Philips GM2317 signal generator (1954)

The GM2317 is an LF generator introduced by Philips around 1953. The model has been sold for more than 11 years and it has gone through a number of improvements during that period. Although my specimen is from 1954, it already has a number of changes that are not documented in the 1953 service manual.

The GM2317 uses the Wien-bridge oscillator principle. The valve line-up consists of Rimlock types:

EF40	pentode as pre-amplifier,
EF40	pentode as second amplifier stage,
EL41	power pentode as output stage,
AZ41	rectifier.

It also has four OA70 germanium diodes as a bridge rectifier for the output meter. The three amplifier valves form a 3-stage non-inverting amplifier. The EL41 is configured as a cathode follower, a low-impedance buffer at the output.

Except for the missing handle. the generator was in good condition. It had an inventory label on the front and a glue residue from a label exposing the text “ELEKTR.MEETT.” which is an abbreviation of “Electronic Measurement Technology (Laboratory)”. After checking the capacitors and slowly starting the generator to let the power supply electrolytics reform, it just worked. I found that the coupling capacitor from the first to the second stage was leaky, but I did not replace it at the time.

I have used it for years until I collected more AF generators to use. The GM2317 wasn't always stable as it should be. My notes say that the amplitude tended to vary and sometimes suddenly the signal was distorted. This may have just been the automatic gain control from the incandescent lamps swinging round the set point. Somewhere around 2015 the output meter stopped working and the tuning knob didn't work any more. So I put it on a shelf to fix it sometime later.

Inspection

In early 2026, after fixing the GM2885 RF generator, I decided to take up the GM2317. I opened the cabinet and dusted the inside. At further inspection, to my surprise the GM2317 contained four incandescent lamps instead of two. The schematic in the 1953 manual shows two incandescent lamps used as the cathode resistor of the first stage. They are part of the negative feedback loop and their power dependent resistance stabilises the amplification and limits distortion. I traced the circuits and drew a schematic. It turned out that the four incandescent lamps are on a different place in the circuit. They are in series with the positive feedback circuit, regulating loop amplification. The same circuit is found in the later GM2317/01 through /04 versions, but those have other types of valves and a number of small additional refinements. So what version did I have?

I checked manufacturing date codes on the capacitors and other components and found dates in 1954.

So was this a GM2317, but different? There was an tabel on the front that said “Et2317/01” but this was merely an inventory label of the lab that had once owned this instrument. The type and serial number label at the back simply read “GM2317” without revision number. It couldn't be a GM2317/01, that has Noval valves and is from 1959. I did some research on Radiomuseum.org, Dutch Forum on Old Radios and the NVHR site and found that while no improved version of the GM2317 with Rimlock valves was documented, there were more GM2317 specimens from the mid-fifties having the same circuitry.

Repair

I started the obvious repairs. The driving cord had snapped. I removed the front to get to the tuning mechanism. I made a cord of 440mm according to the manual. Giving it 3 turns round the tuning know shaft, it fitted.

One of the resistors in the output divider was burned. Its resistance was too high. I replaced it by 2 precision resistors in parallel, to get the right value of 680Ω. When I measured its value and the other resistors, I noticed that the output meter moved only with the ohmmeter leads in one orientation, not when I swapped the leads. So I checked the four rectifier diodes of the output meter and indeed: one of the OA70 germanium diodes was interrupted. I replaced the defective diode and tested the meter. Was OK now.

Then I unsoldered the leads to the power supply electrolytic capacitors to reform and test them. This is necessary because there is a voltage divider (R49+R50) parallel to the + voltage to bias the heaters of the valves. Philips did this to avoid cathode-heater breakdown. The voltage divider varies from version to version but it is there in all GM2317 versions. Anyway, its presence in the circuit would disturb my leakage current measurements so I temporarily removed it. There was another difference from the original schematic: the series resistor R34 was 2k7, not 4k7. This would increase the power supply voltage for the pre-amplifier valves. It was clear that this was original because the ink dots on the solder joints were still in place. The electrolytic capacitors were good after an hour. I also reformed the output capacitor (a 50+50μF electrolytic like the power supply capacitor) and saw it was good up to 220V  but above that its leakage current increased. This was enough for its function so I left it.

When I tested the generator, it just started to work. When switching frequency range, the signal would disappear, come up abruptly, get distorted and settle to a nice stable sine wave.

Checking the voltages it turned out that the power supply voltage was only 220V instead of 270V. I replaced the AZ41 and got 275 V. This is as specified for the GM2317. The GM2317/01 and later have 300V, but those have a different power transformer and an EZ80 rectifier valve.

Now the cathode voltage of the EL41 was a bit low. I saw that the getter mirror on the second EF40 was very thin, possibly indicating wear. I replaced it, getting a slightly lower anode voltage. There was a +1V on its grid so it drew too much current (3&nbps;mA instead of 2). The anode is DC-coupled to the EL41 so it also offsets the cathode voltage of the output valve. The cause was the leaky paper capacitor C11 between first and second stage. I replaced it by a Philips mylar capacitor.

When turning the output level control, it cracked. It was dirty inside. I unmounted it, sprayed a drop of mild contact cleaner inside and turned it a few times. Reattached it and the cracking was gone. I also cleaned the contacts of the range control resistor carrousel. There was some brownish sticky goo on them, exactly like the pollution on the contacts of the GM2885 oscillator I finished two weeks before. I suspect this is dried-out contact grease Philips originally had applied. I cleaned then using isopropyl alcohol first, then ammonia and finally applying fresh contact grease to the contacts. They should be fine for the next 50 years.

There was a last cosmetic aspect to cure: the rim of the frequency scale had small rust specks. The scale itself was looking good. The rust would gradually become worse. So to stop that, I disassembled the scale and its rim, sanded the rim, treated it with metal primer and then creamy white paint. I reattached the rim to the scale and then the scale to the front of the chassis and the indicator to the tuning capacitor.

Then I resoldered the wires to the BNC-connectors and reattached the front. After cleaning, I mounted the knobs and reassembled the cabinet. This was looking good!

To make the device complete again, I attached a replacement handle on top of the cabinet.