HAGUSB input cards for a complete solution.

I'm thinking a factory assembled machine will retail for $1495 plus additional $95 for USB input.  This is for redbook 44.1kHz CD playback only.  Overall, that should be quite a bargain.  This circuit is no slouch, we're using the best components available and are aiming for sonics to compete with the average $10,000 to $20,000 CD player.

The neat thing is that you can buy just a HAGDAC and put into your own CD player or other box.  All it needs is a SPDIF input and +/-6V to +/-12V supplies.  Or buy a HAGUSB card to add USB capability to your existing CD player.  Interestingly, the modularity also opens the door for future DAC cards.  Perhaps one for DVDA or SACD or DAT?  For now, it seems the killer app is for awesome sound from the vast CD libraries out there.

Comments?

jh

[Yoda]

I like the modular converter board approach.  This seems like a wise solution to tackle the uncertainty of future formats and maintain a good analog output section.  Would there be any economy to combining the redbook DAC onto the output board with room to plug in optional Dac boards?  you would order a 1/2 built 1/2 kit...

M

[Adinoto]

Hi Jim,

This modular scheme is realy interesting.
Question about HAGUSB : it will receive USB input and what as output? SPDIF? I2S? How it connected to standard CD or DAC? trough SPDIF coax? If so, then it can be a sepereate stand alone product : Computer-> HAGUSB -> any DAC ?

Regards
adinoto

[hagtech]

I'm thinking the HAGUSB would be SPDIF output.  Input to HAGDAC would be SPDIF.  Or you feed it to another machine that accepts such input.  Neat thing is that it is bus powered by USB, so it is standalone and should connect straight to any DAC.

The HAGDAC might be able to run at other frequencies, not just 44.1k.  It can be set for 48k or 96k, too.  All the parts used can handle it.  The limitation is my PLL reclocking.  I didn't want to run such a wide locking range to keep phase noise to a minimum.  Rather focus on a very small tuning range around 44.1k (~11.3MHz).  Optimize for CD sonics, rather than a do-everything approach.

jh

[orpheus]

Jim-
    Wow, you work pretty fast.  I like your idea of using modular boards.  

I want to say that I am a layman, an enthusiast and not the most knowledgeable one at that.  There are huge gaps in my understanding, so take my suggestions and comments with that caveat.

What is doing the I/V conversion in the design?  Are you planning on having the HAGDAC module output current or voltage?  

I really, really like your idea of optimizing the circuit for redbook CD.  That's the most prevelant format, and it's also the format that needs the most help.  If adding conversion for other formats will impair your pll scheme for redbook, it's not worth it.  

I might have misunderstood what you wrote about the Chime board. When you say that it will include the input/output, do you mean that the digital input and receiver will be on the Chime board, or that it will also act as a line level preamp?  

I would like to see an option for a seperate supply for the digital board.  Could the kit you already designed for the bugle be used for that?  

Have you considered using a digital input transformer?  It would add to the cost, but it might really improve the circuit.

It also might be nice to have at least the option of using a seperate (even a wallwart might be good) supply for the usb card, because the power from inside a computer is probably not as good as a seperate supply.  Not everyone would necessarily be interested, but it would be nice if it was possible.  

I don't know if you already know about this, I apologize if you do. There are some pitfalls with usb audio and windows xp.  Windows uses a built in software called kmixer to process digital audio, and it needs to be bypassed in order to achieve optimum sound.  Here is a link in case you're interested.  

http://www.staudio.de/kb/english/drivers/

Thank you again for your hard work and being so open to receiving comments and suggestions.

-Aaron.

[hagtech]

[What is doing the I/V conversion in the design? Are you planning on having the HAGDAC module output current or voltage?

I/V will be low value resistor.  I really don't like the idea of using the feedback network of an opamp to do the job.  Of course, the compliance of the dac chip is an issue.  So we just keep the voltage signal swing really small.  Hence, the need for an external tube gain buffer (CHIME).  Filter is my own LC Bessel design.  All passive, no feedback.

I can also add opamp gain stages so the HAGDAC spits out both the raw (50mV) signal and one with gain (1V) for those not using the CHIME motherboard.  No, they won't be as good as the tube stage, but very good for SS.

I really, really like your idea of optimizing the circuit for redbook CD.

Thinking, I can also suggest PLL oscillator components that will cover both 44.1k and 48k.  Or even a solution for 96k.  Does anyone want 96k capability?

I might have misunderstood what you wrote about the Chime board.

The CHIME will have input SPDIF connectors and the optional SELECT switch, optional VOLUME control, tube stage, and power supplies.  Basically almost a CLARINET but with the daughter card interface.

I would like to see an option for a seperate supply for the digital board. Could the kit you already designed for the bugle be used for that?

Yes.  But designed for +/-8V outputs.

Have you considered using a digital input transformer?

Started with it in, but realized the RCA jacks I use have a common ground pin for the shields.  The transformer would then not isolate chassis grounds for more than one input.  Also, since we are re-clocking, the input to the receiver should not be as critical.  Instead I am including a special receiver filter for optimum bandwidth - something most designers don't know about.  It goes to digital channel theory (Shannon) and noise issues.  Somewhat non-intuitive, the result.  Bottom line is that it removes jitter inducing noise for optimal BER (bit error rate).  Stuff I learned working on 3GHz communication systems.

So no transformers required.

There are some pitfalls with usb audio and windows xp.

I don't suppose anyone is suprised.  Yes, it will be a can of worms I am not looking forward to.  

Don't think a wall-wart will be any better.  The good news is that the 5V from PC is regulated down to a clean 3.3V for the audio circuit on HAGUSB.  I can get that clean.  Will add output buffer and optional RCA jack for SPDIF, so you can run long cables.  Hey, maybe the board can be built as a dongle?  Will have headphone connector too, as the PCM2704 has built-in 16-bit DACs and headphone amps.  Or connect the heaphone output to input of a line stage.  Maybe HAGUSB doesn't go inside CHIME?  Not bad for $95.  Maybe I'll have to charge more...

jh

[BradJudy]

Hmmm...I don't have any experience with the headphone part of the 2704, but I'm not optomistic about its capabilities (since I own a couple of decent headphone amps).  I suspect it would be better to leave it out of there and consider headphone outputs later in the chain (maybe there can/should be a headphone output daughter board).

[hagtech]

Was thinking the headphone output was only for computer speakers.  Not quality audio.

Or maybe this is a dead end?  A quick Google and it seems there are already plenty of USB -> SPDIF boxes out there.  No need to recreate the wheel, as this portion of circuitry is not too critical sonically.  

jh

[hagtech]

Ya know, I think I'll just put the USB chip down on the CHIME motherboard.  All of the supporting components can be regular feed-thru types, so only one chip is surface mount, and it's not really that small.  Better than a separate board.  Guess I don't want to get into the dongle business.

So CHIME will have USB input along with two RCA jacks for SPDIF.  This gives us the capability we want at lower cost.  The USB section will be on separate ground and power (5V supplied from PC) and coupled to CHIME via transformer at SPDIF.  This gives us the isolation we need to remove computer junk getting into audio system.

jh

[clivem]

The USB section will be on separate ground and power (5V supplied from PC) and coupled to CHIME via transformer at SPDIF.

Jim, why not use the I2S output of the PCM2707 to drive a standalone DAC chip? http://focus.ti.com/lit/ds/symlink/pcm2707.pdf. Converting USB to SPDIF is a halfway solution.

I've played with the 'Guzzler' DIY USB DAC mentioned earlier in this thread http://www.mellowparenting.demon.co.uk/guzzler/usbdac.html and would tend to agree with Gordon Rankin that the Sigma-Delta DAC units that these one chip solutions use are not very good.

Regards

Clive

[hagtech]

Jim, why not use the I2S output of the PCM2707 to drive a standalone DAC chip?

Because that would bypass the re-clocking.  It is easier to mux several spdif-only inputs than mix breeds.  Makes the HAGDAC daughter card much simpler.  If this were to be a USB-only DAC, then I would probably do as you suggest.

jh

[flocchini]

Jim:

This is really shaping up as a great project.

Is differential output still on the board?

Any chance of a Octal version? I am really enjoying the Cornet2 with the 6sN7.

Thanks.

[hagtech]

Is differential output still on the board?

I goofed, the dac chip is single-ended.  So conversion to differential will be off of daughter card.

To keep with "C" series, I will probably do just a single-ended tube stage.  Perhaps a future Trumpet-like product will do the fully balanced thing (not a kit).

jh

[hagtech]

Ok, now we're getting somewhere.  Schematic is done and layout started.



HagDac daughter card is 2.2 x 4.4 inches and fits nicely onto CLARINET style motherboard.  Could also be used as upgrade to existing players?  Architecture is pretty straightforward, but with PLL flywheel for reclocking between digital filter and output converters.  Takes redbook S/PDIF input, upsamples to 24/96, reclocks, with 24 bit PCM conversion into passive I-V filter.  Has both raw (35mV) and opamp buffered (0.78V) outputs.  Just need to supply +/-7V to +/-15V.  Regulators onboard with plenty of filtering and decoupling.  Had to reach deep into my bag of tricks to find ways to improve on what others have done.

There are three control inputs: MUTE (turns on soft mute function), STEEP (selects digital filter rolloff mode), and INVERT (for setting output polarity).  Three outputs available for monitoring status: RX_LOCK (receiver locked to signal), PLL_LOCK (reclocker locked), and METER (an analog output showing PLL tracking in real time).

Getting there...

jh

[Yoda]

Cool!


Matt

[hagtech]

Back to making a separate HAGUSB card.  The chip is an SSOP-28, which is probably too much to ask everyone to solder.  Yeah, there's a simple trick to it, but in the pursuit of simplicity I think it best to offer the USB circuit already assembled.



So once again the CHIME (hey, maybe I should call it a digital processor) has two pre-built daughter cards: HAGUSB and HAGDAC.  I'll offer both separately, too.  The CHIME half-kit should be no more difficult to build than a CLARINET.  

I added an RCA output footprint on HAGUSB to make standalone operation easier.  All it does is convert a USB audio stream from a PC into an electrically isolated S/PDIF.

jh

[hagtech]

Ok, got the board set completed.  Ordering prototypes.  The HAGDAC layout turned out to be quite difficult.  Picture below shows the myriad of low-impedance power planes.  You can only do this with multi-layer boards.  Sure helps keep each circuit's power supply clean and uncorrupted.



Besides the separation of supplies and liberal decoupling, I added passive filtering between chips (digital) to greatly reduce switching transient issues.  I don't see anyone else doing this.  Also other jitter-reducing tricks such as optimized channel bandwidth (Shannon's theorem) should help cut noise.  I can see now why SPDIF transformers improve sonics (hint: it's not just the isolation).  

Curious to see how well the PLL works.  Keep you posted...

jh

[Gbatokai]

Ok, got the board set completed.  Ordering prototypes.  The HAGDAC layout turned out to be quite difficult.  Picture below shows the myriad of low-impedance power planes.  You can only do this with multi-layer boards.  Sure helps keep each circuit's power supply clean and uncorrupted.



Besides the separation of supplies and liberal decoupling, I added passive filtering between chips (digital) to greatly reduce switching transient issues.  I don't s ...

This project is very interesting. It's nice that you apply some fresh thinking here - it's definitely needed. Grounding scheme problems and ground noise is an issue in most DACs I've met, so please keep this in mind.  The ciruitry apparently generates all kinds of garbage that must not be allowed to propagate into the analog sections.

Just to throw some things out there, the AD1865N-K chip used in my Ontech DAC is said to be superior to many or most modern 192KHz chips.  Since you're staying at 44.1 KHz, it won't impose any limitations - so it would be nice if you found the time to check this chip out. The Ontech - which is a new design - also uses the obsolete DIR1701, by the way.

Incidentally, I finished my Frykleaner Pro kit today - now for some cable exercise...