1 is enough for me, thank you 
4 doesn't hurt anyones interests too imho.
Rigisystems Evaluation board looks good, software form pc side is ready to go and pricing is reasonable. Just having some concerns about 2 clocks. I can see 2 crystals near fpga, but second one is probably used for clocking spdif in ? So additional licensing may be needed for fpga source in order to solve this issue.
Couldn't find any open source projects in close to completion state, so if going for any other Xilink + FX2 demo board, everything must done almost from scratch, as only some bits and pieces are available for reference.
4 doesn't hurt anyones interests too imho. Rigisystems Evaluation board looks good, software form pc side is ready to go and pricing is reasonable. Just having some concerns about 2 clocks. I can see 2 crystals near fpga, but second one is probably used for clocking spdif in ? So additional licensing may be needed for fpga source in order to solve this issue.
Couldn't find any open source projects in close to completion state, so if going for any other Xilink + FX2 demo board, everything must done almost from scratch, as only some bits and pieces are available for reference.
1 is enough for me, thank you
Rigisystems Evaluation board looks good, software form pc side is ready to go and pricing is reasonable. Just having some concerns about 2 clocks. I can see 2 crystals near fpga, but second one is probably used for clocking spdif in ? So additional licensing may be needed for fpga source in order to solve this issue.
Couldn't find any open source projects in close to completion state, so if going for any other Xilink + FX2 demo board, everything must done almost from scratch, as only some bits and pieces are available for reference.
well these guys had a private wiki last time i 'eard. They wanted to make it public once its finished. Oh and i meant 2stereo outs to ease up crossover price.
Last edited:
The oscillators on Rigisystems board are obviously for the 44.1 kHz and 48 kHz based sample rates.
With the available 18 output channels you could for an example create an active 3x 3-way + 4x 2-way (=18 channels I2S) 7.1 speaker system if you used the S/PDIF output for the subwoofer channel. If that is not enough you can use multiple boards (supported by the drivers).
With the available 18 output channels you could for an example create an active 3x 3-way + 4x 2-way (=18 channels I2S) 7.1 speaker system if you used the S/PDIF output for the subwoofer channel. If that is not enough you can use multiple boards (supported by the drivers).
I like the board in Mhelin link a lot.
8 x 8 channels audio @ 192KHz ??
24 bits
support for win7 and Apple
asio support
What's not to like about this? Some more details on the clock would be good..
USB Audio - Rigisystems
edit: I've just noticed that the system is modular. That is, it can prove to be the ultimate multichannel, active crossover dream system.
8 x 8 channels audio @ 192KHz ??
24 bits
support for win7 and Apple
asio support
What's not to like about this? Some more details on the clock would be good..
USB Audio - Rigisystems
edit: I've just noticed that the system is modular. That is, it can prove to be the ultimate multichannel, active crossover dream system.
Last edited:
Folks,
Okay, I have been working for a good time on the other side of the fence, so most of the items discussed here (and quite a few not discussed) have passed my hands (and Test systems, including an AP2 - same as JA uses at Stereophile).
Here is my suggestion to all real DIY'ers.
1) Forget I2S, it is in theory a good idea, but if you must isolate the the signal it becomes a nightmare. Use SPDIF, it is quite easy to isolate and the Musiland Monitor 01 USD is cheap and readily obtained and with some mods it makes a swell USB2SPDIF converter.
2) SPDIF Jitter is nowadays not so difficult to solve. We have ready 192KHz capable receivers. Adding a FIFO and a secondary PLL is not hard, as Jos Van Eindhofen has shown.
3) Even without any serious secondary PLL, using the new "precision mode" in the Musiland Driver and feeding the Musiland a precision clock with tight tolerance plus adding the same to a receiver's out in slave mode will lock out jitter.
There are other options, including the XMOS system, others mentioned here and some as of yet unreleased (to the general public and available under strict NDA only) chips, so in the long run the issue we are debating will just go away.
Ciao T
Okay, I have been working for a good time on the other side of the fence, so most of the items discussed here (and quite a few not discussed) have passed my hands (and Test systems, including an AP2 - same as JA uses at Stereophile).
Here is my suggestion to all real DIY'ers.
1) Forget I2S, it is in theory a good idea, but if you must isolate the the signal it becomes a nightmare. Use SPDIF, it is quite easy to isolate and the Musiland Monitor 01 USD is cheap and readily obtained and with some mods it makes a swell USB2SPDIF converter.
2) SPDIF Jitter is nowadays not so difficult to solve. We have ready 192KHz capable receivers. Adding a FIFO and a secondary PLL is not hard, as Jos Van Eindhofen has shown.
3) Even without any serious secondary PLL, using the new "precision mode" in the Musiland Driver and feeding the Musiland a precision clock with tight tolerance plus adding the same to a receiver's out in slave mode will lock out jitter.
There are other options, including the XMOS system, others mentioned here and some as of yet unreleased (to the general public and available under strict NDA only) chips, so in the long run the issue we are debating will just go away.
Ciao T
Last edited:
ThorstenL, first of all, thank you for your input. Not a lot of us here will start to tear of us t-shirts arguing that SPDIF will not yield comparable results if proper measures are implemented. It's just that there are some people, like me for instance , that just feels gratuitousness of whole muxing/demuxing stage of already usable I2S, for a long time now. Others may have their own reasonings, but in the end, this is the path we chose to follow.
I have a couple of questions about your experience with Musiland, that you may be able to answer:
- What is the difference between Musiland clocking of I2S, using it's 1 crystal and would imagine couple internal DCM stacked vs. M2Tech and others, using 2 separate clocks in terms of skew jitter generation ? Maybe you've done any measurements ?
- Maybe you have some suggestions on how this could be solved after isolation in our case ? As synchronous reclocking is impossible and using asynchronous has it's own limitations of effectiveness.
EUVL, price for Rigisistems is 204USD per 1 of 10 m.o.q units. This will be even lower for larger quantities as always. Smaller company then xmos or m2tech, so maybe more willingness for negotiations. If it has 2 clocks, worth contacting imho.
, that just feels gratuitousness of whole muxing/demuxing stage of already usable I2S, for a long time now. Others may have their own reasonings, but in the end, this is the path we chose to follow. I have a couple of questions about your experience with Musiland, that you may be able to answer:
- What is the difference between Musiland clocking of I2S, using it's 1 crystal and would imagine couple internal DCM stacked vs. M2Tech and others, using 2 separate clocks in terms of skew jitter generation ? Maybe you've done any measurements ?
- Maybe you have some suggestions on how this could be solved after isolation in our case ? As synchronous reclocking is impossible and using asynchronous has it's own limitations of effectiveness.
EUVL, price for Rigisistems is 204USD per 1 of 10 m.o.q units. This will be even lower for larger quantities as always. Smaller company then xmos or m2tech, so maybe more willingness for negotiations. If it has 2 clocks, worth contacting imho.
I agree that this would be off the budget badly. Hiface goes for 120EU smth. ?, so if more than that + shipping, better stick with hiface as an option. But we never know for sure if not contacting.
Of course there is an option of buying 10psc. and then sharing drivers and fw image, but if someone from rigisystems is reading this, then we are already doomed
Of course there is an option of buying 10psc. and then sharing drivers and fw image, but if someone from rigisystems is reading this, then we are already doomed
Hi,
It is not so much that part. It is quite important to isolate the PC from the rest. With SPDIF you can run line through a decent transformer or two and still get a signal that a receiver can lock upon (and you handle the jitter in a different way).
With I2S you need to isolate four signals, one of them very fast (usually 256 * Fs = 49.152MHz) and then reclock from the very fast signal. You just move the complexity from one area which is well understood and which is amenable to DIY into one quite a little harder.
I don't disagree. But I am looking at the money you guys want to pay and what else you would need to make this work properly. So I come up with my suggestion. Even the HiFace seems outside your target. And by the time you have fixed the two clocks on that and all else that is sub-optimal you will have paid probably the same money again and then you still need to isolate the I2S signal.
I completely understand what you want and I have implemented in commercial equipment (though not for 192KHz yet - solutions are too immature)
Yes. I come up somewhere around 100...150pS rms (assuming I interpret the trace in the AP2 right) with heavily modified examples. But that is of course a meaningless figure, just as THD is and thye modifications cost a fair bit as well.
Well, you need to use MCK for recocking. Isolation of a clock running from 12MHz to 50MHz without introducing a ton of jitter is quite non-trivial. Most commercial opto and capacitive isolators are hopeless in terms of jitter, Transformers have their own problems.
So I guess you need to look at a Video Line driver designed for HD analog video and try to find a matching isolation transformer (so both have enough bandwidth) and then optimise the resulting circuit to minimise jitter. There are also some dedicated clock distribution IC's that have low jitter and likely matching transformers for communication systems, but similar caveats apply.
I cannot think of much else that will work well enough to be worth bothering with. Hence my suggestion of SPDIF and fixing it's problems.
Ciao T
It is not so much that part. It is quite important to isolate the PC from the rest. With SPDIF you can run line through a decent transformer or two and still get a signal that a receiver can lock upon (and you handle the jitter in a different way).
With I2S you need to isolate four signals, one of them very fast (usually 256 * Fs = 49.152MHz) and then reclock from the very fast signal. You just move the complexity from one area which is well understood and which is amenable to DIY into one quite a little harder.
It's just that there are some people, like me for instance
I don't disagree. But I am looking at the money you guys want to pay and what else you would need to make this work properly. So I come up with my suggestion. Even the HiFace seems outside your target. And by the time you have fixed the two clocks on that and all else that is sub-optimal you will have paid probably the same money again and then you still need to isolate the I2S signal.
I completely understand what you want and I have implemented in commercial equipment (though not for 192KHz yet - solutions are too immature)
Yes. I come up somewhere around 100...150pS rms (assuming I interpret the trace in the AP2 right) with heavily modified examples. But that is of course a meaningless figure, just as THD is and thye modifications cost a fair bit as well.
Well, you need to use MCK for recocking. Isolation of a clock running from 12MHz to 50MHz without introducing a ton of jitter is quite non-trivial. Most commercial opto and capacitive isolators are hopeless in terms of jitter, Transformers have their own problems.
So I guess you need to look at a Video Line driver designed for HD analog video and try to find a matching isolation transformer (so both have enough bandwidth) and then optimise the resulting circuit to minimise jitter. There are also some dedicated clock distribution IC's that have low jitter and likely matching transformers for communication systems, but similar caveats apply.
I cannot think of much else that will work well enough to be worth bothering with. Hence my suggestion of SPDIF and fixing it's problems.
Ciao T
Hi,
I have. At Audio Asylum -- Audio Reviews, Audiophile Forums and Stereo Reviews in the PC-Audio Asylum. Just for musiland and my ID...
It is 192KHz capable and can be purchased cheaply with versions that output SPDIF or that allow I2S to be tapped off. Modifications needed are substantial in both cases, but not hard for a DIY'er. If you read above you notice that I recommend for DIY to simply buy the Musiland and to use the SPDIF Version as this makes insulating the PC from the rest of the audio system easy.
Ciao T
I have. At Audio Asylum -- Audio Reviews, Audiophile Forums and Stereo Reviews in the PC-Audio Asylum. Just for musiland and my ID...
It is 192KHz capable and can be purchased cheaply with versions that output SPDIF or that allow I2S to be tapped off. Modifications needed are substantial in both cases, but not hard for a DIY'er. If you read above you notice that I recommend for DIY to simply buy the Musiland and to use the SPDIF Version as this makes insulating the PC from the rest of the audio system easy.
Ciao T
?
Hi Thorsten,
i would like to have more info about this:
I don't understand well (also for problem with english language), do you mean the same "precision clock" for musiland and the receiver in slave mode?
The musiland use 24MHz and the slaved receiver would need some multiple (128, 256) of Fs so i feel i missed something.
Ciao
Andrea
i would like to have more info about this:
I don't understand well (also for problem with english language), do you mean the same "precision clock" for musiland and the receiver in slave mode?
The musiland use 24MHz and the slaved receiver would need some multiple (128, 256) of Fs so i feel i missed something.
Ciao
Andrea
Hi,
The issue is that if you us a common receiver in "slave mode" you need to feed it a clock that matches that input clock average speed very well.
If you use a clock with a guaranteed frequency deviation of no more than 1ppm for all three clocks (44.1KHz base clock, 48KHz base clock and 24MHz clock for the Musiland) you loose or repeat in the worst case 2 samples per million samples. But jitter is fully locked out.
Of course, you need to select the clock and division ration for the receiver side correctly for this to work.
Ciao T
The issue is that if you us a common receiver in "slave mode" you need to feed it a clock that matches that input clock average speed very well.
If you use a clock with a guaranteed frequency deviation of no more than 1ppm for all three clocks (44.1KHz base clock, 48KHz base clock and 24MHz clock for the Musiland) you loose or repeat in the worst case 2 samples per million samples. But jitter is fully locked out.
Of course, you need to select the clock and division ration for the receiver side correctly for this to work.
Ciao T
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.