Based on your more detailed description, it does appear to be broken in some way. Waiting probably won't help the dac, but maybe you will feel better.
Don't recall if you said how you are powering it?
Also, presume you have double checked all the dip switches, the u.fl clock cable, have tried playing source material from different sample rate families, etc?
Hopefully, you have an oscilloscope?
I do not know if it is a flaw of the battery supply. The smps ps was making such a noise that a few cinch cables that were lying next to the dac on my test bench were receiving it and all the noise was amplified to the speakers. Like you mentioned the noise can be picked up through the air, there are maybe some pcb traces acting like antennas.
Any way, I am not using this smps anymore.
Regarding the AD 797, I want to give the sparkos a second try and have ordered a second one. In the mean time the sound from the opa627 is more fitting ( more smooth) in my system, I might give ad797 another try later on.
I have a ton of capacitors (also film) in stock so I will try the film types to see if it sounds better, but not too much changes and some time to burn in the new components before changing stuff........
Mark, I'll get it on the test bench and scope it out.
Its run off Ian's battery board with a smps powering that. I tried the battery board fronted by a quad pack of 18650s so I know it's not the smps. I threw a pair of lt3045s between battery board and pi, so it's not that 5v rail either.
I have ldo's bypassed on the fifopi, and running in true sync. I'll swap the ufl cable over to a fresh one and check all the dips etc. I've tried 44.1 and 48 families and dsd, played airplay, USB card in pi and various software,iTunes, audivarna etc, the same sound always.
I'll look at it with clean eyes tomorrow. Start with the iv and work my way down.
Maybe it does need something on the avcc.
Its run off Ian's battery board with a smps powering that. I tried the battery board fronted by a quad pack of 18650s so I know it's not the smps. I threw a pair of lt3045s between battery board and pi, so it's not that 5v rail either.
I have ldo's bypassed on the fifopi, and running in true sync. I'll swap the ufl cable over to a fresh one and check all the dips etc. I've tried 44.1 and 48 families and dsd, played airplay, USB card in pi and various software,iTunes, audivarna etc, the same sound always.
I'll look at it with clean eyes tomorrow. Start with the iv and work my way down.
Maybe it does need something on the avcc.
First impressions with supercaps on FIFO Pi
Easy, cheap and positive. Had to share.
1. Put a 1.5F Nesscap 5.4v cap on the PS to clean FIFO Pi. Took days to burn in, but immediate improvement in detail in general across the audible range.
2. The second and most impactful was adding same 1.5F cap right on the 3.3v pins of the 45mHz clock. This cap I burned in for 5 days on a bench supply before install. Add in sound was surprising. Major increase in detail. One of those need to re-listen to your collection to hear new details for the first time kind of events.
I use linear PS, not LifePO4. Powering clock direct, not via FIFO rails. Trickle charging cap with 3 series low noise regs. I suspect putting one on the clock powered by LifePO4 would see similar result.
I will be adding supercaps to both clean and dirty rails of FIFO, and to I2SPCM pcb to see if any further gains are to be had.
Nesscap's are obsolete stock but still sold on digikey. FYI 69mOhm ESR.
Easy, cheap and positive. Had to share.
1. Put a 1.5F Nesscap 5.4v cap on the PS to clean FIFO Pi. Took days to burn in, but immediate improvement in detail in general across the audible range.
2. The second and most impactful was adding same 1.5F cap right on the 3.3v pins of the 45mHz clock. This cap I burned in for 5 days on a bench supply before install. Add in sound was surprising. Major increase in detail. One of those need to re-listen to your collection to hear new details for the first time kind of events.
I use linear PS, not LifePO4. Powering clock direct, not via FIFO rails. Trickle charging cap with 3 series low noise regs. I suspect putting one on the clock powered by LifePO4 would see similar result.
I will be adding supercaps to both clean and dirty rails of FIFO, and to I2SPCM pcb to see if any further gains are to be had.
Nesscap's are obsolete stock but still sold on digikey. FYI 69mOhm ESR.
Attachments
Last edited:
Thanks Wlowes for the tip.
Seems those caps works like a LiPo cell with low enough impedance, cool
Do you feed the Xtals with the external reg of the picture ? If yes, did you have to cut a trace or remove something on the board to disable the embeded reg ?
Did you try the FCA caps without the BG NX ?
Seems those caps works like a LiPo cell with low enough impedance, cool
Do you feed the Xtals with the external reg of the picture ? If yes, did you have to cut a trace or remove something on the board to disable the embeded reg ?
Did you try the FCA caps without the BG NX ?
My pleasure diyiggy!Thanks Wlowes for the tip.
Seems those caps works like a LiPo cell with low enough impedance, cool
Do you feed the Xtals with the external reg of the picture ? If yes, did you have to cut a trace or remove something on the board to disable the embeded reg ?
Did you try the FCA caps without the BG NX ?
I don't know what they bring to the party, but based on several posts, it seemed like they cure something that needs curing around clock circuits.
I do feed the NDK oscillator directly with Ian's TPS7A4700 reg. I removed the power pin from the clock adapter but left the other 3. That means it get's it's own 3.3v supply but shares a ground with FIFO. I should try cutting the ground path as well to see if it is better or worse, but I am unlikely to take that step and leave it to someone more disciplined to try all the permutations. The 3.3v reg is fed by a 4.1v 4700 reg which is fed by a 5v salas reg. I should be getting pretty close to battery power level of noise. This plus other poster experience leads me to suspect that the supercap will equally benefit battery PS's. Dirt cheap and effective.
I have not tried the setup with only the FCA cap. Again my bad. My bet is leaving out the BG will not be a problem. Good thing... they are hard to find.
Pardon my ignorance, I can't find clear info from Google. What is dual mono 9024? Is it a dac?
@ Wlowes,
That s funny, I tried the TI reg chip as well and found it better than the LT3042. Both were external board as your in the picture. My bet is the LT3042 is more difficult when it comes to have good results on an external board. Or maybe the LT works bad with more than the post 4,7 uF decoupling some consider here to be the best decoupling numbrt for this chip...
That s funny, I tried the TI reg chip as well and found it better than the LT3042. Both were external board as your in the picture. My bet is the LT3042 is more difficult when it comes to have good results on an external board. Or maybe the LT works bad with more than the post 4,7 uF decoupling some consider here to be the best decoupling numbrt for this chip...
Right! it is hard to see in the photo because there are wires going from the reg, the BG cap and the supercap.
+of the cap goes to +3.3v pin which in the picture is upper left corner of adapter. This is also the corner where I remove the pin to disconnect VCC from the FIFO.
-of the cap goes to bottom right which is under the cap. I actually just tack onto the copper ground wire going from the reg to the ground pin.
If the TPS7a4700 is better in this application, my use of it was a happy accident. I used it because it was on hand and had worked very well powering the same clock in the acko s03.
There is some chatter on the site that suggests that while the LT3042 measures very well, it is not always perceived to produce best sound around clocks. Only way to tell would be to try them both. Sounds like you have done just that.
I recently had a 8*TDA1387 HAT from Audiophonics on my desk. I was building a HAT tower (the SW part) for a friend of mine.
Just for fun I've been using the FifoPi with the 1387.
I never expected what was happening.
Suddenly instruments sounded like real instruments on my Adam speakers.
I'll soon get my own device delivered. If further tests turn out to confirm my earlier experience, I have to question all these Sabre resp. Delta/Sigma DACs in general.
Yep. Many of us have been in that area with 1543/1541 or R2R projects.
Last edited:
Indeed, I am running FIFOPi with a well sorted 1541a rig. IMHO it is a fine combination. An audio buddy popped by to hear my system. He brought his ESS based OPPO. It has been professionally upgraded with improved power supplies. It was fine, but we were both shocked at how the FIFO/1541a was in a different league. Our impression echoed your description. It is surprising how Phillips was onto something good 35 years ago. It just needed proper implementation that smart guys like ecdesigns and iancanada have figured out.
@sq225917
Regarding the problems you report with the dac: Now that I have mine on the bench and hooked up to a good HPA, I think I am probably hearing about the same as you reported. I didn't remember Katana (with isolator board) sounding that bad, so I removed the FIFO dac and hooked up Katana to the same power supplies (LDO-free linear +-15v and +5v for dac boards, ifi for dirty +5v), and the same HPA. Katana definitely sounds quite a bit less distorted to me, particularly at higher frequencies.
Current status of problem dac: Dual_ES9038Q2M dac AVCC is powered by a NewClassD UWB2 3.3v regulator making AVCC somewhat less suspect for the moment. Output stage opamps still running rather hot, and with less than as much HF filtering as I would prefer. The (presumed) RF heating of the opamps can't be helping SQ. Maybe why some people report preferring transformer outputs(?). Haven't looked into possible clocking issues yet, but currently using NDK 45/49 clocks which are the same types as used in Katana.
Regarding the problems you report with the dac: Now that I have mine on the bench and hooked up to a good HPA, I think I am probably hearing about the same as you reported. I didn't remember Katana (with isolator board) sounding that bad, so I removed the FIFO dac and hooked up Katana to the same power supplies (LDO-free linear +-15v and +5v for dac boards, ifi for dirty +5v), and the same HPA. Katana definitely sounds quite a bit less distorted to me, particularly at higher frequencies.
Current status of problem dac: Dual_ES9038Q2M dac AVCC is powered by a NewClassD UWB2 3.3v regulator making AVCC somewhat less suspect for the moment. Output stage opamps still running rather hot, and with less than as much HF filtering as I would prefer. The (presumed) RF heating of the opamps can't be helping SQ. Maybe why some people report preferring transformer outputs(?). Haven't looked into possible clocking issues yet, but currently using NDK 45/49 clocks which are the same types as used in Katana.
Mark, good to hear my ears aren't failing me. ;-) It's definitely in the higher frequencies 1k+ where it's most obvious. From what you say we can rule out AVCC as the source of the noise.
Strangely my stack crapped out this afternoon. Everything lights up, ESS controller responds to input from Volumio but there's no sound at all on XLR, RCA or headphone out. So either a DAC or output stage issue.
Not the best showing so far.
I have loads of contract work on at the minute so i just can't spare the time to sit down with it on the bench, paying the bills takes priority. Also I have boxes and boxes full of through hole parts I can throw at any project that requires work, with smt I have to order everything in which makes iterative investigation tough.
I wonder if it's as simple as a lack of filtering on the output stage and running at 13.2v that's not helping.
Strangely my stack crapped out this afternoon. Everything lights up, ESS controller responds to input from Volumio but there's no sound at all on XLR, RCA or headphone out. So either a DAC or output stage issue.
Not the best showing so far.
I have loads of contract work on at the minute so i just can't spare the time to sit down with it on the bench, paying the bills takes priority. Also I have boxes and boxes full of through hole parts I can throw at any project that requires work, with smt I have to order everything in which makes iterative investigation tough.
I wonder if it's as simple as a lack of filtering on the output stage and running at 13.2v that's not helping.
Try the controller's master reset option?
Regarding opamp heating, I saw the same effect with my first modded ES9038Q2M dac. My 2nd dac modding effort didn't have the problem, and the only reason I can think of that might account for it is that the output stage is farther away from the dac chip, and the dac chip analog +- outputs for each channel are twisted together and run close to the ground plane on their way down to where the output stage is located. That probably adds a bit RF of filtering. (Twisted wires were sometimes used as 'gimmick' capacitors in radio work.) Maybe helpful to see pic below. Yellow and blue twisted wires are dac outputs for each channel. Loose yellow wires were later connected up to bring Vref to the output stage opamp.
Attachments
Markw4
Just thinking... Any chance you are picking up RF from cell phone or other source? My system is pretty prone to act as an antenna, but I chose to use a lot of unshielded cables, so I expect it. I am noticing I have to turn off my cell phone frequently to rid distortion on one or both channels. Just wondering if there might be a trace tuned to some frequency that is picking up something that then sets the opamps off. Could explain why some people have the issue and others do not? You could put a faraday cage (metal bucket or pot?) over the whole stack and see if there is an effect. Probably a wild idea.
Also, I run a 3b+ Pi which has new and improved wifi. I turn off the wifi. Wonder if it is causing RF to your opamps?
Just thinking... Any chance you are picking up RF from cell phone or other source? My system is pretty prone to act as an antenna, but I chose to use a lot of unshielded cables, so I expect it. I am noticing I have to turn off my cell phone frequently to rid distortion on one or both channels. Just wondering if there might be a trace tuned to some frequency that is picking up something that then sets the opamps off. Could explain why some people have the issue and others do not? You could put a faraday cage (metal bucket or pot?) over the whole stack and see if there is an effect. Probably a wild idea.
Also, I run a 3b+ Pi which has new and improved wifi. I turn off the wifi. Wonder if it is causing RF to your opamps?
Last edited:
First, I am comparing a Dual_ES8938Q2M dac stack to a Katana dac stack, and the degree of difference in SQ is my primary concern. I am pretty sure they can be made much closer to each other in SQ without heroic shielding, but you raise a point I have looked into here before.
Regarding local RF fields, they can cause a little distortion where I am (that I can measure, and which are somewhat dependent on opamps in use), but not to the extent of such heating. There are two main sources of environmental RF in this location: 1) a DECT wireless phone base station that I can power off, and, 2) I am on top of a hill near an airport and overlooking the Sacramento valley for as much roughly 100 miles distant. There are lots of low level RF emissions that I can detect. I can shield pretty effectively by putting devices inside a steel file server case. It is designed to attenuate RF. However, the wireless connections for Volumio might not work with the dacs in there.
IMHO, probably the two biggest noise sources that often aren't controlled for in dacs are power related (grounding, line noise), and from lack of appropriate RF shielding (a seamless steel case is probably best, and is about what Benchmark DAC-3 uses, that and ferrites on all I/O). Please see pic below.
Attachments
Last edited: