Tom, just seeing your new post now. Congratulations on measuring some data! As far as I know, nobody has put Quasimodo on a 500VA Antek, so we don't really know what optimum value it will discover. Perhaps the leakage inductance will be a little higher for the 500 VA 25V secondary than for the 400VA 18V secondary, and the optimum snubbing resistance will change correspondingly, but that is just a guess.
You could of course replace the fixed 10 ohm snubber resistor on your board, with a 50 ohm 25 turn trimpot. Then dial it for least ringing as viewed through your highpass filter. Saves time compared to disconnecting the trafo and attaching it to a Quasimodo. If the least ringing you can get, all the way across the dial, is still comparable to (or worse than) the other candidates, you're all done. If it's lower, but not much lower, you're probably all done. If it's quite a bit lower, now you can recommend values for either a 2-component snubber with ringing X-amount, or a 3-component snubber with ringing far-less-than-X-amount.
By the way, you want to do the test with a hefty big load current being pulled from the supply. You want those diodes to be working hard, slamming on and off vigorously. I recommend at least 60% of max permitted load current
You could of course replace the fixed 10 ohm snubber resistor on your board, with a 50 ohm 25 turn trimpot. Then dial it for least ringing as viewed through your highpass filter. Saves time compared to disconnecting the trafo and attaching it to a Quasimodo. If the least ringing you can get, all the way across the dial, is still comparable to (or worse than) the other candidates, you're all done. If it's lower, but not much lower, you're probably all done. If it's quite a bit lower, now you can recommend values for either a 2-component snubber with ringing X-amount, or a 3-component snubber with ringing far-less-than-X-amount.
By the way, you want to do the test with a hefty big load current being pulled from the supply. You want those diodes to be working hard, slamming on and off vigorously. I recommend at least 60% of max permitted load current
Last edited:
Whoops. I realized I swapped Cx and Cs. It should be 10 nF || 150 nF + 10 Ω for Mark's snubber. My bad. I've included all the results below so they're easier to compare. You can find the test conditions a few posts back. I forgot to mention that this measurement was taken with 8 Ω resistive load on each rail, resulting in roughly 275 W of dissipated power.
Ironically, Mark's snubber performs as well as Rod's. Mark's just uses more parts.
Either way, going snubber-less does not create any ringing. There is no issue to begin with and no fix is needed. I doubt any snubber does much to reduce the energy radiated, though, using any of the snubbers at least reduces the edge rate, thereby, the potential for coupling into sensitive circuits.
My recommendation remains: Either go without a snubber or use the RC snubber. I don't see the point of using six parts when four will do the job.
Tom
Ironically, Mark's snubber performs as well as Rod's. Mark's just uses more parts.
Either way, going snubber-less does not create any ringing. There is no issue to begin with and no fix is needed. I doubt any snubber does much to reduce the energy radiated, though, using any of the snubbers at least reduces the edge rate, thereby, the potential for coupling into sensitive circuits.
My recommendation remains: Either go without a snubber or use the RC snubber. I don't see the point of using six parts when four will do the job.
Tom
Attachments
Last edited:
That may be your go-to analogy, but I think it's fundamentally broken. You're basically arguing that because "dirty signals" exist inside the amplifier enclosure, the amp is somehow intrinsically flawed even if none of the dirt is even remotely measurable on the amp output. Have you ever looked at the mains voltage? It's not pretty. Should we all convert our amps to battery operation to avoid "dirty signals"?
A better analogy would be the birthmark I have between two toes. It has no impact on my ability to walk or function in society. It's completely harmless. To tell you the truth, I can't even tell you between which two toes it is without taking my socks off and searching for it. You wouldn't be able to tell even if you saw me barefoot. So does it matter that it's there? To me it doesn't.
Also, I hope you wouldn't walk up to somebody on the street and suggest they have their unsightly but harmless facial warts removed. I hope we can agree that would be incredibly rude and socially unacceptable (to put it mildly). So why would you do that here?
Tom
They're both underdamped which is why they both ring.
I'm working on finding a loaner AN5225 to try on Quasimodo and dial up a C+RC snubber which correctly damps this actual transformer. Just in case 25VAC 500VA turns out to be appreciably different than 18VAC 400VA.
Are the rectifiers in your PSU, the same ones whose junction capacitance curve appears in #1072? If not, do you mind mentioning the zero bias Cj of the rectifiers you do use? I assume you don't use those 30A Vishays in #1076.
I'm working on finding a loaner AN5225 to try on Quasimodo and dial up a C+RC snubber which correctly damps this actual transformer. Just in case 25VAC 500VA turns out to be appreciably different than 18VAC 400VA.
Are the rectifiers in your PSU, the same ones whose junction capacitance curve appears in #1072? If not, do you mind mentioning the zero bias Cj of the rectifiers you do use? I assume you don't use those 30A Vishays in #1076.
Our posts may have crossed. I had Cx and Cs swapped. My bad. I also just realized I attached the files without annotation. Grr!! I've attached annotated plots for NO snubber, your snubber (10 nF || 150 nF + 10 Ω), Rod's snubber (100 nF + 10 Ω).
I don't see any ringing in any of these measurements. No ringing without the snubber. No ringing with the snubber. No problem. No fix needed.
The diodes are the ones I showed the reverse bias capacitance for some ten posts back. They're 30 A, 1200 V in TO-247 packages by ST Micro.
If you find the values to be significantly different from the 10 nF, 150 nF, 10 Ω I used, just let me know and I'll remeasure.
Tom
Attachments
Why spend you money on expensive audio tweaks, when a glass or mug or two of your favorite adult beverage will make all the difference.
And as they say in the ads, rinse and repeat.
And as they say in the ads, rinse and repeat.
Since I had the measurement setup running anyway, I decided to postpone the IPA consumption for a little bit.
Attached show the Power-86 in three conditions:
- NO snubber
- CRC snubber (10 nF || 10 Ω + 150 nF) - Mark's design
- RC snubber (10 Ω + 100 nF) - Rod's design
Transformer: Antek AN-5225
Load: 8 Ω on each rail (275 W)
Conclusion: No ringing even without snubber. No ringing with Rod's snubber. No ringing with Mark's snubber. No issue. No fix.
I shall drink beer now. I think I'll have one of the Yellow Dog.
Tom
Attached show the Power-86 in three conditions:
- NO snubber
- CRC snubber (10 nF || 10 Ω + 150 nF) - Mark's design
- RC snubber (10 Ω + 100 nF) - Rod's design
Transformer: Antek AN-5225
Load: 8 Ω on each rail (275 W)
Conclusion: No ringing even without snubber. No ringing with Rod's snubber. No ringing with Mark's snubber. No issue. No fix.
I shall drink beer now. I think I'll have one of the Yellow Dog.
Tom
Attachments
That particular transformer, and I suspect, all instances of that same part number transformer from Antek, has got plenty of loss at the secondary resonant frequency. Whether it's "hysteresis loss" or "eddy current loss" or good old "core loss" I don't know. But it's got enough loss at ~120 kHz that it keeps the resonant "Q" down in the 0.5 region and the damping factor up in the 1.0 region. The transformer doesn't ring because it self-damps. As long as you don't change the external capacitance (== choice of rectifier diodes) too much, and as long as Antek keeps building AN-5225s the same way, it'll keep on self damping. External components can ensure that it is wildly overdamped, if you wish. But as-is, it seems to be plenty damped all by itself.
A non-negligible percentage of toroidal transformers appear to have this property. Typical load currents plus typical designed-in diodes (especially discrete diodes rather than bridge rectifiers!) don't exhibit much ringing with the transformers in this special group, sometimes no ringing at all. Self damping gets you all the way to the goal that external damping would normally achieve. Give thanks, you can leave three $0.20 parts off your PCB as a result.
If you're dying of curiosity you can slowly add capacitance across the secondary and notice that eventually ringing starts to appear. This happens when the self damping of the transformer is insufficient to overcome the increasing quality factor "Q" of the resonant circuit. (Q is proportional to sqrt(Ctotal)). It's thoroughly self-damped for small Cexternal, thus no ringing, but it's underdamped for large Cexternal such as 100nF, and ringing appears.
A non-negligible percentage of toroidal transformers appear to have this property. Typical load currents plus typical designed-in diodes (especially discrete diodes rather than bridge rectifiers!) don't exhibit much ringing with the transformers in this special group, sometimes no ringing at all. Self damping gets you all the way to the goal that external damping would normally achieve. Give thanks, you can leave three $0.20 parts off your PCB as a result.
If you're dying of curiosity you can slowly add capacitance across the secondary and notice that eventually ringing starts to appear. This happens when the self damping of the transformer is insufficient to overcome the increasing quality factor "Q" of the resonant circuit. (Q is proportional to sqrt(Ctotal)). It's thoroughly self-damped for small Cexternal, thus no ringing, but it's underdamped for large Cexternal such as 100nF, and ringing appears.
Last edited:
That could be. Thankfully, one of the transformers in my arsenal wasn't one of the self-damped ones. It's a 2x35 VAC, 530 VA toroid. Quite good quality. It came from RS Components some 20+ years ago. P/N: 177-945.
I hooked it up to a Power-86 and tried the three snubber combinations:
- NO snubber
- CRC: 10 nF || 10 Ω + 150 nF (Mark's)
- RC: 10 Ω + 100 nF (Rod's)
Results:
- Some ringing without snubber. The ringing dies pretty quickly. I'm not overly concerned by it, but also don't see any harm in removing the ringing.
- CRC snubber: Dampens out the ringing nicely. Lowers the flyback amplitude dramatically. I could reduce R if I wanted a tighter response, but it looks good as-is.
- RC snubber: Same as for CRC snubber.
My method takes place in situ, with the mains impedance reflected by the transformer, requires only an RC highpass filter (1 kΩ, 10 nF) and an oscilloscope, and requires only two snubber components.
You're right that the components are only 20 cents each, so it's not a lot of money. Still, using three components where two will do the job just as well is just not my cup of tea. Each to their own.
You're certainly free to keep arguing. I'll keep reading. At this point, though, I think I have enough data that I should focus on getting it onto my website (Snubber page) and into the design documentation for the Power-86 and Power-686. :micdrop:
Tom
Attachments
Last edited:
I have a solution, I've included the logo that MJ can use on his avatar 🙂 While he might tPiss people off, he means well, I think. Just like the image of the guy I posted here. Why don't we just call him DIY dad and be done with it.
Cheers,
Cheers,
Attachments
Last edited:
This time with corrected image text.
As far as the discussion goes: I agree wholeheartedly with BrianL's comments from earlier. That said, it was insightful for me to take a second look at snubbers. I now have more than enough scope shots to put a good page together on the topic. Life is good.
Tom
As far as the discussion goes: I agree wholeheartedly with BrianL's comments from earlier. That said, it was insightful for me to take a second look at snubbers. I now have more than enough scope shots to put a good page together on the topic. Life is good.
Tom
Attachments
Last edited:
Hi Tom, I'm actually agreeing with you, not arguing.
Your data from the Antek AN5225 transformer shows
I believe you conclude that: no ringing means no need for a snubber. I completely agree.
Also congratulations for measuring another transformer to see whether it also has the property above. Many, many do not. In particular, we don't yet know whether a 500VA transformer made by Toroidy instead of Antek, exhibits enough self-damping to not need a snubber. I've never measured a Toroidy myself, but I have measured toroidal transformers by Triad, Avel-Lindberg, and Antek. They all rang differently in my testing, FWIW.
Since diyAudio members just love the shiny stainless steel Faraday cages that Toroidy installs over their transformers, it's only a matter of time until somebody pairs a Toroidy with your PSU and asks, snubber or no snubber? I understand that European DIYers pay huge shipping fees when buying from Antek, but much lower shipping fees when buying from Toroidy (in Poland). So it's natural and expected that European buyers of your PSU, probably will buy a Toroidy transformer too.
In situations where the transformer+rectifier does in fact ring, it can be damped with a 2-element snubber (1R+1C) or with a 3-element snubber (1R+2C). One way to size these components is to build the complete PSU, attach a high wattage dummy load, connect the snubber-under-test, turn on the mains power, and measure with an oscilloscope. Which is the approach you took, and it was successful.
Another way is to use a battery powered test jig like the one shown below ("Cheapomodo"), to stimulate the transformer in isolation: connected to nothing else. The complete PSU need not be assembled yet, mains power is not applied, high wattage dummy loads are not required. This even works for 2-element snubbers (1R+1C) although it requires you to know the zero bias capacitance of your rectifier diodes. QM Figure 14 (pages 12-13) shows how.
Don't let the length of this message distract from my main theme, which is: congratulations.
Mark Johnson
_
Your data from the Antek AN5225 transformer shows
Some (Transformer + Rectifier) Combinations Don't Ring At All; they don't need a snubber
Also congratulations for measuring another transformer to see whether it also has the property above. Many, many do not. In particular, we don't yet know whether a 500VA transformer made by Toroidy instead of Antek, exhibits enough self-damping to not need a snubber. I've never measured a Toroidy myself, but I have measured toroidal transformers by Triad, Avel-Lindberg, and Antek. They all rang differently in my testing, FWIW.
Since diyAudio members just love the shiny stainless steel Faraday cages that Toroidy installs over their transformers, it's only a matter of time until somebody pairs a Toroidy with your PSU and asks, snubber or no snubber? I understand that European DIYers pay huge shipping fees when buying from Antek, but much lower shipping fees when buying from Toroidy (in Poland). So it's natural and expected that European buyers of your PSU, probably will buy a Toroidy transformer too.
In situations where the transformer+rectifier does in fact ring, it can be damped with a 2-element snubber (1R+1C) or with a 3-element snubber (1R+2C). One way to size these components is to build the complete PSU, attach a high wattage dummy load, connect the snubber-under-test, turn on the mains power, and measure with an oscilloscope. Which is the approach you took, and it was successful.
Another way is to use a battery powered test jig like the one shown below ("Cheapomodo"), to stimulate the transformer in isolation: connected to nothing else. The complete PSU need not be assembled yet, mains power is not applied, high wattage dummy loads are not required. This even works for 2-element snubbers (1R+1C) although it requires you to know the zero bias capacitance of your rectifier diodes. QM Figure 14 (pages 12-13) shows how.
Don't let the length of this message distract from my main theme, which is: congratulations.
Mark Johnson
_
Attachments
Last edited:
Tom,
I thought it might be good idea for you to consider following SMPS supplier for 686 builds.
HDPLEX 400W AC-DC Internal Adapter with PFC
I had couple of items from them. They are very robust, reliable and compact designs. One of the things things HDPLEX specializes in are low noise SMPS for Audio PC. I thought you may want consider them to build 28V or 36V version as alternative to connex and meanwell. Just a thought ...
I thought it might be good idea for you to consider following SMPS supplier for 686 builds.
HDPLEX 400W AC-DC Internal Adapter with PFC
I had couple of items from them. They are very robust, reliable and compact designs. One of the things things HDPLEX specializes in are low noise SMPS for Audio PC. I thought you may want consider them to build 28V or 36V version as alternative to connex and meanwell. Just a thought ...
Last edited:
Interesting power supply. It seems optimized for use in PCs, though. One could toss them an email to enquire about audio supplies...
I find it interesting that the supply comes with an IEC320-C8 harness (shaver plug) in addition to the more commonly used IEC320-C14 (D shaped connector). There's no mention of the isolation level between primary and secondary. I'd definitely ask about that should I contact them.
Tom
I find it interesting that the supply comes with an IEC320-C8 harness (shaver plug) in addition to the more commonly used IEC320-C14 (D shaped connector). There's no mention of the isolation level between primary and secondary. I'd definitely ask about that should I contact them.
Tom
Hi Do
Any tips on re-forming the LM3886 leads ? I like your space-saving mounting of the Modulus-686 in post #1000
Modulus-686: 380W (4Ω); 220W (8Ω) Balanced Composite Power Amp with extremely low THD
Cheers
Colin
Any tips on re-forming the LM3886 leads ? I like your space-saving mounting of the Modulus-686 in post #1000
Modulus-686: 380W (4Ω); 220W (8Ω) Balanced Composite Power Amp with extremely low THD
Cheers
Colin
Last edited:
Hi Colin,
I had the flatten all the pins, make them all straight as possible, then measure the board height to make sure to bend the pins so they're long enough to be soldered. Please account for thermal pad as well. I used drill bits to space the rows. You lay it flat on the pins touching the case and you do the same for the second row but you need to find a drill bit that will make the proper spacing in between the first and second row.
I don't know if you follow me?
Thanks
Do
Thanks Do
I think I'll get a spare LM3886 or two to practice on to get the right size drill bits to use. As you say, I need to allow for the thermal pad and also determine which size spacer to use for mounting the PCB.
Cheers
Colin
I think I'll get a spare LM3886 or two to practice on to get the right size drill bits to use. As you say, I need to allow for the thermal pad and also determine which size spacer to use for mounting the PCB.
Cheers
Colin
All of the LM3886 projects I've worked on, used metric M3 screws to fasten the IC to the heatsink. I used a 2.5mm drill bit to drill the hole in the aluminum heatsink, and then an M3 tap to create the threads inside the hole. A few drops of Tap Magic help quite a bit. This isn't the kit I used but it does illustrate the sizes:
https://www.amazon.com/VEBE-Adjustable-T-handle-Wrench-Machine/dp/B06X9YV16Z/
If you're using the insulated tab LM3886 then you don't need a shoulder washer for the mounting bolt. If you're using the metal tab LM3886, you do. In my experience the hole size in the LM3886 tab matches the hole size in the tab of a TO-220 packaged power transistor. So you can use shoulder washers sized for TO-220 and they work perfectly on a metal tab LM3886. For example,
1000PCS TO-220 High Temperature Insulation Nylon Washer Bushing Brown Beige RoHS | eBay
https://www.amazon.com/VEBE-Adjustable-T-handle-Wrench-Machine/dp/B06X9YV16Z/
If you're using the insulated tab LM3886 then you don't need a shoulder washer for the mounting bolt. If you're using the metal tab LM3886, you do. In my experience the hole size in the LM3886 tab matches the hole size in the tab of a TO-220 packaged power transistor. So you can use shoulder washers sized for TO-220 and they work perfectly on a metal tab LM3886. For example,
1000PCS TO-220 High Temperature Insulation Nylon Washer Bushing Brown Beige RoHS | eBay