Nice and inspiring work Mark, taking it from ground up this time 🙂 I am getting the replacement OCXO done and will ship early next week. Btw, this unit has a 1000pf pps Cap for bypass installed. Now , that you have the nice soldering gear, you may want to try out from 100pf …
Last edited:
I can see nobody is going to do anything about this unless I do:
Have to assign footprints, then the layout work begins.
Have to assign footprints, then the layout work begins.
Are the Toshiba optocoupler the best (quieter and more linear, fast enough) isolation way over the others isolating technic active chips based too ? I know Andrea is using it as you told but also saw most designers seem to prefer the other way ? It is way above my head, I just ask by curiosity.
They are good to 50MHz. Other than that, shouldn't matter as the reclocking will occur after the optocouplers. They are just for galvanic isolation. If they do a good job of that, its all I can ask.
yup thanks for the feedback, I indeed asked about their isolation bahavior despite it is before the reclocking. Hesitate myself to use it but after a reclocking on a DAC board of mine... . I wanted something to "break" the signal and ground from a big front end, but certainly does arm after a clean reclocking. Btw off topic, sorry for that.
Just a brief note to those folks using sine wave oscillators with their dacs: Found audible problems with some squarers I have. Turns out they are not as perfect as phase noise plots may suggest, assuming phase noise was ever measured. Ferrites appear to be only one of the problems.
An LTC6957-3 based squarer board was found to have audible cross coupling between channels if both 44kHz and 48kHz family oscillators were connected at the same time. Some of the coupling appears to be power rail related but could be other coupling mechanisms are involved too.
Also, some 74AC11004N based squaring circuits were found not be as clean sounding as one using LTC6957-3. Could be the problem in this case is related to layout and or power, rather than being a problem with 74AC11004N itself. Not sure yet.
The above having been said, ultra-low close-in phase noise clocks through very good squaring circuits seem to have clearly audible advantages over simpler and lower cost alternatives (assuming the rest of the dac is ready for better clocking). IOW, IME clocks are not usually among the fist dac problems to go after.
An LTC6957-3 based squarer board was found to have audible cross coupling between channels if both 44kHz and 48kHz family oscillators were connected at the same time. Some of the coupling appears to be power rail related but could be other coupling mechanisms are involved too.
Also, some 74AC11004N based squaring circuits were found not be as clean sounding as one using LTC6957-3. Could be the problem in this case is related to layout and or power, rather than being a problem with 74AC11004N itself. Not sure yet.
The above having been said, ultra-low close-in phase noise clocks through very good squaring circuits seem to have clearly audible advantages over simpler and lower cost alternatives (assuming the rest of the dac is ready for better clocking). IOW, IME clocks are not usually among the fist dac problems to go after.
Last edited:
Hi there Mark,
Top quality work as always. And very interesting.
A couple of questions:
Regarding the use of Omron RF relays vs. ordinary MUX ICs. I assume that you have tried both and noticed that the mechanical relays perform better, right?
Btw, the part number of the relay that you are giving on the schematic is for the latching version of the relay, while it looks like you need the stable version.
Regarding the use of the high speed photocouplers, I see that they need at least 5.5mA to be driven properly. Isn't that a bit much for the I2S lines? Not that it should be a problem for properly buffered I2S signals, but many I2S sources do not really provide buffered signals.
Top quality work as always. And very interesting.
A couple of questions:
Regarding the use of Omron RF relays vs. ordinary MUX ICs. I assume that you have tried both and noticed that the mechanical relays perform better, right?
Btw, the part number of the relay that you are giving on the schematic is for the latching version of the relay, while it looks like you need the stable version.
Regarding the use of the high speed photocouplers, I see that they need at least 5.5mA to be driven properly. Isn't that a bit much for the I2S lines? Not that it should be a problem for properly buffered I2S signals, but many I2S sources do not really provide buffered signals.
Another issue with those photocouplers is the whopping max 20ns (+/-10ns) device-to-device propagation delay skew.
Last edited:
What are typical mA level of those digital frontend signals (here I2S) please ?
At which bandwidth linearity one should cope when reading the datasheets of such ic (isolators ic, octocouplers...)
Is there for instance with some dac chips a slew rate problem at their input stage ? (sorry, perhaps not the rigth words, do not know if SR question has sense)
thanks
At which bandwidth linearity one should cope when reading the datasheets of such ic (isolators ic, octocouplers...)
Is there for instance with some dac chips a slew rate problem at their input stage ? (sorry, perhaps not the rigth words, do not know if SR question has sense)
thanks
That is correct. The correct part number should be in the associated Clock Board Information document. It gives part numbers for RF and non-RF relays that can be used....it looks like you need the stable version.
Regarding the optocouplers, the are what my Andrea Mori reclocker board uses. I don't see a problem with them for use with Marcel's dac, and when driven from PCM2DSD. However, anyone is free to redesign that part of the circuit as they see fit. Also, if initial testing shows there is a problem with the isolators then I will change them out before releasing a full design package.
Moreover, regarding all the open source designs I am sharing, they are offered as-is and without any warranty implied or expressed. If anyone finds a problem or wishes to make improvements, they are welcome to do so. All the designs are prototypes for my own use. I use them to learn from and I hope others will learn from them too or maybe just enjoy what the designs can do for a system.
BTW, the performance tuning resistor for LT1763 can have significant effects on the sound depending on source devices used with the board. In particular the squaring circuits I have draw somewhat more current than most clock modules. Found that raising the resistor value some had a beneficial effect. Users should expect to experiment with that a little.
Last edited:
I believe the relays should have less adverse effect on close-in phase noise than MUX ICs would have. Not all logic ICs perform equally well in that regard. OTOH, simple inverters (particularly unbuffered ones) seem to be one of the best logic devices in terms of maintaining low close-in phase noise.Regarding the use of Omron RF relays vs. ordinary MUX ICs.
Omron Relay part number has been removed from the Schematic attached to post #1. Please see Clock Board Notes document for information on part numbers.
Last edited:
Linearity would be a factor if we were using the optocouplers to pass analog signals. However in this case the intention is to pass digital signals only (on/off, 0 or 1, etc.).At which bandwidth linearity one should cope when reading the datasheets of such ic (isolators ic, octocouplers...)
Output slew rate of the reclocker board is primarily determined by the slew rate of the flip-flop ICs and the series output damping resistors. Some dacs seem to work better with slightly slowed slew-rates and other dacs may sound better with faster slew rates. Output slew rates could be slowed a little if desired by using 100R series output damping resistors instead of 33R.Is there for instance with some dac chips a slew rate problem at their input stage ? (sorry, perhaps not the rigth words, do not know if SR question has sense)
ah, as the substract and electricity is "analog" I believed there was a transition path between the 0 & the 1 ?!
There is a transition path but we only read the state of the optocoupler outputs at selected times, which is to say we only read them after the outputs have settled to a steady state 1 or 0. That's part of what the clock line going to the flip flops is for. It tells the flip flops when to read the state of the optocoupler outputs.
Now, if we read the optocoupler outputs at the wrong time, outside of the setup and hold time requirements of the flip flops, then there is a problem which needs to be fixed. In some cases it could be fixed by introducing a delay into the clock signal the feeds the USB board. Or in other cases a slight delay could be introduced into the clock signal at the reclocker board by passing the signal through one or both clock inverters on the board.
Also, it is for reasons like the above that it is strongly recommended to have a 100MHz (or better), 2-channel oscilloscope if doing dac work. Otherwise its a guessing game to figure out what's going on. A used scope from ebay could work fine, it just needs to have sufficient bandwidth to reasonably display clock signals. IOW, its not very helpful if a slow scope makes square waves look like sine waves. 100MHz is about the minimum bandwidth needed for dac work because we are dealing with RF square waves.
NOTE: In the above explanation, the word "we" should be read as referring to a nonspecific 3rd person. The words "one" or "someone" could be used instead. Sometimes in math and science the word "we" is commonly used for that purpose. e.g. By adding 3 + 5 we obtain 8.
Now, if we read the optocoupler outputs at the wrong time, outside of the setup and hold time requirements of the flip flops, then there is a problem which needs to be fixed. In some cases it could be fixed by introducing a delay into the clock signal the feeds the USB board. Or in other cases a slight delay could be introduced into the clock signal at the reclocker board by passing the signal through one or both clock inverters on the board.
Also, it is for reasons like the above that it is strongly recommended to have a 100MHz (or better), 2-channel oscilloscope if doing dac work. Otherwise its a guessing game to figure out what's going on. A used scope from ebay could work fine, it just needs to have sufficient bandwidth to reasonably display clock signals. IOW, its not very helpful if a slow scope makes square waves look like sine waves. 100MHz is about the minimum bandwidth needed for dac work because we are dealing with RF square waves.
NOTE: In the above explanation, the word "we" should be read as referring to a nonspecific 3rd person. The words "one" or "someone" could be used instead. Sometimes in math and science the word "we" is commonly used for that purpose. e.g. By adding 3 + 5 we obtain 8.
Last edited:
Mark, inspired by your work, I made a simple "lite" version - not as elaborate as your tool of course, but for my set-up with SMA inputs to SinePI a very handy tool, to compare clocks or do functionality tests on the work bench. I picked same setup for the clocks as in my DAC and used straight forward quality components - I will not start listening-tests on this board. It should stay a handy tool, nothing more. It works with DIL14 and DIL8 clocks btw.
In case anyone wants the Gerbers, just send me a PM to avoid long discussions here. I do not want to crash your thread Mark. So, PM it is for all further questions please 🙂
So, THANKS Mark for your great work and sharing 🙂
In case anyone wants the Gerbers, just send me a PM to avoid long discussions here. I do not want to crash your thread Mark. So, PM it is for all further questions please 🙂
So, THANKS Mark for your great work and sharing 🙂
Have you tried to swap SEPQ by a SEP and the ceramic with a Acrylic 😉 ANd listen to if any difference 🙂
....ouos where I shelved my bullet proof jacket 🤐


....ouos where I shelved my bullet proof jacket 🤐


- Home
- Source & Line
- Digital Line Level
- General Purpose DAC Clock Board