|
|
Post by brucew268 on Jul 24, 2023 17:32:24 GMT
Over at DIYAudio, this PSU has impressed a few based on the PSU used in the old Studer900 and are reasily available from China as kits or built up. I've been using one feeding my DAC for the past few years with good results. However, the Chinese board has less than ideal trace sizes so some people have redone the layout for better current/impedance. So I built up one with high quality parts and a CRC filter section on someone's redesigned board and have not quite gotten the results I expected, so thought I'd post here to see what ideas people had on which direction to try first. I've attached the SCHEMATIC HERE and below are the list of primary differences between the two builds: vers 1 Chinese PCB for kits
| vers 2 Enthusiast PCB with much larger traces for pos & neg rail | | DC hookup wire: Pure Silver 18AWG - 5cm | DC hookup wire: OCC copper/PVC 18AWG - 20cm | | Filter: Single cap, 10cm x 1mm silver leads | Filter: CRC + snapins
| | Kemet ALS30… 6800R/63V, ripple 8.2A, ERS 26mOhm (big can) | 2x Epcos B41231 6800R/63V, ripple 5.16A, ERS?
| | Output caps: Nichicon FG 470uF + Mundorf EVO Supreme 0.1uF | Output caps: Epcos B41231 2200uF + Mundorf EVO Supr 0.1uF |
|
| | vers1 SQ: Greater clarity & purity in vocals & those freq, both in leading edge & tone. | vers 2 SQ: Slightly lower noise floor & separation
|
So I'm considering the possible changes to the vers 2 build to retain the lower noise floor and yet get back what I really liked in vers 1. Note that this is the same circuit but with the differences noted above. -Silver vs OCC Copper hookup?
-Hookup wire length?
-Thinner copper hookup wire?
-Use cap with Higher ripple/lower ERS?
-Drop to single filter cap?
-Better filter resistors in the CRC?
-Lower value output electrolytic?
-Nichicon Muse vs Epcos B41231 on output? Ideas? |
|
|
|
Post by misterc on Jul 24, 2023 17:35:34 GMT
I'm guessing the the noise floor of the Chinese one. Try swaping out the dc silver wire first before you try anything else Bruce
|
|
|
|
Post by brucew268 on Jul 24, 2023 17:40:05 GMT
I'm guessing the the noise floor of the Chinese one. Try swaping out the dc silver wire first before you try anything else Bruce Yes, but I'm trying to fix the version 2, enthusiast board! I suppose you're suggesting what I like in the chinese board is actually unnatural sharpness from the silver? The music sure sounds more realistic and natural in balance in the chinese board. |
|
|
|
Post by misterc on Jul 24, 2023 17:49:28 GMT
I have a strong reasoning why, just try that first before anything else it will save a lot of guess work here imho
|
|
Bigman80
Grandmaster
The HiFi Bear/Audioaddicts/Bigbottle Owner
Posts: 16,398
|
Post by Bigman80 on Jul 24, 2023 17:49:30 GMT
I'm guessing the the noise floor of the Chinese one. Try swaping out the dc silver wire first before you try anything else Bruce Yes, but I'm trying to fix the version 2, enthusiast board! I suppose you're suggesting what I like in the chinese board is actually unnatural sharpness from the silver? The music sure sounds more realistic and natural in balance in the chinese board. How much time have you got on it? The new PCB. |
|
|
|
Post by brucew268 on Jul 24, 2023 17:51:13 GMT
Yes, but I'm trying to fix the version 2, enthusiast board! I suppose you're suggesting what I like in the chinese board is actually unnatural sharpness from the silver? The music sure sounds more realistic and natural in balance in the chinese board. How much time have you got on it? The new PCB. 150 hours. It was obvious in the first 100 hours that it needed more. |
|
Bigman80
Grandmaster
The HiFi Bear/Audioaddicts/Bigbottle Owner
Posts: 16,398
|
Post by Bigman80 on Jul 24, 2023 18:02:38 GMT
How much time have you got on it? The new PCB. 150 hours. It was obvious in the first 100 hours that it needed more. Ok, could need even more. |
|
|
|
Post by misterc on Jul 24, 2023 18:30:04 GMT
just built a dc load for it Bruce that is capaable of 12 amps that'll break it in quicker
|
|
|
|
Post by brucew268 on Jul 24, 2023 18:48:09 GMT
150 hours. It was obvious in the first 100 hours that it needed more. Ok, could need even more. just built a dc load for it Bruce that is capaable of 12 amps that'll break it in quicker Yeah, could be. I'll run it a few more days. In this current build, it's throttled beyond 2A and the DAC generally takes 0.75-1A and I'm working it out with Isotek's burn-in CD. Interestingly: - the original circuit in the Studer was spec'd with the output cap of 2200uF but most of the chinese kits spec 470uF.
- in vers 1 there was a noticeably SQ difference between using Pan FC, Elna Silmic, and Muse FG in the 470uF ouput.
- With Chris noting a diff on his AN DAC with different materials and gauge's, I wondered about 20cm vs 5cm of DC hookup.
- The CRC should be a lower noise floor than a single filter capacitor, but when I put a CRC filter in my preamp, various HQ resistors resulted in different sound.
|
|
|
|
Post by misterc on Jul 24, 2023 19:56:06 GMT
I know why it does but if I told you chaps where would be the fun finding out?
Audio is about learning and experimenting that's the fun imho
|
|
Bigman80
Grandmaster
The HiFi Bear/Audioaddicts/Bigbottle Owner
Posts: 16,398
|
Post by Bigman80 on Jul 24, 2023 20:12:23 GMT
I know why it does but if I told you chaps where would be the fun finding out? Audio is about learning and experimenting that's the fun imho Pretty sure he wouldn't be asking if he was having fun, Tony. 🤷 |
|
|
|
Post by brucew268 on Jul 24, 2023 21:00:47 GMT
I know why it does but if I told you chaps where would be the fun finding out? Audio is about learning and experimenting that's the fun imho Pretty sure he wouldn't be asking if he was having fun, Tony. 🤷 Experimenting and learning is interesting, to a point. Not wanting to waste too much time heading the wrong direction or wear out this PCB, or my ears, with my basic soldering kit.
I should be able to easily bypass the CR in the CRC and just feed one cap to see if that makes a positive difference. But noise floor can be a subtle thing not quickly noticed by ear, so that could waste time if it is a tradeoff.
I can change the output cap to a Muse FG 470uF since that works well in the version 1 board, but would be nice to know what the effect on the circuit and SQ of 470uF vs 2200uF, or for that matter the same capacitance but significantly different body sizes. But there is a limit to how much desoldering and resoldering this board will tolerate with my available equipment.
If I really should be messing with different wire gauge/material, the above will be a waste of time. |
|
|
|
Post by brucew268 on Jul 26, 2023 20:32:19 GMT
The plan: 1. I’d used low ESR caps as Output Cap and other caps trailing the Voltage regulator.
| Generally not desirable in a PSU circuit. Change.
| 2. I’d configured a CRC filter in the power capacitor position.
| Test bypassing the resistors to see effect on SQ, freq balance, noise floor. Possibly move back to single power capacitor.
| 3. DC hookup wire is a bit long even though twisted. Effect on impedance?
| Consider redoing layout to allow for very short DC hookup though longer AC primaries.
| 4. Change power cap for much larger ripple rating 5A > 10+A
| x
|
Listen thoroughly after each change before going to next item.  20230726_150339 by Bruce W, on Flickr |
|
Bigman80
Grandmaster
The HiFi Bear/Audioaddicts/Bigbottle Owner
Posts: 16,398
|
Post by Bigman80 on Jul 26, 2023 21:20:22 GMT
The plan: 1. I’d used low ESR caps as Output Cap and other caps trailing the Voltage regulator.
| Generally not desirable in a PSU circuit. Change.
| 2. I’d configured a CRC filter in the power capacitor position.
| Test bypassing the resistors to see effect on SQ, freq balance, noise floor. Possibly move back to single power capacitor.
| 3. DC hookup wire is a bit long even though twisted. Effect on impedance?
| Consider redoing layout to allow for very short DC hookup though longer AC primaries.
| 4. Change power cap for much larger ripple rating 5A > 10+A
| x
|
Listen thoroughly after each change before going to next item. by Bruce W, on Flickr Just my observations: 1) AC wires are too long. Move traffo closer to the IEC socket and make them as short as possible. I'd also make the secondaries as short as possible too. 2) I'd take that Mundorf cap out. I have never been a fan of Mundorf and they never perform as well in PSUs as something like a WIMA DC-Link cap. (IMO) 3) Those three Wurth caps, what are the values? I found that Rubycon ZLG are king of the hill (spec wise) in certain values and Panasonic FM better in others. Wurth had no value better than either in the same Capacitance that I needed. I'd do a full spec op on them and see if you can get something with lower ESR and Higher PSRR 4) That DC wiring look incredibly big...what is the voltage/amp going through it? Have you got a link to the PCB? |
|
|
|
Post by brucew268 on Jul 26, 2023 21:47:50 GMT
Just my observations: 1) AC wires are too long. Move traffo closer to the IEC socket and make them as short as possible. I'd also make the secondaries as short as possible too. 2) I'd take that Mundorf cap out. I have never been a fan of Mundorf and they never perform as well in PSUs as something like a WIMA DC-Link cap. (IMO) 3) Those three Wurth caps, what are the values? I found that Rubycon ZLG are king of the hill (spec wise) in certain values and Panasonic FM better in others. Wurth had no value better than either in the same Capacitance that I needed. I'd do a full spec op on them and see if you can get something with lower ESR and Higher PSRR 4) That DC wiring look incredibly big...what is the voltage/amp going through it? Have you got a link to the PCB? 1. Are there standardised best practices on whether DC or AC wires are more important to be shorter? Usually one can't do both and intuitively I lean towards DC needing to be short. If I recall, from a voltage/power point of view on a macro scale that's true.
3. Your examples are all low ESR which is not desirable after the voltage regs, and will be my first test in this situation.
4. 18AWG/1mm solid, which is used especially in this length in order to minimise impedance thinking overkill shouldn't harm.
|
|
Bigman80
Grandmaster
The HiFi Bear/Audioaddicts/Bigbottle Owner
Posts: 16,398
|
Post by Bigman80 on Jul 27, 2023 6:07:16 GMT
1) Not sure, but i always minimise AC lengths.
3) Ok, fair enough
4) Shouldn't be an issue
|
|
|
|
Post by misterc on Jul 27, 2023 9:25:11 GMT
Impressed Bruce, you are one of the very few I've seen who actually realises that the aluminum casework is anodized and you have scrapped the coating off from a proper earth connection  What is main pass tranny? This looks like it might actually handled 4-6 amps output. Shunt resistors what value & wattage looks like 5W + all that heat lol I can see over a dozen items that can effect the result, however the main one is it is overly complex for what it achieves imho You are correct having ultra low Z caps in that area is counter productive |
|
Bigman80
Grandmaster
The HiFi Bear/Audioaddicts/Bigbottle Owner
Posts: 16,398
|
Post by Bigman80 on Jul 27, 2023 11:30:10 GMT
Impressed Bruce, you are one of the very few I've seen who actually realises that the aluminum casework is anodized and you have scrapped the coating off from a proper earth connection What is main pass tranny? This looks like it might actually handled 4-6 amps output. Shunt resistors what value & wattage looks like 5W + all that heat lol I can see over a dozen items that can effect the result, however the main one is it is overly complex for what it achieves imho You are correct having ultra low Z caps in that area is counter productiveDoesn't that depend on the regulator being used? |
|
|
|
Post by misterc on Jul 27, 2023 12:00:47 GMT
You are correct having ultra low Z caps in that area is counter productive Doesn't that depend on the regulator being used? Many factors Oli, trace width/length/regulator output pass tranny/ straight voltage reference/compartitor style, decoupling/ capacitor size/ board stack up/where a dual rail etc
Personal I do not use low Z caps post regs
|
|
Bigman80
Grandmaster
The HiFi Bear/Audioaddicts/Bigbottle Owner
Posts: 16,398
|
Post by Bigman80 on Jul 27, 2023 12:31:54 GMT
Doesn't that depend on the regulator being used? Many factors Oli, trace width/length/regulator output pass tranny/ straight voltage reference/compartitor style, decoupling/ capacitor size/ board stack up/where a dual rail etc
Personal I do not use low Z caps post regs
Fair enough. I shall consult the encyclopedia and see what I think, considering I ONLY use these types of caps. |
|
|
|
Post by misterc on Jul 27, 2023 12:58:13 GMT
One of the many bonuses of having TDR in real time is that you can see the effect of various components in the circuit path on instantiations impedances and how & why this effects. Width traces are grouped into to two categories, greater trace widths the lower Z but slower transmission times (high dielectric constant) narrow the width the higher the Z but faster transmission times. its about measured impedance characteristics. the we need to look at how close the transmissions lines are together in relation to distributing 'aggressor noise' from one transmission line to another depending on signal/power/powers/veers and placement. For genuinely quiet circuitry TDR is pretty much essential. Interesting fact of the day with signal pathways (where circuit board traces/cables or complete circuits) these are ALL transmissions lines and all signals electrical travel at the speed of light roughly 186K miles per second so if we have equal dielectric constants and impedance pathways they pretty roughly the signal travels at 150mm per nanosecond
|
|
|
|
Post by brucew268 on Jul 31, 2023 17:08:27 GMT
Changes: - Before changing out the low ESR caps trailing the voltage regulator, Tony recommended to put film 0.1uF caps (Wima) across the leads of the first and last electrolytics.
- Rather than shortening the AC primaries and secondaries I just put a proper twist on them.
- I also noted that I had a filter resistor on both the pos and negative traces between the power caps, and the filter resistor is generally only used on the power trace of each rail, so I replaced the neg one with straight wire.
Fifty hours in and it sounded rather impressive... until realising that it's all higher frequency stuff and the mids and lows lack body, weight, and texture -- so not as frequency balanced and thus not as musically engaging. Listening again to the less pimped build on the Chinese small trace board, it confirms that it has in presentation what the newer one lacks. BTW: I did take the opportunity to give better attention to the wire twists on the Chinese kit PSU as well, with good results. Since Wima caps can take a bit to break in, I gave it another 50 hours on the Isotek burn-in disc, totalling 100 hours on the Wima film caps and 250 hours on the Epcos electrolytics. Same result, mids and bass lacking body and balance. So, I’ll probably change out the final (output) low ESR electrolytic 2200uF with a Nichicon 470uF and replace the small low ESR Wurth caps with Nichicon too and hear how it sounds. BTW: over at the DIY enthusiast site I asked about low ESR vs “normal” caps in the positions trailing the voltage regulators and generally got the answer that I could just put a film cap following them in order to keep the ESR from going too low. Or I could use “General Use”, Long life, or High Reliability caps. Whether the film cap often used at output for a better sense of space and imaging fits that purpose or not is unclear and I’m treating them differently. I did think the additional Wima caps across the leads benefitted.  20230727_143423 by Bruce Warren, on Flickr  20230729_114756 by Bruce Warren, on Flickr BTW: the older chinese board below  20230727_164406 by Bruce Warren, on Flickr |
|
Bigman80
Grandmaster
The HiFi Bear/Audioaddicts/Bigbottle Owner
Posts: 16,398
|
Post by Bigman80 on Jul 31, 2023 21:23:39 GMT
Funny isn't it.
Flog the posh one on DIY and keep the cheapy.
👍
|
|
|
|
Post by brucew268 on Aug 3, 2023 17:15:06 GMT
Funny isn't it. Flog the posh one on DIY and keep the cheapy. 👍 I’m not dead yet! I decided to bypass the filter resistor between the power capacitors and have a listen. Voila! Body and balance returned to the mids and bass! However, now the highs were sharp and grating. Just in case it was the 5cm of wire used as the bypass, I left the DAC playing for 24 hours on the burn-in disc. Yep, now all the sharpness was gone.  Studer900 PSU better board2 by BruceW, on Flickr Next I changed the output electrolytic from Epcos 2200uF to Nichicon Muse FG 470uF and let it play for an hour before listening. Better again but probably needs some hours but now I’m on the right track and will make a few more changes over the next week to see what improvements I can get, and compare to the build using the Chinese board |
|
Bigman80
Grandmaster
The HiFi Bear/Audioaddicts/Bigbottle Owner
Posts: 16,398
|
Post by Bigman80 on Aug 3, 2023 17:26:29 GMT
Funny isn't it. Flog the posh one on DIY and keep the cheapy. 👍 I’m not dead yet! I decided to bypass the filter resistor between the power capacitors and have a listen. Voila! Body and balance returned to the mids and bass! However, now the highs were sharp and grating. Just in case it was the 5cm of wire used as the bypass, I left the DAC playing for 24 hours on the burn-in disc. Yep, now all the sharpness was gone. Studer900 PSU better board2 by BruceW, on Flickr Next I changed the output electrolytic from Epcos 2200uF to Nichicon Muse FG 470uF and let it play for an hour before listening. Better again but probably needs some hours but now I’m on the right track and will make a few more changes over the next week to see what improvements I can get, and compare to the build using the Chinese board Well done for sticking with it, Bruce. I'd have binned it by now 😂😂 |
|
|
|
Post by misterc on Aug 3, 2023 19:12:09 GMT
This is the joys of R&D Oli I'm sure there is more to come Bruce
|
|
|
|
Post by brucew268 on Aug 17, 2023 8:24:11 GMT
The new build with better traces went from trailing the SQ of the Chinese board to now surpassing it.
This is what was required:
Capacitor-Resistor-Capacitor: obviously the CRC topology did not work well feeding this circuit as the high quality resistor still noticeably degraded the sound. Removed.
Output capacitor: Got rid of the Epcos B41231 here because it had low ESR and wanted something a bit more euphonic or at least more neutral. Replaced with Nichicon FG. Nice sound SQ bump in spaciousness and staging.
Rectifiers: replaced with Wolfspeed. This brought a significant increase in detail across the board but sounded a bit disorganised and incoherent even after 48 hours which was concerning. The HF leading edges were slightly plasticky at first but tamed with time. I didn’t know if these rectifiers would last due to the tradeoffs in SQ.
Wiring: changed layout to minimise length of wires. The main one that was too long was the output from the board to the DC jack 20cm down to 3cm. For the others, I just made sure they were twisted fairly tightly and uniformly. I think the drastic shortening of the DC output wire helped with the weight and body of the mids and lows and the dynamics, though I was forced to change these at the same time as the rectifiers so can’t be sure which did what.
Power capacitor: Got rid of the Epcos B41231. They sounded pretty decent until you have something good to directly compare them to. But I also wondered if the Wolfspeed rectifiers and the Epcos power cap were not playing well together leading to the incoherence of too much detail in my face and not organised musically. I replaced it with a Kemet big can with screw terminals having at least double the ripple rating and mounted on its side with very short leads -- and that sound problem disappeared. Both the Kemet sound and the ripple are significant factors, and the final item that fixed the various sound problems I’d been experiencing... in concert with the above. It did take 24 hours (only!) of burn in to get that sound and reasonable to assume it will continue to improve.
Board: It is still unclear whether the PCB with much thicker traces actually improved the SQ, but then I never quite had the parts choices equalling each other. If I’d had the patience to skip the Wolfspeed rectifiers until the end, I might have been able to directly compare. Ah well. The Chinese PCB can now be had on the ‘bay as version 1.2 with reasonably improved traces and will be a good base for building up.
Purchased as a complete kit with their parts it’s £15, as a bare board it’s £5.40 and good parts will cost you close to £45 for a single rail. It might be thought of as a complex circuit and a lot of parts, but the Jung-Didden Super Regulator will cost £65 in parts plus the bare board, $25USD shipped, but that’s two rails pos/neg.… which I’ll probably try in my preamp this winter.
I still have a few more things to do to further improve the sound on this PSU but will wait on those until the new enclosure arrives that is slightly larger and somewhat echoes the look of the DAC that this PSU serves.
|
|
|
|
Post by misterc on Aug 17, 2023 8:53:57 GMT
Good stuff Bruce, you now starting on the road to understand how and why these items are really just influentical, the only downside there are just so many combinations to discover. Keep up the good work.
Actually a nice little project when you have finished these is an adjustable passive load, get those psu's wemt run in, in no time!
|
|
|
|
Post by brucew268 on Aug 17, 2023 9:01:54 GMT
Good stuff Bruce, you now starting on the road to understand how and why these items are really just influentical, the only downside there are just so many combinations to discover. Keep up the good work.
Funny, I was just coming back on to add an edit: Thank you Tony for your advice and suggestions at various points along the way, which we helpful as well as generous. |
|
Bigman80
Grandmaster
The HiFi Bear/Audioaddicts/Bigbottle Owner
Posts: 16,398
|
Post by Bigman80 on Aug 17, 2023 10:50:54 GMT
The new build with better traces went from trailing the SQ of the Chinese board to now surpassing it. This is what was required: Capacitor-Resistor-Capacitor: obviously the CRC topology did not work well feeding this circuit as the high quality resistor still noticeably degraded the sound. Removed.
Output capacitor: Got rid of the Epcos B41231 here because it had low ESR and wanted something a bit more euphonic or at least more neutral. Replaced with Nichicon FG. Nice sound SQ bump in spaciousness and staging.
Rectifiers: replaced with Wolfson. This brought a significant increase in detail across the board but sounded a bit disorganised and incoherent even after 48 hours which was concerning. The HF transients were slightly plasticky at first but tamed with time. I didn’t know if these rectifiers would last due to the tradeoffs in SQ.
Wiring: changed layout to minimise length of wires. The main one that was too long was the output from the board to the DC jack 20cm down to 3cm. For the others, I just made sure they were twisted fairly tightly and uniformly. I think the drastic shortening of the DC output wire helped with the weight and body of the mids and lows and the dynamics, though I was forced to change these at the same time as the rectifiers so can’t be sure which did what.
Power capacitor: Got rid of the Epcos B41231. They sounded pretty decent until you have something good to directly compare them to. But I also wondered if the Wolfson rectifiers and the Epcos power cap were not playing well together leading to the incoherence of too much detail in my face and not organised musically. I replaced it with a Kemet big can with screw terminals having at least double the ripple rating and mounted on its side with very short leads -- and that sound problem disappeared. Both the Kemet sound and the ripple are significant factors, and the final item that fixed the various sound problems I’d been experiencing... in concert with the above. It did take 24 hours (only!) of burn in to get that sound and reasonable to assume it will continue to improve.
Board: It is still unclear whether the PCB with much thicker traces actually improved the SQ, but then I never quite had the parts choices equalling each other. If I’d had the patience to skip the Wolfson rectifiers until the end, I might have been able to directly compare. Ah well. The Chinese PCB can now be had on the ‘bay as version 1.2 with reasonably improved traces and will be a good base for building up.
Purchased as a complete kit with their parts it’s £15, as a bare board it’s £5.40 and good parts will cost you close to £45 for a single rail. It might be thought of as a complex circuit and a lot of parts, but the Jung-Didden Super Regulator will cost £65 in parts plus the bare board, $25USD shipped, but that’s two rails pos/neg.… which I’ll probably try in my preamp this winter.
I still have a few more things to do to further improve the sound on this PSU but will wait on those until the new enclosure arrives that is slightly larger and somewhat echoes the look of the DAC that this PSU serves.
What was the required output voltage and Amps for this, Bruce? |
|
