In a 3-wire serial digital audio stream, including 3-wire I2S, the audio channel data is transmitted sequentially on only one of the three wires. To do this, the channels are what’s known as, ‘time-multiplexed’. Meaning, channel-pair data for a given sample do not arrive simultaneously. They arrive one, after the other. In other words, one channel arrives briefly delayed relative to the other. This is a physical fact. However, this then enables the channel-pairs to efficiently share usage of a single wire. In order to time align (to synchronize) the conversion of the channel-pairs, the time-multiplexed signal must first be ‘de-multiplexed’. Which requires the first channel to arrive, which is typically the left channel, to be held (registered) somewhere to wait for arrival of the right channel data. Such holding may be accomplished by several means, one of which is to hold the left channel‘s data in an external array of shift-registers, as it is here. There are other potential means to achieve this, such as ‘stopped-clock’ operation, but those are outside the scope of this discussion.
To summarize: Serial digital audio interfaces are data sequential. Channel-pair data arrives alternately. Not simultaneously (meaning, without relative delay). They inherently entail a brief arrival delay between channel-pair data of a given sample. That delay is eliminated by holding the first channel data somewhere until the second channel data finishes arriving.
To summarize: Serial digital audio interfaces are data sequential. Channel-pair data arrives alternately. Not simultaneously (meaning, without relative delay). They inherently entail a brief arrival delay between channel-pair data of a given sample. That delay is eliminated by holding the first channel data somewhere until the second channel data finishes arriving.
Last edited:
This is simply wrong. SDIF-2, SDIF-3 and Dual AES are simultaneous. As for demux'ing in a serial input multichannel dac, that is what the serial to parallel register is for.
Last edited:
C’mon, rfbrw. You are being argumentative for it’s own sake. I know from your many past posts that you are much smarter than this. Firstly, as you certainly know, those are Transport-to-DAC inter-box interfaces. In the same general domain as an S/PDIF connection. We have been talking about an already decoded data stream as presented directly to the DAC-chip right on the PCB. Secondly, those interfaces you cite utilize multiple connecting cables (2) to achieve high-speed parallel data transfer, incurring no relative channel delay as a result. I’m really quite surprised that there seems to be controversy over the simple fact that typical 3-wire digital audio data interfaces feature time-division channel-pair multiplexing. My argument has simply been that the delay is inconsequential for speaker listening. Now, I strangely find myself debating whether the delay exists at all.
I rather hoped to see your long experience and knowledge utilized to help guide those here who would certainly benefit from it.
Last edited:
It is not that the delay does not exist, it is that it does not exist as an issue in the general case. It only affects a narrow class of devices with separate left and right inputs and no digital filter. For the others it simply isn't an issue.
Yes, and that narrow class of devices are NOS DACs, which are the subject of the thread.
I'll summarize it for everyone: My DACs have channels without delay, because one channel contains the buffer to wait for the other channel and both channels are then latched together 😎
Yes. IIRC, they even support a mode of I2S which is 32 bits per frame, rather than the usual 64 bits. Enabling twice the input sample rate.
But the question is why bother with two dacs in the package along with hardware to deserialize both channels into the same time frame especially given the 4Fs SAA7220 which would make the delay even smaller ? Once had a CDP-101 and having only one dac was the least of its problems, not that I ever noticed. That being said I'm with Philips on this one. The delay must go.
To avoid an IC digital filter, such as the SAA7220, and simply go NOS, of course. Why it is that many audiophiles prefer NOS is a different discussion that I don’t wish to engage in right now. Suffice to say, that I hear most of the positive subjective characteristics of NOS that they do. Although, I also hear it to exhibit certain flaws.
Last edited:
The digital filter is a side issue. The point is that they went from the mono TDA1540 to the dual channel TDA1541 adding the deserializing hardware. Seems to me at no point did Philips think the delay was acceptable unlike Sony.
No, because each channel data is latched in each DAC's channel register simultaneously. At the end of a frame.
@miro1360
We tested the @mikorist DAC PCM63P-Y yesterday. We had 5 DIY DAC devices in total. Pavouk DAC 1865, Pavouk DAC AD1865 passive I/V with tube stage, Miro DAC AD1865 with Sparkos, DDDAC PCM1794 and Miro PCM63 DAC with Sparkos. Except for DDDAC which had a WaveIO USB/I2S board, the others were with a JLS USB/I2S board. The Miro PCM63 beat everything in the test, it really is one powerful machine.
Factory DACs in that price range are not even worth testing, it's simply a waste of time.
The other DACs on the test also sound great, and I could live with any of them for a while, but the PCM63 is definitely the best.
And I said that after DDDAC I won't make another DAC, and now I have to start again and make PCM63. And I just made that cheap Pavouk AD1865, to have a spare device. 🤣
We tested the @mikorist DAC PCM63P-Y yesterday. We had 5 DIY DAC devices in total. Pavouk DAC 1865, Pavouk DAC AD1865 passive I/V with tube stage, Miro DAC AD1865 with Sparkos, DDDAC PCM1794 and Miro PCM63 DAC with Sparkos. Except for DDDAC which had a WaveIO USB/I2S board, the others were with a JLS USB/I2S board. The Miro PCM63 beat everything in the test, it really is one powerful machine.
Factory DACs in that price range are not even worth testing, it's simply a waste of time.
The other DACs on the test also sound great, and I could live with any of them for a while, but the PCM63 is definitely the best.
And I said that after DDDAC I won't make another DAC, and now I have to start again and make PCM63. And I just made that cheap Pavouk AD1865, to have a spare device. 🤣
Last edited:
How about noise induced by shift registers?
if JLsound usb has same noise level for generating pcm and i2s signal.
it will be beneficial for pcm signal go into DAC as that doesn’t need to go through shift registers.
just my guess! Can guru give me a hints?
thanks
if JLsound usb has same noise level for generating pcm and i2s signal.
it will be beneficial for pcm signal go into DAC as that doesn’t need to go through shift registers.
just my guess! Can guru give me a hints?
thanks
That option exists at JLS. We haven't tried whether that sounds better or worse. I would not deviate from the Miro solution, because I currently have the option to have SPDIF/TOSLINK inputs in addition to USB. Then I have to have PCM63 with I2S input, that means 6 shift registers.
I don't care what deviation I2S has when this DAC plays better than anything else in this world. Not just a DAC, but anything - including a turntable. I'm not interested in technology. But the final result in my ears. Maybe is Voodoo magic - I don't care at all what is....🤣
Last edited:
AD1865 vs PCM63- my view
Bass area better AD1865.
Mid and high Better PCM63.
The differences are not big, but they are audible.
Bass ... AD1865 has more of it, more present, tighter, more precise. PCM63 has it, but after AD, something is missing.
Mids and highs ... somehow more natural and realistic, voices and instruments got something that AD doesn't have. Salted it with magic dust. The cymbal on the Jakob Bo - Gefion got something I can't describe. There was another song where I heard some other sounds that I didn't hear and then returned it. Now, as good as it is, it doesn't work for a bad production. It is more noticeable with PCM63 when something is not OK.
Stage ... no complaints, I'd just have to listen to it and compare. I honestly have no idea.
Conclusion ... if I hadn't heard the PCM63 and compared it directly with the AD1865 I would think it is the same. Well, I heard it. Since I have boards, chips, and I'm missing a few capassitors, resistors ... I'll make it.
miro1360, thanks again for the excellent dacs