The third try, or Revenge of the PCM1704

The cheap DAC designs are done and passing from the stage of my thoughts. I now turn to another absurdly expensive DAC! This time I will design the whole thing in public, on this forum, with feedback from the prophets, elders, and acolytes thereof.

To make sure this DAC is actually different from the previous efforts, I will not use an asynchronous sample rate convereter. At least, I will not use the CS8420. I might think about the AD1896.

For the digital input I shall use the Burr-Brown/Texas Instruments DIR1703. The claimed output jitter is 75ps, and it can receive 32-96kHz SPDIF signals. Physical interfaces will be transformer-coupled coax and Sharp GP1FA551RZ fiber optic.

I think I will use the DF1704E filter and PCM1704U-K DAC again. It's a good combination, even if the pins are all in bad positions. Perhaps the oversampling could be switched on or off at will.

Let's see how the budget is working out so far:

DIR: $3
Filter: $16
DACs: $152
BNC connector: $9
Pulse trafo: $15
Fiber receiver: $2
PCB: $40
On-board regulators: $20
Reset pulse generator: $3
24.576MHz oscillator: $8
I/V BJTs: $13
Output capacitors: $160 or ??
Output XLR connectors: $12
Output RCA connectors: ??
Misc. passives: $20?
Power supply rectifier: $18
Power supply trafo: $18

So far: $509. Not too bad. More than the el-cheapo, less than buying it from Mark Levinson. This is still in the planning stages. Let me hear your best ideas!
 
Four DACs: two for left and right each. The PCM1704 makes this easy for us with the INVERT input.

How many windings do you propose? I think each pair of DACs should share the same power rail, to ensure that power noise is common mode. So, perhaps a winding for each channel of DACs, a winding for each channel of I/V conversion, and another one for the rest of the digital circuitry? That's four dual-secondary trafos. That wouldn't be too expensive but you now require a lot of extra rectifiers as well.

Nevertheless I like your idea.
 
I went a bit overboard but I used separate windings for: input receiver, oscillator, filter, ea. DAC has 4 separate supplies (LT1085 and LT1036 , separate digital and analog) for a total of 8 supplies just for DACs with separate windings. Plus four windings for output stage. Total: 16 windings-8 trafos. Now ea. windings has 4 soft rec. rectifiers for a total of 64 diodes. I also used something like 24 caps for prereg. (off the PCB).

I didn't go balanced but I paralelled PCM 1704 (so I'm also using 4 of them). I also included output switched pot, so no preamp required. This was my first DAC and when I started building it I didn't have a clue about digital. Now I know a bit more and I'm ready for another, absurdly expensive DAC.;)

I will be participating in a project.
 
Here is a pic. The trafos on left are stacked one on top the other, so there are 6 of them. I also use split bobbin type, because they are better.;) Right beside trafos there are 20 Sprague Extralytic caps under the PCB (so no see). Rectifier diodes are directly attached to the small trafos so no see either. On the right side under another PCB there are 8 more big filter caps.
 

Attachments

  • 12.jpg
    12.jpg
    94.3 KB · Views: 2,267
alvaius said:
jwb, if you are going for broke, why not use a DSP to do digital filtering and give yourself complete control over the process as opposed to relying on the filter you buy? Not much cost adder, but certainly some time adder.

Who is going to write the software? As soon as I get a DSP, I suddenly need microcontrollers and EEPROMs. I write software for a living. I don't want to write software as a hobby, too.
 
Why not use the DF1706? It has 192kHz input. You can let the AD1896 do 4x OS and leave the other 4x to the filter.

Also, how do you get the DIF1703 to work properly with only one crystal? If you use a 24.576 mHz crystal, it's not a supported multiple of 44.1 kHz. Wouldn't you also need to employ a second crystal to handle 44.1 and 88.2 kHz sample rates? 11.2896 would work nicely at 256 and 128 fs respectively. How would one even implement a multclock system? Sorry, I'm kind of new to this.

Stu
 
alvaius said:
EEProm yes, but why would you need a microcontroller?

Perhaps I don't. As soon as you mentioned DSPs, I assumed I would need a microcontroller to boot one up and so forth. But looking at the data for the SHARC, it looks like it can boot itself from an EEPROM.

Anyway I'm not interested unless someone has some software laying around already.
 

rfbrw

Member
2001-10-26 11:51 pm
.
If one plans to make use of all the options available on the DF1704 a microcontroller will be necessary. In order to provide the oversampling bypass a CPLD or a FPGA will be needed if the pcb is to reasonably compact. From there it is relatively small step to rolling your own digital filter in the FPGA. This also opens up the option of PLL or FIFO based retiming.

ray.

p.s. $160 for output caps?????
 
maczrool said:
Why not use the DF1706? It has 192kHz input. You can let the AD1896 do 4x OS and leave the other 4x to the filter.

Also, how do you get the DIF1703 to work properly with only one crystal? If you use a 24.576 mHz crystal, it's not a supported multiple of 44.1 kHz. Wouldn't you also need to employ a second crystal to handle 44.1 and 88.2 kHz sample rates? 11.2896 would work nicely at 256 and 128 fs respectively. How would one even implement a multclock system? Sorry, I'm kind of new to this.

Stu

I was under the impression that the DF1703 generates the required system clock based on its interal PLL and VCO, nevermind the rate of the master clock input. I believe the master clock input is used only to drive the SpAct when the 1703 is in PLL mode. The datasheet claims that the SCKO will be 256f<sub>s</sub> (or 128, 384, 512) regardless of the crystal frequency. If not, I am a dunce. The DIR1703 datasheet is extremely vague.
 
rfbrw said:
If one plans to make use of all the options available on the DF1704 a microcontroller will be necessary. In order to provide the oversampling bypass a CPLD or a FPGA will be needed if the pcb is to reasonably compact. From there it is relatively small step to rolling your own digital filter in the FPGA. This also opens up the option of PLL or FIFO based retiming.

If all I want is a FIR filter, I hardly see why I need to implement it myself in a DSP or FPGA. A million other people, nearly all of them smarter than myself, have already done that. I can even buy them in hardware: the AD1896 implements a 64-tap FIR filter for 8:1 interpolation.

I'm inclined to use the AD1896 for interpolation and disregard the DF1704 altogether. The Analog part has much better specifications. So the incoming signal would be reinterpolated to 96kHz, which for CD audio would be 2.2x oversampling and 2x for DAT and friends. And the Analog datasheet seems to imply very low jitter on the output.

Then I would need some logic to split the left and right word into the different DACs. Is there a canonical way to do that?

The chipset would then be changed to DIR1703->AD1896->(logic)->PCM1704.
If it was AD1896->DF1704->PCM1704, that would be way too much like my previous DAC, except without the CS8420.

I can't seem to quickly locate a supplier of AD1896. Anyone know?
 
Pioneer-standard apparently stocks them when they feel like it. They run around $20.00. Here's the link http://MyPioneer.com/onestop/partde...t_number=AD1896YRS&sSearchID=11834763&rnd=801

About the DIR1703, that's not at all how I took it to work. From other posts I have read I would be inclined to think differently. I'll ahve to go back and read the doc on it. I hope you are right though. I have a few and would like to use them as you suggest. I have a couple PCM1704s and DF1706s on order.

Stu
 
maczrool said:
About the DIR1703, that's not at all how I took it to work. From other posts I have read I would be inclined to think differently. I'll ahve to go back and read the doc on it. I hope you are right though. I have a few and would like to use them as you suggest. I have a couple PCM1704s and DF1706s on order.

If you needed a different crystal for 44.1 and 48kHz, it wouldn't be a very useful chip. I can't imagine it works that way. And from a careful reading of the datasheet, I believe it can do any frequency up to 96kHz with a 24.576MHz clock input.
 

rfbrw

Member
2001-10-26 11:51 pm
.
jwb said:

I'm inclined to use the AD1896 for interpolation and disregard the DF1704 altogether.

snip snap snip

Then I would need some logic to split the left and right word into the different DACs. Is there a canonical way to do that?

Well, I would ditch the AD1896 as well as the DF1704/6 and go with something along the lines of the Meitner IDAT but tastes differ as they say.
I do not know whether it was canonical or not but I have connected a CS8412 in I2S to two PCM1702 with no interchannel delay. In its most compact form using the "stopped clock method" it fit in a 44-pin XC9536 CPLD and would require only minor mods to work with the PCM1704.

ray.
 
Your DAC ?

Hello, Peter

are You happy with this DAC (PCM 1704; DF1704 )?
I understand that it is from broadhurst project.
How much costs the parts today, what you think about it?
You know my interst for to make a good level DAC-may be this is alternative to TDA 1541 superdac ?
What plans do you have for the future--new DIY DAC ?
What is for you by this DAC a problem ? or more important is to find all time better and better ideas. ( I myself made all the time upgrades for my tube amplifier: PP EL34 triode mode class A )

best regards

hifiwolf

Peter Daniel said:
Here is a pic. The trafos on left are stacked one on top the other, so there are 6 of them. I also use split bobbin type, because they are better.;) Right beside trafos there are 20 Sprague Extralytic caps under the PCB (so no see). Rectifier diodes are directly attached to the small trafos so no see either. On the right side under another PCB there are 8 more big filter caps.
 
This is the finest DAC I had and heard, period. The TT doesn't even come close (and I had Maple Noll air bearing, Sumico Blue Point Special and Ono as preamp). I replaced my Technics X1000 (which was CAD 8,000) and it's like day and night.

Lot's of tweaking and mods went into this design, but the key elements are separate supplies (including transformers) for everything, good oscillator and parallel PCM 1704K chips (also with separate digital and analog supply).

I started from an oversampling DAC and my next move is to try NOS designs. I will mod my marantz CD94, then built TDA 1543 and then try AD1865K. Last two DACs will be fed direct I2S from my CD Pro 2 transport, because some people claim that SPDIF is the root of all evil.;)

As for I/V conversion I will definitely try Pass D1 stage. I'm surprised nobody's experimenting with it yet (or at least nobody's talking about it);)