Actually, the last time I swapped out a clock I did it a little differently than before. I removed the old one as usual with Chip Quik alloy.
Then rather than trimming away any ground traces around the solder pads, I observed that all the green solder mask was still in good condition and unscratched, so far so good. Then I put a little solder on the side of the tip of the iron and slid it along one of the solder pads on the bottom of the new clock module. It left a little bead or a bump of fresh shiny leaded solder with a good amount of flux still on it on the bottom of the clock solder pad. I did that for each of the four solder pads on the clock and made sure each little bump of solder was the same height for each pad.
After redoing any that needed it until they all looked the same, then I just held the new clock carefully in place on the dac board with my fingers and touched the tip the iron to the side of the ground pad on the clock to melt the solder I put on the bottom (ground is the one directly across from the clock output to the dac). I then looked to make sure the clock pads were perfectly centered on the dac board solder pads, which they weren't quite. So I reheated the side of the ground pad again and moved the clock slightly into place.
Once it was exactly centered, I heated the other pad sides one at a time. Doing that of course melted the solder I left in place on the bottom of the clock pads earlier, so I didn't need to add any solder at this stage.
As I went around and heated and flowed each one, I checked with a DVM using sharp tiny probes to make sure each pad was soldered and no shorts on any of the pads to ground. It went pretty quick and easy that way. Usually clocks give me more soldering anxiety than any other solder job, but this was the easiest way I found to do it yet.
The reason for anxiety with these clocks is because the new clocks are a little bigger than the old ones. They just exactly fit on the board pads, but only if the clock is perfectly centered.
However, I noticed in Serge's thread that people seem to be able to change the clocks okay. We haven't had anyone yet who said they ran into a problem and had to Chip Quik off the new clock in order to redo it more well centered.
If by any chance it were to become necessary to Chip Quik off a new clock, just be sure to use a small amount so it doesn't get into the slots in side of the clock cover. That might be hard to clean out, so go slow and careful and use the least amount possible would be my advice.
Last edited:
Trying to get setup to do harmonic distortion compensation adjustment. What I found out so far is that the peak level coming out of the computer needs to be less than -6dBFS. Setting it to -7dBFS works fine here. (This is with a 1kHz test tone and 48kHz sample rate, which probably has some influence on the exact distortion threshhold.) Any more output level than that and distortion starts to increase greatly, presumably because of clipping in the multiple layers of SRC between the AK4137 and ES9038Q2M ASRC.
The above seems more or less consistent with my subjective observation that setting the master volume slider for the XMOS sound output device in Windows sound control panel to about 92 out of 100, or so, works well to minimize distortion.
Unfortunately, I seem to be having a problem finding freeware I can use to look at harmonic distortion.
Spectrum Lab used to be a good program, but doesn't seem to work at all well with Windows 10 drivers, and or with multi-channel sound devices (more than one stereo pair).
Arta, in its freeware form is limited to 16-bits which is okay since I have a notch filter and some low distortion post-filter gain. However, the Arta signal generator will only let me turn down the pink noise output to -50dB. There doesn't seem to be any way to turn it off. I can turn on signal averaging in the spectrum analyzer view to reduce noise, but it only drops the generator noise so much. I guess I might be willing to pay for an Arta license for $99, but there is no clear information about what changes with the licensed version. I assume 24-bit mode must be enabled, but who knows what the the signal generator will or won't allow in the way of settings. So forget that.
Besides, I would like to find some freeware anyone in our group could use. I can't see paying $100 for 5 or 10 minutes of one time use to adjust two distortion settings in a dac.
The above seems more or less consistent with my subjective observation that setting the master volume slider for the XMOS sound output device in Windows sound control panel to about 92 out of 100, or so, works well to minimize distortion.
Unfortunately, I seem to be having a problem finding freeware I can use to look at harmonic distortion.
Spectrum Lab used to be a good program, but doesn't seem to work at all well with Windows 10 drivers, and or with multi-channel sound devices (more than one stereo pair).
Arta, in its freeware form is limited to 16-bits which is okay since I have a notch filter and some low distortion post-filter gain. However, the Arta signal generator will only let me turn down the pink noise output to -50dB. There doesn't seem to be any way to turn it off. I can turn on signal averaging in the spectrum analyzer view to reduce noise, but it only drops the generator noise so much. I guess I might be willing to pay for an Arta license for $99, but there is no clear information about what changes with the licensed version. I assume 24-bit mode must be enabled, but who knows what the the signal generator will or won't allow in the way of settings. So forget that.
Besides, I would like to find some freeware anyone in our group could use. I can't see paying $100 for 5 or 10 minutes of one time use to adjust two distortion settings in a dac.
Last edited:
Rightmark is useless as far as I can tell. I don't want and can't really use an automated measurement system. I just need to generate a 24-bit 1kHz sine wave and look at it with a real time FFT spectrum analyzer that has signal averaging to reduce random noise in the spectrum. Some choice of window functions would be helpful.
Last edited:
Hi Markw4,
I posted few RMAA results few months ago
I'm curious to compare your results against mine (modded es9038q2m and tone board)
Occip
I didn't do any measurements with RMAA. I am looking for a 24-bit signal generator and real-time spectrum analyzer. RMAA doesn't appear to have that. It only seems to have off line file based spectrum analysis. At least, that's what it looks like. There are check boxes for it to run pre-programmed measurements, nothing that looks like it will let me drive.
Last edited:
I have been following this thread with interest. Thanks.
Mark (and others) you may be interested in the ES9038Q2M DAC I spotted on Aliexpress recently -
HIFI DAC decoder player ES9038Q2M USB input decoder XMOS U208 chip receive Bluetooth 5.0 supports DSD512-in Digital-to-Analog Converter from Consumer Electronics on Aliexpress.com | Alibaba Group
The output stage looks more comprehensive than most other DACs of this type I have seen before, with many through-hole capacitors, allegedly by WIMA.
But what I found particularly interesting is the at the input: it includes an AK4113 s/pdif receiver, an XMOS U208 USB receiver, and a CSR8675 bluetooth receiver. I2S outputs from these 3 chips all go to a XILINX XC3082A chip, and then to a second XILINX chip which I can't identify, before being passed to the DAC chip itself.
Now I have read that the I2S output from the CSR8675 blutooth chip has so much jitter that most DACs are unable to sustain consistent lock. So I think the XILINX chips must be doing some form of reclocking or stabilisation. And I also wonder if these XILINX chips are doing some upsampling, too?
In terms of DAC power supply and DAC output stage, I'm not sure how this particular ES9038Q2M implementation stacks up against the mods under discussion in this thread,
But maybe the XILINX stuff is of interest?
In any case, this DAC seems to offer a decent number of features for under US$200.
Mark (and others) you may be interested in the ES9038Q2M DAC I spotted on Aliexpress recently -
HIFI DAC decoder player ES9038Q2M USB input decoder XMOS U208 chip receive Bluetooth 5.0 supports DSD512-in Digital-to-Analog Converter from Consumer Electronics on Aliexpress.com | Alibaba Group
The output stage looks more comprehensive than most other DACs of this type I have seen before, with many through-hole capacitors, allegedly by WIMA.
But what I found particularly interesting is the at the input: it includes an AK4113 s/pdif receiver, an XMOS U208 USB receiver, and a CSR8675 bluetooth receiver. I2S outputs from these 3 chips all go to a XILINX XC3082A chip, and then to a second XILINX chip which I can't identify, before being passed to the DAC chip itself.
Now I have read that the I2S output from the CSR8675 blutooth chip has so much jitter that most DACs are unable to sustain consistent lock. So I think the XILINX chips must be doing some form of reclocking or stabilisation. And I also wonder if these XILINX chips are doing some upsampling, too?
In terms of DAC power supply and DAC output stage, I'm not sure how this particular ES9038Q2M implementation stacks up against the mods under discussion in this thread,
But maybe the XILINX stuff is of interest?
In any case, this DAC seems to offer a decent number of features for under US$200.
Hi linuxfan,
Thank you for bringing the unit to our attention. However, it is not new. When they first came out I ordered one to check out. If you use the thread search tool and look for 'bluetooth' and or 'su5' you should be able to find everything we have on it.
Bottom line though would be to strongly avoid it. Very difficult to fix into something with decent sound quality. Mine is still in pieces. Just an overwhelming number of problems.
If you are interested in a good dac, I would strongly recommend the one we are modding. The latest one I built with the new AVCC and output stage sounds really good I have to say. I find it to be a true pleasure for listening to music. Haven't really proven it, but I think the optional film caps for AVCC filtering may contribute to its silky smoothness. Of course, it is extremely detailed and accurate too, with great dynamics and deep punchy bass. You will hear things in plain old CDs that you never knew existed before. No fatigue, not clinical at all. The one crazy thing about it is I am the only person in the world to have the whole thing fully up and working.
Also, it can be used with RPi and Volumio to play streaming music if you want. RPi can also be used to tweak dac control registers for best sound quality which does help if the build is clean enough to get the jitter to a minimum. While that would not include bluetooth, it would do a lot of what some people like in a music player. It can also be used as a general purpose dac for mixing records, watching movies, anything people use dacs for.
The only thing at all not to be happy about it is that nobody else has gotten very far with modding one all the way yet. Hopefully, by the end of the year we will start to see some builds come to completion. When that happens I know we will have some very happy owners.
Thank you for bringing the unit to our attention. However, it is not new. When they first came out I ordered one to check out. If you use the thread search tool and look for 'bluetooth' and or 'su5' you should be able to find everything we have on it.
Bottom line though would be to strongly avoid it. Very difficult to fix into something with decent sound quality. Mine is still in pieces. Just an overwhelming number of problems.
If you are interested in a good dac, I would strongly recommend the one we are modding. The latest one I built with the new AVCC and output stage sounds really good I have to say. I find it to be a true pleasure for listening to music. Haven't really proven it, but I think the optional film caps for AVCC filtering may contribute to its silky smoothness. Of course, it is extremely detailed and accurate too, with great dynamics and deep punchy bass. You will hear things in plain old CDs that you never knew existed before. No fatigue, not clinical at all. The one crazy thing about it is I am the only person in the world to have the whole thing fully up and working.
Also, it can be used with RPi and Volumio to play streaming music if you want. RPi can also be used to tweak dac control registers for best sound quality which does help if the build is clean enough to get the jitter to a minimum. While that would not include bluetooth, it would do a lot of what some people like in a music player. It can also be used as a general purpose dac for mixing records, watching movies, anything people use dacs for.
The only thing at all not to be happy about it is that nobody else has gotten very far with modding one all the way yet. Hopefully, by the end of the year we will start to see some builds come to completion. When that happens I know we will have some very happy owners.
Last edited:
Hi Mark,
Sorry for being silent.
I was pretty busy at work for the last 3 weeks...
I received at the end of last week the ak4137, the xmos and my new 1.07 board.
AK4137 is working ok except for some dsd files. (btw which filter do you use?)
Does the xmos need a psu or is it feed by the usb cable?
(Till now, I always used I2s)
I got my new clock I will try to solder it on a old 1.04 board I have in order to train myself and not ruining the new board.
I'm still waiting for the prototype boards.
Do you have a mouser reference for the copper foil?
It looks that there is hundred of them and I don't know which one to buy.
Hope the mods went well.
I have several pages of the thread to read to catch up.
Last edited:
Hi Terry,
XMOS is powered by USB.
Thing about soldering your new clock to your old 1.04 board, if it works that means you are going to be upgrading the 1.04 board for all the mods then, if that is what you intend. You certainly don't want to solder the new clock to the old board, then unsolder it, the re-solder it to the new board.
You could remove and replace the original clock on the 1.04 board for practice if you want. Maybe the most important thing when unsoldering an old clock to take it gently and try not to scratch up the solder mask on the dac board while doing it since that helps the solder only stick where you want when it comes time to solder a clock back on as the next step. By the way, I did write some more recently about clock replacement soldering.
I don't have a Mouser reference for copper foil. I bought some using ebay and some from amazon.
Regarding AK4137 and playing DSD files, I haven't tried it and that could be a limitation of using it. It certainly seems to help with PCM files. Maybe there is some way to bypass AK4137 when playing native DSD files, haven't looked into it but maybe will have to.
XMOS is powered by USB.
Thing about soldering your new clock to your old 1.04 board, if it works that means you are going to be upgrading the 1.04 board for all the mods then, if that is what you intend. You certainly don't want to solder the new clock to the old board, then unsolder it, the re-solder it to the new board.
You could remove and replace the original clock on the 1.04 board for practice if you want. Maybe the most important thing when unsoldering an old clock to take it gently and try not to scratch up the solder mask on the dac board while doing it since that helps the solder only stick where you want when it comes time to solder a clock back on as the next step. By the way, I did write some more recently about clock replacement soldering.
I don't have a Mouser reference for copper foil. I bought some using ebay and some from amazon.
Regarding AK4137 and playing DSD files, I haven't tried it and that could be a limitation of using it. It certainly seems to help with PCM files. Maybe there is some way to bypass AK4137 when playing native DSD files, haven't looked into it but maybe will have to.
Last edited:
thank you! good to know
I was about to solder it on the old board for testing it.
I read your post about the clock. Very interesting.
When I unsoldered the clock on my old board it was pretty clean.
I will try to do the same on the new.
Problem with dsd files is fixed.
The problem was the rpi and SoX.
Mark,
I'm using Arta with EMU1212M. I wasn't aware of the bit depth restrictions of
the freeware version. I'll have to check that. It surprises me because I was
able to do decent loop through verification of 1212M performance and it
appeared to meet the spec.
This was a while ago, I'll see if I can get time to check this out again.
T
Hi Terry,
In the sound card setup the unlicensed version has a grayed out setting for wav bit depth, which is locked at 16-bits. I don't know if it only applies to in relation to reading and writing wav files. If only files, then the signal generator and spectrum display depth may be 24-bit.
In the sound card setup the unlicensed version has a grayed out setting for wav bit depth, which is locked at 16-bits. I don't know if it only applies to in relation to reading and writing wav files. If only files, then the signal generator and spectrum display depth may be 24-bit.
OK, I don't remember that, will check it out.
In the meantime, have a look at this one:
AUDio MEasurement System download | SourceForge.net
cheers
T
An update on the new modded dac: Got the harmonic distortion compensation to work with Arta running in demo mode (freeware mode).
Found this 2nd version of the dac didn't need any C2 adjustment at all, but it has some low level odd harmonics that do affect the sound and that did benefit from some adjustment of C3.
C3 was minimized at a setting of -157, however that left C3 perhaps a little bit too much lower than C5, and it didn't sound subjectively best there. So, after some more listening tests we decided we liked -150 best for C3. (By the way, the default settings of HD compensation in the various dac registers is zero compensation, and HD comp 'disabled.' First thing is therefore to enable it before trying any adjustment settings.)
The C3 compensation difference between the default of 0 and a final setting of -150 used for this dac is pretty noticeable to us as we listen on the system here. I can only say it is definitely worthwhile to make use of the HD compensation capability if at all possible, as I think Allo found during their testing of Katana, and of course as Benchmark does for DAC-3. Same finding here. The effects are audible, and there will be a setting that sounds best.
By way of comparison, the first dac I modded came out with C2 = 5 and C3 = -7, IIRC. Quite a big difference from this second dac. The adjustment was audible as it affected complex music signals, even in that case.
My current plans include building an SMD version of the new schematic output stage and using it to upgrade the 1st modded dac. The AVCC circuit is already up to date, except no film caps. It uses a 10uf SMD X7R ceramic for LTC6655 filtering and organic polymers for opamp output filtering. We will see how it sounds with the new output stage, and if it seems like it might benefit from the addition of film caps for AVCC. Don't know when I will get started on that though.
Regarding Arta, I am operating the dac in PCM mode for making HD compensation adjustments. DSD noise floor is a bit too high to see the harmonics well even with signal averaging, but DSD for music playback still sounds best.
For the HD comp adjustments, used a 1kHz notch filter I made and which is handy to have around for various projects. I can describe how to build one and post some pics sometime. After the filter I go into an add-on low distortion preamp because my best ADC is calibrated for +24dBU FS, IIRC. It needs a lot of signal which means most people should not need an additional preamp like I am using. Without a notch filter, it is likely a sound card ADC will add a lot of its own HD from the high level 1kHz tone. With that frequency attenuated enough, there is no significant additional distortion contribution from the ADC data acquisition chain, so only dac distortion is seen in the FFT spectrum.
Don't know why more harmonic distortion in this dac as compared to the last one. Could perhaps have something to do with better gain matching and tighter tolerance resistors used in the 1st dac output stage. Had some .01% 10k resistors in stock that got used for the differential summing stage, and maybe the SMD I/V resistors just happened to be more well matched by luck. Haven't measured them to check.
Found this 2nd version of the dac didn't need any C2 adjustment at all, but it has some low level odd harmonics that do affect the sound and that did benefit from some adjustment of C3.
C3 was minimized at a setting of -157, however that left C3 perhaps a little bit too much lower than C5, and it didn't sound subjectively best there. So, after some more listening tests we decided we liked -150 best for C3. (By the way, the default settings of HD compensation in the various dac registers is zero compensation, and HD comp 'disabled.' First thing is therefore to enable it before trying any adjustment settings.)
The C3 compensation difference between the default of 0 and a final setting of -150 used for this dac is pretty noticeable to us as we listen on the system here. I can only say it is definitely worthwhile to make use of the HD compensation capability if at all possible, as I think Allo found during their testing of Katana, and of course as Benchmark does for DAC-3. Same finding here. The effects are audible, and there will be a setting that sounds best.
By way of comparison, the first dac I modded came out with C2 = 5 and C3 = -7, IIRC. Quite a big difference from this second dac. The adjustment was audible as it affected complex music signals, even in that case.
My current plans include building an SMD version of the new schematic output stage and using it to upgrade the 1st modded dac. The AVCC circuit is already up to date, except no film caps. It uses a 10uf SMD X7R ceramic for LTC6655 filtering and organic polymers for opamp output filtering. We will see how it sounds with the new output stage, and if it seems like it might benefit from the addition of film caps for AVCC. Don't know when I will get started on that though.
Regarding Arta, I am operating the dac in PCM mode for making HD compensation adjustments. DSD noise floor is a bit too high to see the harmonics well even with signal averaging, but DSD for music playback still sounds best.
For the HD comp adjustments, used a 1kHz notch filter I made and which is handy to have around for various projects. I can describe how to build one and post some pics sometime. After the filter I go into an add-on low distortion preamp because my best ADC is calibrated for +24dBU FS, IIRC. It needs a lot of signal which means most people should not need an additional preamp like I am using. Without a notch filter, it is likely a sound card ADC will add a lot of its own HD from the high level 1kHz tone. With that frequency attenuated enough, there is no significant additional distortion contribution from the ADC data acquisition chain, so only dac distortion is seen in the FFT spectrum.
Don't know why more harmonic distortion in this dac as compared to the last one. Could perhaps have something to do with better gain matching and tighter tolerance resistors used in the 1st dac output stage. Had some .01% 10k resistors in stock that got used for the differential summing stage, and maybe the SMD I/V resistors just happened to be more well matched by luck. Haven't measured them to check.
Last edited:
Interesting findings Mark, nice work.
You should post your results on the Blowtorch thread (HaHa)
. They are very low levels however, we did tests years ago introducing low
levels of H2 / H3 in an amplifier and found the same thing. We also had
similar results WRT H3 needing to be there, some amount of H3 is not a bad
thing.
Did you have any way of cancelling any harmonics above H5 or testing
audibility of them?
So I take it you acquired a data sheet and signed the NDA?
Do you have any FFT's with distortion cancellation?
One thing I have been wondering with Sabre DAC is if you can run pure
voltage OP and use the cancellation to linearize it. This opens up more OP
stage possibilities.
T
Only H2 and H3 can be tweaked. So, nothing affected above that. However, I am curious why no H2, but a succession of odds. Will plan on looking at it a little more to see if I can affect it at all.
Yes.
I forgot to grab a screen shot, but will try to get one next time I'm in there. Besides the odd harmonics, other imperfections I see are some 120Hz and a few harmonics above that which look to be line frequency noise. Don't know if they are from the dac or from the measurement system. In addition, the averaged noise floor looks higher below 1kHz. Don't know why, if it is an artifact of averaging, something from the notch filter, maybe from the 1/f noise corner of one or more of the LDO regulators for low voltage rails, or whatever else. So, there are some things that could be investigated and potentially improved, especially to the extent they may be real. However, it is the odd HD that I am currently most interested in, even though it is pretty low. It's still not as low as the 1st dac, and I would like to know to know why.
All the above having been said, it still sounds very good, no reason not to get to work on one. If I happen to find anything that can make it sound slightly better I will post here, but I wouldn't count on anything more turning up.
Yes. Occip did that and posted a big-ish number he found was optimal for H3 compensation. However, no one has compared the sound of doing that with the sound of doing a full blown current mode output stage with filtering, etc. According to ESS, current mode will give the least distortion. Don't know if they tried what occip did, but don't know why they wouldn't.
Last edited:
@Terry, Occip posted re his 2nd board coming in the same as the 1st for HD compensation for H3. None used for H2, IIRC. This is with the original voltage output stage, but with an AVCC mod: https://www.diyaudio.com/forums/digital-line-level/314935-es9038q2m-board-211.html#post5502201