Building the ultimate NOS DAC using TDA1541A

[-ecdesigns-]

Hi philpoole,

thanks for your reply [post#493]

MCK (as indicated in the universal I2S interface diagram) is the master oscillator output of the CD transport (16.9344 MHz typical), so it's higher in frequency than BCK (2.8224 MHz typical). Since timing for SPDIF is derived from this master clock by the DAC / servo chip, they have a fixed phase relation (they run in sync). By keeping SPDIF interconnection very short, jitter is already much lower than when used with the common COAX or toslink cable. So when reclocking BCK, WS and DATA with this 16.9344 MHz masterclock, you don't get the typical asynchronous reclock problems. So reclocking certainly makes sense here and can effectively reduce residual jitter on the I2S signals, regardles if they are comming directly from the DAC / servo chip or from the CS8412 outputs.

When reclocking BCK, WS and DATA (that may still contain some jitter), these can only change on the master clock transient, so they run in perfect sync with the master clock. Jitter value now only depends on the master clock stability.

So using a low jitter (external) master clock can reduce jitter even further when correctly applied.

What about I2S jitter? Is the I2S output by the transports DAC / servo chip really jitterfree? we just take it for granted it is. Reclocking it with the master clock can be revealing.

[tubee]

Reclocking it with the master clock can be revealing.

Within the 304mk2 (have it from Bernard!) i modded all clock signals, and dem, they all are devided from 1 source, a 11.2896Mhz Kwak7. Yes, after some days listening, its unbelievable what an important mod this can be. The original clock outputs of dig. filter and decoder are left unused.
There still is some slight sibilance, but i will cure that. I assume a Bessel filtering could be better than the standard 3rd octave butterworth, created around opamp. Bessel has a little roll-off in highs, and better impulse handling. This time (i did in a non-os cdp) i don't want to remove the analog filter. Two beads on dac output extra could help too.

EC: what can you say about the cristal mod only, is it improved ?

[rfbrw]

I find it amazing that one would come up with the audio equivalent of a Heath Robinson design when all one need do is change a tap or two.

[maxlorenz]

No need to go from 48 to 64. Just use different shift register taps.

Thanks again, rfbrw. You will make me (us) study

Somewhere I read that I2S is not so good for long distance (>1.2m) transmision. It is probably better for a one box CDP, I guess.
Or, does the use of RS422 drivers solve this?

I am 100% convinced now transporting I2S using the differential interface, outperforms SPDIF by far.

I'm not crazy, then!

Regards
M

[Terry Demol]

Originally posted by -ecdesigns-
Hi all,

The clock upgrade and the universal I2S interface produced an improvement that really shocked me, I never expected such an improvement was even possible. The improvement was verified by other people that listened to this setup and could directly compare it with the "old" SPDIF setup. Also note that this interface outputs the PHILIPS FORMAT that is needed for the octal D-I DAC, so no further conversion is necessary. If your transport already outputs the philips format, the CS8412 is not necessary of course.

Here is what I personally can conclude from measurements and listening tests:

Never ever transport SPDIF over a longer cable (>20cm), it will have a very bad effect on sound quality. I am afraid, a similar effect would occur using USB, as this signal also carries both data and synchronization. In fact the PCM2706 datasheet already indicates a SpAct PLL circuit is used to recover a low jitter audio clock from the USB packet data.

The good news however is when the SPDIF / USB cable is kept very short and no transformers are used, things look quite different. So it's of paramount importance to put the SPDIF / USB receiver as close to the source as possible, then transport the generated I2S signal by using the differential interface. When reclocking can be applied as well, I2S performance can be obtained.

I am 100% convinced now transporting I2S using the differential interface, outperforms SPDIF by far.

ECD

I recommended Guido Tent's XODAC PLL for a good reason, it has
very high attenuation of SPDIF interface induced jitter from a
very low corner frequency.

Personally I run a slaved transport with master oscillator in DAC.
This is theoretically even better than I2S because the XO is -right
at the dac- reclocking all signals.
Also the interface is less important.
We also run our own design clock, but that is another story.

Cheers,

Terry

[Voltaic]

I don't want to hijack, so I'll keep this short:

-ecdesigns- you don't have your email address public, but I'd like to get in touch with you about your sonic resonators, if you could plz contact me, I'd really appreciate it.

in regards to this thread; it's incredibly interesting, I havent had time to read the entire thing just yet, but I'm printing it just now for my bedtime reading material.

cheers,
Daniel

PS: if you want to put me up for a few days, I'd love to come out for a listening session, can't afford to fly to the Netherlands for 8 hours, and fly back to the UK though

[-ecdesigns-]

Hi all,

I have a lot of catching up to do with answering posts, but as I already indicated, I have been very busy lately.

weisi [post#480]

Thanks for your compliment regarding this thread, it really motivates me!

Terry Demol [post#462]

After a lot of testing, I am convinced SPDIF is not the format you want to send over a interlink, but when connecting the SPDIF receiver really close to the transport and use the differential I2S interface to run the signals across the interlink, it's a complete different story.

But I am still very interested in the transformerless differential receiver you described as it can be used for the differential I2S interface as well. However, multiple differential I2S are to be used in the octal D-I DAC to switch between various digital sources like CD player, sattelite receiver and USB from the mac. So we need many differential receivers.

cheers, John

stefanobilliani [post#469]

The balanced I2S interface is in the universal I2S interface schematics, different transmitters / receivers have to be compared for obtaining best results. Currently I am using 4 X DS8922, 70mV hysteresis, 12nS delay, 500pS skew.

Maxlorenz [post#471]

The CDPRO-2M can be connected to the universal I2S interface, problem solved. Changing taps in the timing chain can also be done, read [post#482] carefully. But then only the sony format can be used, so with multiple inputs on the octal D-I DAC using both sony and philips format you have a problem (USB interface PCM2706 uses the philips format). So I like to stick to the philips format.

a333bt [post#483]

SPDIF is no longer an issue as I now use I2S for the interlink. I don't know yet when the PCB's can be manufactured, as I still have to design and test some more modules (I2S input switch, analog output switch, system controller and power supplies). But perhaps I could already start with ordering PCB's for the analog mainboard and it's modules soon.

rfbrw [post#487]

I referred to the sony and philips format

rfbrw [post#489]

Read post #486 carefully, also have a look at the schematics. The CS8412 is soly used to get a philips format from a transport outputting the sony format. The circuit works beyond expectations, It has resulted in a major improvement that can be noticed within seconds. It's in huge contrast with the existing SPDIF interlink I used previously.

duderduderini [post#490]

Welcome to this thread, the cdmpro2 seems to use the sony format, so the universal I2S interface [post#486] can be used to convert to philips format, however I want to improve it, so there will be some modifications. I plan to provide circuit boards (also for the power supplies and tube output stages) and housing parts (if desired) for this project, however supplying electrical components is a problem due to the RoHS directive.

[-ecdesigns-]

Hi all,

Still catching up answering posts,

tubee [post#491]

Yes if a philips format is available on the transport, the CS8412 can be left out since it is only used with transports that output the sony format. The setup using only the clock upgrade already improved sounquality, but I think it's improvement was partially covered up by the SPDIF interlink.

Bernhard [post#494]

The universal I2S setup (with or without CS8412) reclocks the I2S signals so they run in perfect sync with the master clock. If a low jitter master clock is used, and the laser-pickup circuitry only provides a digital bitstream (with or without servo jitter). Would this jitter have any effect in this setup as the I2S signals jitter now only depends on the master clock?

philpoole [post#495]

Yes mechanical servo noises can be a problem, I once compared a philips SACD1000 with my old twin DAC setup. Man that player sure was noisy, one could hear all servo activities in minute detail to a point it became very irritating. Noise increases when the laser pick-up frame touches or is firmly mounted to the metal housing, The metal housing then amplifies these vibrations. Similar problems occur with mounting harddisks. Using a setup Isolating the laser pick-up mechanism noise from the chassis could reduce servo noise quite a bit.

rfbrw [post#503]

I only used the CS8412 to obtain the philips format from my sony player, it's simple and it does the job, so what's the problem? The octal D-I DAC will have multiple inputs. Input devices may have sony or philips format, so one can integrate a complex timing-chain supporting both formats and (automatically) switch between these formats, or one could only adapt the device(s) outputting the sony format and exclusively use the philips format in the octal D-I DAC. I like the second option.

maxlorenz [post#504]

I am using 3 meters of network cable (3 twisted pairs are used), and the sound quality is way better than the existing SPDIF setup. The differential interface seems to work very well with I2S. I will do some "extreme" testing with 5, 10, 25, 50 and 75 meters to see how it affects BCK timing and sound quality. I also need to find the best differential transmitters and receivers, so some more testing has to be done.

No you're not crazy, I2S makes a world of difference, I should have used it much sooner.

Terry Demol [post#505]

I must admit I was sceptical about these clock upgrades, just like I was about using tubes. But now I heared what difference only a simple external clock module can make, I fully agree with you

I read about the slaved transport on the internet, some time ago. Very interesting, indeed it's logical to have the master clock as close to the DAC as possible to reclock the I2S signals. So basically, correct me if I'm wrong, you will have to feed the 16 MHz master clock back to the transport for it's internal servo mechanism and I2S timing. If so, won't this affect servo stability / servo noise? Transporting a 16 MHz clock across a interlink seems quite critical. But I am still very interested in how it improves sound quality.

[-ecdesigns-]

Hi Voltaic,

Thanks for your compliments with regards to this thread, I really appreciate it.

I just added my email address, so you (and other diyaudio members) can contact me for a listening session.

I think it's important now for readers to get some feedback from people who actually listened to both the octal D-I DAC and sonic resonators.

[rfbrw]

[QUOTE]Originally posted by -ecdesigns-
[B].
rfbrw [post#489]

Read post #486 carefully, also have a look at the schematics. The CS8412 is soly used to get a philips format from a transport outputting the sony format. The circuit works beyond expectations, It has resulted in a major improvement that can be noticed within seconds. It's in huge contrast with the existing SPDIF interlink I used previously.

Originally posted by -ecdesigns-


rfbrw [post#503]

I only used the CS8412 to obtain the philips format from my sony player, it's simple and it does the job, so what's the problem? The octal D-I DAC will have multiple inputs. Input devices may have sony or philips format, so one can integrate a complex timing-chain supporting both formats and (automatically) switch between these formats, or one could only adapt the device(s) outputting the sony format and exclusively use the philips format in the octal D-I DAC. I like the second option.

The CD Pro2 and the CS8412 both have an I2S option. Wlth the CD Pro2 BCK is 48Fs and 64Fs with '8412 but that is an aside. The numbers for the CS8412 aren't that great and as I see it, it will always be better if one can do without it. It is trivially easy to switch between taps to meet the need of different input formats but then again I wouldn't use discrete logic.