[Top] [All Lists]

Re: [Amps] HV Diodes / engineering pragmatism

To: "'Manfred Mornhinweg'" <>
Subject: Re: [Amps] HV Diodes / engineering pragmatism
From: "Leigh Turner" <>
Date: Wed, 15 Oct 2014 10:26:44 +1030
List-post: <">>
Manfred, you have made some most excellent riposte comments here...bravo!

I concur with them, and your pragmatic no-nonsense engineering design
philosophical outlook.

Your savvy and insightful posts often comprising long essays to this forum
are always ones that I look out for and enjoy reading.


-----Original Message-----
From: Amps [] On Behalf Of Manfred
Sent: Wednesday, 15 October 2014 5:48 AM
Subject: Re: [Amps] HV Diodes


 ** IMO there is a balance point between using obsolete diodes and overkill.

Are you suggesting that the 1N4007 is obsolete? I just checked, and it's
still in full production by several companies, such as Diodes Inc,
Fairchild, ON Semiconductors, Vishay, and a few more. It's a highly popular
diode, massively used, and despite having been available for a long time, I
haven't seen any signs of it becoming obsolete.

To me, a device is obsolete when something better becomes available at the
same price, or something as good becomes available at a lower price, and
thus the sales of the original device fall, making the manufacturers
discontinue it. As far as I know, this hasn't happened to the 1N4007. But it
has happened to almost all tubes...

So, I would look for a balance point between reliability and cost. Given the
low cost involved, indeed 1N5408 diodes might be that sweet spot, despite
being huge overkill. My point was to illustrate that technically the 1N4007
diodes should be fine for legal limit ham amplifiers using bridge
rectifiers, but a bit tight for those using voltage doublers. There was no
intention to force any of you to use them! So, please don't eat me alive...

 >> Let's assume a pretty big amplifer, solid legal limit, CCS, which is
 >> than any ham needs. The power supply might deliver 3500V at 0.8A. Each
 >> diode string in a bridge rectifier will then see a peak voltage that
 >> reach 4000V in the event of line overvoltage, and an average current of
 >> 0.4A at full output.
 > **  That is an obsolete assumption with RTTY and data modes having many
 > users these days. Even AM linear is pushing it.

Sorry, Carl, I don't understand you here. What exactly did you mean by
"obsolete assumption"? If I base my calculation on a 1500W CCS amplifier, in
what way would this be obsolete, or too weak for a ham running RTTY? In fact
it still has some headroom, because no ham RTTY operation is continuous in
the long term, even RTTY bulletins might last only 10 to 20 minutes!

Or did you mean that 0.4A average current per diode is too low, for an
amplifier running 1500W in RTTY? It isn't! If that amplifier runs on 3500V
and 0.8A, giving 2800W input and 1500W output, meaning a 53% efficiency,
then each diode in the bridge rectifier carries 0.4A average current. That's
a fact, not an assumption! The actual average current will be slightly
different, depending on the amplifier's efficiency, and the voltage it works
at, but will hardly get much above 0.5A, unless you have a lousy efficiency,
run illegal power, or use tubes that need a very low supply voltage.

The current will be peaky, due to the capacitive filter, but the average
will still be just that low value. The peaks might be around 3A, but such
repetitive peaks are within the safe operating area of a 1A diode, as long
as the average current is low enough.

 > ** I detest when someone who should know better starts championing the
 > lowest denominator.....the cheapskates that infest the hobby at the amp
 > level. Keep it to QRP where nobody gets hurt (-;

ROFL. Seems that we two represent opposing poles! I absolutely love getting
the most bang for the buck, and building legal limit amplifiers that cost
under 500 dollars, total. And which work well, too. Remember that old adage?
"A piece of engineering is the product of material and brains. The more you
use of one, the least you need of the other." I like using lots of brains,
and little material!

And yes, I built QRP equipment for the first several years of my ham career.
The 2SC1969 was my preferred output device, but I also sometimes ran a few
2N2222 in parallel to deliver one watt... I made many intercontinental
contacts using power levels under 5W, into simple dipoles or quarter wave
verticals. Mainly on ten meters. And many of them in FM! When the other guy
has good ears, you don't need a kilowatt. On the other hand, a kilowatt gets
heard better in many situations. As an old QRPer, I do know that!

 >> Any voltage transients will be clamped to the capacitor voltage.
 > ** Nope; just some will be clamped.

Do you mean that your filter capacitor is selective, and will clamp some 
transients but not others?

 > Others will possibly blow the diodes if
 > there isnt a path provided around them.

In a bridge rectifier, like in a voltage doubler, there is always a path
THROUGH diodes, in the forward direction, and that protects the diodes which
are reverse biased at that moment. Only if you have a center tapped
transformer with one diode string at each end, do I see a chance for large
voltage spikes to appear on the diodes. But such a configuration would be
weird in a modern amplifier. You might find it in very old amplifiers that
used tube rectifiers.

> Very fast and high spike transients will have a repetition rate that could
> be considered RF and we all know how well electrolytics like RF.

Yes, the electrolytics and associated wiring do have some significant 
inductance, that will limit their ability to conduct very fast spikes.
That's true, although many people believe the effect is larger than it
really is. But then, your transformer has a significant leakage inductance,
which is in series with any transients. This prevents such fast transients
from appearing on the diodes!

Let's imagine a simple model: Power line, transformer, rectifier bridge,
filter capacitor. The capacitor has some inductance, the transformer also
has some leakage inductance, and the power line has additional inductance,
between the place the transient is generated, and your amp. For the sake of
high frequency components in the spikes, this circuit reduces to line
inductance, in series with leakage inductance, the transformer's voltage
ratio, the diode bridge, and the output of the bridge shorted by the
capacitor's series inductance. So, the transient voltage appearing on the
diodes is basically dependent on the ratio between all these inductances. We
really need to put numbers to them, to calculate anything meaningful.

I took a few capacitors and measured their series inductance. A modern 330µF

400V electrolytic with short leads measured 31nH. That's _nano_henry! A
smallish 100µF 350V electrolytic showed 49nH, measured through long leads.
The additional inductance comes from the leads, not the capacitors. And then
I took a big, old oil filled capacitor of 10µF, 4000V, and it measured 72nH.
So, when you put 8 electrolytics in series, with typical wiring, you can
expect maybe 0.4µH total series inductance, while a single big oil capacitor
will be somewhat better.

Then I measured the total leakage inductance of a transformer intended for a

light duty 1200W amplifier. It has a 1760V secondary. It measured 660mH of 
equivalent total leakage inductance, on the secondary side. That sounds like
a whole lot, but is quite normal for a transformer like this. Instead a
good, generously sized, well designed transformer for a 1500W CCS amplifier
would probably have roughly half that much leakage inductance, or maybe one
third. It would be interesting if you, Carl, or anyone else who has several
suitable transformers at hand, would measure their leakage inductance. It's
done simply my shorting the primary and then measuring the inductance of the

Now, if you have a voltage divider composed by several hundred
_milli_henries on one side, but less than one _micro_henry on the other side
of the bridge rectifier, any spike voltage will divide by a few hundred
thousand times! In this way, the combination of the filter capacitor and the
transformer's leakage inductance should be a total, absolute, massive
protection against diode damage from transient overvoltage!

In practice, of course, things are never as good as one wishes. The fact is
that the leakage inductance of any transformer is paralleled to some degree
by inter-turn capacitance. And that capacitance conducts fast transients
very well. But at any frequency you wish to test, the mix of leakage
inductance and stray capacitance will always have some inductance left. How
much, depends on the exact design of the transformer. In any case, during 30
years in electronics, and being involved in high voltage supplies in
scientific applications and in industrial, noisy environments, I have never
found a transient large enough to cause significant voltage overshoot across
a bridge rectifier connected to an electrolytic capacitor, behind a
transformer. To me, this seems to be a non-issue.

Instead of you have a center tapped transformer and two diode strings, then 
there is no clamping effect like in the bridge or the voltage doubler,
because the transformer leakage inductance isolates one diode string from
the other. In that situation the diodes can indeed see significant voltage
spikes. Instead of fixing that by using lots of diodes, or capacitance in
parallel with them, I would rather avoid that configuration altogether. And
almost all designers seem to think like I do, in this regard.

 > A single spike may not cause damage but a string
 > of them will raise havoc to diodes and caps.

That's true, in several ways. One is that a lot of spikes, during RX, might 
charge up the filter cap to a voltage well above nominal, and all that
appears across the diodes. Another is that if there are spikes across the
diodes that are large enough to make the diodes go into avalanche, then the
diodes will have to dissipate the power, storing the pulse energy as heat in
their thermal mass, and then bleeding off this heat slowly. Any repeated
avalanches can of course overheat a diode much easier than a single event.

 > Adding a single 4700pf across each string and another to ground at the
 > output provides a non destructive path for whatever gets thru or around
 > the transformer. An oscilloscope will easily show the effect.

Yes, that can be very useful, if any spikes manage to get through the 
transformer. But if a choke-input filter is used, those little spike
capacitors would be essential! Or at least a single small cap across the
bridge rectifier's output, before the choke.

 > **  I just measured the secondary DC resistance of a couple of
 > dating over 50 years and all using SS doublers or 866A's and
 > 3B28's (-;
 > NCL-2000 doubler 14 Ohms
 > Hunter Bandit 2000C doubler 16 Ohms same EI ratings as above to 3-400Z's
 > Clipperton L doubler 5 Ohms and a well known "thumper"
 > SB-220 doubler 13 Ohms
 > Amp Supply LK-550ZC FWB 30 Ohms 3x 3-500Z, or 3 x 3CX800A7's in the
LK-800C and other QRO versions. A 46# Dahl C Core very silent
 > Command Technologies 2500 Magnum FWB  31 Ohms C Core, not a Dahl
 > Drake L4/L7 doubler 8 Ohms
 >>From a commercial water cooled pulse amp 7000VAC 2.5A FWB 78 Ohms. For my
 > "dream amp"
 > Ameritron AL-811H  44 Ohms

This data is interesting! I find those resistance values very low, though. 
Almost too good to be true. My own transformer for that ICAS 1200W amplifier

shows 92 ohm!

 > As you can see there is a fairly wide range of design philosphy across
many of them. Obviously the Ameritron was done as cheap as possible for
mainly a SSB audience, 45uF of C helps.

Yes, a lot boils down to cost versus performance. But not all. When you have
the task of designing a 2000V, 1A transformer, to mention any numbers, and
you get a certain budget for it, there are many different ways of doing it.
Which one a designer should choose, depends to a large extend on additional
data, which is dearly needed to optimize the design! For example, one is the
duty cycle. A transformer optimized for being the most efficient when
delivering 2000V 1A continuously, will be different from one optimized for a
service in which it delivers 200V 1A for one hour, and then stays plugged in
but idling for one week. The latter transformer will use a lower flux
density, thus more turns of a thinner wire, or will use a smaller core and a
larger winding package. This comes from the fact that core power loss occurs
continuously, while winding power loss is directly proportional to the
square of the load current.

Most manufacturers of ham amplifiers think ICAS, and in fact the lower end
of ICAS, that is, at most 50% transmit duty cycle, and no more than about
60% average current during transmit, relative to the maximum current. Such
an amplifier will require a reduction of power in RTTY, but will run full
power in SSB, CW.
Other amplifiers are "high end ICAS", they can transmit RTTY at full power
for several minutes, but not 24/7. Of course there are also CCS amplifiers
sold to hams, but these are unnecessarily large, heavy, and expensive, in my
humble opinion. The fact is that I won't transmit a key down carrier at full
power for a week - nor should any ham! So we don't need CCS. High end ICAS
is enough!

But I digress...

 > Those are DC ohms and not AC reactance and without knowing all the design
 > details Im not making any guesses as to flux leakage, etc.

To arrive at the true resistance that's acting in series, you have to
measure both the primary and secondary resistances, apply a factor according
to the transformer's voltage ratio, and then add them. For example, if you
measure 20 ohm on the secondary, a half ohm on the primary, and the
transformer is 240 to 2000V, then you have a ratio of 8.33. You have to
square this, obtaining an impedance ratio of 69.44. So, you can calculate
your equivalent secondary total resistance as

0.5 * 69.44 + 20 = 54.72 ohm,

or the equivalent primary total resistance as

0.5 + 20 / 69.44 = 0.788 ohm

In practice, it's often accurate enough to assume that the transformer was 
designed with about the same amount of loss in each primary and secondary,
so that you can measure either of the two resistance, and multiply it by
two, to arrive at the total resistance seen from that side.

The voltage drop of a transformer is given basically by the series
combination of these resistance, with the leakage inductances. Of course,
the total equivalent leakage inductance can be calculated by the same
method, after measuring both, or just one.

 > The cheapy SB-220 diode is basically a 1N4005 string of 14 and is the
 > weakest link in the PS and those diodes are known for going up in smoke.
 > Spikes dont help nor do gradually leaking filter caps.

The SB-220 uses a voltage doubler. Indeed in that case the current is too
close for comfort to the limit of the 1N400x series, and that's why in my
NCL-2000 I used 1N5408 diodes. The NCL-2000 comes with diodes rated at 750mA
in the schematic, and 1A in the parts list. Probably these are also stacks
of 1N400x diodes. This current capability is marginal for an amplifier using
a voltage doubler, but it's fine for one using a bridge! The diodes in those
amps go up in smoke because they work too hot from the current and slowly
degrade, and not because of any spikes.

 > The 1N400x series is actually rated at a 8.3ms peak 30A surge over a half
 > cycle.

Datasheets are a wonderful thing. Indeed the table of maximum rates says
exactly that, but the graph relating number of cycles to allowable current
gives the 30A level for a full cycle, not a half cycle! That would make it
about 35A for a half cycle, extrapolating the curve. Which one is true, I
don't know...

I checked the datasheets from Fairchild and Diodes Inc, and both contain the

same inconsistency!

 > And all ratings are at 25C ambient

Except the current rating of 1A, which is for 75°C ambient temperature, and
the leads held to that temperature at 9mm from the case.

 > resistive or inductive load ONLY.
 > For a capacitive load derate current by 20%.

Indeed the current rating has to be derated for capacitive loads. But where
did you get that 20% figure from? Obviously the true derating factor depends
on the capacitance value relative to voltage, current and frequency, and the
source impedance of the line and transformer. In some cases the derating
factor can be so small that we can ignore it, while in other cases it could
be as high as 50%! 20% makes sense as a typical, average derating factor,
but it's somewhat dangerous to appy it, instead of calculating the actual
factor that counts in the particular situation at hand.

A typical example of this happens in the abovementioned amplifiers using 1A 
diodes in voltage doubling supplies. It's marginal, but might survive for
many years as long as the filter caps are small, like the 10µF total of the
NCL-2000. When after some decades these filter caps die a natural death, and
a well-meaning but unknowledgeable ham replaces them by caps with the same 
physical size but a much larger capacitance rating, the necessary derating 
factor for the rectifiers increases too, and the small diodes die soon after
the capacitor retrofit.

 > ** Most 1N4007's are floor sweepings of no particular tolerance or QC.
 > mainly in LV consumer goods where a little leakage is OK. The 1N5408 and
 > bigger are commercial/industrial quality and subject to QC and tolerance.

Now where did you take that from??? ROFL! Come on! It's the exact same
companies that make 1N4007 and 1N5408 diodes, and I don't see why the
quality standards should be different! Both types are used in lots of
consumer and industrial equipment. The difference is their current ratings,
and thus their size, capacitance, leakage, cost, etc. Not their quality!

Did you perhaps buy a bag of of ultracheap 1N4007's on some web site, which 
turned out to be re-labelled 1N4004's ?

 > IMO continuing to promote the 1N4007 for larger amps is doing a 
 > misservice to those that are not familiar with the details.

I have no intention of promoting any specific part. The only reason to write
my post a few days ago was to try and make people see what the real working 
conditions of those diodes are, to help them decide what diodes they want. I

even wrote quite clearly that I have absolutely no problem with anyone
prefering to use larger diodes than necessary - that's the freedom everybody
has! I just wanted, and still want, to increase precisely this familiarity
with those details, among the general ham population, and amplifier users
specially! That's also why I'm taking the time to write this long message,
instead of simply hitting the "delete" key.


> ##  Commercial broadcast HV supplies will typ use TRIPLE the piv rating  
> for each leg of a FWB.  The theory here is.... the MOVs across the 240
> vac input, or 208vac, 3 phase input will not start to clamp until the ac 
> line V has doubled.  For a 3500 vdc no load B+ supply, use 10 kv piv per 
> leg. 

That makes a lot of sense, for power supplies without a filter capacitor 
connected directly to the rectifier. With that cap, I think triple
overrating is overkill. But then, of course, the cost is so low, that
probably they do this for peace of mind.

> ### A 1N5408  runs pretty damn warm to hot  with 1A  CCS  flowing, when 
> I tested em for bias use.... using a variable dc power supply + resistor
in series
> with the string of 1N5408s.   A 6A10  runs warm to hot with just 2A CCS
> And that?s with full lead lengths on each end of each diode. 
> Try running 1A  CCS  through a 1N4007, and see hot hot it gets. 

Using these diodes with full lead length is foolish. The data sheet
specifies current rating with 9mm leads, and that should be taken as an
absolute maximum length. When I use those diodes near their full current
ratings, I make the leads short, and solder them to nice big circuit board
areas. That way they work very much cooler than with long leads!

Also these diodes, like all silicon semiconductors, can take some
significant heat. In fact they are rated for 1A, with 75°C ambient
temperature. At that ambient temperature, they will burn your finger even at
zero current! Working at 1A in a 75°C environment, the silicon will be at
roughly 120°C, which is hot enough to instantly vaporize any water, and
cause nasty burns, but is safe for a pretty long time for the diodes. It
still leaves some thermal headroom for spikes!

So don't assume that a diode that burns your finger is working too hot.

> ###  Nice try.  My dahl xfmr has a 6 ohm dc resistance across its 5200 vac
> winding, and only  3 ohms across the 2600 vac tap.  Pri resistance is just
> .002 ohm.

That sounds like a transformer of roughly 20kW, maybe more. Indeed with such
a transformer you need diodes MUCH bigger than what we are talking here, and
at least a 15kW power tube. At that point surely we aren't talking ham radio

anymore! When I write on this forum, I have ham amplifiers in mind. That's
1500W output, strictly, with no concessions. If anybody wants to run more,
that's his problem, but I think it shouldn't be assumed that we are talking
broadcast transmitters in a ham forum!

And the resistance ratio of that transformer clearly suggests that the
primary is for 120V. Do you really run such high powered amps from 120V? I

> ##  If you cant afford to buy 1N5408s  or 6A10s.... you shouldn?t be in
this hobby. 
> Even if you are in the poor house,  you can still afford something better
> than the $.0990  1N4007 pos diode. 

I have to say that I find this quite a bit rude. First, ham radio is
supposed to be a hobby for people interested in radio, without any
exclusions based on wealth. Secondly, I'm not too poor to afford a few
1N5408's, but I'm enough of an engineer, in addition of a ham, to try to
figure out when it makes sense to use a larger part, and when a cheaper,
smaller, lighter, is plenty. And with ham legal limit amps I drew the line
at the power supply configuration: Bridge rectifiers feeding 1.5kW amps at a
3kV level can use 1A diodes, while voltage doubler circuits would be
marginal with them, and thus should use larger diodes.

And I also wrote that it's perfectly fine with me if some fellow ham elects
to use larger diodes than technically required. Be it just for peace of
mind, or because he has them by the hundreds in his junkbox, or because he
finds them cheap enough to not lose a thought on it, or because he wants his
amp to survive the next several nuclear wars including EMPs. It's his
business, not mine, but on this forum we can hopefully all express our views
of technical matters, without being thrown out of the hobby!

Personally I have an issue with people who believe that heavier is always 
better, and that more expensive is always better, and that rules of thumb
work better than actually calculating, but I respect their right to live as
they please, and enjoy themselves!

> ##  How cheap can you get ?

There is no limit! ;-)  When you get the hang of it, there is a huge thrill
in doing things as cheaply as you can, and at the same time BETTER than some
other people manage by throwing lots of money at it!

> Blow up some 1N4007s and you will be cursing that you didn't use a real 
> diode.

I have been using them for 36 years, and can't remember having ever blown up

any! It's strange, actually, as I have sometimes blown up some other
components. I have been using bigger diodes too, almost for the same time,
in circuits that need them. The largest I have used in my own designs were
rated for 400A, and the smallest for 15mA. It's not like one size would fit
all applications.


Visit my hobby homepage!
Amps mailing list

Amps mailing list

<Prev in Thread] Current Thread [Next in Thread>