LT3042 and LT3045 are very similar apart from the current capability. Based on datasheets LT3042 has slightly better specs but in practice implementation details have more impact.
IMHO and IME, its not simply about having LT3045 or LT3042. They are low noise, but not necessarily ideally linear. For Andrea oscillators and Acko oscillators, IME its more about the linearity of reasonably low noise power supplies, and about the squaring circuit and its voltage regulation. Please see the following thread for more info: https://www.diyaudio.com/community/threads/general-purpose-dac-clock-board.413001/
BTW, I have kept in touch with Andrea Mori and Roberto with regard to my findings. They know about the above thread. Acko does too.
BTW, I have kept in touch with Andrea Mori and Roberto with regard to my findings. They know about the above thread. Acko does too.
Last edited:
Mark, I went to that thread but I could not find much about the power supply used there. What would you suggest to use with Andrea's clocks + FIFO?
swak, I do not use Andrea's FIFO any more. I don't use his dacs either. Reason is the ferrites and nonlinear caps in all the power supplies and the squaring circuits. Again, Andrea and Roberto know all about my findings, and they know more than I have posted here in the forum.
Right now I only use asynchronous USB with I2SoverUSB. The rest of the dac system is entirely synchronous so I don't need a FIFO. If I wanted to add SPDIF/TOSLINK support then I might consider using an Iancanada FIFO_Pi or else an ASRC. If using a FIFO_Pi it would be because I could more easily isolate it from the rest of my system than if using Andrea's FIFO. IOW, I don't trust either FIFO to be clean at this point, so I would design to allow for that.
If I wanted to use Andrea's R2R dac board (which I have here), then I would probably modify it somewhat similar to what I did with the experimental Andrea DSD dac I was using here.
Right now for my own use, it is a MarcelvdG RTZ DSD dac board, with the latest official PCM2DSD firmware, and with an unobtainium prototype transformer output stage. It Uses my own squaring board design, and Andrea or preferably Acko oscillators (the particular RTZ dac doesn't work optimally with clock doublers). Power supplies for the oscillators are simple LT1023 or LT1024 regulators (or pre-regulators) with aftermarket trimpots. The squaring circuit has three onboard LDO regulators, and is pre-regulated by an external LT1023 or LT1024 power supply board.
Right now I only use asynchronous USB with I2SoverUSB. The rest of the dac system is entirely synchronous so I don't need a FIFO. If I wanted to add SPDIF/TOSLINK support then I might consider using an Iancanada FIFO_Pi or else an ASRC. If using a FIFO_Pi it would be because I could more easily isolate it from the rest of my system than if using Andrea's FIFO. IOW, I don't trust either FIFO to be clean at this point, so I would design to allow for that.
If I wanted to use Andrea's R2R dac board (which I have here), then I would probably modify it somewhat similar to what I did with the experimental Andrea DSD dac I was using here.
Right now for my own use, it is a MarcelvdG RTZ DSD dac board, with the latest official PCM2DSD firmware, and with an unobtainium prototype transformer output stage. It Uses my own squaring board design, and Andrea or preferably Acko oscillators (the particular RTZ dac doesn't work optimally with clock doublers). Power supplies for the oscillators are simple LT1023 or LT1024 regulators (or pre-regulators) with aftermarket trimpots. The squaring circuit has three onboard LDO regulators, and is pre-regulated by an external LT1023 or LT1024 power supply board.
Last edited:
I get it. Could you still suggest something that, as you suggested above, works well (linearly enough) with those clocks?
IIUC, Acko is using TP7A470x regulators for his clock oscillators. IMHO those sound better that LT104x even though they aren't as low noise.
I have also been using LT1763 and LP5907MFX-3.3 regulators, but with the performance somewhat tuned using load resistors to ground. For LT1763 there is a low noise mod for them that dddac mentioned in a post in the clock board thread. With both of these regulators I use MU caps for input and output. They aren't as low ESR as other caps, but IME they are more linear. IMHO the linearity can be more important in some cases than low ESR and ESL. Depends on the particular circuitry. For very high speed digital logic, or maybe even a dac clocked at 100MHz, it might turn out to be very different.
I have also been using LT1763 and LP5907MFX-3.3 regulators, but with the performance somewhat tuned using load resistors to ground. For LT1763 there is a low noise mod for them that dddac mentioned in a post in the clock board thread. With both of these regulators I use MU caps for input and output. They aren't as low ESR as other caps, but IME they are more linear. IMHO the linearity can be more important in some cases than low ESR and ESL. Depends on the particular circuitry. For very high speed digital logic, or maybe even a dac clocked at 100MHz, it might turn out to be very different.
I agree with Mark - you can read the post on the LT1763 mod here:
http://blog.dddac.com/dddac1794-mk3-on-board-regulator/
http://blog.dddac.com/dddac1794-mk3-on-board-regulator/
@dddac, IIRC, I did try many years ago with increased bypass cap on the LT1073 (100n, I think) but load regulation or output accuracy @3.3V fixed type was poor. Contacted LT support and they advised the bypass cap was too large and when changed back to 10n , all good. Interesting to see yours all fine with even larger cap🤔
Last edited:
You are Right, i published the measurements…. I did not test accuracy, as it is not important for the application, output impedance has improved. Load regulation as in sharp load jumps are also not so relevant for a constant DC current application imo
With respect to ideal linearity, do you all have a view whether the Salas Ultrabibs would fit that profile vs. the low noise LT3045? Mark, you noted you compared your supplies against Tentlabs shunts in the past.
Shunt regulators can sound good and have useful Z-out vs frequency characteristics. However, the final regulators for some clock related circuitry should arguably be close to the load devices and on the same ground plane. IOW, a final regulator on a separate board probably would be too far away for best results. Whether or not a RF LDO (LDO with wide regulation bandwidth extending up into RF frequencies) is needed, IMHO depends on the overall power distribution system (PDS) design. Regulators don't necessarily have to be as fast as the circuitry they supply power to, and for very high speed circuitry that's impossible anyway. (There are no RF LDOs with regulation bandwidth extending up into high GHz frequencies. Therefore the PDS for those systems is designed to work well with slower regulators. IOW, it's all doable.) Also, bypass caps peppered around a board to lower power plane impedance sometimes can only be operated above self-resonance, but they still help keep PDS impedance low anyway. Power plane capacitance to adjacent ground plane layers also helps to control PDS impedance as seen by load devices. https://www.sigcon.com/Pubs/straight/resonance.htm
My only point here is there is more than one way to design a PDS for some given load circuitry. Different designers may have their own preferences in that regard. I use what I think is good, which is what I can speak to most directly.
Last edited:
That discusses digital logic where operating above resonance is not an issue. But as that same paper states analog circuits are different. Clocks and clock buffers are not digital logic gates so applying design principles suited for digital logic gates may not be the optimal approach.
I tried a few different approaches to power supply with Andres clocks and IMHO found it not all that impactful to sound quality. So in my opinion you would be hard pressed to find a difference in sound quality between LT3045 vs LT3042 in this application.
FWIW, I use a TPS7A4700 that I had on hand from another project with his 5MHz DRIXO.
FWIW, I use a TPS7A4700 that I had on hand from another project with his 5MHz DRIXO.
^Agree about DRIXO oscillator relative insensitivity to PS. However, existing Andrea squaring circuits have issues with power, and with overall design. Same for the existing FIFO board clocking related stuff. Sorry, but that was determined by a series of experiments in the lab. Again, I already shared all the details with Andrea's colleague, Roberto. I'm pretty sure Andrea knows about too, despite his extended illness.
However, other clock modules such as Crystek and Iancanada clocks do show sensitivity to PDS design. Same for clock buffers and reclocking flip-flops.
That said, the above do not appear to be especially sensitive to phase of PDS bypass caps. IOW, they are not analog filters that need precision phase response in the filter sections (also to be clear, phase response is not the same thing at all as oscillator phase noise).
However, other clock modules such as Crystek and Iancanada clocks do show sensitivity to PDS design. Same for clock buffers and reclocking flip-flops.
That said, the above do not appear to be especially sensitive to phase of PDS bypass caps. IOW, they are not analog filters that need precision phase response in the filter sections (also to be clear, phase response is not the same thing at all as oscillator phase noise).
Last edited:
Good to know. He did suggest to me getting the supply "as clean as possible" so I always assumed he was hinting this is critical.
I am very sorry to hear this, I hope the best for him.
I have the DAC lite so I need the FIFO and I use his 5/6 MHz clocks. FWIW, I am pretty satisfied. From what you have written here and on the thread you referred to it seems the issues you've found can't be corrected (mods) and that you have concentrated on DSD DACs.
One of the last things I remember reading about his DSD solution (the balanced version) is that they were using the slowest clocks and several doublers, the reason being that the results were better that way. It seems you have started with higher freq. clocks instead.
I might try DSD at some point (out of curiosity) and If I do, I will certainly take a close look at all you have contributed.
Last edited:
The problems can be helped to some extent. I removed all the ferrites from the power filters on the boards where I could find them. Also removed nonlinear caps in board power inlet filters. On the DSD dac boards I removed all the ceramic caps and all the tantalum caps. I knew I was taking a chance, but it had gotten to the point where Marcel's RTZ dac was sounding clearly better, so I had to find some solution or give up. I replaced all the tantalums with Nichicon UKZ bypassed with .01uf Wima MKS, and replaced all the ceramics with MU caps. Left existing Panasonic SMD film caps in place. That almost brought Andrea's DSD dac boards up to the sound of Marcel's RTZ dac. However, there was still some problem, which I suspect was due to clock skew from the clock buffers driving the output flip-flops. Couldn't fix that without a board redesign so decided to decommission the dac.
Problem with Andrea squaring circuits was a little different for each type. Ferrites and nonlinear caps were a problem, but the problems went deeper too. Board redesigns would have been necessary in both cases. However, in one case the squaring ICs were EOL, so no point trying to fix that one.
For the MCLK signal used to drive the DSD dac boards, it worked best if I sent my own clean copies of MCLK to the dac boards. The clock signals going into the FIFO board where then only used internally to that board.
The issue with doublers is mostly limited to the case of Marcel's RTZ DSD dac. It doesn't have enough FIFDAC taps to be sufficiently insensitive to doublers. Thus, for best results Acko 22/24MHz clocks are required. However, there there have been some issues with those too. I have been working with Acko to get that resolved to our mutual satisfaction, which I think we figured out how to do. Also, Acko now has clocks with internal squaring available.
Last edited:
I did within my capabilities make my supply "as clean as possible". But based on experience it was not something I felt worth obsessing over. In practice that meant I did not just pick up a switching power supply. I built a reasonable linear supply using stuff on hand... a good Plitron transformer, Panasonic caps, a LM317 pre reg, multiple clclc filters and the final TPS7A4700 has Blackgate caps which some claim makes an improvement. It's been powered for a few years and it makes music that pleases me ever since. 🙂