Crysteck clock panel
This is a clock panel planned and soldered by Mzperx. He is a great member of Hungarian diyers community. The 3.3V of the clock is provided by an ADM7150ARDZ ultra low noise LDO. The panel works with external 5-15 V DC. The Crystek CCHD-950-25 is from Mouser. I am going to install the clock panel soon. There is a 33R at the otput.
This is a clock panel planned and soldered by Mzperx. He is a great member of Hungarian diyers community. The 3.3V of the clock is provided by an ADM7150ARDZ ultra low noise LDO. The panel works with external 5-15 V DC. The Crystek CCHD-950-25 is from Mouser. I am going to install the clock panel soon. There is a 33R at the otput.
Attachments
Last edited:
What about this DIYinhk boards combo, for a complete DAC based on the ES9038:
DAC: ES9038PRO XMOS DSD DXD 768kHz USB DAC with Bit-perfect volume control and SPDIF input - DIYINHK
DSD: Isolated XMOS DSD DXD 384kHz high-quality USB to I2S/DSD PCB with ultralow noise regulator - DIYINHK
Power supply: 0.8uV Ultralow noise DAC power supply regulator 3.3/5/7V 1.5A*x2 - DIYINHK
I2S interface:
FIFO Reclock Jitter Eliminator - DIYINHK
I'm not sure why or what for would I need this:
1.3"OLED and rotary encoder volume control PCB - DIYINHK
No idea how I would select the inputs between DSD, coaxial and optical. That's something I might ask DUYinhk.
Basic bill till now: $230.
What else would I need or what is not necessary? What case might hold all these boards? I'm planning on a separate box for the transformer, to keep noise low.
DAC: ES9038PRO XMOS DSD DXD 768kHz USB DAC with Bit-perfect volume control and SPDIF input - DIYINHK
DSD: Isolated XMOS DSD DXD 384kHz high-quality USB to I2S/DSD PCB with ultralow noise regulator - DIYINHK
Power supply: 0.8uV Ultralow noise DAC power supply regulator 3.3/5/7V 1.5A*x2 - DIYINHK
I2S interface:
FIFO Reclock Jitter Eliminator - DIYINHK
I'm not sure why or what for would I need this:
1.3"OLED and rotary encoder volume control PCB - DIYINHK
No idea how I would select the inputs between DSD, coaxial and optical. That's something I might ask DUYinhk.
Basic bill till now: $230.
What else would I need or what is not necessary? What case might hold all these boards? I'm planning on a separate box for the transformer, to keep noise low.
I'm not getting your math..
The XMOS and DAC combo is $249 itself. (with the DAC board including a soldered 9038PRO chip)
Add $60 for the 3.3 V supply,
$20 for the FIFO reclocker,
$25 for the display and encoder(may be used to select inputs..)
and you are already at ~$355.
You also need a Bipolar 12V supply, which looks to be another $60, for a total of $415.
Transformers for the power supplies are not mentioned so add another $20 each for them (guestimate) (2). Connectors and switches will cost you another $30, plus add at least $100 for a case(I'm lowballing this one).
Grand total now is $585USD.
The particular case will have to be determined by yourself using the boards and other components themselves to do a physical layout, to find what case size will be appropriate.
The XMOS and DAC combo is $249 itself. (with the DAC board including a soldered 9038PRO chip)
Add $60 for the 3.3 V supply,
$20 for the FIFO reclocker,
$25 for the display and encoder(may be used to select inputs..)
and you are already at ~$355.
You also need a Bipolar 12V supply, which looks to be another $60, for a total of $415.
Transformers for the power supplies are not mentioned so add another $20 each for them (guestimate) (2). Connectors and switches will cost you another $30, plus add at least $100 for a case(I'm lowballing this one).
Grand total now is $585USD.
The particular case will have to be determined by yourself using the boards and other components themselves to do a physical layout, to find what case size will be appropriate.
Fair enough, thank you. Even though it might not seem like it, it does help me to understand how things are coming along for other folks so long as I get any replies at all. If silence, it may be that people just don't care and I am wasting my time.
Hey Mark, would the copper foil you are talking about be something like this:
EMI Copper Foil Shielding Tape Conductive Self Adhesive Barrier Guitar | eBay
Thank you for the update on the clock. Looks like a nice little board.
Hope you won't mind if I say that with 100MHz RF, I always feel more comfortable if it is soldered to the dac board right next to the dac chip. But, if enough of the right care is taken, it is probably possible to get very nearly the same results with a clock a short distance away.
Also, with the growing popularity of ufl connectors, I sometimes wonder that we still pipe around 100MHz RF on pin header connectors like it was only 10MHz.
The other thing that I have been wondering about recently and I would like to find out more about has to do with they DIY community love affair with LDO RF regulators and their noise density graphs. I have to suspect that it is the marketing department of LDO manufactureres that wants those graphs included in data sheets and scope traces of actual peak to peak noise omitted. While the noise density at every frequency shows a lot of 1/f noise perhaps from the RF transistor manufacturing processes they now run, noise density still looks very low even at pretty low frequencies like 10Hz. However, we don't really care about noise density at individual frequencies, we care about integrated noise over the band that can affect operation of our circuits. If we integrate the area under the noise density curves we get something much more suggestive of what peak to peak noise actually looks like. Also, IIRC, the Chinese LT3045 module makers used to show a photo of a couple of unlabeled graphs in their ads. One showed a noise curve that rose to about -70dB at LF and other trace was flat all the way down to LF. Well, once I found a bigger, higher resolution copy of that picture, and it turned out -70dB was noise out of the LT3045 down at 10Hz, and the flat graph was showing the accuracy of fit of an equation modeling the measured noise performance (in other words, meaningless). Okay then, -70dB noise out of a power supply might be okay for a circuit that has another 60dB - 80dB PSRR at 10Hz, but not for something with less or no PSRR. The 'no PSRR' category would include AVCC.
But, what about phase noise at 10Hz from a Crystek oscillator? I don't know how RF LDO noise affects that or what the Crystek PSRR down at 10Hz is. Sure would like to see some comparative measurements though.
Last edited:
Yes, thank you. They also sell it at Amazon. The stuff is very thin, but usable. Also found out that Scotch Tape makes rolls of adhesive backed copper foil in multiple thicknesses and widths. However, thick and wide can get expensive for a roll. The thin stuff is better than no ground plane, that I know.
To keep the ball rolling, I wanted to mention that I've been collecting the various bits and pieces to do the upgrades you have mentioned on the Chinese board. So far I've got the following:
LT3045 3.3v regulator
"Budget" AK4137 SRC board
Arduino Due and DimDim's Universal Signal Isolator board for use with his HiFiDuino code to hack the 9038Q2M's registers. (TFT display still on the way)
$7 logic analyzer you mentioned in the thread (another rabbit hole to go down for sure)
CCHD-575-50-100 clock crystal
I also have a 15V "ClassA" power supply that should work for the analog section.
I've also got an E-Mu 0204 ADC/DAC that has pretty decent specs that I'm thinking of trying to do some measurements of the board with RMAA, to make some mea surements along the way. It's probably worse from a noise perspective than a well designed 9038Q2M DAC, but it might be a good starting point, particularly if the Chinese board is particularly bad to start with.
LT3045 3.3v regulator
"Budget" AK4137 SRC board
Arduino Due and DimDim's Universal Signal Isolator board for use with his HiFiDuino code to hack the 9038Q2M's registers. (TFT display still on the way)
$7 logic analyzer you mentioned in the thread (another rabbit hole to go down for sure)
CCHD-575-50-100 clock crystal
I also have a 15V "ClassA" power supply that should work for the analog section.
I've also got an E-Mu 0204 ADC/DAC that has pretty decent specs that I'm thinking of trying to do some measurements of the board with RMAA, to make some mea surements along the way. It's probably worse from a noise perspective than a well designed 9038Q2M DAC, but it might be a good starting point, particularly if the Chinese board is particularly bad to start with.
Aside from possible questions about cost, I know nothing about their quality. I know I would like to see proper good quality parts laid out well in well-designed circuits, and have it be on mulitlayer boards with not only ground planes but power planes too.
I can't tell anything at all about any of the above considerations from pictures or other information on the seller website. The only way to find out for sure if the devices are what I would consider to be good would be buy them and see what I get, or maybe to have someone else buy them and then take close-up, in-focus photos of both sides of all the boards, and make up parts lists too. That should be enough information to figure how good of a deal the purchase was, and how well the dac could likely be expected to perform.
In case people are not following iancanada's long (and I do mean long) running thread, he is working on a Pi based board set (he calls it FifoPi) which incorporates many of his previous design ideas, including galvanic isolation. It uses Crystek clocks as well. He has some good ideas on power supply requirements.
At some point, he may turn it into a group buy. See this post:
Asynchronous I2S FIFO project, an ultimate weapon to fight the jitter
Whit
At some point, he may turn it into a group buy. See this post:
Asynchronous I2S FIFO project, an ultimate weapon to fight the jitter
Whit
With a notch filter and some clean make up gain you can see static, noise-averaged, FFTs @1kHz or whatever frequencies you want to make filters for. I made a notch filter with the active electronics on a small motherboard, with plug-in daughter-board frequency selection components. I only made a 1kHz plug-in, but it works okay giving a good 40dB of attenuation at that frequency. The individual Q2M dac harmonics are all down at -120dB or lower, so most sound cards aren't up to the task without filtering. After the filter I use a Grace Designs m101 mic preamp for some make up gain to drive the ADC at a reasonable level. The mic preamp doesn't add any measurable distortion I can see at the low gains where I run it.
Last edited:
Depends on how it turns out, what it you get, what it costs, etc.
Turns out it is not necessary to use a FIFO to minimize jitter, although that is one way. Crane Song Solaris doesn't use a FIFO and they have gotten jitter down to 45fs (.045ps) from 10Hz to 20kHz. They appear to be using a architecture similar to Benchmark DAC-3, which appears to use successive stages of jitter reduction associated with upsampling. They are also using some nearly unobtainium clocks.
Other than jitter, there also appears to be some reason to suspect ESS dacs sound better with DSD or some variation thereof, maybe bypassing some internal dac filters. My ES9038Q2M here sure sounds better if digital audio is first upsampled and converted to DSD. Don't know if iancanada is into trying things like that.
Last edited:
Hi Mark, from your earlier comments in the thread re ESS and DSD I thought to audition the effect.
I've taken a few of my favorite tracks and up-sampled them to 256DSD using the free Teac editor..(Works for Windows, Mac, and surprisingly well in Linux using Wine...Ubuntu 18.04)
TEAC Hi-Res Editor | OVERVIEW | TEAC - International Website -
The improvements are certainly audible although player dependent in magnitude... ( Snakeoil OS, wtfplay, on x86 hardware Moode, Volumio on RPi)
Pleasing enough that I've ordered an AK4137 src board. Thanks for the info on this upgrade..!
Snakeoil OS
wtfplay-project.org
moodeaudio.org
I've taken a few of my favorite tracks and up-sampled them to 256DSD using the free Teac editor..(Works for Windows, Mac, and surprisingly well in Linux using Wine...Ubuntu 18.04)
TEAC Hi-Res Editor | OVERVIEW | TEAC - International Website -
The improvements are certainly audible although player dependent in magnitude... ( Snakeoil OS, wtfplay, on x86 hardware Moode, Volumio on RPi)
Pleasing enough that I've ordered an AK4137 src board. Thanks for the info on this upgrade..!
Snakeoil OS
wtfplay-project.org
moodeaudio.org
Last edited:
I would tend to agree that the FIFO solution is a bit of a brute force solution to the jitter problem. The good news about it is that it can solve the problem for all input types in one fell swoop. Witness Ted Smith's designs which resulted in the PS Audio DirectWave DACs. He upsamples and converts absolutely everything to DSD and uses a one-bit converter (which is hyper sensitive to jitter, BTW). One downside is the compexiity of the firmware needed - which puts it a little out of reach of the ordinary hobbyist.
Another drawback happens with multi-room synchronized players. Individual FIFOs can have differing playback delays.
If you only care about I2S, then I tend to like the Allo master-DAC approach which is a neat, cost-effective solution to the problem. The DAC clock serves as both the decode clock and the source clock, so I have strong hopes for the Katana when they settle down on the design tweaks.
iancanada lists his design goals in this post: Asynchronous I2S FIFO project, an ultimate weapon to fight the jitter
Clearly, he will be supporting DSD and the ability to roll XOs. Based on his track record, I think his designs and components will be useful to those interested in advancing the DIY SOA.
Cheers
Whit
Right. Supporting DSD playback is fine, but I was thinking more along the lines of converting all audio input sources, files, SPDIF, TOSLIK, AES, to 11.2MHz DSD on the fly in a Spartan 6, or something more like that. I sort of suspect Benchmark DAC-3 might be doing something like that, but using somewhat lower output sample rates in order to be able to use 30MHz dac clocks with lower jitter than is possible from 100MHz clocks. Perhaps it could be the DSD is 8x oversampled to bypass some internal filtering, or maybe not. Haven't actually gotten around to looking into 128_fs mode for DSD to see if it is possible.
Regarding Benchmark, I think for one of their earlier dacs they said they were sending it data in a format similar to DSD-wide. Don't know if they still do that or not. They said the reason for doing it was that it operated the dac chip modulator in the most favorable oversampling mode for minimizing distortion.
I do know I am upsampling and converting all audio including 16/44 CD wav files to 11.2MHz DSD on the fly using AK4137 for use with my ES9038Q2M dac. It sounds best that way of all the options I have. Reducing DSD mode DPLL bandwidth as low as possible while still retaining stability in the Q2M helps sound quality some additional amount too. Harmonic distortion compensation in Q2M is the last digital improvement I find I can make. My point regarding iancanada Q2M dacs was unless he has tried all those things I just mentioned at the same time, I would wonder if he is truly SOA with Q2M. Taking it to the next step might involve replacing AK4137 with a custom solution which would give control over the final interpolation filtering.
Regarding Benchmark, I think for one of their earlier dacs they said they were sending it data in a format similar to DSD-wide. Don't know if they still do that or not. They said the reason for doing it was that it operated the dac chip modulator in the most favorable oversampling mode for minimizing distortion.
I do know I am upsampling and converting all audio including 16/44 CD wav files to 11.2MHz DSD on the fly using AK4137 for use with my ES9038Q2M dac. It sounds best that way of all the options I have. Reducing DSD mode DPLL bandwidth as low as possible while still retaining stability in the Q2M helps sound quality some additional amount too. Harmonic distortion compensation in Q2M is the last digital improvement I find I can make. My point regarding iancanada Q2M dacs was unless he has tried all those things I just mentioned at the same time, I would wonder if he is truly SOA with Q2M. Taking it to the next step might involve replacing AK4137 with a custom solution which would give control over the final interpolation filtering.
Last edited: