How to Tame reverse recovery spike of the UF4007?

[CpuZapper]

Is this necessary in the ST-35 power supply?
I have been thinking of two methods to accomplish this and could use some advice from someone who has tried this on the ST-35 power supply.

1: The reverse recovery spike filter or RRSF.

The only drawbacks is it can generate more heat and more components.

2: The UF4007 (First) in series with a CREE 2A 600v (Second) diode.

From what I read two diodes in series will increase the voltage rating, but note two Cree's in series will not work in this manner.

Is the Cree protected and safe to use like this?
If so it would only add a $1.05 per Cree diode to block the reverse recovery spike.

What do you think? Hope someone can tell me if I'm on the right track.

Thanks, egilbert.

[erichayes]

Hi Egilbert,

For a linear power supply running off of 50 or 60~ mains, RRSFs are totally unnecessary. Thing is, everybody's using switching supplies to run their ultrafast computers and whatnot, so not much is written about the plodding, ancient power supplies we use with tube amps. If UF diodes weren't virtually the same price as 1Ns, I'd still be using the latter with a .001µF cap across them.

My 100 watt monoblock prototypes use three UF5408s in series on each leg of a standard full wave center tapped, choke-input power supply. I'm achieving .0027% THD @ 1W. Believe me, if the power supply wasn't sterling, there'd be no way in Hell I could get those numbers.

[CpuZapper]

Hello erichayes, with so many people using a RRSF on other sites for tube amps, I thought why not look at some type of spike suppression. I have used 1N's with a .001uf cap as well but have read you can cause bigger trouble like this if you don't use a resistor for a proper snubber and then it must be tuned to the circuit if I understand the articles that I have read. I'm learning as I go with no way to test my results so I ask questions.

What do you think of the UF4007 in series with the Cree, is it safe to use in this situation without any drawbacks? I would like to try this if it is safe and won't cause additional side effects.

egilbert.

[Gingertube]

I still routinely wire a 10nF 2KV disc ceramic across the UF diodes.
It might be "belts and braces" approach but the additional cost is insignificant.

The thing to watch is that any reverse recovery spike reflects back through the transformer into other windings and so if you have a bias supply or a rectified heater winding etc. these MUST be given similar treatment, that is: don't just do the HV circuit and neglect other secondaries.

Cheers,
Ian

[erichayes]

The beauty of DIY is that you can do any damn thing you want to, and be held accountable to no one. If you want to give the Cree a go, do it and report back to us. Just remember to fuse everything so you don't wind up looking like a Wile E Coyote "after" picture.

Whenever a piece of new technology comes along (or old technology is reinvented), some audio "authority" discovers it and proclaims it to be the greatest thing since Rice and Kellogg developed the dynamic loudspeaker*. My compatriot in New Hampshire has been known to use the term "snake oil" for a lot of these discoveries, and it's why I frequent this forum exclusively.

I enjoy dealing with open minded people who aren't afraid to actually DO some experimentation. Conversely, I am tired of reading the opinions of stuffed shirts who won't listen to anyone but themselves regarding audio. I got trained to listen by two of the best ears on the planet--Bud Wyatt's-- and I can assure you that most of the golden ears on the other forums can't tell an oboe from a soprano sax.

To reiterate, don't be afraid to try 1Ns, UFs, Cree diodes, snubbers, or any combination thereof. The worst that can happen is you'll blow a fuse.


* Q: What's a conventional tweeter cone called?
A: A Kellogg Rice Crispy, of course.

Please, ripe tomatoes only.

[erichayes]

Hi Gingertube,

In the colonies "braces" are known as "suspenders", but the analogy is universal.

Damn, I love the way you guys corrupt the Queen's language.

[CpuZapper]

Thanks all, I think I will try the better Cree's and spend a little more to be safe. I never considered that the spike would travel back into the windings effecting other areas, just one more reason to try some sort of spike suppression or better diodes.

I will post back my results but it will be some time before I get all my parts. I hope I end up with something worth posting pictures of.

egilbert.

[EWBrown]

In the colonies "braces" are known as "suspenders", but the analogy is universal.

Damn, I love the way you guys corrupt the Queen's language.

Avago at these:


http://www.upfromaustralia.com/aussieslang.html

As I remember, the late and great Steve Irwin used a lot of these pjrases in his "Croc Hunter" TV show. The title phrase was used repeatedly in one particular show Yellow_Light_Colorz_PDT_01

Then these is this memorable bit of "strine" seen written on an outhouse wall:

"May all your chooks turn into emus, and kick your dunny down"


/ed B in NH in the land upover

[kheper]

If price is no consideration you could use
hexfreds or other fast, soft recovery diodes.

http://www.vacuumtube.com/Products.htm##Fast%20Recovery%20Diodes

Here is an ultra-fast, soft-recovery dual diode.
Available for $1.30 at digikey.

http://www.fairchildsemi.com/ds/IS/ISL9K8120P3.pdf

[WA4SWJ]

Hi Guys,

I agree with Eric. Does any SCIENCE exist here? Who says "reverse recovery spikes" have any effect on anything? Of course reverse recovery time exists in semiconductor diodes and is well documented in data sheets, but is there any definitive data showing that this is any kind of a problem in vacuum tube audio amplifiers? Anybody done some analysis with oscilloscopes and spectrum analyzers and the like to see if this causes aberrations in the output of the amplifier? You get the picture I think. I'm not trying to be a smart a** or be critical of the question. It's a legitimate question to ask. But it just ain't necessarily so just because one of the bloviating audiophools (Ed's term I think) says it's so. Where's the science? My guess is that the very faint amount of hum that some of us have in our amps would completely mask anything that might be caused by reverse recovery spikes.

I have to admit too that I don't read too many other sites on these subjects due to the lack of science. Point us to something - I'd sort of like to read about this even if it is floobydust (remember that?). I'll try a search and see what comes up.

It would be interesting to find out what your results are. Let us know, especially if you have some equipment to make some analysis. That would be interesting for sure. Again, I'm not slamming anyone here. It's an interesting question. Let's see the data though.

My guess (not based on any data so you can call it an opinion) is the distributed inductance and inter-turn and inter-winding capacitances that come into effect at high frequencies in the power transformer itself completely remove any effects of this. Remember your Fourier analysis of fast rising signals and the effect of these distributed inductances and capacitances have on fast rising signals? Think low pass filter. What do those do to a fast rising signal? They round the edges significantly - you can make sort of good sine waves from square waves with the right low pass filter. This would completely remove any high frequency effects due to reverse recovery in the rectifier diodes. Just some thoughts (opinion). Maybe someone has some data on this that compares all these boutique diodes.

This is an interesting thing though. Let's see what some true science can produce. It might even be there somewhere on the net. Thanks for the soapbox.

Best regards,

[kheper]

Maybe someone has some data on
this that compares all these boutique diodes.

I do not know about the 'reverse recovery
spike phenomena'. I doubt that it is the cause
of blown power trannies, or we would have
heard about it by now. But, INx and UFx
diodes have long forward times, much
longer than tubes. So, a hexfred is not a
bad idea; It is actually cost saving, given
the prices for NOS 5AR4s. I checked the
price for the Fairchild ultra-fast, soft-recovery
diode. It is now $1.07. Not exactly a
'boutique' in expense.

[dhuebert]

floobydust

I do remember that! Maybe we should ask Bob Pease this question, he has given the audiophools some good blasts in the past. And when he worked for Philbrick he designed tube based operational amplifiers! So if anyone could offer some rational thinking, it'd be Bob.

My thinking on this one tho, is that these current reversals might contribute a little noise in the ground plane and not much more.

Don

[EWBrown]

From my perusals through Botlehead's forum, it looks like RRSF is more applicable in low level signal applications like phono and line stages, and less relevant in power amplifiers, where this tachnique begins to approach the essence of snakeoildom. Yellow_Light_Colorz_PDT_16 Yellow_Light_Colorz_PDT_04

Their approach was something like 0.001 / 1600V ceramic disc caps across the power trannie secondary, then a 10 ohm resistor before the two uF4007s, with a second 0.001 / 1600V uF ceramic disc cap from anode to anode.

Ceramic disc caps have only two legitimate audio applications, and neither is in the actual signal path. RRSF is one, and DC bypassing is the other.

Same applies for a 4 diode bridge, cap across secondary, two 10 ohm resistors, a second cap at the resistor AC inputs to the bridge.


HTH

/ed B in NH

[Gingertube]

Below is something I posted to another forum talking about diode reverse recovery spikes. This was in response to a request for an explanation.
I hope you find it useful and not just more confusing.

I have replaced standard 1Nxxx type diodes with Ultrafast Soft Recovery types in at least 8 tube amps and I have found it always makes an improvement.


Skip to the last 3 paragraphs if you don't want to wade thru' any of the physics - although I tried to keep it simple.

Whenever you have a P and N type semiconductor material the P type has an affinity for electrons (called holes - somewhere for electrons to go) and the N type wants to give electrons up for each of them to attain their most stable states. This has to do with the properties of the doping elements.

When you bung em together to form a PN Junction in a diode (Anode and Cathode respectively) or a transistor junction electrons flow from the N to the P creating a "depletion region" around the junction - that is its depleted of "holes" and electrons. The electron flow is terminated when the electric potential created by this flow of electrons (charge) is sufficient to resist any more flow. The depletion region will have a certain width

The junction (depletion region) will therefore has an electric field across it and hence has capacitance.

This junction capacitance has a Reverse Recovery charge (Qrr). That is, before a diode can turn off this stored charge has to be removed (in technical jargon - by recombination of majority carriers). In simpler terms - Current has to flow to remove this charge. Therefore as the AC voltage reverses on a rectifier diode there will be a pulse of current in the wrong direction before the diode actually switches off. This puts a noise spike on the power supply rail which has very high current rise time (di/dt). Filter capacitors are not very good at removing (shunting to ground) this noise spike because they have a small amount of inductance and inductors resist current change. Therefore you end up with a noise "SPLAT" on the power supply rail.

The differences in Silicon, Ultrafast Silicon, Schottky and Silicon Carbide Schottky of interest when using them as rectifiers is how fast they can turn off and how large this Current "SPLAT" is before turning off. This is obviously directly related to their Reverse Recovery Charge (Qrr). Small capacitors across each diode can help to absorb this splat and for some time they were in fact required by Electromagnetic Interference (EMI) standards for all commercial products (at least in Europe - I don't know if this is still true).

SKIP To HERE:
So here is the useful (representative ONLY) Qrr data:
Standard Silicon Diodes Qrr approx 500 nC (nano-Coulombs)
Ultrafast Silicon Diodes Qrr can be down to 100 nC
Schottky Diodes Qrr 50 to 70 nC
Silicon Carbide Schottky Diodes Qrr <20nC

Oh - To answer the original question - the other thing you need to know is that the noise spike associated with the current "SLAT" makes your amp sound seriously crap and that its edge speed is often fast enough to radiate into anything that looks like remotely like an arial (resistor leads, connecting wire etc). It can make the sound very "harsh" and is typically noticed mostly in the smoothness and detail of the high frequency "top end".

I rabitted on a bit here because the concept of junction capacitance is very important not just in rectifier diodes but also in transistors and is something amplifier designers need to deal with. If you want to understand things like EARLY Effect in transistors its a concept you'll need to know.

Cheers,
Ian