GS-23b 23cm Amplifier
Click on callsign for e-mail

GS23B / 4CX1600U 1KW PA for 23cm

I needed an amplifier for EME use and was searching for TH327 or YL1050 at a decent cost without any luck. I had a couple of GS23Bís and decided to try them on 1296MHz trying to reach 1KW output. The 4CX1600U data sheet from Svetlana specifies the tube as good for 1300MHz with reduced ratings.

Click for Zoom

The amplifier described here is a good alternative to YL1050 and has very similar performance. I strongly recommend modifying the tube for water-cooling. This will reduce or eliminate thermal drift which is a common problem in amplifiers at this frequency. This amplifier operates with grounded screen grid so the cathode is -500V in relation to ground. This means that you must use a floating plate power supply. G3SEKís tetrode board can be configured for this type of operation. Please read his web page at:

Test Results for GS-23B Tubes

The first five tubes to be tested yielded the following results:

    Tube #1: Good tube. Water cooled, No thermal drift
    Tube #2: Marginal tube. Fan cooled, <15% thermal drift
    Tube #3: Marginal tube. Water cooled, some thermal drift
    Tube #4: Great tube! Fan cooled, No thermal drift, very stable
    Tube #5: Unusable tube! Fan cooled....

Measurement Tube #1Tube #2 Tube #3Tube #4 Tube #5
Plate voltage3050V3050V3050V3050V3000V
Plate Current1020mA900mA1000mA900mA1000mA
Plate input3100W2745W3050W2745W3000W
Screen voltage500V500V500V500V500V
Screen Current23mA24mA25mA22mA20mA
Screen dissipation (Max 12W)11.5W12W12.5W10W11W
Grid voltage-38V-37V-40V-32V-38V
Grid current27mA24mA25mA30mA32mA
Grid dissipation (Max 1.5W)1.0W0.88W1.0W1.14W0.96W
Heater voltage5.7VAC5.7VAC5.7VAC5.7VAC5.7VAC
Drive power80W85W90W50W100W
Plate dissipation
(Max 1650W air-cooled)

NOTE:  Increasing the screen voltage to get higher plate current, thereby lowering plate load impedance, results in a better match of the tube to the cavity in this design. Raising the screen voltage to 525VDC with tube #1 installed in the PA, holding plate voltage at value shown, power output increased to 1350W @ 39% efficiency.

Differences Amongst Tubes

The difference in gain and efficiency between tubes #1 and #2 existed before I converted #1 for water-cooling. Water-cooling did eliminate the thermal drift for tube #1 so I would recommend it to anyone pushing the limit.

Thermal drift is measured as reduced output by internal changes in the tube causing de-tuning of the cavity over 1 minute key-down.

Tube #2 reached max screen dissipation at 850W output so that is where I stopped.

Tube #3 did not perform as well as tube #1. The following was observed:

    Thermal drift. A sharp de-tuning of the cavity after about 10 seconds key-down. It will recover to normal after a few seconds idle.
    Date code = 1985
    Lower efficiency.
    Lower internal capacitance. Had to increase the tuning capacitance for resonance.
    Higher grid current.
    Lower gain.
    Max stable output without above problem is 600W @ screen dissipation 4.5W

The difference between tubes is to be expected in 1296MHz PAs since even a very small mechanical change will de-tune a cavity at this frequency. Anyone using GI7B, GS9B, 2C39, or 7289 triodes, all rated for 3000 MHz, experience the same problem. YL1050 and TH327 PAs also exhibit the same variations at 23cm and it is considered normal. Everyone should be aware of this fact before they get started. In fact, after testing over 20 tubes, It appears that only one of somewhere from 5 to 7 are good for 23cm, so testing GS-23b tubes for 23cm suitability is a necessity. Even so, those tubes which do not work at 1.3GHz (see discussion below) should operate at 432MHz, no problem!

Why Tubes Vary in 23cm Performance

The problem is that radiated heat from the plate plus the dissipated power in the screen will mechanically change the screen, resulting in a change in plate/screen capacitance. The ONLY way to minimize this problem is to have GOOD (preferably water) cooling to eliminate the heating from the plate to the screen. It all has to do with the mechanical design and assembly of the tube. If the grid cage and the screen cage windows are perfectly lined up the resulting screen current will be very low and stability improved. A tube that shows high grid and screen current doesn't have the grid & screen cage windows perfectly lined up and some accelerated electrons hit the screen instead of passing through the window.

Many tubes have the cathode in the bottom and then a flat grid above it and then a flat screen above the grid and so on. The GS23B has all the elements inside the tube in a coaxial configuration. The cathode on GS23B is a cylinder in the middle and then the grid is another cylinder outside the cathode and so on. This makes for a very compact tube with large element surfaces and low inductance. This is a very common design in large tetrodes.

A tube that responds well to drive and has normal grid and screen current but low efficiency probably has a small variation in the distance between the elements around the circle. Any variation in this distance will accelerate the electrons in a non-symmetrical fashion around the circle and they will arrive out of phase. There is nothing we can do about that from outside the tube. This behavior is frequency dependent since the phase error gets smaller at lower frequency. This is why a tube like this would work well at 70cm but not at 23cm.

Testing Conditions

Overall, maximum plate dissipation is a function of cooling, so it should not be a problem to exceed it as long as the temperature is kept low.

         Tube max data:
         Max Screen dissipation  =  12W
         Max Control dissipation  =  1.5W
         Max Plate current  =  1.2Amps
         Max Cathode current  =  1.8Amps

None of the above critical parameters were exceded. These parameters are based on continuous carrier so it might be possible to get even more output in intermittent ham use, but I am not going to test it.

I have stress-tested the amp with tube #1 at 1200W output and everything looks good. I increased the drive-power to 100W and the plate current went from 1020mA to 1120mA. The efficiency increased. I got 200W more output for 300W more plate input. This is to be expected since I am getting further away from the idle current of 200mA. I set my CW-keyer to 20 seconds of code and 60 seconds pause with 100W drive. I operated the amp for more than 96 hours like this with no problems occurring. I am exceeding the screen dissipation by 10% under these conditions, but I expect that the ratings on this tube are very conservative.

PA Construction - Materials

The PA shown here and used for testing was manufactured using brass and bronze. It did not perform as expected until it was silver-plated. I strongly recommend, therefore, avoiding these materials to make parts for this amplifier unless you are going to silver plate it. To check further, I manufactured two more copies of this PA, one of aluminum and another of unplated brass and bronze. The output from both of these amps as initially constructed was 300W lower output than the prototype at the same plate input power. After silver-plating the brass-bronze amp, power increased to be identical to the prototype output. Clearly, a PA made using brass-bronze requires silver plating for optimum performance. Bronze bushing material was chosen for the anode cavity and grid sleeve because it has dimensions which simplify fabrication of these parts. It appears that only if made of copper would this PA not require silver plating. Aluminum should be avoided!


Click for Zoom

The aluminum PA; looks great, works poorly - wasted time!

PA Construction

Water Cooled GS-23b

Preparing the GS-23b to be water cooled.


Click for Zoom

Completed amp with water cooled tube.

23cm GS-23b PA

23cM GS-23b PA parts & construction details.

Water Cooled GS-23b

Description & pictures of a water cooling system.

home back home
go to the top
Best Viewed: MSIE 4.0 & higher, or Netscape 3.02 & higher, and screen set to 800x600 pixels
Last Updated: 20 April 2002
Feedback: Paul S. Goble, III, ND2X
Copyright © 2001 & 2002 by ND2X and KD5FZX, all rights reserved