This article should be an obligatory read upon joining the forum by now. Sooner or later the thread says "read Jim Hagerman's article" and there's that.
I still want to see same measurements done for all the fancy diodes though. Many claims have been made, not many of them have ever been backed up.
I can see how soft recovery could be a nice thing in general.
Ultrafasts? Meh.
I still want to see same measurements done for all the fancy diodes though. Many claims have been made, not many of them have ever been backed up.
I can see how soft recovery could be a nice thing in general.
Ultrafasts? Meh.
Good article. It reflects what I learned in college... 
Note that the values for the leakage inductance and circuit capacitance may also be found experimentally.
1) Using an oscilloscope, measure the frequency of oscillation (f_osc) at switch/diode turn-off.
2) Add snubber capacitance (zero series resistance) until the measured frequency of oscillation is reduced to f_osc/2.
3) The secondary circuit capacitance, C_sec, may now be calculated as C_sec = C_added/3, where C_added is the snubber capacitance added in step #2.
4) Calculate the secondary leakage inductance as L_sec = ((1/(2*pi*f_osc))^2)/C_sec.
Now follow Mr. Hagerman's article from the "Determining Values" section. Above procedure takes all leakage and parasitic inductances as well as parasitic capacitances into account. It doesn't require any special tools (unless you consider an oscilloscope to be special). It does, however, require an assortment of caps and a bit of experimenting.
The observant reader may wonder where the factor of three comes from in C_sec = C_added/3. That's quite simple...
Recall, f_osc = 1/(2*pi*sqrt(L_sec*C_sec)). Now add cap until f_osc(new) = f_osc/2 --> f_osc(new) = 1/(2*pi*sqrt(L_sec*(C_sec+C_added))) = 1/(2*2*pi*sqrt(L_sec*C_sec)).
-->
2*2*pi*sqrt(L_sec*C_sec) = 2*pi*sqrt(L_sec*(C_sec+C_added))
-->
2*sqrt(L_sec*C_sec) = sqrt(L_sec*(C_sec+C_added))
-->
4*C_sec = C_sec+C_added
<->
C_sec = C_added/3
~Tom
Note that the values for the leakage inductance and circuit capacitance may also be found experimentally.
1) Using an oscilloscope, measure the frequency of oscillation (f_osc) at switch/diode turn-off.
2) Add snubber capacitance (zero series resistance) until the measured frequency of oscillation is reduced to f_osc/2.
3) The secondary circuit capacitance, C_sec, may now be calculated as C_sec = C_added/3, where C_added is the snubber capacitance added in step #2.
4) Calculate the secondary leakage inductance as L_sec = ((1/(2*pi*f_osc))^2)/C_sec.
Now follow Mr. Hagerman's article from the "Determining Values" section. Above procedure takes all leakage and parasitic inductances as well as parasitic capacitances into account. It doesn't require any special tools (unless you consider an oscilloscope to be special). It does, however, require an assortment of caps and a bit of experimenting.
The observant reader may wonder where the factor of three comes from in C_sec = C_added/3. That's quite simple...
Recall, f_osc = 1/(2*pi*sqrt(L_sec*C_sec)). Now add cap until f_osc(new) = f_osc/2 --> f_osc(new) = 1/(2*pi*sqrt(L_sec*(C_sec+C_added))) = 1/(2*2*pi*sqrt(L_sec*C_sec)).
-->
2*2*pi*sqrt(L_sec*C_sec) = 2*pi*sqrt(L_sec*(C_sec+C_added))
-->
2*sqrt(L_sec*C_sec) = sqrt(L_sec*(C_sec+C_added))
-->
4*C_sec = C_sec+C_added
<->
C_sec = C_added/3
~Tom
I have been
experimenting with the power supply on my TPA B-II DAC. First, used a stock parts Placid Bipolar power supply, this uses a standard, slow, Fairchild bridge rectifier. Then, I substitued a bridge built from discrete SMD schottkys from ON semi. The schottkys revealed a little more detail, but seemed to be producing some artifacts as well (perhaps RFI?) a little electronic edge to vocals and horns-the circuit uses .1 uF film caps across the transformer secondaries, but no snubbing directly across the diodes.
Next, I built a small sub board, using a full wave bridge of ON semi MSRF860G diodes (these are very soft recovery, medium speed parts) with a CRC arrangement (3300 uF-R47-3300 uF). Now I am getting fantastic results. Super quiet, lots of detail, and liquid sound with no electronic edge or glaze. I am pretty sold on these MSR soft diodes. I also used no snubbers, and no caps across the transformer secondaries either. It seems to me these super soft diodes do not need the extra circuitry, and this may be an advantage as well (especially for those without scopes) as one can avoid the potential of resonance inducing circuits via extra caps. Full disclosure: the transformer is an EI type, 56 VA.
experimenting with the power supply on my TPA B-II DAC. First, used a stock parts Placid Bipolar power supply, this uses a standard, slow, Fairchild bridge rectifier. Then, I substitued a bridge built from discrete SMD schottkys from ON semi. The schottkys revealed a little more detail, but seemed to be producing some artifacts as well (perhaps RFI?) a little electronic edge to vocals and horns-the circuit uses .1 uF film caps across the transformer secondaries, but no snubbing directly across the diodes.
Next, I built a small sub board, using a full wave bridge of ON semi MSRF860G diodes (these are very soft recovery, medium speed parts) with a CRC arrangement (3300 uF-R47-3300 uF). Now I am getting fantastic results. Super quiet, lots of detail, and liquid sound with no electronic edge or glaze. I am pretty sold on these MSR soft diodes. I also used no snubbers, and no caps across the transformer secondaries either. It seems to me these super soft diodes do not need the extra circuitry, and this may be an advantage as well (especially for those without scopes) as one can avoid the potential of resonance inducing circuits via extra caps. Full disclosure: the transformer is an EI type, 56 VA.
@barrows:
It doesn't strike me as an entirely fair comparison.
The ability of radiated RF noise to interfere with other circuitry drops with frequency. Different diodes have different capacitance, so they will re-tune the LC frequency. So the difference in might not be due to intrinsic properties of the diodes themselves.
Eva put up photos of scope traces in another thread showing the addition of snubbers cleaning up the waveform. Especially if you have a scope to verify the results, IMO you should calculate correct snubber values for the different diodes and compare them under these optimised conditions.
I'm not sure that any diode operates entirely without exciting the LC resonance, but it seems to me that ringing could still be induced by external sources of interference, so snubbers would be in order anyway.
That's just how I figure it.
It doesn't strike me as an entirely fair comparison.
The ability of radiated RF noise to interfere with other circuitry drops with frequency. Different diodes have different capacitance, so they will re-tune the LC frequency. So the difference in might not be due to intrinsic properties of the diodes themselves.
Eva put up photos of scope traces in another thread showing the addition of snubbers cleaning up the waveform. Especially if you have a scope to verify the results, IMO you should calculate correct snubber values for the different diodes and compare them under these optimised conditions.
I'm not sure that any diode operates entirely without exciting the LC resonance, but it seems to me that ringing could still be induced by external sources of interference, so snubbers would be in order anyway.
That's just how I figure it.
No 'scope here,
so anytime I can avoid adding potentially resonance inducing components I am all for it. Certainly if I had a good 'scope, and I wanted to go deeper into this for any reason (sound problems) I would do so. But right now the sound with this component is great. Note that the PS is feeding a shunt regulator, so it may not be as sensitive to incoming noise as a simple three pin linear IC regulator might be.
The results are good enough to make me very impressed with the MSR style diodes in this application. I have also tried Qspeeds and Stealths, but both of these have higher voltage drops, and no apparent sonic advantages over the ON MSRs-once again, via listening tests.
The MSRs have great specs for soft recovery, that and the fact that they sound better is good enough for me.
I would be happy to see some measurements in circuits from those who have good 'scopes as well, and, perhaps, the results of using them with snubbers. Just wanted to share my experience, and am not claiming my experience is definitive.
so anytime I can avoid adding potentially resonance inducing components I am all for it. Certainly if I had a good 'scope, and I wanted to go deeper into this for any reason (sound problems) I would do so. But right now the sound with this component is great. Note that the PS is feeding a shunt regulator, so it may not be as sensitive to incoming noise as a simple three pin linear IC regulator might be.
The results are good enough to make me very impressed with the MSR style diodes in this application. I have also tried Qspeeds and Stealths, but both of these have higher voltage drops, and no apparent sonic advantages over the ON MSRs-once again, via listening tests.
The MSRs have great specs for soft recovery, that and the fact that they sound better is good enough for me.
I would be happy to see some measurements in circuits from those who have good 'scopes as well, and, perhaps, the results of using them with snubbers. Just wanted to share my experience, and am not claiming my experience is definitive.
At the first time this "diode question" I saw raised, an article in The Audio Amateur with a Gary Galo suggested supply, Rick Miller had made some test, measuring diodes and listening to supplies with them, and the best sounding at the time were the fast/slow decay types.
Back then GI types were the chosen ones, but I don't think GI is no more. Gagy Galo suggested some Vishay Hexfreds, but they are a bit too much (8A) for low current supplies, like on preamps or CD players.
Schottkies did not sound as well as those mentioned, and later reports seem to show that do well in digital circuitry, but not as well in analogue.
Back then GI types were the chosen ones, but I don't think GI is no more. Gagy Galo suggested some Vishay Hexfreds, but they are a bit too much (8A) for low current supplies, like on preamps or CD players.
Schottkies did not sound as well as those mentioned, and later reports seem to show that do well in digital circuitry, but not as well in analogue.
Hi !
thanks a lot for the very much interesting thread.
I still wonder if anyone has direct experience of different diodes in low voltage low current applications.
What would be your first choice for a power supply of a dac ?
Moreover, I cannot locate soft recovery diodes.
If I search on ebay.com fast and ultra fast types pop up immediately, but slow does not provide any result.
I have bought some SB/SR5100 to try out but maybe soft recovery type are better for less RFI generation ?
I sse from datasheets that normal regulators have no suppression of RFI at all.
Thanks a lot again.
thanks a lot for the very much interesting thread.
I still wonder if anyone has direct experience of different diodes in low voltage low current applications.
What would be your first choice for a power supply of a dac ?
Moreover, I cannot locate soft recovery diodes.
If I search on ebay.com fast and ultra fast types pop up immediately, but slow does not provide any result.
I have bought some SB/SR5100 to try out but maybe soft recovery type are better for less RFI generation ?
I sse from datasheets that normal regulators have no suppression of RFI at all.
Thanks a lot again.
Let us not forget that many silicon Schottky diodes these days are something else.
Trench barrier rectifiers, superbarrier rectifiers, might matter to you, it might not...
All real silicon Schottkys are in parallel with a silicon Zener guard ring diode. If you
put too much reverse voltage or too much forward current, the guard ring conducts.
Few people think the forward current, but if you are pushing more than 0.7V through
an undersized Schottky, then a forward biased guard ring probably carried some if it.
And guard ring recovery will not be soft.
Nothing wrong with a real Schottky in Silicon Carbide if you don't mind the higher
price and higher forward drop. No idea if they have guard rings...
Trench barrier rectifiers, superbarrier rectifiers, might matter to you, it might not...
All real silicon Schottkys are in parallel with a silicon Zener guard ring diode. If you
put too much reverse voltage or too much forward current, the guard ring conducts.
Few people think the forward current, but if you are pushing more than 0.7V through
an undersized Schottky, then a forward biased guard ring probably carried some if it.
And guard ring recovery will not be soft.
Nothing wrong with a real Schottky in Silicon Carbide if you don't mind the higher
price and higher forward drop. No idea if they have guard rings...
Last edited:
Hi ! thanks a lot for the very valuable advice.
I see in the ds
so these seems the key characteristics to look for.
I guess you check the results with listening tests ?
I have learned that the real issue with normal regulators is the high Hz noise. They let all pass through unsuppressed.
Instead for low Hz noise they are quite excellent.
Then I learned of this high Hz diodes switching noise and I guess this is a real issue.
I looked in the web but I have not been able to find a spectrum of this noise to understand level and position in Hz.
But above let's say 10kHz any noise is an issue with the usual regulator.
From what I read here I think that the right snubbing circuit should be the key to solve the issue and it could even take care of some noise incoming through the transformer.
And this could also allow to use quite normal, realiable and cheap diodes.
I will look for info about snubbers then.
Thanks a lot again.
Last edited:
I noticed I had typed "RCR", had meant to type CRC.
I have used the same values and evaluated the results through listening to the same music on the same system for an extended period before changing, so far the results have been good.
I use film/foil capacitors .01uf paralleled with a .1uf + 18ohm, as close to the diode bridge as possible.
I have used the same values and evaluated the results through listening to the same music on the same system for an extended period before changing, so far the results have been good.
I use film/foil capacitors .01uf paralleled with a .1uf + 18ohm, as close to the diode bridge as possible.
Has anyone tried old germanium diodes? They have low forward voltage ("sensitive") and do not switch on/off as steeply than silicon diodes. They clip softer and at lower voltage. Also the junction capacitance is very small.
Previously I used Vishay HFA08SD60S HexFred diodes in my DAC and anode supply of the preamp. They made the sound brighter than regular silicon diodes. Quite hifi sounding but not very relaxed. My goal was to tweak the rectification to sound more like tube rectification but that dod not happen with HexFred, more like the opposite.
Yesterday I changed them to 1N91 germanium diodes. They have "gold bonded" juction and forward voltage of ~0,1-0,12V. They are also "shielded" as the metal cover is in direct connection with the cathode of the diode. They might not radiate as much as non-shielded diodes.
These germanium diodes sound very good right from the start. They are not bright like HexFred, darker/bassier and relaxed top end, good low level detail in acoustic music. I'm waiting for them to break in some more.
An externally hosted image should be here but it was not working when we last tested it.
Previously I used Vishay HFA08SD60S HexFred diodes in my DAC and anode supply of the preamp. They made the sound brighter than regular silicon diodes. Quite hifi sounding but not very relaxed. My goal was to tweak the rectification to sound more like tube rectification but that dod not happen with HexFred, more like the opposite.
Yesterday I changed them to 1N91 germanium diodes. They have "gold bonded" juction and forward voltage of ~0,1-0,12V. They are also "shielded" as the metal cover is in direct connection with the cathode of the diode. They might not radiate as much as non-shielded diodes.
These germanium diodes sound very good right from the start. They are not bright like HexFred, darker/bassier and relaxed top end, good low level detail in acoustic music. I'm waiting for them to break in some more.
Actually I don't even bother with changing the diodes on anything, have found that a decent snubber arrangement takes care of anything (and more) that might be gained by swapping the diodes.
The only reason I would practically see for better rectifier diodes would be to keep the parts count down in an oem design.
The only reason I would practically see for better rectifier diodes would be to keep the parts count down in an oem design.
In my experience, PN diode generally sounds somehow noisy while unipolar device like schottky or super barrier one sounds clean. I found similar difference between bipolar transistors and fets in amplification circuits so it may not be related to switching but noises caused by static electrical current.
However, I am not saying schottokey is best because there often are some tonal coloration. It may be because of some metal to create schottky junction I dream up.
There are few documents regarding noise of semiconductors.
However, I am not saying schottokey is best because there often are some tonal coloration. It may be because of some metal to create schottky junction I dream up.
There are few documents regarding noise of semiconductors.
cool thanks, I just bought about 27x BYV-27 50's. BYV27-50 Qty27 Ultra fast low-loss controlled avalanche rectifiers PH NOS 2A | eBay
Pretty cheap.
Pretty cheap.
I know this is a very old thread, but just assembling some notes getting back to speed. (pun intended) I had never heard of silicon carbide diodes. After my time in the lab.
It makes some sense to me that in some signal path circuits, choice of diode technology may effect the sound. Good old 1N914/1N4148 for example maybe surpassed by new technology in various limit and protection functions, but already 4nS. Newer parts with lower noise for places where they serve as a reference. Sure. Better thermal or noise characters than an LED for a CCS maybe? So some current part numbers would be helpful.
As far as a rectifier goes, unless the filter bank and regulator are totally a POS, I see no way they can make any sonic differences. I get the transformer ringing argument, but a proper snubber is a better solution. Is all this just more audio cool-aid?
Sorry, but much of the preceding description of sonic character is more than non-sense. I don't suggest anyone is making things up, but it is well understood game our brains play if we expect a change, we will notice one real or not. Then better the story, the more we wil notice. If it is your own idea, it is even more pronounced. We are funny creatures indeed.
It makes some sense to me that in some signal path circuits, choice of diode technology may effect the sound. Good old 1N914/1N4148 for example maybe surpassed by new technology in various limit and protection functions, but already 4nS. Newer parts with lower noise for places where they serve as a reference. Sure. Better thermal or noise characters than an LED for a CCS maybe? So some current part numbers would be helpful.
As far as a rectifier goes, unless the filter bank and regulator are totally a POS, I see no way they can make any sonic differences. I get the transformer ringing argument, but a proper snubber is a better solution. Is all this just more audio cool-aid?
Sorry, but much of the preceding description of sonic character is more than non-sense. I don't suggest anyone is making things up, but it is well understood game our brains play if we expect a change, we will notice one real or not. Then better the story, the more we wil notice. If it is your own idea, it is even more pronounced. We are funny creatures indeed.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.