Uh...XMOS USB is an I2S source, as is RPi.
Perhaps you are speaking of some other difference?
Last edited:
Costs
I got good results too with the MI 10000 mAH power bank. And they are way cheaper, if you can do the frequent charging and connecting !!
I got good results too with the MI 10000 mAH power bank. And they are way cheaper, if you can do the frequent charging and connecting !!
Last edited:
Well, let me think about that for a moment....
If I have RPI with a USB disk, my music passes though that one USB connection, is converted to I2S by RPi and sent to the dac.
If I have a PC with a SATA disk and use a USB board, my music passes through that one USB connection, is converted to I2S by the USB board and set to the dac.
....seems to me I2S itself is equally direct either way
I tend to focus on the music replay chain from the point where the driver processing occurs that converts the .WAV or .FLAC file (or whatever) into the signal that will eventually be the music TO the point where that signal enters the DAC (typically in I2S or the equivalent DSD format).
Things that occur before that point, while in my experience they DO have an impact on the ultimate SQ, seem to have a much less crucial impact. So I focus on the process that performs the actual conversion of the data file into a music signal and from there to the input of that signal into the DAC.
I also tend towards simpler chains with minimal transforms as that is what has produced the best results for me in the past. That doesn't preclude more complex chains (such as USB), but in my experience, getting them to produce good results takes a lot more effort, HW, and cost.
In the case of an RPi I2S output, the driver processing is of course happening on the RPI and results in the I2S signal. In computer processing reality (I used to be a mainframe Assembly Language programmer years ago!), it is REALLY NOT simple, but from a systems prespective, it seems so.
That I2S output might go through an Isolator (which will add some jitter due to how they work) and hopefully after that a reclocker like a Kali (which is designed to reduce jitter), and then onto the DAC chip. This, BTW, is where a lot of I2S done right happens... you want good signal integrity from the output of the reclocker (if you use one) or the RPi (if no isolator or reclocker), which means short connections (3" max IMHO, but in my experience the shorter the better) with good, hopefully impedence-controlled connections like u.fl.
IF one can't keep that connection to the DAC at that 3" or shorter length AND do it right, then I think either I2S over LVDS (ala' PS Audio I2S over HDMI or Twisted Pear Teleporters) or USB is better.
Of the 2, the I2S over LVDS is the simpler transform AND what I would use if I have to. The process uses chips to convert the I2S signal to a balanced one at the transmit side and reverse that at the receive side. Just like Isolators, these transforms add jitter, so again best to use a reclocker at the output side just before the DAC.
USB is a much more complex situation. Honestly, I am less well versed in the details here, I know enough to be dangerous, but I am no expert here. My UNDERSTANDING is that transmitting I2S via USB involves the following chain:
USB output driver -> USB transmit OpSys processing -> USB transmit HW processing -> USB cable -> USB Receiver (I believe XMOS does both the USB physical receiving and conversion process on-chip) -> I2S.
Then because this process produces both noise and jitter, on better USB->I2S HW setups you will have built-in isolators and reclocking before the I2S output.
Because of the rigor put into developing and implementing USB, this is definitely a bit-perfect process. BUT not necessarily a transparent process from an audio perspective.
There have been a lot of developments to improve USB audio transmission. The move from Isosychronous to Asynchronous USB processing was a significant one and marked a watershed in USB Audio quality. BUT other things in the transmission still mattered such as the USB cable. Then there was an awakening of understanding some of the physical processing that occurred during USB. I follow audio engineer John Swension a lot and found his posts on this very illuminating. Some of this is on the Bottlehead audio forums and some on the now Audiophile Style (used to be Computer Audio) forums. John had just designed a USB & S/PDIF input DAC for Bottlehead and went to some significant lengths to make both inputs immune to the quality of the sending HW and intervening cables. He details those lengths in posts on the Bottlehead forum... search there to find that info.
AND he found that even with those techniques, those things still mattered in the ultimate SQ, just a bit less than before. He got down into the actual physical processing happening at the USB receiver end and found some mechanisms that impacted the output SQ. Those posts were among the early ones John posted on Audiophile Style... search there to find them. As he began to understand those mechanisms, he came up with a way to reduce their impact on the resulting SQ out of a USB input DAC and that resulted in the Uptone Audio Regen which I believe was the first active USB cleanup device. This started an explosion of similar and otherwise different devices, some with isolation, some with regeneration, and so on. That eventually resulted in full processing endpoints with Ethernet inputs and USB outputs starting with the Sonore MicroRendu (which John also had a hand in). AND that was followed by improved models from Sonore and a host of other competing units. AND there have always been scads of various USB cables including some that cost as much as one of the amps I have in my systems (and they ain't cheap!).
AND if you follow some of the threads on Audiophile Style, all of that still hasn't made USB perfect yet.
I2S as an output music signal only works REALLY WELL when you can position the DAC close to the generating device. I do DIY DACs and can do that, so it is great for my purposes. MOST computer audio involves DACs in a box with an input that you connect to a remote sending device (computer with USB output, CD drive with S/PDIF, various specialized conversion boxes, whatever). In this world USB is more universal (AND I2S over LVDS/HDMI is sometimes used, but comparatively rare). BUT what it takes to make it really work well IMHO is a lot more effort and cost than it takes to get a good I2S signal into a closely-coupled DIY DAC.
Hence my preference for I2S. BUT YMMV and I2S is not the solution for the masses. AND USB can be made to work quite well nowadays.
This seems more like my 50 cents this time!
Greg in Mississippi
Things that occur before that point, while in my experience they DO have an impact on the ultimate SQ, seem to have a much less crucial impact. So I focus on the process that performs the actual conversion of the data file into a music signal and from there to the input of that signal into the DAC.
I also tend towards simpler chains with minimal transforms as that is what has produced the best results for me in the past. That doesn't preclude more complex chains (such as USB), but in my experience, getting them to produce good results takes a lot more effort, HW, and cost.
In the case of an RPi I2S output, the driver processing is of course happening on the RPI and results in the I2S signal. In computer processing reality (I used to be a mainframe Assembly Language programmer years ago!), it is REALLY NOT simple, but from a systems prespective, it seems so.
That I2S output might go through an Isolator (which will add some jitter due to how they work) and hopefully after that a reclocker like a Kali (which is designed to reduce jitter), and then onto the DAC chip. This, BTW, is where a lot of I2S done right happens... you want good signal integrity from the output of the reclocker (if you use one) or the RPi (if no isolator or reclocker), which means short connections (3" max IMHO, but in my experience the shorter the better) with good, hopefully impedence-controlled connections like u.fl.
IF one can't keep that connection to the DAC at that 3" or shorter length AND do it right, then I think either I2S over LVDS (ala' PS Audio I2S over HDMI or Twisted Pear Teleporters) or USB is better.
Of the 2, the I2S over LVDS is the simpler transform AND what I would use if I have to. The process uses chips to convert the I2S signal to a balanced one at the transmit side and reverse that at the receive side. Just like Isolators, these transforms add jitter, so again best to use a reclocker at the output side just before the DAC.
USB is a much more complex situation. Honestly, I am less well versed in the details here, I know enough to be dangerous, but I am no expert here. My UNDERSTANDING is that transmitting I2S via USB involves the following chain:
USB output driver -> USB transmit OpSys processing -> USB transmit HW processing -> USB cable -> USB Receiver (I believe XMOS does both the USB physical receiving and conversion process on-chip) -> I2S.
Then because this process produces both noise and jitter, on better USB->I2S HW setups you will have built-in isolators and reclocking before the I2S output.
Because of the rigor put into developing and implementing USB, this is definitely a bit-perfect process. BUT not necessarily a transparent process from an audio perspective.
There have been a lot of developments to improve USB audio transmission. The move from Isosychronous to Asynchronous USB processing was a significant one and marked a watershed in USB Audio quality. BUT other things in the transmission still mattered such as the USB cable. Then there was an awakening of understanding some of the physical processing that occurred during USB. I follow audio engineer John Swension a lot and found his posts on this very illuminating. Some of this is on the Bottlehead audio forums and some on the now Audiophile Style (used to be Computer Audio) forums. John had just designed a USB & S/PDIF input DAC for Bottlehead and went to some significant lengths to make both inputs immune to the quality of the sending HW and intervening cables. He details those lengths in posts on the Bottlehead forum... search there to find that info.
AND he found that even with those techniques, those things still mattered in the ultimate SQ, just a bit less than before. He got down into the actual physical processing happening at the USB receiver end and found some mechanisms that impacted the output SQ. Those posts were among the early ones John posted on Audiophile Style... search there to find them. As he began to understand those mechanisms, he came up with a way to reduce their impact on the resulting SQ out of a USB input DAC and that resulted in the Uptone Audio Regen which I believe was the first active USB cleanup device. This started an explosion of similar and otherwise different devices, some with isolation, some with regeneration, and so on. That eventually resulted in full processing endpoints with Ethernet inputs and USB outputs starting with the Sonore MicroRendu (which John also had a hand in). AND that was followed by improved models from Sonore and a host of other competing units. AND there have always been scads of various USB cables including some that cost as much as one of the amps I have in my systems (and they ain't cheap!).
AND if you follow some of the threads on Audiophile Style, all of that still hasn't made USB perfect yet.
I2S as an output music signal only works REALLY WELL when you can position the DAC close to the generating device. I do DIY DACs and can do that, so it is great for my purposes. MOST computer audio involves DACs in a box with an input that you connect to a remote sending device (computer with USB output, CD drive with S/PDIF, various specialized conversion boxes, whatever). In this world USB is more universal (AND I2S over LVDS/HDMI is sometimes used, but comparatively rare). BUT what it takes to make it really work well IMHO is a lot more effort and cost than it takes to get a good I2S signal into a closely-coupled DIY DAC.
Hence my preference for I2S. BUT YMMV and I2S is not the solution for the masses. AND USB can be made to work quite well nowadays.
This seems more like my 50 cents this time!
Greg in Mississippi
Last edited:
Since the is really an Allo thread, I better keep this short: For those wanting to try USB I2S done right, there is a very will designed one with galvanic isolation, reclocking, and that comes with NDK SDA clocks. Supports native DSD512 and below. It can also run from external clocks. Very low jitter in my comparison listening tests vs other vendor solutions, so its the one I am using for AK4499 development. I2SoverUSB - I2S over USB Audio
Thank you very much Greg,
That certainly is much more than 50cents to me
I know very little about it and appreciate that you take the time to educate us. I think I can use ufl since it's available in kali. Let me absorb what you said and think this through carefully. I also think it might be worth the effort. No venture no gain.
Mark. It's an Allo question since I either would wait for the new usbridge signature or go with kali. Both obviously have its merits and downfall. I hope you don't mind
Thank you and best regards
Anthony
That certainly is much more than 50cents to me
I know very little about it and appreciate that you take the time to educate us. I think I can use ufl since it's available in kali. Let me absorb what you said and think this through carefully. I also think it might be worth the effort. No venture no gain.
Mark. It's an Allo question since I either would wait for the new usbridge signature or go with kali. Both obviously have its merits and downfall. I hope you don't mind
Thank you and best regards
Anthony
Easy question , hard answer
RPI and Kali-i2s is a very robust solution .
Yet you still need a music source (be it ethernet or USb sticks).Please note that all transfer mechanisms (ethernet and usb) have error correction
Kali in the path of i2s lignes cleans nicely the noisy RPI i2s lignes but its capped at 384Khz/PCM and some DOP
Usbridge sends the data through USB . Xmos USB is using asinc USB transfer (so no error correction ) but our implementation is very robust and hub IC is the closest to the PCB edge (shortest path) to the Xmos USB so technically they should be very similar. Xmos i2s lignes are very clean ( optimized by us). Of course the Xmos allows for theoretical speed of 768Khz and DSD 512
I hope it answers your question
Thank you so much for sharing your tweaks!
Didn't go as far as you did, but I can attest that even my modest modifications were able to squeeze more performance out of the great design:
1. Removed the stock cables and soldered on decent Canare terminated with gold-plated Type-C plugs.
2. Replaced the generic 3-in-1 switch with a gold-plated-copper one from Aucharm.
3. DIY Oyaide power cord.
4. Applied Japanese "Fo.Q TA-52" material for vibration control.
Wurth engineers are still working on Nirvana transformer..
I think Nirvana has around 74% efficency and Shanti (not an exact figure) at around 58%
@cdsgames,
Some very good points. AND the max sampling rate limitations of I2S from an RPi is one that I often ignore as I currently don't do upsampling and almost all of my source materials is ripped CDs. USB supports much higher rates!
@markw4, thanks for that post. The JLSounds unit is a very good one, as are their other offerings (I have some here). OTOH, many of the people reporting SQ improvements from use of the better USB endpoints (like the Allo, Sonore, and SOTM ones) and USB cleanup devices (such as the Uptone Audio, SOTM, Schitt, and others) have USB DACs with well-implemented USB interfaces such as these.
@AnthonyA asked for my asssessment of I2S done right. I provided my view. AND a lot of that is based on how I personally do digital audio where direct I2S works well. BUT for every I2S-connected DAC in the world, I suspect there are 100 USB-connected DACs or more. USB won in the marketplace a LONG time ago.
AND back to the subjects of this thread, I think the new USBBridge Signature and Revolution DAC will be worthy additions (and very cost-performance leaders) to this marketplace. AND probably sell a lot more units than the also very worthy Katana.
Very much looking forward to hearing them in my setups and seeing how they perform against my 'tweaked and optimized' Katana setup.
Greg in Mississippi
P.S. Amp? Interesting!
Some very good points. AND the max sampling rate limitations of I2S from an RPi is one that I often ignore as I currently don't do upsampling and almost all of my source materials is ripped CDs. USB supports much higher rates!
@markw4, thanks for that post. The JLSounds unit is a very good one, as are their other offerings (I have some here). OTOH, many of the people reporting SQ improvements from use of the better USB endpoints (like the Allo, Sonore, and SOTM ones) and USB cleanup devices (such as the Uptone Audio, SOTM, Schitt, and others) have USB DACs with well-implemented USB interfaces such as these.
@AnthonyA asked for my asssessment of I2S done right. I provided my view. AND a lot of that is based on how I personally do digital audio where direct I2S works well. BUT for every I2S-connected DAC in the world, I suspect there are 100 USB-connected DACs or more. USB won in the marketplace a LONG time ago.
AND back to the subjects of this thread, I think the new USBBridge Signature and Revolution DAC will be worthy additions (and very cost-performance leaders) to this marketplace. AND probably sell a lot more units than the also very worthy Katana.
Very much looking forward to hearing them in my setups and seeing how they perform against my 'tweaked and optimized' Katana setup.
Greg in Mississippi
P.S. Amp? Interesting!
I will share pics and info before the end of the week
Sorry all but I have my hand full with the new projects ..
Amp anyone ?
Amp??
im interested.. is it the monoblock Volt D?
Much interest in Revolution DAC here too
A few interesting details he shared here(scroll down):
VOLUMIO soundcheck invitation for Hamburg, Germany : Diy Volumio Projects - Page 18
Matt
Revolution DAC PCB was sent to PCB house for 10 units of samples (we hope those are the last samples before production).We should get them in 2 weeks . From those we will send samples for SQ testing to the usual "golden ears"
Everything works great , THD is 126.8 , supercaps ... on opamp rails... (never been done before), no asrc , pre IV filtering (and it was not an easy task to keep the great THD while filtering the DAC output)
AMP , I have already a good sample of the tpa3255 and parallel I am designing a ground breaking discrete class D and will chose between the 2 based on SQ testing.
Everything works great , THD is 126.8 , supercaps ... on opamp rails... (never been done before), no asrc , pre IV filtering (and it was not an easy task to keep the great THD while filtering the DAC output)
AMP , I have already a good sample of the tpa3255 and parallel I am designing a ground breaking discrete class D and will chose between the 2 based on SQ testing.
Last edited: