Well, I was overly optimistic when I said the change of C5 1500 uF to 1 uF
(on the CDC-245 CPU board) eliminated all of the noise. After using the 92M
's extensively I observed that some popping and hissing remained on the 1st
LO when the receiver was off tuned a kHz or two from a strong signal. This
was especially noticeable when using ECSS (which I seldom use).

Recently I noticed there were two 5V lines, and that they were easy to
disconnect from the 92M 5V power supply. I did that and powered the 92M 5V
line from an external power supply. I could hardly believe it, but the
remaining popping and hissing noise went away. As I said, there are two 5V
lines, one to the front panel, and one to the mother board (and all the
plug-in PC boards). So I powered each of the two 5V lines separately from
the external power supply. It turned out that the noise problem was due to
one of the plug-in PC boards. It was not difficult to provide 5V power from
the external power supply to each of the plug-in boards, one at a time, by
unsoldering one end of the 5V line choke on the plug-in board. This
isolated the noise problem to the CGA-78 Loop 3 & BFO board.

At first I considered buying a 5V regulated power supply and bolting it to
the rear panel. That would have been the simplest approach in terms of time
required to modify the 92M. I measured the current drawn by the CGA-78
board 0n its 5V line, and it was about 620 mA., while the entire plug-in
board 5V line drew about 1 amp. The internal 92M 5V power was derived from
the 92m 24v nominal (25.8V) line. There were no other windings on the power
transformer for the lower voltage lines. This is because the 92M is
designed to be powered off 24V DC. If I used an external 5V regulated power
supply, the 92M could no longer be powered by 24V DC. So I decided to try
adding another 5V line to the 92M using a 3 pin 1 amp 7805 regulator.
Connecting the regulator directly to the 24V nominal line did not seem like
a good idea because of current foldback and power dissipation
considerations. According to the 1993 ARRL handbook a 12 volt difference
between input and output is the maximum acceptable for full current rating.
The 92M nominal 24V line is typically 25.8V, which gives a 20.8V difference.
The minimum acceptable voltage difference (dropout voltage) is about 2.5V,
so I decided to use power resistors to drop the nominal 24V line to about 8V
at the input of the regulator. (Here it might be better to use a 20V, a
15V, and a 10V regulator in series bolted to the side panel as a heat sink.)
Most of the excess power is dissipated in the power resistors (four 100 ohm
10 watt in parallel). These must be clamped to the chassis or they will run
too hot, 100° C or more.

The voltage regulator must also be heat sinked. At first I used a finned
heat sink, but the 7805 ran at 62° C, closer to the 70° C maximum rating
than I wanted. So I bolted it to the chassis rear, which also forms part of
the rear panel. Now the 7805 runs about 45° C, less than the 92M power
supply, which is about 52° C as measured on the heat sink fins through the
side panel with the room at 23° C ambient. At turn on, the input voltage is
about 7.8V, just slightly higher than is needed to meet the dropout voltage
condition. After about 30 minutes warm up, the 5V regulator input voltage
is about 9.2V. More details of the mod are given in the pictures below.
BTW, now I can hear the close-in LO spurs (which is normal) that I couldn't
hear before because they were below the noise.