glt,
maybe we should keep the technical stuff to this thread as it's easier for me to post pics & the audience on headfi aren't that interested in DIY, I feel?
You raised a question about the Musiland outputting only 88.2Khz or 96KHz clocks over I2S no matter what the original sample rate of the audio file being played!. I did some scope shots of the I2S MCLK (system clock) line that feeds my ES9022 DAC. People will have to help me interpret these pics as I'm a newbie with a scope.
So what I did was put the red probe on the MCLK line & the Green probe on ground. The Y axis was set at 2V/Div & X axis at 20ns/div.
So here'e the shots:
- first one is with nothing playing - I make peak to through about 5 divs & so peak to peak just > 10 divs = 220nS = 44.1KHz?
- next is 44.1KHz file playing - does this look a bad shape? I don't have any R in series with the signal lines - again just > 10 divs = 220ns = 44.1KHz
- next is 88KHz - peak to peak slightly > 4 divisions = >80nS = ?KHz
This has me confused
continued on next post
maybe we should keep the technical stuff to this thread as it's easier for me to post pics & the audience on headfi aren't that interested in DIY, I feel?
You raised a question about the Musiland outputting only 88.2Khz or 96KHz clocks over I2S no matter what the original sample rate of the audio file being played!. I did some scope shots of the I2S MCLK (system clock) line that feeds my ES9022 DAC. People will have to help me interpret these pics as I'm a newbie with a scope.
So what I did was put the red probe on the MCLK line & the Green probe on ground. The Y axis was set at 2V/Div & X axis at 20ns/div.
So here'e the shots:
- first one is with nothing playing - I make peak to through about 5 divs & so peak to peak just > 10 divs = 220nS = 44.1KHz?
- next is 44.1KHz file playing - does this look a bad shape? I don't have any R in series with the signal lines - again just > 10 divs = 220ns = 44.1KHz
- next is 88KHz - peak to peak slightly > 4 divisions = >80nS = ?KHz
This has me confused
continued on next post
Attachments
Last edited:
For 44.1:
If you have a period of 220nsec, Frequency is 1/200nsec = 0.0045X10^9 or 4.5MHz. Assuming 24 bit data, two channels is 48 bit, data rate is 4.5/48=93750 Hz or 96K
for 96K Input, period is 80nsec, f=1/80nsec = .0125X10^9 or 12.5MHz. Assuming 24 bit data, 2 channels is 48 bit, data rate is 12.5/48= 260K... That's too much...
Try measuring the LRClock (the left-right channel clock) line. That runs much slower. Also can you force 24 bit from the player?
In the MINI:
Pin3 is LRCK
Pin4 is BCK
Pin5 is Data in
Pin6 is MCK (Is this what you measured?)
LRCK tells left channel and right channel
BCK clocks the data into the DAC
If you have a period of 220nsec, Frequency is 1/200nsec = 0.0045X10^9 or 4.5MHz. Assuming 24 bit data, two channels is 48 bit, data rate is 4.5/48=93750 Hz or 96K
for 96K Input, period is 80nsec, f=1/80nsec = .0125X10^9 or 12.5MHz. Assuming 24 bit data, 2 channels is 48 bit, data rate is 12.5/48= 260K... That's too much...
Try measuring the LRClock (the left-right channel clock) line. That runs much slower. Also can you force 24 bit from the player?
In the MINI:
Pin3 is LRCK
Pin4 is BCK
Pin5 is Data in
Pin6 is MCK (Is this what you measured?)
LRCK tells left channel and right channel
BCK clocks the data into the DAC
Last edited:
jkenny, last time I uses a scope was like 30 years ago!. (Thanks for doing this)
Anyways, there is still hope. According to the DAC spec (your DAC) these are the specifications for the MCK
so if we assume 128fs,
Then for 44.1 MCK is 5.6 MHz and for 96K, MCK is 12.2 MHz which is very close to your results. But to make sure, we need BCK and LRCK
The wave pattern looks awful! but maybe is your scope
Anyways, there is still hope. According to the DAC spec (your DAC) these are the specifications for the MCK
so if we assume 128fs,
Then for 44.1 MCK is 5.6 MHz and for 96K, MCK is 12.2 MHz which is very close to your results. But to make sure, we need BCK and LRCK
The wave pattern looks awful! but maybe is your scope
Last edited:
Decided to take out my state-of-the-art 40 buck scope and some measurements myself. The highest frequency I can measure is the LRCK which coincides with the sample rate.
As it turned out, the Musiland sends native sample rate.
But based on jkenny measurements of the SCK, the system clock, the frequencies are not supported by higher end DACS such as the Wolfson WM8741 according to the spec. Music comes out but maybe that is why some of the upsampling doesn't work in the Wolfson DAC
In any case, it seems matching I2S is not so trivial. One must ensure that the system clock rates are supported in order to take full advantage of the DAC. Thus maybe spdif is a better solution until a more generic I2S solution is developed.
Here are some pictures:
The little scope is pretty handy and accurate within the resolution of the device
As it turned out, the Musiland sends native sample rate.
But based on jkenny measurements of the SCK, the system clock, the frequencies are not supported by higher end DACS such as the Wolfson WM8741 according to the spec. Music comes out but maybe that is why some of the upsampling doesn't work in the Wolfson DAC
In any case, it seems matching I2S is not so trivial. One must ensure that the system clock rates are supported in order to take full advantage of the DAC. Thus maybe spdif is a better solution until a more generic I2S solution is developed.
Here are some pictures:
The little scope is pretty handy and accurate within the resolution of the device
I've done some more work & put a local clock on the Sabre DAC - effectively this makes it re-clocking but done the ESS way i.e. I think they can do this because they use Time Domain jitter reduction & this is just another clock timing everything so it becomes a synchronous clock. So I'm not routing the I2S MCLK signal through to the DAC but using the local clock instead.
Anyway, there seems to be an improvement in sound using Crystek clocks. First I tried an 80MHz clock but got background noise & a bit of an edginess to the music. Then I tried a 20 MHz Crystek - no noise, dead silent & music played beautifully except when trying to play 24/192KHz where there's some ringing (it seems to me).
It would appear some that somewhere between 24MHz & 80MHz there is an optimal clock speed!
Now, I have a question about using a Sabre DAC in this way - does the quality of the clock in the transport have any sonic influence?
I know that people have reported differences, in using Sabre DACs, between different sources so can anybody explain what is happening?
I have I2S from a Musiland to Sabre DAC & wondered if I2S from HiFace (with lower jitter clocks) would make any sonic difference in this case as it's being re-clocked?
Anyway, there seems to be an improvement in sound using Crystek clocks. First I tried an 80MHz clock but got background noise & a bit of an edginess to the music. Then I tried a 20 MHz Crystek - no noise, dead silent & music played beautifully except when trying to play 24/192KHz where there's some ringing (it seems to me).
It would appear some that somewhere between 24MHz & 80MHz there is an optimal clock speed!
Now, I have a question about using a Sabre DAC in this way - does the quality of the clock in the transport have any sonic influence?
I know that people have reported differences, in using Sabre DACs, between different sources so can anybody explain what is happening?
I have I2S from a Musiland to Sabre DAC & wondered if I2S from HiFace (with lower jitter clocks) would make any sonic difference in this case as it's being re-clocked?
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.