The diodes are probably breaking down. You need to have them rated at
2.8X of the voltage for the peak to peak from the AC. I would use
diodes rated to at least 3000 V to 4000 V minimum in this case because
the no load voltage is over 900 V. You can use 4 of 600 V in series to
get the proper voltage rating. I would also put caps at about 0.005 uF
/ 5 kV rated across each diode. This would be for HF noise suppression.

You should find out the required current needed for the supply load. The
diodes should be rated to at least 3X the required current at minimum.
This will allow for the inrush when the power supply is started up. The
electrolytic filter caps have to get charged up.

In series with each leg of diodes in series, I would put a 10 ohm 5 Watt
resistor in an average tube power amp supply.

If you use 1 Amp rated diodes, you can easily find diodes rated to 1000
Volts. This would allow for 900 Watts maximum load at 1000 Volts. Using
3 in series on each leg of the rectification path would be very adequate
for a 1000 Volt no load source.

I would not re-use the present diodes at this time. Most likely they may
be damaged even if they read okay.

Before going to all this trouble, make sure you find the cause of the
original rectifier failure. It may be age, but a short or something
pulling too much current should not be ruled out.

If you want to go more authentic and you have the space, get an 8 pin
octal socket, and a 5U4-GT tube. Knock out the holes in the chassis and
install the tube rectifier assembly. You will need a high voltage
isolated floating 5 Volt 2 Amp supply just for the heater. The heater is
tied hot to the cathode of this tube. You can then feed the AC 900 Volts
plate to plate, and the rectified plus source would be on the cathode.


--

Jerry G.


"EricM" wrote in message
...
I have replaced two old Federal/ITT selenium rectifiers in an
amplifier power supply with new silicon diodes; the original part
numbers are 103H4AX1 and 104B1AX1. I can't find any reference to
either part on the web - anywhere - ITT can't even provide
information. Since one of the legs of the circuit supplies the plate
voltage (600V) I used 600V 10A diodes (NTE5815HC) to be safe. Problem
is, there is a time-delay relay that closes to engage the 600V
circuit, and without it attached to the amp; all tube supplies,
biasing voltage, etc. are normal. If hooked up, once the relay closes
- the main power fuse blows. I'm guessing I need a dropping resistor,
but it would be nice to have the original data on the selenium parts
to be able to figure out the value. The 600V and the -38V biasing
voltage are derived from the same part of the power transformer. The
-38 side works, but the 600V side reads upwards of 927 volts without a
load. Using a Variac, the plate voltages were above 600V at about 70
percent. Is there a way to find the original specs on these selenium
parts? Any leads would be greatly appreciated!