NOT QUITE NEVER AGAIN, THE GROUP BUY DAC BUILD

[jrling]

Wow Greg - that's quite some summary! I will need to re-read it again to get full value for money.

I have only peripheral experience to feed back.

I don't have a Soekris DAC or Rpi like you. I run WTFPlay off a conventional PC board which accepts a 12V DC input direct. (See my signature for details).
I used to run the PC board with a Maplin XM21X AC conventional PSU with good results as others have here, but it is pretty cheap & cheerful (unlike your plethora of excellent AC supplies mentioned).

I then got into LiFePO4 batteries. Only went for the best A123 26650s, so have no experience with other brands. Three banks (3P4S) outputting 12.1V to get a little headroom for voltage loss in connections. No other balancing components. And float-charged them with the Maplin 24/7 (thanks to nige2000). Immediate noticeable step up in SQ.

Then I got into Ultracaps. Again went for the best Maxwells 350F/2.7V, so have not tried Nesscaps. I have 5 UltraCaps in series running at 12.1 VDC (58F effective capacitance) with only 1k resistors across each one as balancing method. Everyone said I would get imbalance, but I check occasionally and never had any issue with A123s or Maxwells balance after a couple years running constantly. The Maxwells are float charged by the A123s and Maplin (all in parallel effectively).

And a more noticeable improvement with them in the PSU. Once initially charged, I leave the whole PSU chain float charging 24/7 and so, just throw the PC switch and I am off to the races.

I only mention all this, as it is my experience with the PSU of a conventional PC player NOT the DAC (which is a very good one but with conventional built-in AC PSU which I don't dare touch). The SQ improvement from hot-rodding the source player PSU is on my experience just as important as the DAC. I read that your Rpi is powered by a good quality AC PSU. I would highly recommend trying float-charged UltraCaps on it; the Maxwells of course at 2.7V max rating are perfect for 5V with pairs in parallel. One pair for testing should be well good enough.

Whilst the purity of a DAC PSU is of course most important, I figure from my experience, that the purity of the source PSU can have as much negative and positive impact on the signal conveyed along the chain even with supposedly foolproof noise filtering devices! I have gone for the minimalist approach with as few PSU components as possible and it really shines through in the listening experience.

Lastly, I must give full credit to frd1996 for the superb minimalist WTFPlay SQ which I have not found to be bettered.

Jonathan

P.S. I need to go away and put my hair-shirt on.
PPS For the full-fat LiFePO4 experience, I am in the low foothills compared to nige's chip amp beast.

[gstew]

Jonathan,

Thanks for the very informative response. Also thanks for the positive comment on my post... mostly I just get TLDR!

You reminded me of some areas that I've had on my list to try, but haven't as yet...

- Powering my player RPi from an Ultracap-output supply. Up to now my best power solutions for RPis were either a pair of 7V-set Uptone Audio LPS-1.2 paralleled by an MPAudio dual 5V 3||LT3045 board into a stock RPi or the same front end, but with the MPAudio board output at 6V feeding a modified RPi using LDOVR.com's RPi Mezzanine board supplying linear-regulated 5V/3.3V/1.8V rails and removing the onboard DC-DC converter chip. I had thought I'd first do that when Allo's new USBBridge Signature, a more-extensively linear regulated RPi substitute became available. BUT with your jog, I tried powering my stock RPi with the new Allo Shanti, a dual 5V (+3A & +1A) supply with a direct-connected dual 3.3F Ultracap pair at the output. It immediately beat my standard AC-connected supply handily (a K&K Audio low-votage supply, an LT3083-regulator board with some diode and cap mods). I then tried a direct-connected dual 350F Ultracap pair charged by the K&K... better than the K&K alone, different than the Shanti alone, but not necessarily better. Finally I used the Shanti to float-charge the pre-charged to 5V 350F Ultracap pair and that's an overall winner. I do need to try that with an LPS-1.2 as the charger, we'll see if that beats the Shanti (not sure it will... will report).

- Charging Ultracaps from LiFePO4. I've seen positive comments before, yours jogged me to move this to the top of the queue. What I'm planning to do is try this combo for powering the 3.3V rails of Ian's GB DAC and FiFoPi. My plan is to pre-charge the Ultracaps to 3.3V before connecting them to the batteries, as I'll be using Ian's LiFePO4 power supply board and it does not support float charging. Again, will report.

I also was happy to hear of your experience with a string of >2 Ultracaps balanced by only 1K R's. I remember some comments about this (here?) and have been cautious about trying it. I will now.

Thanks again for your response & comments.

Greg in MIssissippi

P.S. I've also been following WTFPlayer and if I ever get back into motherboard computer audio, it is on my list. My issue is that I've been working to stay out of the USB and S/PDIF digital connection world and only use closely-connected I2S DACs. While the ESS Juli@ provided a means to do that in the motherboard world, it is now very old tech and the PCI-E version is near impossible to source (and few MBs today have PCI connectors). The Pink Faun I2S Bridge is the only game I know of for motherboard I2S (anyone know of another option?) and as I remember is limited to 192 and no DSD (neither of these are fatal to the way I play music today, but I'm not going to drop 325 euros on a new device that doesn't include these). So I've stuck with RPi (and soon equivalents) and SDCard players as sources.

P.P.S. I'm also only using A123 26650s, dealing with possible issues of non-pedigree cells is not worth it for me. OTOH, IF I ever attempt to build a monster LiFePO4-powered amp like Nige's, I'll learn to deal with them!

[jrling]

Wow again! Great write-up of your endeavours.

Whilst I fully admit that I have not tried any of the items you list and so cannot give definitive opinions, my views are therefore based more on my theoretical approach which is very much to start off with the absolute minimal components in the chain and try and see if they work without issue.

On that basis, I offer further comments on your experiments:

Finally I used the Shanti to float-charge the pre-charged to 5V 350F Ultracap pair and that's an overall winner. I do need to try that with an LPS-1.2 as the charger, we'll see if that beats the Shanti (not sure it will... will report).

I am not surprised that combo came out best, as it is the most direct and uses 350F UltraCaps rather than 3.3F in your first example which in series only gives 1.67F capacity, compared to the series Maxwells @ 175F.

I would try the 1K balancing resistors on your pair of Maxwells. Apart from safety, keeping the pair balanced has to be a good thing.

Re running the Maxwells off LiFePO4 batteries -

What I'm planning to do is try this combo for powering the 3.3V rails of Ian's GB DAC and FiFoPi. My plan is to pre-charge the Ultracaps to 3.3V before connecting them to the batteries

I wasn't sure whether you meant charging the UltraCaps to 3.3V in parallel or 1.75V each in series? As the Maxwells are rated to 2.7V, taking them to 3.3V could be fatal and running them at 1.75V each would work, but I think would be sub-optimal.

My experience using LiFePO4 feeding Maxwell 350F UltraCaps is that the PSU feeding them is not critical. I use a basic PSU which costs £40 new and is nothing special at all, to float charge my PSU. You can easily test this for yourself, by charging everything up with your choice of PSU, play some music and switch off the PSU and see if there is any improvement to SQ. Or any deterioration when you switch on the charger. I cannot detect any myself with my PC board PSU.

PPS - Not only do I keep to A123 LiFePO4 brand, but I also stick with Maxwell UltraCaps. There are one or two other US-made brands of SuperCaps that seem good quality (Eaton for instance), but I have seen tests of various brands on YouTube with lamentably bad results mainly from Chinese brands.https://www.youtube.com/watch?v=OglAziNcTxA

I would say that the brand of LiFePO4 that Nigel uses for his chip amp seem to be very good quality and no issues. Made in Europe.

[randytsuch]

I've been building a supercap ps for a little project.
Did some research this weekend, and I think 1k is too large a value for the balancing resistor for these caps.
Its supposed to draw a lot more current then the cap leakage current

Leakage is .75ma Note that alternate BCAP0360 P270 S18 has same leakage
Aim for around 10x more, or 7.5ma or 0.0075 a with balancing resistors
For a 3.3V power supply, each cap sees 3.3/2 = 1.65V
v=ir r =v/i = 1.65/.0075 = 220 ohm max recommended resistance.
Can go smaller. Smaller will equalize voltage faster which is better. Disadvantage is higher power and higher current.

If use 1000 ohm resistor, i = 1.65/1000=.00165 a =1.65ma p = .0027w Just about 2x more than leakage
If use 100 ohm resistor, i = 1.65/100=.0165 a =16.5ma p = .027w
if 75 ohm resistors, i = 1.65/75=.022 a = 22ma p = 0.036w
if 50 ohm resistors, i = 1.65/50=.033 a = 33ma p = 0.054w

On voltage, IMHO there is no real downside to having 1.65V on a 2.7V cap. On the other hand, if you put lower voltage on a battery, then you charge it less and it won't work as well. For 3.3V lifepo4, 3.3v is perfect but I never liked have two 2.5V batteries to make a 5V supply. Using two supercaps should work much better.

Randy

[jrling]

Hi Randy

Thanks for this detailed and informative post with supporting calculations.
TBH I asked a pro I trust and he said 1K. I also saw another hi-fier who had built a string like mine using 1K.

I have found the string of 5 Maxwell BCA 350 caps with the 1K resistors to be balanced to within a hundredth of a volt any time I check them. So I left it at that!

I am keeping them float charged 24/7, so the variation in voltage is minimal.

On voltage, IMHO there is no real downside to having 1.65V on a 2.7V cap. On the other hand, if you put lower voltage on a battery, then you charge it less and it won't work as well. For 3.3V lifepo4, 3.3v is perfect but I never liked have two 2.5V batteries to make a 5V supply. Using two supercaps should work much better.

Totally agree that 2.5V on a 3.3V LiFePO4 is sub-optimal and right at the bottom of their rated working voltage and Ultracaps at 2.5V are perfect for 5V.

One downside of running the UltraCaps @ 1.65V instead of their nominal 2.7V rating, will reduce their capacity (linearly) by 55% from 350F to 192.5F.

Jonathan

[randytsuch]

So FYI, my previous post was based on information from this link
https://www.tecategroup.com/products/ul ... or-FAQ.php
The part on balancing is a ways down the link

I never heard of this company, but they make supercaps, and what they say about passive balancing makes sense.
So I just applied ohms law to the max leakage spec of the cap, and came up with 220 ohms.
If you adjust the leakage for voltage (ratio of 2.7 vs 1.65), which seems reasonable, R changes to 360 ohms.

This only becomes a factor if the leakage is different for the two caps in a pair. As long as the leakage is close, you wont see much variance.
I calculated that a 8% leakage difference will result is 0.1v diff on the caps. So if the leakage is within a percent, then you will get a millivolt diff or so. And that's based on worst case leakage, and you should never see the worst case numbers, pretty likely you will get 1/2 of worst case or less.

In a way, using a lower resistance balancing r is masking an issue. If you have two caps with different leakage, you won't know. I don't know if different leakage means other parameters are also different, so I don't know if this matters either.


One last thing to add, on batteries vs caps. Caps are linear for stored energy vs voltage, batteries are not, you need to be close the the recommended V to get the stored energy, it drops off quite rapidly as V goes down. That's why 2.5V on a 3.3V battery is really not good.

[jrling]

Interesting article and very informative. I haven't heard of Tecate either, but one of their products uses Maxwell SuperCaps!

Whilst theoretically all of the variations you mention could make a difference, I suspect that in the grand scheme of things they are absolutely infinitesimal to the SQ. Or putting it another way, the variations in other components in the audio chain would be orders of magnitude more in their actual measurement variance and the SQ effect.

To get to within 1/100 V on each of five SuperCaps in my series chain with passive balancing, seems like a first class result to my amateur ear.
The combo of LiFePO4 float charging Maxwell 350F SuperCaps takes a lot of beating at any price.

Happy Days

Jonathan

PS One measurement neither of us have mentioned, is ESR. Again SuperCaps are really low ESR (even in series which adds ESR for each cap).

[gstew]

jrling and randytsuch,

Thanks for your informative posts. They are invaluable!

Some comments...

- All of my UC setups use or are based on Ian Canada's UCPi boards. I have 3 of them here, which are introduced in post 163 and details including schematic are in post 187 of this thread:

https://www.diyaudio.com/forums/power-s ... ply-4.html

His circuit uses 1K balancing resistors across the 2 Ultracaps and 2 325uF/2.7V or 350uF/2.7V parts in series for use at either 3.3V or 5V. BTW, I would never take these the 2-UC string above 5.2V or so! My DIY versions use the same Maxwell 350uF/2.7V parts as I use on his boards. I omit the input series choke and the jumper-controlled on-off along with the indicator LED's and associated parts. For the DIY P-T-P units, I use the standard RC car/airplane connectors that I use on most of my higher-current supplies and some 18g solid copper wire as the main connecting parts with the intention of keeping the paths short and fat (sorta like me!).

I posted some pix of my build process here along with Ian's schematic:

You'll see I used Kapton tape liberally to insulate possible short points and later will post pix of how I insulated/shielded them against ANY easy potential shorts... I consider UC's used in this manner for experienced users (like present company) only due to the massive instantaneous current available from a short... hence a lot of care to make sure I don't allow those shorts to happen!

Oh and physically damped them, because that's what I do on almost all of my builds. In a couple of the pix you'll see 2 3-pin sockets... one is for a charging regulator (I'm using Ian's LT3042 units here) and one is for a voltage monitor (I'm using these from Amazon:

https://www.amazon.com/dp/B00YALV0NG/re ... h3Ch7saQ9D )


- Randytsuch's balancing analysis is VERY useful. I will be replacing the 1K resistors with lower values. I did find some imbalance in about a third of the DIY P-T-P units I built. I don't know if that was due to inherent leakage differences in the parts or if soldering the heavy copper wire to the terminals damaged them somewhat. I was more careful when soldering with the 2nd batch I made. The highest imbalance was almost .5V when I first charged the unit... leaving in on charge that diminished to <.1V after a couple of days. The others that were imbalanced started out <.1V, typically about .05V. AND the rest were almost exactly in balance.


- So I know about using 1K balancing resistors for 2-UC setups. AND now I know (from you jrling) that scheme will work with longer strings of UCs for higher voltages and (from you randytsuch) that lower values are likely needed. THANKS both of you!!!


- I know I'm going to be the odd man out here, but at least with UCs I have heard differences when using them with different charging supplies AND between being float-charged and just running them down...

1. I mention an almost imperceptable difference (to my ears and in my setups) float-charging Ian's UCPi units with his LT3042 regulator and an Uptone Audio LPS-1.2 set to 7V supplying the raw voltage in my 8/16 post above. These were powering a modified Allo Katana.

2. I also describe in that post how I tried to 'phatten' up the bottom end of the UCPi experience by charging the UCs to the target voltage, then jumpering the charging regulator and feeding them from an MPAudio 6||LT3045 board set to 5V. This DID 'phatten' up the bottom end, but with a loss of definition that had me go back to the UCPi fed from the LPS-1.2 setups.

3. I mention a more significant audible difference float-charging a UCPi with either an Allo Shanti or my modified K&K Audio low voltage supply in my 9/8 post above. This was powering a player RPi.

4. Last night I was trying a UCPi powering a modified RPi (using an LDOVR.com Mezzanine board to replace the on-board DC-DC converter chip for the 3.3V and 1.8V rails). I first tried it float-charged with a Shanti... "good, but not as good as I expected". Then I thought I tried it with the K&K Audio supply. "Hmmm better! That's odd because it's opposite what I got in my other system." Then I saw that the voltage was down to 4V (interesting that the RPi was still running ok). I had forgotten to power-up the K&K supply, so it was running just off the UCs. It was getting late at that time and I didn't get a chance to compare it with the K&K working, will do that in the next few days.

Ian also said he preferred the UCPi with no float-charging setup attached. He just recently sent a short description of how the UCs are working in different mods when being float-charged versus run free (and I think I've seen the same info posted on this board in the past). His assessment is that running free is better. My experience running them float-charged with LPS-1.2 is that the difference I hear is so small (if it even exists) that I have no issue running them that way for my most critical uses.

Float-charging them from LPS-1.2's is an expensive proposition that most won't try... I only do because I bought a bunch of them over the 1st 2 years they were out as they were then my best supplies. I'm VERY happy I have found a new use for them!

I know Ian is looking at a non-float-charging setup for his eventual UC supply. Me, I'm sticking with the LPS-1.2's for float-charging since I have them here!

AND I've only run LiFePO4 cells direct with non-float-charging setups, so I have no experience with them in float-charged configurations.


- On the terminology of Supercaps versus Ultracaps, I don't know if there is an industry common-usage... what I've seen is the terms are used interchangeably.

I've used the term 'Supercap' to describe parts such as the .33F/5.5V Panasonic caps (look like small backup batteries) that Joe Rasmussen mentioned in his controversial DIY Audio threads on 'Alternative DAC Filtering' AND that Allo has used in both their Boss and Katana DACs. These parts are fairly high ESR and are just supplying some additional bulk energy storage. My experience adding them to DAC's and associated digital processing cards (isolators and reclockers) is they add some heft to the bottom end and body to the mids, but take about 1-2 weeks to break in and lose a sandy texture in the highs and upper mids). IMHO, these are worth adding when using more conventional supplies... not sure yet on components that are powered via direct-connected LiFePO4 or UCs.

I've used the term 'Ultracap' to describe parts that are generally (though not always) 1F or higher capacitance and generally low ESR. Because of the fairly wide-band low ESR, they tend to become the dominent source components when used in a power supply with other caps or raw supplies. Though the Allo Shanti shows how when you combine smaller (a pair of 3.3F/2.7V UCs) with a substantial electrolytic filtering component (almost 100,000uF split between the two rails), you can hear the UCs dominating in the middle-to-upper regions of the spectrum and the ECs the lower regions.

I'm not saying this to ask anyone to conform to my useage, but just to explain how I use those terms.


- Finally on the SQ impact of different brand UCs, I did mention in my 8/16 post how I heard an upper-mid/lower-treble emphasis using the NESSCAP parts that I do not hear with the Maxwell parts. I DO have one LPS-1.2 that also uses a similar-line NESSCAP part and also hear a bit of that same emphasis using it powering components close to the DAC or the DAC (so I generally use it on network components). I WON'T be using NESSCAP parts again in the near future. Allo uses AVX parts in their Shanti AND I don't hear any frequency abberations with them. Others use NESSCAP parts and haven't reported what I heard in those two instances... maybe I'm more sensitive to that or that a certain manufacturing process caused that that may not have been in use earlier. BUT based on what I heard (and the cost of a single 350F/2.7V part), I won't be getting any NESSCAP parts again soon!

ENOUGH! NOW past the TLDR limit.

Again, thanks for your great posts full of great info jrling and randytsuch!

Greg in Mississippi

[gstew]

I did further comparisons of a float-charged pair of 350F/2.7V UCs powering both a modified RPi and a stock one in my two audio setups this weekend (continuing point #4 above).

The results were 'interesting'. Not in that there were not good results, but they weren't always what I expected. LOL, this is a microcosm of my experiences with power supplies!

First, it was VERY nice that powered by the UC pairs, I could quickly swap different float-charging supplies while the setups ran without any issues.

I did hear a difference between running free and float-charging. It was not so large as to make me not want to float-charge them from a convenience standpoint. AND powering an RPi, you are only going to get 10-15 minutes running free, so that's not really an option.

The dual 7V or 9V LPS-1.2 with the 5V MPAudio board was the cleanest and most detailed in the mids-treble AND the closest sonically overall to having the UCPi run free. BUT musically, I preferred the Shanti and/or the modified K&K to that setup. They added a subtle bit of heft in the lower mids and upper bass that added PRAT and tunefulness.

My results were interesting with the Shanti and the modified K&K. In my upstairs setup, the K&K won. In my downstairs one, the Shanti won. It was not a huge difference between the two in either system, but a clear preference for one over the other in each. Note that I used two different Shanti and modified K&K, one each in each system, so they were different units and each has slight differences from the one in the other system. AND of course the systems, while sharing some identical equipment and similar equipment where different, DID have different DACs, are physically different gear, and in different rooms with different levels of sound absorbing materials on the surfaces.

Also the RPi player SBC in the downstairs setup is stock and the one upstairs is one I modified to replace the DC-DC-converter generated 1.8V & 3.3V with linear-regulated rails.

In the upstairs setup I was running one of Ian's GB DACs powered by his LiFePO4 supply. It also has a 5V output that is regulated from the DC power input and uses an LT3042 with a pass transistor. Before I tried the UC pair buffer, I didn't like it as much as the bare Shanti or bare modified K&K. With the UC pair it stayed behind them. Still, adding that UC buffer moved it up significantly over it without the UCs and at least above the modified K&K alone, possibly both.

I also did a quick trial powering Ian's GB DAC and connected FiFoPi with the 2 3.3V float-charged 350F/2.7V UCs fed from either the 2 3.3V rails of Ian's LiFePO4 supply or from a pair of 7V LPS-1.2. I preferred the LPS-1.2 charged setup to the LiFePO4 one. I didn't hear anything obviously wrong using the batteries, but also didn't hear anything to encourage further use of this setup. These UCPi were new and had not been used before. I left them connected to the pair of Uptone Audio LPS-1.2 to keep them charged up. I'll let them run for a week or so, then try this again.

I pre-charged the UC pairs to the target voltages using current-limited LT3042 regulators. When using a supply with higher than the target voltage (like the 7V LPS-1.2) I left those installed. When using same-target-voltage float-charging supplies, I removed the regulators and jumpered where they were. Somewhat of a pain, but seriously sonically worthwhile.

Of course, these are my results with my setups, my preferences, and my ears. YMMV!

Greg in Mississippi

[jrling]

Greg - I admire your determination and persistence. Impressive.
Perhaps now you could have some time to listen to some music without worrying about the PSU?! You must have a great result (from just about every permutation you have tried).

Best
Jonathan