Thanks @wlowes, I just repeated the same experiment and found the same thing. The only reason I repeated it was because the DAC wouldn't initialise without the other clock in place. I put it back, powered up, warmed things up and repeated...same thing. Definitely sounds better. I'm sure Ian said that the board was able to be used with only one clock. I'm using it to generate the mclk for the dac boards so maybe dacs are trying to initialise before Ian's board switches to the clock that's there. I'll have to take a look at the manuals again.
@gebla
You need to feed spdif from the TV straight to your DAC. If using Ians ESS dac then there is an spdif input on there to which you can wire your own optical port. You need to disable auto mute on the ess controller .
In other news ....a few updates and observations.
Ian kindly replaced the ESS controller for my 2nd streamer build for my brother. This one went into an aluminium chassis, runs Pi less. Chromecast Audio optical into ReceiverPi and runs fifo on 1x NDK SDA for 44mhz family. Running 1.5f supercap via LT1963 to the fifo clean , another 1.5f over the clock itself and a 1.5f fifo dirty powering that and the receiver.
So there is room for some improvement to the psus there. DAC has 3 separate rails. Running the Chinese 600:10k output transformers with 510r across the secondaries as per Ians recommendations for the Lundahls.
My brother thinks this build sounds better than the 1st one. Main differences being heavy aluminium chassis and the passive IV.
I don't really understand the transformers! Initially I had then wired wrong. Took ages to realise. I had them wired backwards so they were stepping down instead of up I think. They are 1:4 I believe. This would explain the major lack of output level we experienced. Strangely though I had also wired the SE outputs using the +/- terminals, no ground reference. Still worked! When I sorted this lot out and wired correctly I also wired in XLRs aswell as the RCA. The ground wire to the XLR seemed to induce a hum out of the RCA so they are disconnected at the moment. My brother uses the SE output anyway.
One observation is that the system will not work from the BBC Sounds App. So all the music is cast from the Google Chromecast. Mainly Spotify which works fine. Radio Paradise app works fine. When we try to cast from the BBC app the receiverpi acknowledges the optical input but then the fifo won't lock. I don't understand this. Can the signal from the BBC app be different/lower quality.? Wierd.
Getting my brothers finished has given me some time to have a look at mine again. Being in my bro's possession for nearly a year and powerd 24/7 the first thing I noticed was the controller display is loosing pixels and looking a bit tired. Anyone else seen this?
I've also realised i can save a psu as I was powering fifo dirty, and receiver separately. Which in turn powers the Rpi in my build. So I've saved some space and removed then separate psu I had to the Receiver. It now is powered via gpio from the Fifo. I know I can remove an inductor on the Reciever for a separate psu and possible SQ improvement. Having not removed this my additional psu was wasted.
I am plucking up the courage to add some 310F caps to the clock (s). I do leave it powered up 24/7 so once charged they shouldn't add any inconvenience.
On clocks....I ran with one for a while but at the moment I have 45mhz Crystal and 49mhz SDA. I was hoping to use maybe just one and upsample everything to that family. But for some reason when I play from Spotify though either Moode or Volumio in the Rpi, despite the software being set to 300 odd Mhz it always comes out on the display as 44. It won't upsample a Spotify feed.
Have been trying Tidal MQA and also Amazon HD. I don't know the best way to get these higher bitrate streams into the system. You can't 'cast tidal masters. It downsamples back to 'hifi '. So I am going to try a Diyinhk XMOS board and plug my android phone into this and see what happens! Likewise with Amazon HD if I try to cast in their 'ultra HD which is 24bit 48k flac to the chromecast into the receiver it still displays at 44 on the controller .
I'm very green with USB so have no idea if this will work but hey thats part of the fun isn't it?!
Last point I shall make here in this long post! Have built these streamers with blue potted Talema and Block transformers. Initially dead quiet but 'they'....I havent taken the time to identify which exactly as there are a few in there....have developed mechanical noise...a hum. Doesnt seem to come through the music but they are now humming away quite a bit and are warm. They used to be silent and cool?!
You need to feed spdif from the TV straight to your DAC. If using Ians ESS dac then there is an spdif input on there to which you can wire your own optical port. You need to disable auto mute on the ess controller .
In other news ....a few updates and observations.
Ian kindly replaced the ESS controller for my 2nd streamer build for my brother. This one went into an aluminium chassis, runs Pi less. Chromecast Audio optical into ReceiverPi and runs fifo on 1x NDK SDA for 44mhz family. Running 1.5f supercap via LT1963 to the fifo clean , another 1.5f over the clock itself and a 1.5f fifo dirty powering that and the receiver.
So there is room for some improvement to the psus there. DAC has 3 separate rails. Running the Chinese 600:10k output transformers with 510r across the secondaries as per Ians recommendations for the Lundahls.
My brother thinks this build sounds better than the 1st one. Main differences being heavy aluminium chassis and the passive IV.
I don't really understand the transformers! Initially I had then wired wrong. Took ages to realise. I had them wired backwards so they were stepping down instead of up I think. They are 1:4 I believe. This would explain the major lack of output level we experienced. Strangely though I had also wired the SE outputs using the +/- terminals, no ground reference. Still worked! When I sorted this lot out and wired correctly I also wired in XLRs aswell as the RCA. The ground wire to the XLR seemed to induce a hum out of the RCA so they are disconnected at the moment. My brother uses the SE output anyway.
One observation is that the system will not work from the BBC Sounds App. So all the music is cast from the Google Chromecast. Mainly Spotify which works fine. Radio Paradise app works fine. When we try to cast from the BBC app the receiverpi acknowledges the optical input but then the fifo won't lock. I don't understand this. Can the signal from the BBC app be different/lower quality.? Wierd.
Getting my brothers finished has given me some time to have a look at mine again. Being in my bro's possession for nearly a year and powerd 24/7 the first thing I noticed was the controller display is loosing pixels and looking a bit tired. Anyone else seen this?
I've also realised i can save a psu as I was powering fifo dirty, and receiver separately. Which in turn powers the Rpi in my build. So I've saved some space and removed then separate psu I had to the Receiver. It now is powered via gpio from the Fifo. I know I can remove an inductor on the Reciever for a separate psu and possible SQ improvement. Having not removed this my additional psu was wasted.
I am plucking up the courage to add some 310F caps to the clock (s). I do leave it powered up 24/7 so once charged they shouldn't add any inconvenience.
On clocks....I ran with one for a while but at the moment I have 45mhz Crystal and 49mhz SDA. I was hoping to use maybe just one and upsample everything to that family. But for some reason when I play from Spotify though either Moode or Volumio in the Rpi, despite the software being set to 300 odd Mhz it always comes out on the display as 44. It won't upsample a Spotify feed.
Have been trying Tidal MQA and also Amazon HD. I don't know the best way to get these higher bitrate streams into the system. You can't 'cast tidal masters. It downsamples back to 'hifi '. So I am going to try a Diyinhk XMOS board and plug my android phone into this and see what happens! Likewise with Amazon HD if I try to cast in their 'ultra HD which is 24bit 48k flac to the chromecast into the receiver it still displays at 44 on the controller .
I'm very green with USB so have no idea if this will work but hey thats part of the fun isn't it?!
Last point I shall make here in this long post! Have built these streamers with blue potted Talema and Block transformers. Initially dead quiet but 'they'....I havent taken the time to identify which exactly as there are a few in there....have developed mechanical noise...a hum. Doesnt seem to come through the music but they are now humming away quite a bit and are warm. They used to be silent and cool?!
Last edited:
Do you, or anyone else knows, if NAA (HQPlayer) works for RPI with I2S output? I am thinking about making a I2S streamer with the Ian Canada reclocker + I2S transport board (and maybe the isolator).
I will only need one clock since I up-sample to DSD (the one for 44.1khz and multiplier) so can spend a little more on that one, but which is the best clock to use for the reclocker?
Anything else to consider?
I was a little confused, since an RPI is used as base with Ian Canadas modules, so I send an email to Ian to make sure. The answer: an RPI4 with the overclocking module (FIFOPI Q3 ULTIMATE) and the I2S transport will support the following over I2S:
PCM up to 768KHz
DSD up to DSD1024
DoP up to DSD256
PCM up to 768KHz
DSD up to DSD1024
DoP up to DSD256
Magnus, this is exactly what I am doing as I found I2S to sound better than USB. I have a Pi3 with a fifo pi on top but I've only ever managed to get PCM up to 384k. It looks like Ian is saying that we should be able to output 768...I wonder which driver he's using and whether he is stipulating a pi4 because the Pi3 is not up to the processing or whether the driver isn't available?
In terms of which clock is best, I haven't found anything better than the Pulsar OCXOs that were available a few years ago. I still use them and they are great. There's a great project that is in GB now (I think) that is well worth a look: The Well Tempered Master Clock - Building a low phase noise/jitter crystal oscillator
But the real question is, will a RPi4 with Ian Canadas modules sound better on I2S than for example I2S output from a Matrix SPDIF 2 USB?
Building oscillator boards is to much DIY for my taste, so if I go the RPI path I will use a pin-compatible clock to Ian Canadas reclocker board.
I guess the only 'real' difference is that SPDIF(Toslink that is) provides galvanic isolation. Whether that makes a difference and sounds better to your ears, only you can determine. I've been living with a fully exploited Ian DAC for a couple years now, and it's earned permanent residence in my main listening system. I don't have another DAC that bests it. It's a sonic joy to listen to. Pure, articulate, smooth and absolutely non-fatiguing. Everything I listen to comes over I2S via my network.
USB and SPDIF
In my journey it was my understanding that SPDIF is inherently inferior to I2S despite the galvanic isolation. But I have no direct experience to support that belief. OTOH, eliminating USB (on a WAVEIO board, of which I was a big fan) to the FIFO approach was a significant improvement. I always was suspect of USB somehow degrading the system. It seemed confirmed by the fact that the USB cable had an impact on sound quality.
redjr
I think its noteworthy that you find this setup bests all challengers you have seen to date. I'm happy with the result and find no need to chase improvements. It just seems logical that this branch of DIY could easily beat all but the most expensive commercial DAC's. Up to a certain pricepoint we use the same technology DAC, but we have the freedom to dramatically overspec power supplies, clocks and output stages as we voice our DAC. It would be interesting to hear a list of commercial DAC's we have compared to the Iancanada stack. I recall he had mentioned the much vaunted Benchmark DAC as one that did not compete. One member brought a highly enhanced Oppo DAC. Nice, but no contest.
In my journey it was my understanding that SPDIF is inherently inferior to I2S despite the galvanic isolation. But I have no direct experience to support that belief. OTOH, eliminating USB (on a WAVEIO board, of which I was a big fan) to the FIFO approach was a significant improvement. I always was suspect of USB somehow degrading the system. It seemed confirmed by the fact that the USB cable had an impact on sound quality.
redjr
I think its noteworthy that you find this setup bests all challengers you have seen to date. I'm happy with the result and find no need to chase improvements. It just seems logical that this branch of DIY could easily beat all but the most expensive commercial DAC's. Up to a certain pricepoint we use the same technology DAC, but we have the freedom to dramatically overspec power supplies, clocks and output stages as we voice our DAC. It would be interesting to hear a list of commercial DAC's we have compared to the Iancanada stack. I recall he had mentioned the much vaunted Benchmark DAC as one that did not compete. One member brought a highly enhanced Oppo DAC. Nice, but no contest.
I have the RME ADI-2 FS DAC currently, I have a feeling Ian Canadas DAC stack will at best be a sideway move (but less expensive). More likely it won't sound quite as good.
But its also a question of sound signature and preferences, ESS chips tends to be a little more forward and edgy compared to AKM chips.
But its also a question of sound signature and preferences, ESS chips tends to be a little more forward and edgy compared to AKM chips.
Nothing you can't amply adjust upon your preference acting on dac output stage, something you can't do on commercial dacs...
I have built some ESS dacs sounding edgy and forward and some other ESS dacs sounding warmer and smoother than AKM dacs too, only difference were a few components in the output stage.
Easy and common preconcepts are always simplistic, in diy hifi.
After direct listening, i think my diy 9038pro dac sounds no worse (or mostly better) than many AKM 4490/4493/4497 dacs, so maybe no worse than RME ADI 2 too. Surely not more "forward" or "edgy"...
Last edited: