I admit I'm not fully on-board with the use of Schottky diodes in a linear power supply. Perhaps this subject has already been hashed out elsewhere, so please direct me if this subject is old news.
I keep seeing references to the fact that a "soft" turnoff characteristic is desireable to limit EMI radiation, buzz in the power rails, and ringing in the transformer secondary windings. Looking at the general data sheet descriptions for Schottky's, it is obvious that when compared with the typical leisurely reverse recovery of a Silicon diode, a Schottky has technically no reverse recovery time at all (or very little depending upon the die arrangement, etc).
So, on the face of it, the standard silicon power rectifier diode (1N4007, etc) would have a much "softer" reverse recovery because the turn-off event takes place much more slowly. If a Schottky has essentially zero recovery time, then things happen rather quickly, i.e. the current stops flowing almost immediately. Any time a physical or electronic event happens abruptly (step function), you get a powerful tendency toward oscillation, ringing, harmonics, and the like. Looking at it this way, it seems that linear audio power supplies would do much better with standard silicon power diodes.
And yet, lots of audiophile designs specify power Schottky's despite the lower reverse voltage ratings (now superceded by SiC Schottky's), and especially in spite of the well-known high reverse current/temperature relationship. Obvioulsy we need fast diodes in places where things are taking place at high frequencies and in switching applications. But for 50/60 Hz rectification? What am I missing?
Thank you in advance,
RestAssured
I keep seeing references to the fact that a "soft" turnoff characteristic is desireable to limit EMI radiation, buzz in the power rails, and ringing in the transformer secondary windings. Looking at the general data sheet descriptions for Schottky's, it is obvious that when compared with the typical leisurely reverse recovery of a Silicon diode, a Schottky has technically no reverse recovery time at all (or very little depending upon the die arrangement, etc).
So, on the face of it, the standard silicon power rectifier diode (1N4007, etc) would have a much "softer" reverse recovery because the turn-off event takes place much more slowly. If a Schottky has essentially zero recovery time, then things happen rather quickly, i.e. the current stops flowing almost immediately. Any time a physical or electronic event happens abruptly (step function), you get a powerful tendency toward oscillation, ringing, harmonics, and the like. Looking at it this way, it seems that linear audio power supplies would do much better with standard silicon power diodes.
And yet, lots of audiophile designs specify power Schottky's despite the lower reverse voltage ratings (now superceded by SiC Schottky's), and especially in spite of the well-known high reverse current/temperature relationship. Obvioulsy we need fast diodes in places where things are taking place at high frequencies and in switching applications. But for 50/60 Hz rectification? What am I missing?
Thank you in advance,
RestAssured
Others here are more knowledgable about this than I, but my take is that the benefit of utilizing Schottky diodes is based in their low junction capacitance itself, not on their fast commutation enabled by a low junction capacitance. The benefit of which is an raising of the resonance frequency of the parasitic tank circuit formed by the power transformer's leakage inductance and the diode capacitance. I've no idea why that might make for better perceieved sound. Mark Johnson seems to be our resident expert on such matters. I suggest seeking him for for his opinion.
Last edited:
Check for this, "Ideal Diode Bridge" based on LTC4320
LT4320/LT4320-1 - Ideal Diode Bridge Controller - Linear Technology
LT4320/LT4320-1 - Ideal Diode Bridge Controller - Linear Technology
Also checkout this article :
https://diyaudioheaven.wordpress.com/tutorials/power-supplies/rectifiers/
https://diyaudioheaven.wordpress.com/tutorials/power-supplies/rectifiers/
> standard silicon power rectifier ...would have a much "softer" reverse recovery because the turn-off event takes place much more slowly.
No. It "stays on" when it should be OFF. It pulls the power rail the wrong way. This leads to large current spikes, small dips in main cap voltage, and big kick-back in power transformer winding.
I do tend to feel that 1N4007 should be "good enough". But the advocates of no-stored-charge diodes are not deluded, just detail-oriented.
No. It "stays on" when it should be OFF. It pulls the power rail the wrong way. This leads to large current spikes, small dips in main cap voltage, and big kick-back in power transformer winding.
I do tend to feel that 1N4007 should be "good enough". But the advocates of no-stored-charge diodes are not deluded, just detail-oriented.
Thank you! I think sesebe has it. I thought I had missed one of my courses in school. My next question would be why there are so many dual parallel Schottky devices available, when many technical papers prove that current does not divide equally between them? To me, it seems like dual parallel diodes is the same thing as having dual parallel fuses. Why play around with unpredictable elements in your designs, when it is hard enough to get stuff working in the first place?
Referring to the ringing of the transformer secondary, I find it very interesting that a snubber network can be found with some clever and simple circuitry. It seems to me that using the Ringer circuit would be more effective once the diode bridge (minus the extra capacitors) was connected to the transformer. That way, the snubber values would take into account any added capacitance from the diodes. Just a thought.
I forget why hooking capacitors accross diodes in a 50/60 Hz bridge might be necessary or desireable. It is/was best practice when designing high voltage supplies where diodes have to be stacked in series. The capacitors in that kind of circuit prevented blowing the particular diode in the stack that was the fastest to switch. That type of failure mode in low voltage bridges is rare, and how would you prove it anyway once the part failed?
Referring to the ringing of the transformer secondary, I find it very interesting that a snubber network can be found with some clever and simple circuitry. It seems to me that using the Ringer circuit would be more effective once the diode bridge (minus the extra capacitors) was connected to the transformer. That way, the snubber values would take into account any added capacitance from the diodes. Just a thought.
I forget why hooking capacitors accross diodes in a 50/60 Hz bridge might be necessary or desireable. It is/was best practice when designing high voltage supplies where diodes have to be stacked in series. The capacitors in that kind of circuit prevented blowing the particular diode in the stack that was the fastest to switch. That type of failure mode in low voltage bridges is rare, and how would you prove it anyway once the part failed?
> why there are so many dual parallel Schottky devices available
Because the standard switcher supply uses the CT winding 2-diode rectifier. So the world needs diode-pairs by the millions.
> papers prove that current does not divide equally between them?
The CT-2-D form, each diode carries half the total current.
If people are using them otherwise, then we have to wonder how bad the unbalance can be, and if it matters. Two diodes made the same day, possibly on the same wafer, tend to match pretty well. With modern process, I would expect 5mV mismatch. 20mV difference is 2:1 of current. So 5mV difference is maybe 20% difference? So one could be carrying 1A and the other 0.84A. Two 1A diodes makes a 1.84A diode. Not double, but not pointless.
> why hooking capacitors accross diodes
Early diodes could have large mismatches in junction area and diffusion depth, thus C and PIV. Modern diodes seem to be much better matched. Also we can often over-buy (1000V for a 400V job) without much higher price.
Because the standard switcher supply uses the CT winding 2-diode rectifier. So the world needs diode-pairs by the millions.
> papers prove that current does not divide equally between them?
The CT-2-D form, each diode carries half the total current.
If people are using them otherwise, then we have to wonder how bad the unbalance can be, and if it matters. Two diodes made the same day, possibly on the same wafer, tend to match pretty well. With modern process, I would expect 5mV mismatch. 20mV difference is 2:1 of current. So 5mV difference is maybe 20% difference? So one could be carrying 1A and the other 0.84A. Two 1A diodes makes a 1.84A diode. Not double, but not pointless.
> why hooking capacitors accross diodes
Early diodes could have large mismatches in junction area and diffusion depth, thus C and PIV. Modern diodes seem to be much better matched. Also we can often over-buy (1000V for a 400V job) without much higher price.
Simply dont use parallel caps to modern diodes in rectifier bridges. Outdated practice leading to nowhere. Uh oh, generalization in front of you.
I like Schottky diodes in 50 Hz rectifier circuits as they have lower Uf so lower dissipation. Leakage current (except for silicon carbide Schottky diodes) is higher compared to "normal" diodes though. Lower reverse voltage was an item but 100V types exist for many years. Never bothered about the leakage current and used them with confidence. I use them in low voltage, relatively higher current applications combined with LDO's so I can choose a lower transformer voltage and keep total dissipation somewhat lower.
They have the dubious fame of being the diodes with the most wrongly spelled name
I like Schottky diodes in 50 Hz rectifier circuits as they have lower Uf so lower dissipation. Leakage current (except for silicon carbide Schottky diodes) is higher compared to "normal" diodes though. Lower reverse voltage was an item but 100V types exist for many years. Never bothered about the leakage current and used them with confidence. I use them in low voltage, relatively higher current applications combined with LDO's so I can choose a lower transformer voltage and keep total dissipation somewhat lower.
They have the dubious fame of being the diodes with the most wrongly spelled name
Last edited:
The article I wrote about diodes for linear power supplies (including Schottkys) is here. Linear Audio magazine charges EUR 0.99 for a soft-copy.
In my opinion the Abstract and the Conclusions can be summarized thus:
In my opinion the Abstract and the Conclusions can be summarized thus:
- 48 different diodes were tested in a high sensitivity jig, with and without CRC snubbers. These included slow, fast, ultrafast, Schottky, Silicon Carbide, HEXFRED, soft recovery, and Super Barrier diodes.
- All 48 diodes rang. Even the Schottkys, even the HEXFREDs. They all rang. Ringing amplitude is tabulated: the worst diode's ringing amplitude was 20X greater than the best diode's.
- (The very worst-performing diode, when used with a snubber), performed much better than (the best-performing diode with no snubber).
If you are facing crushing poverty I would be glad to subsidize 50% of the price of the Linear Audio article (price is EUR 0,99 -- here is the purchase link). Just post a copy of your receipt from Linear Audio showing your name and the date, right here in this thread for all to see, and then PM me your PayPal userid. When I see both of those I will send you EUR 0,50 by PayPal immediately.
Mark
thank you for your Input
would you please clear - you are talking to snubber for transformer secondary or parallel caps to each diode
thank you for your Input
would you please clear - you are talking to snubber for transformer secondary or parallel caps to each diode
The "old" method. The ringing frequency wil lower, but there will still be ringing.
Read the Quasimodo thread or the Lineair Audio article (Mark Johnson).
A good snubber includes a resistor that does the real damping. C+RC snubber is preferred on each transformer secondary.
Why trust other people to tell you the correct answer? They might be strongly opinionated, stubborn, and completely wrong. How would you ever decide whether a forcefully stated opinion is correct, or incorrect?
Instead, what a non-lazy person could do, to find out the correct answer, is experiment. Build and test and measure. Worship at the altar of data!
To show an example of the brave experimenter's approach, I have attached below a little test PCB (49mm x 99mm) that I built last month. Its purpose is to study the current waveforms in transformer secondaries, with and without different kinds of snubbers. You could design and build a similar board, which includes sockets for parallel capacitors across each diode, and then measure with / without those parallel capacitors. Voila, now you would KNOW whether parallel capacitors make things worse, make no difference at all, make things a little better, or make things a LOT better.
I've attached a .zip archive of the Gerber CAD files for my PCB layout, in case that helps you plan the layout of your own board. (If you're wondering why schematic #2 includes the oddball resistance value "1.44 ohms", that's the actual DVM-measured value of the resistor I used. It was marked 1.5 ohms.)
_
Attachments
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.