Scott... I should add this info on filters (in case you didn't already know it), that DACs use different filter sets based on the sampling rate of the input signal and typically use a less onerous & obtrusive / better-sounding filter when you feed it a higher sampling rate. So even being able to upsample to 192 could possibly bring some benefits, again after some fiddling with filters. BUT AFAIK, with the PCM51xx chips, the biggest bang is brought by upsampling to 352/384 with the right filters.
Greg in Mississippi
Greg in Mississippi
And final thought before I forget it... to separate the 3.3v for the DAC digital and clocks from the 3.3v from the Pi so you can provide a separate feed, you have to cut 2 sets of traces (3 in total) from the 2 3.3v connections from the Pi. One of these pins feeds the power plane for the DAC digital and the other the power plane for the clocks.
SO, if one has a hankering to feed the little beast with 4 separate power feeds, adding the fourth to provide really clean power to the clocks (LiFePO4 anyone?), this becomes REALLY easy once you've cut those traces.
While this is pretty low on my set of experiments for this card, it IS on the list.
Pix likely tomorrow or Tuesday.
Greg in Mississippi
SO, if one has a hankering to feed the little beast with 4 separate power feeds, adding the fourth to provide really clean power to the clocks (LiFePO4 anyone?), this becomes REALLY easy once you've cut those traces.
While this is pretty low on my set of experiments for this card, it IS on the list.
Pix likely tomorrow or Tuesday.
Greg in Mississippi
Considering that Greg figured most progress would be made on AVDD mods,
another simple option would be to take out the 0R and to hook up a 3.3V LiFePo to AVDD right after the reg.
The absolute max input voltage on AVDD is 3.9V according to the DS.
The usually slightly higher than 3.3V voltage of the LiFePos wouldn't harm -- I guess.
I'm really tempted to try it. It's just another 10 minute exercise. 🙄
I something like that should work as permanent solution, a relay for syncing the different supplies during power-on cycle might be required.
another simple option would be to take out the 0R and to hook up a 3.3V LiFePo to AVDD right after the reg.
The absolute max input voltage on AVDD is 3.9V according to the DS.
The usually slightly higher than 3.3V voltage of the LiFePos wouldn't harm -- I guess.
I'm really tempted to try it. It's just another 10 minute exercise. 🙄
I something like that should work as permanent solution, a relay for syncing the different supplies during power-on cycle might be required.
Last edited:
Some updates and answers...
First, @Soundcheck, on why I did not feed 3.3v direct from the reg to the AVDD, I am concerned about the additional wire and trace length that will add. My experience has been that digital power is really path length critical and that would add over an inch to the path. I still may try it in the future, but my gut-feel is that keeping the on-board regulator there is probably a good idea.
Second, I let it burn-in more today and then tried an idea that occurred to me last night.... I mentioned earlier in the thread that I have separate clean and dirty AC paths in my setups. I went back and forth yesterday on plugging the Pi's PS into either side... both had some strengths, but I finally settled on it into the dirty side. But as I was shutting down the system last night I saw that I had another choice, on the clean side, but before the PS Audio P10 regenerator. AND after today's additional burn-in, I tried it there and it's the best location. That tells me I need to do better noise filtering on the dirty-side power supplies and their AC cords. But with it setup this way, it has gotten another good jump in quality. Nope, it doesn't challenge either of my Soekris DAM DAC setups or the highly modified/abused Sony HAP Z1-ES. But it is definitely giving the EVS-modified DVD player a run in many aspects, sounding very tuneful and bouncy with good separation between instruments/voices, but lacking a good bit of the inner detail and 'ambience' of the more sophisticated spread (remember, it started out as $45 add-on board!)
Here's some pix... and I put a couple more in OPC's Wire board thread in the Vendor's Bazaar section.
Next steps... first, give it a bunch of burn-in time. My experience with MLCC caps (most of the ceramics on the board) is that a couple of weeks of continuous running is a minimum. And do more noise-filtering on my LMS server and network setup on the dirty side. Then some SW and configuration tweaks and maybe some additional filtering caps, maybe some of the .33F caps Joe Rasmussen recommended in his recent Sigma-Delta DAC mod thread (after I confirm they won't break the regulators!). And if I still think it holds promise over time, I'll try feeding the clocks with another separate 3.3v and MAYBE direct 3.3v for the AVDD.
I'm not expecting to make a world-beater out of this inexpensive little DAC. But I am curious on where I can take it, given its humble, but actually pretty sophisticated beginnings!
Soundcheck, I'll be very curious on how it goes if you try the battery on the AVDD setup. See the 'Soekris DAM DAC' thread on the 'Tir Na HiFi' forum for some useful info on using LiFePo4 cells with gear such as this.
More after a few days of burn-in.
Greg in Mississippi
First, @Soundcheck, on why I did not feed 3.3v direct from the reg to the AVDD, I am concerned about the additional wire and trace length that will add. My experience has been that digital power is really path length critical and that would add over an inch to the path. I still may try it in the future, but my gut-feel is that keeping the on-board regulator there is probably a good idea.
Second, I let it burn-in more today and then tried an idea that occurred to me last night.... I mentioned earlier in the thread that I have separate clean and dirty AC paths in my setups. I went back and forth yesterday on plugging the Pi's PS into either side... both had some strengths, but I finally settled on it into the dirty side. But as I was shutting down the system last night I saw that I had another choice, on the clean side, but before the PS Audio P10 regenerator. AND after today's additional burn-in, I tried it there and it's the best location. That tells me I need to do better noise filtering on the dirty-side power supplies and their AC cords. But with it setup this way, it has gotten another good jump in quality. Nope, it doesn't challenge either of my Soekris DAM DAC setups or the highly modified/abused Sony HAP Z1-ES. But it is definitely giving the EVS-modified DVD player a run in many aspects, sounding very tuneful and bouncy with good separation between instruments/voices, but lacking a good bit of the inner detail and 'ambience' of the more sophisticated spread (remember, it started out as $45 add-on board!)
Here's some pix... and I put a couple more in OPC's Wire board thread in the Vendor's Bazaar section.
Next steps... first, give it a bunch of burn-in time. My experience with MLCC caps (most of the ceramics on the board) is that a couple of weeks of continuous running is a minimum. And do more noise-filtering on my LMS server and network setup on the dirty side. Then some SW and configuration tweaks and maybe some additional filtering caps, maybe some of the .33F caps Joe Rasmussen recommended in his recent Sigma-Delta DAC mod thread (after I confirm they won't break the regulators!). And if I still think it holds promise over time, I'll try feeding the clocks with another separate 3.3v and MAYBE direct 3.3v for the AVDD.
I'm not expecting to make a world-beater out of this inexpensive little DAC. But I am curious on where I can take it, given its humble, but actually pretty sophisticated beginnings!
Soundcheck, I'll be very curious on how it goes if you try the battery on the AVDD setup. See the 'Soekris DAM DAC' thread on the 'Tir Na HiFi' forum for some useful info on using LiFePo4 cells with gear such as this.
More after a few days of burn-in.
Greg in Mississippi
Attachments
I've been running several projects with LiFePos over a couple of years. I still have a bunch of them around.
Usually they were better than 2-stage regulated stock stuff. The batteries were also better then my 3.3V TeddyRegs.
There's a guy feeding the Soekris DAC with several LiFePos and at least 5 inch wires attached and is still reporting good results.
The extra trace length and battery noise competes against shared supplies and multi level regulation and endless ground loops, poor tracing....
What I never liked about batteries was the associated charging hazzle and the more complex power-on setup .
#####
BTW. I see you're still running these crappy RCA jacks and the output filters?? Can't believe it.. You should take them out, or just pass them by. These are a bottlenecks and limit your progress.
###
It would be kind of irritating to me, if this 40€ DAC+20€ in mods would be able approach the Soekris DAC performance level, which under certain conditions even seems to outperform a Yggdrasil DAC.
#####
And I think, from what I read, the sound characteristics you're looking for are usually associated to the Sabre based PI DACs. Perhaps the Hifiberry is a bet on the wrong horse - at least from a DAC chip perspective.
Usually they were better than 2-stage regulated stock stuff. The batteries were also better then my 3.3V TeddyRegs.
There's a guy feeding the Soekris DAC with several LiFePos and at least 5 inch wires attached and is still reporting good results.
The extra trace length and battery noise competes against shared supplies and multi level regulation and endless ground loops, poor tracing....
What I never liked about batteries was the associated charging hazzle and the more complex power-on setup .
#####
BTW. I see you're still running these crappy RCA jacks and the output filters?? Can't believe it.. You should take them out, or just pass them by. These are a bottlenecks and limit your progress.
###
It would be kind of irritating to me, if this 40€ DAC+20€ in mods would be able approach the Soekris DAC performance level, which under certain conditions even seems to outperform a Yggdrasil DAC.
#####
And I think, from what I read, the sound characteristics you're looking for are usually associated to the Sabre based PI DACs. Perhaps the Hifiberry is a bet on the wrong horse - at least from a DAC chip perspective.
Last edited:
Just listening to the Mamboberry on my desktop Adam A5X. Temporary power supply is an Anker battery.
I'm running the kernel 4.4.2 with Clives blck_integer_patch and resample to 96khz PCM.
That should get me into Integer mode.
On LMS I usually run PCM streams, no flac!!
That requires the Marcoc1712 patched squeezelite. That patch enables squeezelite to read streaminfo from PCM headers. That's required to run upsampled PCM streams (conversion rule 2) on LMS. Without that patch you need to run upsampled flacs (first conversion rule), which is IMO the 2nd best option. Running similar conversions locally on MPD based PI2 systems generates some decent load. My system still runs at 0%-0,7% according to top!
I also wouldn't recommend to do the resampling locally on the PI2 with squeezelite.
My /etc/squeezeboxserver/custom-convert.conf conversion rules:
These are showing my preferred upsampling settings.
Note:
I use the Ubuntu supplied sox binary! I copy that one to the LMS server directory:
/usr/share/squeezeboxserver/Bin/x86_64-linux/ . LMS is not known for providing the most up2date binaries!! On older/outdated LMS installations (e.g. certain LMS NAS installations) above rules might not work!
Quick resume:
My motivation to fiddle around with LiFePos on the HifiBerry just went down the drain.
And it's just the Anker supply attached to the MB/PI2 combo.
Thx a lot for sharing the patch Clive.
I'm running the kernel 4.4.2 with Clives blck_integer_patch and resample to 96khz PCM.
That should get me into Integer mode.
On LMS I usually run PCM streams, no flac!!
That requires the Marcoc1712 patched squeezelite. That patch enables squeezelite to read streaminfo from PCM headers. That's required to run upsampled PCM streams (conversion rule 2) on LMS. Without that patch you need to run upsampled flacs (first conversion rule), which is IMO the 2nd best option. Running similar conversions locally on MPD based PI2 systems generates some decent load. My system still runs at 0%-0,7% according to top!
I also wouldn't recommend to do the resampling locally on the PI2 with squeezelite.
My /etc/squeezeboxserver/custom-convert.conf conversion rules:
These are showing my preferred upsampling settings.
Code:
flc flc * *
# F
[sox] -D -q -t flac $FILE$ -t flac -e signed -C 0 -b 24 - rate -v -I -b 98 96000
flc pcm * *
# F
[sox] -D -q -t flac $FILE$ -t wavpcm -e signed -b 24 - rate -v -I -b 98 96000Note:
I use the Ubuntu supplied sox binary! I copy that one to the LMS server directory:
/usr/share/squeezeboxserver/Bin/x86_64-linux/ . LMS is not known for providing the most up2date binaries!! On older/outdated LMS installations (e.g. certain LMS NAS installations) above rules might not work!
Quick resume:
My motivation to fiddle around with LiFePos on the HifiBerry just went down the drain.
And it's just the Anker supply attached to the MB/PI2 combo.
Thx a lot for sharing the patch Clive.
Last edited:
Hi,
soon ill be receiving my PCM5102DAC, is it good idea to power up the DAC using isolated AC DC 5v regulator, instead of using 5V on GPIO ?
thanks in advance.
soon ill be receiving my PCM5102DAC, is it good idea to power up the DAC using isolated AC DC 5v regulator, instead of using 5V on GPIO ?
thanks in advance.
LOL, I don't think we'll get there with this little DAC card. I listened more last night AND spent some time going back to my EVS-modified DVD player and the R-Pi->Soekris setup. While the modified HFBD+P did show some good traits yesterday and last night, it is clearly a level or two below the DVD player in overall quality based on a head-to-head matchup. After what I heard earlier I was expecting it to be closer, but was not to be.
Again, I'll give it some burn-in time and see where it ends up, but based on what I heard last night, I don't think it will come up enough in performance to get to the level of that DVD player. And the R-Pi->Soekris DAM was clearly a level or two above the EVS-modified DVD player.
If the pretty extensive work I've already done on the HFBD+P didn't get it closer, I don't think we'll get it there, no way no how.
LOL... I know their setups look pretty hap-hazard. Still, they are pretty tweaky people who seem to listen carefully and are very excited with their results using LiFePo4 cells on their computer sources and Soekris and other DACs.
You may not have caught this, but what they are doing is using the cells as a local buffer for a low-spec supply. So for example on a 3.3v feed, they will setup a good, but nothing-special 3.3v supply and put an LiFePo4 in parallel with it, leaving the 3.3v supply on all the time. This eliminates on-going charging hassle, but the turn-on/turn-off is a little trickier and the cells must be disconnected when off to keep them from discharging (and permanently damaged).
For truth, it still looks like too much hassle and breaks too many conventions for my tastes. So my current path is to try and get my setups working as well as possible with mains power, then try some selected battery feeds (using the Ti Na HiFi configurations) to see if how close I came. If really close, I'll stick with mains. If not and if further tweaking of my mains setups doesn't get me sufficiently close in performance, then I'll consider moving over to battery.
Actually none of the setups I'm comparing to the HFBD+P use Sabre-chips... the DVD player uses the PCM1752 (what appears to be a custom chip for Panasonic as it and its datasheet is not generally available, but is probably similar to other TI chips of that period), the Sony HAP Z1-ES uses dual mono PCM1795 chips, and of course the Soekris.
I did try a TP Buffalo-32 for a month or so some years back when I built one up for a friend... while it did some things well and caused me to make some changes in my then-AK4399-based DAC, ultimately the Buff-32 implementation was not for me.
I did replace that AK4399-based setup with an EUVL ES9022 DAC card and used that for many years. I still have it around, but after mods, the Sony HAP Z1-ES beat it handily.
So actually, I am mostly compared the HFBD+P's PCM5122 with other TI DAC chip-based setups.
I DO have some Sabre-based DAC card setups to try using better I2S sources (and in synchronous-clocking mode) than the really-tweaked motherboard with linear ATX & modified Juli@ card I'd used on the EUVL ES9022 card, but none of them are running yet. AND I also have AKM, Cirrus, and Phillips TDA1541 & TDA1543 setups to try too! I am definitely NOT a Sabre fan-boy!
OTOH, maybe the ES9023-DAC-chipped MamboBerry will show more promise after some equivalent mods and I'll fall in love with it? We'll see after it arrives.
Yah, you are right they are still there. I DID replace the ceramic chip caps on the HFBD+P with PPS SMD film (hard to see that in the pix), which I've found to be a significant upgrade for this application.
OTOH, ALL of the other setups, the modified DVD player, the Sony HAP Z1-ES, and my Soekris setups are using stock or near-stock (as far as configuration) filter setups (still even using the ceramic filter caps on the Soekris cards) and ALL have equally crappy RCA jacks (The DVD player has the stock nickel-plated jacks, the HAPZ has the stock gold-plated, but nothing special stock jacks, and the R-Pi->Soekris is using a pair of cheap gold-plated jacks I pulled out of a Njoe Tjoeb modified CD player when I upgraded the jacks in that unit.
I DO agree that good jacks make a difference and when I get to the final configurations, my setups all get the pricey WBT Next-Gen RCAs for single-end or good Neutrik or Vampire XLR's for balanced. But none of the units I'm comparing against have anything better than the stock HFBD+P jacks, so I think I'm getting a good apples-to-apples comparison. If the modified HFBD+P can't best these other units with their poor-quality filters and jacks, I doubt it will after removing the filters and upgrading the jacks.
Finally, I'm looking at setting up a bench burn-in setup for this modified HFBD+P.... mainly because the Soekris is so much better I don't want to take it out to break in this lesser DAC.
Later!
Greg in Mississippi
Last edited:
Greg, Maybe you should get another board to test with... once you separate the power you may not need so many bypass caps.
I'm getting pretty good results just separating the AVDD and using linear power on that, and then linear power thru the pi GPIO (I removed the 0Ohm resistor)
Also I tweak the pi, use a ZERO, set LMS to convert everything to PCM (even mp3), with picoreplayer I set a huge stream buffer...(420MB) so the entire song fills the RAM (no streaming), also you can disable the HDMI port (saves 28mA).
Anyway I'm researching using supercapacitors for each of the sections (and for other projects) to avoid expensive linear power supplies.
Basically I'm gonna hit this from another angle--
I'm getting pretty good results just separating the AVDD and using linear power on that, and then linear power thru the pi GPIO (I removed the 0Ohm resistor)
Also I tweak the pi, use a ZERO, set LMS to convert everything to PCM (even mp3), with picoreplayer I set a huge stream buffer...(420MB) so the entire song fills the RAM (no streaming), also you can disable the HDMI port (saves 28mA).
Anyway I'm researching using supercapacitors for each of the sections (and for other projects) to avoid expensive linear power supplies.
Basically I'm gonna hit this from another angle--
Rolled my own dac for the Zero. It uses two 3.3V regulators, has the option of powering the usb from the Zero, and the pcm5142 from an auxiliary input. Pogo pins work great for data transfer.
I used a modded mini2496 dac(ak4396) until now and the zero dac is just as good if not better!
Scott,
Thank for the reply AND tips. I'm really looking forward to hearing how your work with supercapacitors comes out.
On my adding bypasses, honestly I've only added 4 caps to the DAC card... a 560uf at the input of each separate supply (nothing special AFAIK on the value, just the largest capacity / smallest size I had in my boxes) AND the 33uf between the clocks. I added the other 33uf because I was replacing a 10uf MLCC with 3 stacked 1uf PPS caps and wanted to get up or above the original capacitor's value. Everything else were replacements, mostly replacing MLCC with PPS SMD film caps.
I again want to thank Soundcheck & you for your Pi configuration & setup suggestions. I will try them next after I get a bench break-in setup running. One significant architectural difference between this setup and the Soekris (or many other high-end-ish DACs) is the HFBD+P does not have isolation and reclocking after receiving the I2S on the DAC board. While I think that would be a great addition to a new version of this DAC, I'm not planning to try to cobble that in here. BUT my experience has been that in the absence of isolation and reclocking, computer source SW, HW, & configuration play a larger role, so there may be some significant improvements in SQ available applying Soundcheck's & your configuration & setup suggestions. Honestly, while I've applied Soundcheck's LMS server setup suggestions, my Pi is running a stock PiCorePlayer load at this time.
I'm pretty much a neophyte on Linux tweaking, so I may ask some dumb questions here... sorry, always feel free to provide links where I might read and learn if appropriate.
A question for Soundcheck, you, and anyone else poking around in here... does anyone know of anyone doing any other HW or configuration changes to the Pi that might be worthwhile to try here?
And finally, Scott, did you find going to the Zero to be an improvement over the std Pi?
Greg in Mississippi
P.S. In my setup I only play FLAC or occasionally WAV files, with the FLAC decoded on the LMS server.
P.P.S. And I'm not planning to get another HFBD+P card at this time. I can go back to the stock-ish HFBD+P two-supply setup very quickly with a couple of solder joints if I want to compare again. AND I have an IQAudio card here and a MamboBerry on the way. That will be enough Pi DAC HAT cards for me unless I get a breakthrough with this or the MamboBerry and start getting significantly better sound out of them. As for the IQAudio card, we'll see if I even give it a try. I'm a bit discouraged on the results so far. I might try the IQAudio card stock, but am leaning towards just putting it back in a box and putting it up for sale. With the results I'm getting so far with a card that DOES provide a better reference clock to the Pi, I don't have high hopes for one that uses the stock Pi MCLK. I decided not to even give the Durio Audio cards a try for that same reason.
Hi there.
PiCorePlayer wouldn't be my 1st choice. Though I guess you guys won't have many more options.
BTW.
The streaming buffer set to 450MB is not the full story.
On the PI2 I recommend to try
"-b 50000:800000"
with 50MB for the streaming buffer and 800MB for the output buffer.
That's much more effective!!
squeezelite runs in bulk-decoding mode this way. And not just in bulk-streaming mode.
That means all conversions will be executed to the end at once and stored in the output buffer. You'll see a load in the range of 20% for a couple of seconds and then the load drops down close to 0%.
800MB equals around 17 minutes of 96khz PCM in 32bit.
Enjoy.
PiCorePlayer wouldn't be my 1st choice. Though I guess you guys won't have many more options.
BTW.
The streaming buffer set to 450MB is not the full story.
On the PI2 I recommend to try
"-b 50000:800000"
with 50MB for the streaming buffer and 800MB for the output buffer.
That's much more effective!!
squeezelite runs in bulk-decoding mode this way. And not just in bulk-streaming mode.
That means all conversions will be executed to the end at once and stored in the output buffer. You'll see a load in the range of 20% for a couple of seconds and then the load drops down close to 0%.
800MB equals around 17 minutes of 96khz PCM in 32bit.
Enjoy.
@ Greg
I consider onboard modifications much more efficient than Super/Ultra-cap installations.
Two of my friends built SuperCap supplies. I wasn't that much impressed.
I consider onboard modifications much more efficient than Super/Ultra-cap installations.
Two of my friends built SuperCap supplies. I wasn't that much impressed.
@Soundcheck,
Dang, I was hoping the supercaps might be a magic bullet!
Thanks for that info and your previous post.
I've been trying to figure out why I'm not getting better sound than I am... the DAC chip should be capable, I'm getting amazingly good sound from the R-Pi feeding a Soekris using the EXACT SAME LMS server and raw DC power supplies, I do give them time to warm back up after a swap before I get critical (it takes 20-30 minutes). All I can come up with are:
1. Needing more break-in (probably true, but I don't expect this to produce miracles).
2. Needing SW/Configuration tweaking (I need to try some of the suggestions you've offered, likely this coming weekend).
3. Maybe the clocking is really not working as expected (I need to see if I can catch that on my very poor scope).
4. Maybe a Pi-to-Pi variation (easy enough to check, but I seriously doubt it! I'll try swapping system SD cards too).
5. Maybe the power feeds are not as good as I expected (this is the first time I've used that LT3042 board, so I don't have a personal history of success with it yet... I can wire up some ADM715x boards which have worked well in the past and try them. Also try feeding the AVDD directly with 3.3v, bypassing the on-board regulator).
6. Maybe the Pi needs more HW tweaks such as shutting off the HDMI (plausible, that's why I asked if either you or Scott knew of anyone molesting the PI with on-board mods).
All of these are possible. None in my experience will turn a pig's ear into silk purse. What am I missing?
Greg in Mississippi
Dang, I was hoping the supercaps might be a magic bullet!
Thanks for that info and your previous post.
I've been trying to figure out why I'm not getting better sound than I am... the DAC chip should be capable, I'm getting amazingly good sound from the R-Pi feeding a Soekris using the EXACT SAME LMS server and raw DC power supplies, I do give them time to warm back up after a swap before I get critical (it takes 20-30 minutes). All I can come up with are:
1. Needing more break-in (probably true, but I don't expect this to produce miracles).
2. Needing SW/Configuration tweaking (I need to try some of the suggestions you've offered, likely this coming weekend).
3. Maybe the clocking is really not working as expected (I need to see if I can catch that on my very poor scope).
4. Maybe a Pi-to-Pi variation (easy enough to check, but I seriously doubt it! I'll try swapping system SD cards too).
5. Maybe the power feeds are not as good as I expected (this is the first time I've used that LT3042 board, so I don't have a personal history of success with it yet... I can wire up some ADM715x boards which have worked well in the past and try them. Also try feeding the AVDD directly with 3.3v, bypassing the on-board regulator).
6. Maybe the Pi needs more HW tweaks such as shutting off the HDMI (plausible, that's why I asked if either you or Scott knew of anyone molesting the PI with on-board mods).
All of these are possible. None in my experience will turn a pig's ear into silk purse. What am I missing?
Greg in Mississippi
Here are a few of my PI2 modifications:
* I'm turning off HDMI since day one. 😉 I'm in line with scramer with that one.
* I'm also modestly overclocking the device.
* Beside that I'm using the "performance" governor. (or even turn power management completely off in the kernel)
* Don't use those cheap heatsinks with even cheaper adhesive foil. These foils are crap.
Either you go without heatsink or u better use Arctic Silver thermal adhesives.
* Applying some extra nearby low ESR capacitance on each power rail (1,8-3,3-5V) should stabilize the PIs operation. Especially when having quite some capacitance bolted to the piggyback DACs and overclocking in place this mod should be considered.
* I don't use the uUSB port for powering the PI. I just use it for testing purposes only.
* Obviously PI powered peripherals (WLAN,USB disks, etc) should be avoided.
* USB cables lead all kind of mess into the PI, better avoid them.
* Use ethernet cables without shield or just cut the shield. Otherwise these lead even more mess into the
PI then the USB cable.
* I'm turning off HDMI since day one. 😉 I'm in line with scramer with that one.
* I'm also modestly overclocking the device.
* Beside that I'm using the "performance" governor. (or even turn power management completely off in the kernel)
* Don't use those cheap heatsinks with even cheaper adhesive foil. These foils are crap.
Either you go without heatsink or u better use Arctic Silver thermal adhesives.
* Applying some extra nearby low ESR capacitance on each power rail (1,8-3,3-5V) should stabilize the PIs operation. Especially when having quite some capacitance bolted to the piggyback DACs and overclocking in place this mod should be considered.
* I don't use the uUSB port for powering the PI. I just use it for testing purposes only.
* Obviously PI powered peripherals (WLAN,USB disks, etc) should be avoided.
* USB cables lead all kind of mess into the PI, better avoid them.
* Use ethernet cables without shield or just cut the shield. Otherwise these lead even more mess into the
PI then the USB cable.