Non oversampling 2xPCM1702 feed by I2S experiment results

Well, I finished an experiment during the weekend that consisted of Philips CD610 (model from the beginning of 90's) used as a transport and I2S taken from the players servo SAA7310. I also replaced the clock with a four pin in can unit from ELFA. I don't have any verified audiophile clock on hand, but this ought to be better than the original oscillator inside the SAA7220 digital filter.

The I2S was fed through some logic IC's into 2 BurrBrown PCM1702 DAC's. I own a big thanks for interface schematic to my frind Sulo. The DAC chips were fed with +/- 5V created with LM317/LM337 regulators sepparately for digital and analog. Before regulators the supply had LC filtering - 10H indictor and 10000uF of cap.

The analog stage of the DAC - well there wasn't any. At first I fed the signal straight from the DAC output pin to my passive attenuator (10k) and then into JLH'96 amps. As it turned out the DAC clipped with 0dB sinewave with this configuration, although it was very soft tubelike clipping. So by experimenting the final value of the resistor from output pin to ground was 680 ohm. This still showed a little distortion with 0dB signal but it was about 0.1% of second harmonic so I didn't go any lower. There rarely is any information on cd to reach 0dB anyway.

The only techical remark till this was that the output of the DAC without any active stage is about 400mV - a little low:)

Enough of backround... what did it sound like?


WAUUU, if one would put it into one word. This configuration had some qualities unheard of before from any commertial or DIY project. Only player coming close is my frinds project with CD-PRO transport and TDA1541 + tube stage.

The most struck the air and freedom of the soundstage. I had always felt that most players, even if they play for example vocals quite well when there are few instruments on backround loose the beauty when there is a lot of instrument and action going on. This setup seemed to have no such problem at all. There was such ease all the music was played that it was quite astounding.
Also the human voice, both male and female sounded the best I have ever heard. There was no kind of hardness in the vocals that's usually there.

All the little nuances and details seemed to be just in the right place giving a wonderful presentation of Art Ensemble of Chicago live recording I use to evaluate such qualities. And the soundstage was deeep. I have tweaked my Marantz CD67 for about two years gradually and tought I had come to quite a good level, but this showed me I was anywere near the potential on the information on CD....


All things considered, this was a real eyeopener (or should I say earopener). It seems that there still is something that goes wrong in digital part of the player besides Jitter. This kind of nonoversampling seems to be so much better for the ear than all the present day conventional approaches though the BurrBrown DAC's here are great.... Anyway I really recommend to try out something similar and hear for yourself :)

My friend Sulo, who is also a very experienced audiophile, was also astounded by the results. I will go to the local HiFi shop in near days and let the guys hear it out too, but it seems it is not my imagination, but that needs to be confirmed.....

Hi ergo ... always nice to hear someone else his experience... thanks foor the info ...
I've been tinking on a new DAC project for some time now and was wondering is a 16bit or 24bit crystel reciever and a PCM1702 could form a simple but high quality DAC?

How did you get your chips... any suggestion how I could getb them....

Greetings and happy listening!

Are you sure the signal from the decoder was not already oversampled? The CD-621 (i.e. one generation after the 610/620) uses the SAA7345 decoder to do 4x oversampling, hovever the stop band suppression is very poor so I suppose this is either a very short filter or one with very poor arithmetic wordwidth.
All subsequent decoder chips contain the SAA7345 core.

You could verify by checking the frequency of the digital signals.

I suspect the current Marantz CD4000 and 5000 use the same approach. The Marantz home page lists the SAA7378 as decoder, no digital filter chip and the TDA1545 resp. 1549 is a convertors. Neither DAC has built-in oversampling capability, so I suspect the decoder is used to do the oversampling.

Still, I'd be interested in the schematic to separate I2S into to separate signals (saves me some brain cells...).

Yes I'm sure there is no oversampling. The original schematic had SAA7220 digital filter that did the 4X oversampling and TDA1543 DAC. It is also verified as I checked the bit clock and it's 2.8x MHz...

I'll try to input the schematic into some editor in a few days, at the moment it's on the corner of some piece of paper :)


As for CD4000 and CD5000 there indeed the oversampling is done in decoder IC. The decoder IC output mode is configurable and thats done by microcontroller. Unfortunately my knowledge doesn't allow me to backengineer the code to change it to non oversampling mode. If any of you guys are able to do such a thing I think the CD4000 and CD5000 would be fantastic basis for non oversampling players. If also the supplies and clock are improved I think it could be absolute wonder....

Anyway if anyone hacks the player some day drop me an e-mail. It could well be possible to make a little business by providing DIY DAC kit for CD4000 and CD5000

To Jocko,

What digital filters you consider to be good and have you tried many of them. As there seems to be huge differences whether there is a filter or not I suppose there should also be differences between filters. Do you have any experience and comments you could share?


PS. If possible also mention the sources of IC's you suggest.... These can be very hard to find in small quantities.

I done several with the NPC5843. It accepts I2S with little trouble, and has NO ultrasonic garbage; unlike that wretched SAA7220.

Also built some with the DF1704..... a real pain to wire by hand, but I'm working on a solution to that.

NPC parts used to be easy to get. I'm trying to get a better source. Apparently, there is a distributor, but they don't stock anything.

The B-B stuff is easy to get from Digi-Key. And cheap.

Except the good DACs.


I am confident that oversampling will sound better. However, given that there are now so many enthusiasts for none-oversampling, I thought I'd rig up something to get a first hand impression.

Have you experienced sonic differences between NPC, DF1700 and DF1704/06?

As for the soldering: the freeware version of Eagle ( will allow you to make double sided boards up to 100x80 mm². It contains a footprint for the SSOP28 case. If you want to, you could even design a SSOP28 to DIP28 adaptor PCB...

Soldering a SSOP28 is pretty easy if you use solder paste and a hot air blower. I have even successfully re-placed a SSOP28 on a home-etched PCB. Only one wire pealed, meaning I had to hand solder a thin wire.


Only used the NPC with the AD1862......oh wait......I did one PCM1702......

As far as I can tell, the DF1700 is the same thing.

The DF1704 version was full 24-bit.....being fed by good old "Redbook" I2S. The lucky guy who owns it likes it better than his SACD.

The '1706 looks like a '1704 without pull-up/down resistors and 3.3 V operation. So I'm not using it.

I have capabilties to make PCBs, but I made the prototypes by hand.

So.......why the obsession with zero-oversapmling? Anything worth doing is worth doing right, and that approach doesn't qualify in my book. Is it due a problem with getting parts in Europe?



2001-10-26 11:51 pm
Two things.
What dac did you use with the DF1704? The other
thing is your preference for digital filters. Is it based on the belief that if the numbers are right the sound must be right and if you dont it like the too bad. Or is it down to listening ?


ps. DF1700 => SM5813
Here is the interface schematic I used for I2S conversion



  • i2s interface for bb.gif
    i2s interface for bb.gif
    21.2 KB · Views: 3,549
little flaw

Hi Ergo,

I believe there is one cosmetic flaw: the L and R signals will not appear at the outputs at the same time. Rather, there will be a skew of 23 microseconds. I doubt that this is audible, it corresponding to one speaker being advanced by 7 mm, but if it does not hurt my purist ear, at least it hurts my purist eye...


Hi Ergo....

I hope you can help me out a bit... I just got myself a couple of PCM1702 chips.. I really really want to get more into digital design and was aiming for a zero oversampling DAC with the PCM1702 .. I'm not aiming for High End performance (yet) ... now what do I need besides the PCM1702?

Could you explain a bit or maybe give me a link to a Digital Audio introduction?

Thanks for this great thread! It is very informative stuff for me...

I2S Philips format conversion to Sony/Burr-Brown

Hi All,
The Analog Drevices Application 207 note takes care of that using a dual D-type flip-flop and a quad NAND gate.
The two stereo channels separation is accomplished by the so called "stopped clock" technique i.e. the bitclock is stopped during loading the data to the DAC of one channel (don't remember which one). Both channels are latched simultaneuously as they share the same LATCH. No timing difference between left and right channel.
Unfortunately the Application Note is no longer on the AD website but I can send the PDF file by email if in need.


2001-10-26 11:51 pm
I am sort of in the middle with this o/s no o/s debate as I am still not sure. The thing is the non o/s camp are no less adamant that they are right. Whatever,my way out is finish off my dac with a digital filter (NPC5843) that can be switched in and out of circuit.

Re: AN207
AN 207 IS available on the ADI website at
Alas AN207 will not work with the PCM1702 or PCM63 as is.
No problem with being open-minded. I just can't see the point.

I personally don't like the sound of tubes, but I understand the appeal. I suspect the non-o/s crowd really doesn't like digital, which is fine. Then they should stick to vinyl and tubes, and we will all be happy.

See other posts today about slew-rate & op-amps in I/Vs. Might explain my adamant postion.