Questions about Fast, Ultra-Fast, Hyper- Recovery Diodes

[bob p]

I'd like to ask for advice about using Fast / Ultra-Fast / Hyper-Fast recovery diodes for power supply rectification. I'm primarily interested in deploying them in full wave or bridge circuits.

Can anyone tell me what type of diode best fits the bill? For SS rectification I'd normally use 1 or 2 1N4007 in series, bypassed with a 0.01 uF ceramic cap.

Can anyone tell me how the ultrafast UF4007 would compare in this sort of application? Does it need to be bypassed?

Is there any added benefit in using Hyperfast diodes like the Fairchild RHRP8120? (Mouser # 512-RHRP8120).

Thanks!

[Le Basseur]

I don't know about the Fairchild diodes you mentioned (they're unavailable where I live) but I'm using Philips' BY448 diodes.Maybe the're not "the fastest" out there but they're reliable and available (and no,I don't bypass them as I didn't see nothing critical when 'scoped the upgraded PS's).
Really,what's the catch in chasing for a ultra/hyper-fast diode in a guitar amp's HT supply (no pun intended,just curious)?.
I'd be more worried about,say,a LT heater supply for a DHT triode,where the recovery speed/spikes/overall components' quality REALLY counts.
Of course,I could be wrong...
Regards,

[bob p]

oh, i guess i should clarify that my interests are not limited to guitar amps. the question is more related to a hifi project. we don't have a hifi forum here, so i posed the theory question in the theory & design section for guitar amps.

even though my question is hifi motivated, i will probably end up using at least UF4007 in my guitar amps. given the moderate price difference, its not worth keeping a supply of 1N4007 around.

to reopen the topic, if someone can help me out with the differences between hyperfast, ultrafast, and regular diodes, i'd appreciate it.

[dai h.]

sometimes you see diodes bypassed with ceramics usually, about 2200pF up to 0.047uF to reduce the switching noise, and it was my understanding that with fast recovery diodes, you don't need to use them. Subjectively, I tried some fast recovery (UF4007?), and at the time it seemed to make the transients too quick, but maybe there were other things I did that didn't help such as increasing filter cap values. I think I tried some in a BBE 422A sonic maximizer, and what I recall is that while it seemed to help give more detail, the sound seemed weird and didn't sound as nice to listen to, so I changed them back to ordinary diodes. OTOH though, I've read many posts where people didn't seem to note any particular change.

[R.G.]

The need for fast rectifiers comes from the fact that slow rectifiers allow substantial energy remaining in the localized inductances and capacitances to continue ringing while the diode turns off. This produces a hash of RF oscillation that radiates and conducts to everything, and gets rectified back to an audible pulse by the rest of the circuit.

The definitive (to me at least) info on diode ringing is http://www.hagtech.com/pdf/snubber.pdf where the author discusses the optimum way to snub a ringing rectifier diode.

I see several ways to cope with diode turnoff ringing:
- use a really fast diode
- tune the local ringing to a lower frequency with a good RF cap like a ceramic
- snub it with an RC snubber ala Hagerman
- use MOSFETs as synchronous rectifers; this is common in switching power supplies.

The practical methods are always a fight between the cost of fast diodes and the overhead of RC snubbers.

[bob p]

thanks for the excellent link, RG. that's exactly the type of information that i was looking for.

so far i've been using the "C Snubber" method of bypassing a 1N4007 with a 0.01uF ceramic cap, and i've probably been getting results like those that Hagerman shows in Figure 11.

although Hagerman's paper suggests that the RC Snubber method (Fig 10) would be better, or that the RC + C Snubber method would be better still (Fig 12), I've been thinking about abandoning the 1N4007 diode altogether. My stock is running out, so I'm thinking now is the time to find out if I'd be better off replacing them with something else.

The one point that remains unclear after reading Hagerman's paper is how faster diodes like the "ultrafast" UF4007 and the "hyperfast" Fairchild compare to the plain-Jane 1N4007.

are the "ultrafast" or "hyperfast" diodes significantly better than the 1N4007 in not producing the dreaded RF hash that requires us to start using snubbers in the first place? i'm hoping that since you've recommended the use of UF4007 over 1N4007 in another thread, that you might have some experience in this area.

i don't mean to put you on the spot, but by any chance do you have a handle on the relative performance? essentially, I'm trying to decide whether I'd be better off continuing to use the 1N4007 and building snubber circuits to tame the hash, or whether using a better diode like an ultrafast or a hyperfast would allow me to forego building the add-on snubber circuits.

any insights that you might have would be appreciated.

[gdgross]

I use UF40007s exclusively.

I have no idea if they are any better, but they might be, and, in my mind, the marginal price difference is not enough to bother with 1N4007s.

edit - that is, exclusively for power supply rectification. I still used 'normal' diodes in pedals and such.

[R.G.]

The one point that remains unclear after reading Hagerman's paper is how faster diodes like the "ultrafast" UF4007 and the "hyperfast" Fairchild compare to the plain-Jane 1N4007.

Me too. That's why I leaned on my switching power supply design career experience. You notice when the ringing occurs? It's when the diode is turning off. We had major disasters back in the day trying to get diodes to turn off fast enough to not make huge EMC problems or to simply go up in smoke from the dissipation. Normal rectifier diodes saturate their depletion regions full of minority carriers during forward conduction. When they are reversed, there has to be enough reverse current to sweep all of that out, and then the diode slams off. I think it's the buildup of reverse current and the slam off that excites the ring. There's a lot of energy in the slam. by contrast, fast diodes have a reverse recovery time, but it's so fast that they turn off before the parasitics can ring.

i don't mean to put you on the spot, but by any chance do you have a handle on the relative performance? essentially, I'm trying to decide whether I'd be better off continuing to use the 1N4007 and building snubber circuits to tame the hash, or whether using a better diode like an ultrafast or a hyperfast would allow me to forego building the add-on snubber circuits.

I'd reduce that to a matter of money. If the UF diodes are cheaper in money and time than snubbers go that way. If snubbers are cheaper once you figure in the parts, labor to put them in and space they take up, go that way.

Better yet, order yourself a few UF's, and make up a test case with UFs and the more-optimal snubbers and listen.

I know the answer for switching power supplies, but for audio it's different.