Subsequent changes:
It's sounding better and better. A slight periodic high frequency crackle is present. Wondering if the root cause is noise originating from PC source?
- built universal shunt to power the ProtoDAC
- resected USB-A wire to enable discrete PS to power the XMOS 384 Combo input
- moved LM317 PS + LT3042 over to feed the input board
- input voltage for second PS provided from #2 primary / secondary
It's sounding better and better. A slight periodic high frequency crackle is present. Wondering if the root cause is noise originating from PC source?
Did you remove L1 from the USB board, and power it from clean +5v? Or maybe made a modified USB cable to bring in clean +5v power?
Why don't you have a dedicated ground wire for each of the I2S signals? Sharing one ground wire for all of them is bad at RF frequencies.
BTW, we can tell you are sorta learning as you go here. Lots of large capacitors, voltage regulators dangling in the air and connected by wires that are too long, etc.
Why don't you have a dedicated ground wire for each of the I2S signals? Sharing one ground wire for all of them is bad at RF frequencies.
BTW, we can tell you are sorta learning as you go here. Lots of large capacitors, voltage regulators dangling in the air and connected by wires that are too long, etc.
Best if the final voltage regulators are on the same board and same ground plane as the load. Best of every load board has its own dedicated voltage regulators on their own dedicated transformer windings.
A regulator like LT3042 needs be very, very close to the load because it is trying to regulate the output voltage at RF frequencies. Also, layout matters and so does the type of capacitor used for Cset. It needs to be a very low leakage tantalum type. Some of those ebay and or aliexpress boards have the cheapest, worst X5R or lessor, ceramic caps on the board. Only the LT3042 chip is any good, the rest of the stuff is often junk. That's how they can get the prices so low.
For I2S signals, each I2S line should have a dedicated ground wire loosely twisted together with it and grounded at both ends (try to keep them as short as possible), or else use coax such as u.fl cables for each I2S signal.
Also, USB boards often radiate EMI/RFI from the tops of the big chips. May be best of the bottom side of the USB board (the ground plane) is pointed towards the dac and other sensitive electronics. Sometimes you can move around the USB board aiming the top at other boards and see if the sound changes.
Best if you have a schematic and pics of how you have hooked everything together. The more we know about the details the more we can advise on. Also, you might want to consider better clocking than what you have now. Every dac needs a time reference and a voltage reference. The clocks are the time reference. In addition, I have become pretty anti-ferrite for dac board power distribution systems. Usually its possible to design around ferrites so they are not needed. IME the sound can get cleaner if we try to avoid them when they are not really shown to be needed.
A regulator like LT3042 needs be very, very close to the load because it is trying to regulate the output voltage at RF frequencies. Also, layout matters and so does the type of capacitor used for Cset. It needs to be a very low leakage tantalum type. Some of those ebay and or aliexpress boards have the cheapest, worst X5R or lessor, ceramic caps on the board. Only the LT3042 chip is any good, the rest of the stuff is often junk. That's how they can get the prices so low.
For I2S signals, each I2S line should have a dedicated ground wire loosely twisted together with it and grounded at both ends (try to keep them as short as possible), or else use coax such as u.fl cables for each I2S signal.
Also, USB boards often radiate EMI/RFI from the tops of the big chips. May be best of the bottom side of the USB board (the ground plane) is pointed towards the dac and other sensitive electronics. Sometimes you can move around the USB board aiming the top at other boards and see if the sound changes.
Best if you have a schematic and pics of how you have hooked everything together. The more we know about the details the more we can advise on. Also, you might want to consider better clocking than what you have now. Every dac needs a time reference and a voltage reference. The clocks are the time reference. In addition, I have become pretty anti-ferrite for dac board power distribution systems. Usually its possible to design around ferrites so they are not needed. IME the sound can get cleaner if we try to avoid them when they are not really shown to be needed.
From eyeballing your pic of the LT3042 module, Cset looks to be C5 and Rset, R2 (33k in your pic post #14). Definitely Cset shouldn't be a class 2 ceramic as it looks to be, this destroys the low noise virtues of the LT3042. Worth checking that the LT3042 module output voltage is what you think it should be - R2 will be ~50k for 5V out.
Hi Jim,
I'm not sure what would cause the loss of the USB connection identifier with the AmyAlice in place. My thoughts were more focused with comparison of the ProtoDAC powered by an SMPS via the AmyAlice with the Amanero independently powered compared to an all-linear setup through the SuperReg. In my various implementations of SMPS with AmyAlice cards I really haven't had a first-rate linear setup to yield a good comparison, just good, unverified, outcomes. Clearing your current situation up will lead me toward building a similar setup whereby I would attach that combination to my RPi via USB rather than the direct I2S stacking connection. The hope there is that it would be as good as the various reclocker solutions used stacked between RPis and ProtoDACs
Skip
I'm not sure what would cause the loss of the USB connection identifier with the AmyAlice in place. My thoughts were more focused with comparison of the ProtoDAC powered by an SMPS via the AmyAlice with the Amanero independently powered compared to an all-linear setup through the SuperReg. In my various implementations of SMPS with AmyAlice cards I really haven't had a first-rate linear setup to yield a good comparison, just good, unverified, outcomes. Clearing your current situation up will lead me toward building a similar setup whereby I would attach that combination to my RPi via USB rather than the direct I2S stacking connection. The hope there is that it would be as good as the various reclocker solutions used stacked between RPis and ProtoDACs
Skip
Mark/Richard/Skip - your kind thoughts read and acknowledged. Will proceed as follows and report again:
Thanks again,
Jim
- Remove the offending LT3042 board. It may be well doing more harm than good.
- Remove the simple LM317 power supply, and replace it with another 5V universal shunt, matching the most recent addition.
Thanks again,
Jim
Next step(s): finished the second Walt Jung universal shunt PS, this to power the Amanero 384 combo clone. Rectifier working, new shunt measures 4.94vdc output. LED lit on shunt PS.
Then, I connect the modified USB cable which supplies the 384 combo input PCB with the 5v and the shunt PS board’s LED dies.
When I disable the power through the cable the shunt LED illuminates.
I don’t understand why this occurs.
Thinking I might be better off removing L1 on the combo board and powering the board directly, but I don’t know with certainty where to place the + and - 5vdc
Then, I connect the modified USB cable which supplies the 384 combo input PCB with the 5v and the shunt PS board’s LED dies.
When I disable the power through the cable the shunt LED illuminates.
I don’t understand why this occurs.
Thinking I might be better off removing L1 on the combo board and powering the board directly, but I don’t know with certainty where to place the + and - 5vdc
Sounds like the USB board needs more startup current than the voltage regulator can supply?
However, if you do want to remove L1, the +5v goes to the load side pad of L1 (not the USB connector side). The ground for the 5v power supply can be soldered to ground plane on the back of the USB board. Okay to scape off some solder mask and tin the copper to make a place to solder on a power supply ground wire.
All the above having been said, if the USB board is powered from a linear +5v power supply, it probably doesn't have to be a super good 5v supply. There are voltage regulators on the USB board that will drop some or all of it down to 3.3v anyway.
The best thing to do would probably be to get rid of the Amanero and or Amanero clone and get a JL Sounds I2SoverUSB instead. Configure it to run from two mutually isolated (isolated from each other and from ground) 5v supplies. One supply for the dirty (USB bus side), and one supply for the clean side where the clocks are located and which connects to your dac. SQ should probably improve right away. (In that case please note there is a link on the bottom of I2SoverUSB that needs to be cut if the dirty USB side will be powered from a local isolated linear power supply.)
However, if you do want to remove L1, the +5v goes to the load side pad of L1 (not the USB connector side). The ground for the 5v power supply can be soldered to ground plane on the back of the USB board. Okay to scape off some solder mask and tin the copper to make a place to solder on a power supply ground wire.
All the above having been said, if the USB board is powered from a linear +5v power supply, it probably doesn't have to be a super good 5v supply. There are voltage regulators on the USB board that will drop some or all of it down to 3.3v anyway.
The best thing to do would probably be to get rid of the Amanero and or Amanero clone and get a JL Sounds I2SoverUSB instead. Configure it to run from two mutually isolated (isolated from each other and from ground) 5v supplies. One supply for the dirty (USB bus side), and one supply for the clean side where the clocks are located and which connects to your dac. SQ should probably improve right away. (In that case please note there is a link on the bottom of I2SoverUSB that needs to be cut if the dirty USB side will be powered from a local isolated linear power supply.)
Last edited:
The current through the shunt must be greater than the current through by the load. What scheme did you use and how much current is needed for the Amanero?
The maximum current I got from Walt's shunt, at least for now, is 500mA(on the load).
Last edited:
@Markw4 and @grunf thanks very much to you both for your input, super helpful.
Re: inadequate current, first move was to change to a larger transformer, hoping that would correct the issue with the XMOS 384 Combo board. I can see 600mA out of the individual shunt on that supply (unloaded). Which resulted is a slight flickering of the LED on the shunt board, but still no recognition of the board on the Win11 desktop PC sound setting. Swapping out the modified USB cable to standard, it would play again, but still with that slight periodic high frequency crackle. I can't find any apparent current specification, could be an imperfect board.
Which led to swapping out the input board for another know good one, but this one non-combo, XMOS. This one would apparently power up as measured by the shunt LED, but still no sound or recognition from the PC settings software. Interestingly, swapping to a non-modified USB cable (powering the input pcb from the PC), I've resolved the crackle. So that's solved in some sense, though I really do wish to pursue separate power for the input card.
Which leads to likely acquiring the JL Sounds board. I do have one of those running with a Miro AD 1862 so understand the quality and isolation benefits.
Re: inadequate current, first move was to change to a larger transformer, hoping that would correct the issue with the XMOS 384 Combo board. I can see 600mA out of the individual shunt on that supply (unloaded). Which resulted is a slight flickering of the LED on the shunt board, but still no recognition of the board on the Win11 desktop PC sound setting. Swapping out the modified USB cable to standard, it would play again, but still with that slight periodic high frequency crackle. I can't find any apparent current specification, could be an imperfect board.
Which led to swapping out the input board for another know good one, but this one non-combo, XMOS. This one would apparently power up as measured by the shunt LED, but still no sound or recognition from the PC settings software. Interestingly, swapping to a non-modified USB cable (powering the input pcb from the PC), I've resolved the crackle. So that's solved in some sense, though I really do wish to pursue separate power for the input card.
Which leads to likely acquiring the JL Sounds board. I do have one of those running with a Miro AD 1862 so understand the quality and isolation benefits.
Last edited:
Hi, Just checking in. I'm interested to hear if you had gotten the JL Sounds board. It looks to be a very good choice for this project.
Skip
Skip
Skip -
Thanks for the nudge. Yes the JLS board arrived though I have yet to install it. While waiting for its prolonged delivery, I swapped the 384 Combo input for a coax in IS2 SPDIF. This way I could pair the ProtoDAC with a CD transport with digital coax out (Magnavox 650) in the near field desk system. This setup sounds great, as pleasing as the 1541 Vlad tube DAC in direct comparison.
With a laptop as a source the AD1862 / JLS is also great though the same pre/ACA/holey basket setup.
I may leave the ProtoDAC as is for now and use the new in box JLS with yet another AD1862. For that I want to try Salas shunts and Dark LED I/V. For that awaiting the shunt group buy.
Thanks for the nudge. Yes the JLS board arrived though I have yet to install it. While waiting for its prolonged delivery, I swapped the 384 Combo input for a coax in IS2 SPDIF. This way I could pair the ProtoDAC with a CD transport with digital coax out (Magnavox 650) in the near field desk system. This setup sounds great, as pleasing as the 1541 Vlad tube DAC in direct comparison.
With a laptop as a source the AD1862 / JLS is also great though the same pre/ACA/holey basket setup.
I may leave the ProtoDAC as is for now and use the new in box JLS with yet another AD1862. For that I want to try Salas shunts and Dark LED I/V. For that awaiting the shunt group buy.
Poor aesthetic belies excellent sound. I had a minute at the office for a listen to Sonny Rollins.
I'm not bothered by the aesthetic, we all board like this when trialing. 🙂
However, the i2s connections are too long and should, ideally, be each accompanied by an earth in proximity.
Even more worrying is your 'at the office' comment ! That lash-up is fine on a test bench but should never be used elsewhere... just shuddering looking at that mains wiring.
However, the i2s connections are too long and should, ideally, be each accompanied by an earth in proximity.
Even more worrying is your 'at the office' comment ! That lash-up is fine on a test bench but should never be used elsewhere... just shuddering looking at that mains wiring.
I2S connections are RF. Even when clock frequencies are fairly low, maybe a few tens of MHz, the fast, steep risetimes of the edges imply frequency components up into the hundreds of MHz if not higher. IOW, its really RF!
Thus use coax cable, or at least use short runs of loosely twisted signal and ground wires, with each twisted pair given some separation from the other twisted pairs. For reasons involving impedance matching at RF frequencies its also important that LVCMOS I2S connections be kept as short length as possible.
Thus use coax cable, or at least use short runs of loosely twisted signal and ground wires, with each twisted pair given some separation from the other twisted pairs. For reasons involving impedance matching at RF frequencies its also important that LVCMOS I2S connections be kept as short length as possible.
Hi All, new to diyAudio, and about to get into my first ProtoDAC build. Is it appropriate to add to this @JKiriakis thread, or should I start my own?