bernerus / tetrodeboards

The Tetrode Boards by Ian White, GM3SEK. Redesigned with KiCad.

TetrodeBoards

The Tetrode Boards by Ian White, GM3SEK. Redesigned with KiCad.

In 1998 Ian White, G3SEK, nowadays GM3SEK designed control and protection circuits for RF power amplifiers that use tube tetrodes. These boards were sold for many years by the designer but were retired around 2017 and are no longer available.

The manuals are still available however, and that documantation is excellent.

After my retirement as a system developer I found time to return to the VHF DXing that I used to do as a student in the late 1970's. By then one of the things I built was a VHF power amplifier, described by Fred Merry, W2GN, which was used even for some EME contacts until around 1987 when a storm took my antenna. Since then, the amplifier was only used occasionally.

Now, 37 years later, the ham world has changed quite a lot. Nowadays computers make the waveforms and almost every ham has a spectrum analyzer built into their radio. My signals from the old PA didn't look very good there.

Actually it is probaly at miracle that the PA and its power supply worked at all now. There were lots of electrolytic capacitors in the power supply that were surplus already in the mid 1970's, and when I got reports on hum on my signal, I checked these caps. I had only about 8 uF left and the G1 and G2 voltage varied too much to be happy with.

That's when I found out that these tetrode boards could help, but since the boards were no longer sold, I decided to try to re-engineer them using KiCad. This time using methods available today. Thus this design uses SMD mounted components and professionally produced boards. I have some boards left, so anyone who likes to try building on these boards can have them for the expense of postage. Beware though that they have issues. Check the issues tab at GitHub first.

A crucial component in Ian's design is the 748 op amp. These are no longer manufactured, but they are available at eBay. Luckily for me, I found a few 748 IC:s leftover from the happy '70s in my old junkbox.

Christer Bernérus, SM6FBQ. JO67BQ 2024-04-04.

tetrodeboards's Issues

Warmup time and R20

The warmup time with the default seems to be 2 minutes.
That's a long time and lot all tubes need to have that long time to warm up.
To shorten the warmup time, decrease the value of R20.

G2-Control board: U1 connected backwards

This is by far the worst issue. The LM317 voltage stabilizer U1 was turned the wrong way on the schematic compared to Ian's original. Usually signals go from left to right in a schematic, and that's what I assumed, however in this case VIN is to the right
and VOUT to the left.
The electrical rules checker of course warned about this but I didn't run it. :-(

The consequence is that this error has of course migrated to the physical board, and U1 is here a SMD component so the fix requires surgery that connect the pads of R1 and R2 being closest to each other to be isolated from U1 pin 3 and instead
be connected ti U1 pin 2. Likewise the 30V rail that starts at U2 pin 7 snd that makes a clockwise loop around R4 and must be isolated from U1 pin 2 and instead be connected to pin 3.

I made a cut in the 30V rail track about 3 mm from U1 pin 2, and scraped the copper clean on about 1mm along the other side of the cut, i.e. the side still connected to U2 pin 7. Tinned the free copper and soldered a tinned piece of cutaway resistor lead between U1 pin 3 and the tinned copper.

On the other side of U1 I cut track leading to pin 3 immediately at the edge of U1 and soldered a cutaway resistor lead between the pad of R1 that is connected to R2, to the rear side of U1 pin 2.

See this picture.

Q6 connected backwards.

The MOSFET Q6 that controls the T/R Relay was originally turned upside down in the schematic.
This caused the Drain and Source pins to be interchanged on the original PCB.
Due to the pinout of the IRF9530 this cannot be fixed by mounting the transistor reversed.

Too small holes for the BY228 diodes on the G1-ALC board.

To save space I replaced the double 1N4007 diodes on Ian's boards with
the more sturdy avalanche diodes BY228 which is good for 1200V and 3A.

However these diodes are so sturdy that they come
with 1.35 mm leads. I knew this but somehow in the process, the change in hole size got lost,
so to be able to fit the BY228 diodes, I first cut the leads to the approximate needed length.
Then mounted each end in a drill and while slowly turning the diode with the drill
I used a Dremel round sanding band at medium speed to sand down the other
end lead diameter to about 1mm (or until fitting the hole) for a length of approximately 3 mm.
This reveals its copper core but that only matters until it is soldered in place.

G2 trip current too high

I wanted the G2 trip current to be 30 mA. That could not be achieved, It turned out that the value of R16 (1k5) was too high and caused the optocoupler to only deliver a few milliamps at 30 mA IG2. Changing R15 from 82R to 120R did not help.
I put another 1k5 in parallell which made it possible to set the 30 mA limit somewhere in the middle range of RV2. In the next version I will change R16 to 820R.

Missing traces on the G1-ALC board

There are two missing traces on the G1-ALC board.
This was due to me not running the electrical design checker on the schematic and a grid mismatch.

The negative ends of C102 and R102 are not connected. They should both be connected to the CATHODE rail.

As for the capacitor, I just bent the negative lead towards the cathode of D113 and soldered it there on the rear side of the board.

As for R102 I bent the unconnected lead, the one closest to J102, onto the anode end of D103 snd soldered it there, also on the rear side.

See also:

G1-ALC board. Heat sink too close to capacitor

The heat sink that is cooling Q101 is too close to capacitor C101.
There is room enough for the heat sink itself but not really for the screw
that bonds to the transistor.
This causes the heatsink to lean over to C103 which doesn't look nice,
and the only way to mount the heatsink is to first screw Q101 together with
the heat sink and then solder the transistor, or alternatively to mount
C103 after mounting the heat sink.

D6 can reside on the G2-Control board

The zener diode D6 was excluded from the G2-control board because it was drawn in the breakout box together with Q2 and R12.
However both nets it is connecting to are also available on the board, thus it can be placed there somewhere, simplifying the
off-board circuitry.

G2-Control board: Awkwardly turned Q1

Q1 on the G2-Control board is turned such that the heat sink is facing to the inward of the board. Then it comes very close to
the capacitor C2. It should be turned the other way round so that the heat sink could be soldered directly to the grounded edge of the board, provided that Q1 is mounted isolated from the heat sink.

R26 must connect to the source of Q6.

In the original schematic, I connected the hot end of R26 to +12V. This does not work when using 24V feed to the relays.
To make it work, the hot end of R26 must be connected directly to the source of Q6.

Connectors turned the wrong way

The silkscreen graphics in many places show the curly end turned outwards.
Don't mount them like this, turn the curly end inwards, otherwise the screws
will be turned inwards making it harder to measure voltages on the screws.

The silkscreen text showing what signal is connected to each pin still applies,
but pin numbering does not.

This issue applies to both boards.

The attached picture shows a screw connector inserted in a socket having the curls inwards.

R123 Missing

R123 is missing from the initial schematic. Consequently also on the initial PCB.

J107 and J108 should change places on the G1-ALC Board

J108 is the interconnect receptacle for the G2-controö board. It would look better
if J108 changed place with J107 because then the interconnect would line up
with the interconnect on the G2-Control board,

Q4 turned backwards

The silkscreen print turns Q4 backwards. This was actually caused by selecting wrong schematic symbol which numbered the pins in the wrong direction for the 2N4403. The workaround for the initial board is just to mount Q4 the other way round, i.e. against
what the silkscreen print show.

Wrong value for R128

The value for R128 should be 22k, not 1k5 as it is in the initial release.