>Well, so far the only significant reason that has been offered, by Steve K.,
>is that for fast transients, the rectifier string "appears to be a string of
>small capacitors." True enough, and indeed small they are. Down in the 10pf
>range.
>
>Now. lets consider this in context. These rectifiers will typically appear on
>the secondary side of a large transformer that is not exactly a shining
>example of high-frequency transformer design. Although I have never gone
>looking for it, I would not expect to see much high frequency energy on the
>secondary side. Especially peaks in the range of many kilovolts.
>
Why not?
It is quite common to see mains spikes of 1kV and more on the *primary*
side of the transformer. According to ANSI/IEEE statistics, "medium
exposure" sites in the USA can expect several hundred 1kV-peak mains
spikes per year, and even 5kV-peak spikes at about 1 per year.
Plate modulation transformers and output transformers for tube audio
amplifiers will transmit energy up to at least several tens of kHz, and
there is nothing very different about mains transformers of comparable
size. Therefore there is every reason to expect very large
high-frequency spikes reaching the diode strings, from time to time.
>In order to get anywhere with this line of reasoning, we need to adopt a (or
>some) generic models of transient pulses and look at the possible results. I
>think it is futile, however, because of another practical factor. In a
>typical full-wave rectifier, one leg of the rectifier is always in
>conduction.
This is not true for the capacitor-input supplies that we use, so the
rest of the argument doesn't hold up.
In a typical capacitor-input supply, both strings of diodes may be
non-conducting for almost all of the time! In the first quarter-cycle,
the "forward" diodes will not start to conduct until the applied AC
voltage swings up above the DC voltage on the capacitor. The diodes
conduct and charge the capacitor until the AC voltage reaches its peak,
but just over the peak the diodes become reverse-biased again and stop
conducting.
Therefore the diode current is a short, sharp pulse. The duty cycle of
this pulse is very low in the RX condition, because very little DC
current is being drawn. When transmitting, the duty cycle increases at
zero-signal anode current, and increases further as you draw more DC
current. But in linear amplifiers the maximum current you can draw is
limited by the sag in the DC voltage, so the on-time of the "forward"
diodes is very unlikely to be greater than about 15% of the full mains
cycle.
All the rest of the time, no diodes are conducting at all, so they
provide no damping to the circuit.
>
>The basic technical issue at work here is still hanging wide open, and I
>maintain my challenge to anyone to offer something we can chew on. That is,
>how do you get all the rectifiers in the string into simultaneous conduction?
>Bear in mind that when reverse conduction happens, it does so at the nominal
>PIV+. The diodes do not go into an irreversible SCR-type avalanche where the
>voltage drop essentially goes to 0. So even if a transient happened to kick
>the string into conduction, as soon as the transient dies away, we have
>sub-PIV conditions again.
>
That's correct. If the avalanche current is limited, all modern diodes
will avalanche gracefully, and recover when the voltage falls below the
PIV (approximately).
And that is precisely the reason for the "no shunt components"
philosophy! If one or more diodes do go into reverse-bias avalanche, the
current is limited by the rest of the string which has *not* avalanched.
Shunt Rs and Cs may do more harm than good because they allow more
reverse current to flow.
The need for shunt components is a hangover from the old days when
reverse-biased diodes would arc and fail short-circuit, which then
unzipped all the other diodes in the string. Also diodes were expensive,
so strings were as short as possible, and designers used Rs, Cs or both
to try and keep them alive.
Today it's different, because diodes are cheap and it's easy to install
lots of PIV.
>Enough for now. Let's try to put this myth to rest once and for all, or else
>give it a dignified birth certificate.
This is no baby. We're still catching up with a technical change that is
old enough to have children of its own.
These "children" issues are related to the EMC performance when you
don't have capacitors across the diodes... but before we talk about
that, we all need to get past this PIV thing first.
--
73 from Ian G3SEK Editor, 'The VHF/UHF DX Book'
'In Practice' columnist for RadCom (RSGB)
New e-mail: g3sek@ifwtech.co.uk
New website: http://www.ifwtech.co.uk/g3sek
|