On Thu, 07 Aug 2003 01:08:19 +0000, Fred Nachbaur wrote:
Russ W. Knize wrote:
On Wed, 06 Aug 2003 12:52:06 -0400, Steve Robertson wrote:
If you're hell-bent on building your own, then I would suggest a CLC
design to get a higher HT. I don't think regulation is going to be that
big a deal unless you are driving the thing way past class A. And that,
my friend, is going to be so loud that you will likely not hear the
distortion nuances introduced by a bit of power supply sag.

I have been considering the Pi filter as well. The only problem is that
with this transformer, I am looking at 700VDC which is a bit much. They
do have a 510V @ 400mA that I could buy two of, which would give me more
like 660V, which is the max for the 8417s. The only thing that bothers me
about the Pi filter is the EMF generated by leakage flux due to the
current spikes required to drive the first cap. Perhaps I am just being
paranoid. Is this really an issue with properly bypassed hexfreds? Should I go
with a tube rectifier? Any thoughts?

Well, both the capacitor-input (pi) and choke-input (L) filter
topologies have their advantages and disadvantages. You've nailed the
main disadvantage of the pi filter; those high-current spikes can be a
bugger. They can even cause audible buzzing on transformers that aren't
potted. But if you're building your power supply in a separate
enclosure, you should be alright.

I may end up having to do this anyway. Is there a "rough figure" to the
amount extra capacity a transformer should have to keep the buzzing under
some degree of control?

On the L filter side, it can be a nuisance because of mechanical humming
of the filter choke. Furthermore, since such cores are generally gapped,
there will be a lot of 60 Hz. magnetic fields in the vicinity. This is
not normally a problem in line-level amplifiers, but can be a bear to
correct for in high gain (phone, mic or instrument) amplifiers, pentode
input stages seem especially susceptible to this.

I have read this as well. How are Hammond's chokes in this regard?

The other thing to be aware of is that the regulation issue isn't "cut
and dried." The L (choke-input) filter does have better regulation over
most of its operating range, staying pretty close to about 0.9 times the
RMS voltage from the transformer. Except at low currents, where it
skyrockets to up to 1.4 times the RMS value!

Right, that is an issue that bothers me, but I rarely see it addressed.
Sometimes I see shunt resistors put in, but that is wasteful.

If you design for a C-input filter, you'll generally end up with about
1.2 times the RMS voltage under load. So a 510v transformer will give
you about 600 volts under load, or about 715 volts with no load (as
during warmup). To do the same thing with a choke-input filter, you'd
need a transformer of about 665 volts RMS. That means, with no load, the
output voltage will be a whopping 930 volts!

On the other hand -- your transformer's current rating can be a lot
lower for a given DC current output, using the "L" filter topology. This
is because the peak currents are a lot lower. Power loss (and therefore
transformer heating) is proportional to the square of the current
integrated over a cycle, so capacitor-input (pi) filters tend to be a
lot more demanding of power transformers.

Vacuum tube rectifiers are, for me anyway, just a convenient way of
adding series resistance to capacitor-input filters (in order to limit
peak currents, and therefore keep both mechanical and induces buzzes
down to a dull roar). There are people who claim that rectifier tubes
have a "sound", but I could never hear a difference between a tube
rectifier and a silicon rectifier with the same series resistance. The
main difference is, as I say, the convenience of having the "resistor"
socketed and above chassis, where it can vent its heat reasonably elegantly.

Speaking of series resistance, how does one calculate Q for a Pi filter?
I know how to do it to critically dampen an LC filter, but I have not seen
anything for a PI filter. Playing with Duncan's PSU Designer shows that
the Pi filter does resonate, though it takes less C and R to squelch it. I do
like the fact that I can get away with less overall C in a Pi filter to
achieve low ripple and low Q. It allows me to afford higher quality caps.

For choke-input filters, I don't see any advantages to vacuum diodes, in
fact I consider them a liability. In one case I wanted to use some 866AX
merc rectifiers just because they looked cool, but the hassle wasn't
worth the trouble. When the filament transformer arced over, I said
"enough of that nonsense!" and changed to a convenient, trouble-free,
compact, and inexpensive silicon diode string. Problem solved.

Yes, I was following your formerly 300A amp project. Was kinda looking
forward to the blue glow, but I agree about the annoying filament
requirements of the 866As. I have also ruled them out for another project
in the back of my head....