The work required to replace the 3-1000z with a GS35b consisted of:
- Remove the 3-1000z socket and filament transformer
- install gs35b socket, install gs35b filament transformer
- fashion clamps for plate, filaments (out of copper coated plumbing hangers and screws -- home depot)
- plug holes in sheet metal from anode cooling riser
- install W4ZT's bias circuit
- folks have recommended NOT to feed bias to the gs35b through the filament center tap, so I wound a new choke and fed the bias right to the cathode
- added a 'safety choke' across the output in the event of blocking cap failure

At this point, it became possible to 'debug' the changed amp; since I never saw it running in the first place, I thought it prudent to test every part of it. I first burned in the tube with filament voltage only for about 48 hours. The power supply originally had a variac on the plate transformer primary, and the manual warns against exceeding a max DCV of 3600v. I calculated that at full 240v primary (which I have), the DCV would be 4100v, exceeding the rated voltage of the capacitor stack. Putting a small variac on the primary, I verified that the power supply was seemingly good. After checking the B+ input line on the RF deck for shorts with an ohm meter, slowly applied HV to the amplifier circuit. No shorts, however the grid current meter immediately started to show weird stuff - like negative grid current. I traced this down to a blown grid meter shunt -- it was a 1.0 ohm 3 watt resistor gone bad. The replacement with a metal-encased precision power resistor is sort of fitting with the rest of the project. Re-application of the HV, and meters were ok. Next was to set the BIAS at operating HV. When I brought the HV up to 3400 or so, heard a large POP out of the power supply; disconnected it, replaced/rebuilt the capacitor bank, because it HAD to be that. Nope. Replaced the rectifier string, and THEN the PS was happy. While I was at it, increased the number of caps in series to 10, bringing the max V to 4500v. Refused the 220v primary with 5 amp fuses (normally 15) for testing.

To adjust the tube idle bias, I brought the HV to expected operating V of 3600v, then with no signal, keyed the amp; the point is to set 100ma of no-signal bias. Was able to do that. Then, removed HV (disconnected primary from the plate xfrmr), and used an MFJ antenna analyzer to adjust the tuned input networks for each band (this required keying the amp). Re-applied HV, and got ready to try to amplify some RF. Attached a dummy load, built a small keying isolation circuit (the key line is about 12v when floating). Hooked it up to the rig, and keyed with about 5 watts. Quickly tuned for resonance, and got about 20 watts out! Cool! Started to increase the drive, mindful of my 6amp variac on the plate transformer primary. One of the HV primary fuses vaporized, and the variac made a groaning noise. After diagnosis, seems that I blew a pass transistor in the bias circuit. Tony W4ZT suggested 0.1 or 0.2 uf bypass across the bias circuit and cutoff resistor to eliminate any low (non-RF) frequency instability, and to make sure that no RF gets into the bias circuit. A couple of days later and I'd replaced the transistor, bypassed the bias circuits with 0.1uf caps, as well as procured a 10KVA variac via ebay, replacing the anemic 6A one. Got everything assembled again -- set the no-load HV to 3700v, set the bias to 100ma, and began testing again -- this time, no undue pops, snaps... I was able to drive the amp to over 800 watts! At that point I was pretty excited, but also getting concerned about heat -- the original blower for the 3-1000z didn't seem to push much more air than say an SB-220 does. Everything I've heard about the GS-35b tubes says they like good ventilation. I had a Dayton 156cfm blower, but it was physically bigger than the original blower -- however, if I disassembled the amp again, I could cut a hole in the RF-deck's bottom plate, mount the fan underneath the RF deck, patching the holes in the chassis to maintain pressurization. It was the right thing to do -- LOTS of air moves through the amp now, but the blower makes quite a racket. Someday I'll put a temperature-controlled relay and slow the fan down with a resistor. Until then, I have some well insulated headphones.

Once the temperature issue was worked out, I could test in earnest. I tested on 80 through 10 meters, and found that it's relatively easy to tune and load. The tube also glows under keydown with a nice blue tint. Typical operating conditions for this amplifier are:
20 meters - 99 watts drive - 3600v no load, 2900v at 900ma plate current - 1490 watts output into a dummy load (measured using a recently calibrated LP-100 digital watt meter). Subtracting the 99 watts of drive, I calculate about a 53% efficiency - there's some room for improvement.

On the air tests with CW and SSB were done, comparing the quality and bandwidth of the received signal under no-amp and amp conditions, and it's seemingly clean.

Further work will be to install a HV start delay (90 seconds minimum) circuit, maybe slow down the blower to quiet it during receive periods. The knob mismatch and rust spots on the front panel seem more fitting after using the amp a bit. I doubt those will change.

Sorry I don't have a better picture of the blue glow, but I'm a chicken about having exposed HV.