A) These tests were performed while the cabinet had not yet been renovated.

However, we can test the radio, without the cabinet.
I used a mains power supply, the door with the antenna and the HS.
Here is the whole of the installation of test:




Lamps have been plugged in the following manner:




Power supply has been connected on the connector with clips:




(-C is set to -7V, +45 is connected to +40 V, heating is -4V)

The loop antenna has been connected to the inputs of radio with crocodile clips.




And the HS is connected to the 2 wires outputs (note the direction!)





And here are the results: it does work well! The operation is very instable: the settings are difficult, scratching, wheezing ...
I also tried to swap lamps, to see what happened, and, in general, it's worse, except for a few combinations:
- You can swap the IF and the Detector (-B and -D)
- You can swap the RF and AF # 1 (-A and -D)
- You can swap the IF and the AF # 1, with a stronger level (-A and -D)
- We can swap LO and AF # 1, settling -C at -5V (-A and -B)

The optimum setting is:
- Battery = 1
- Volume = 4 to 6 (except for the last 2 permutations: 1 to 2)
- "France-Bleu" (864kHz): Selector 1 = 25 to 30, Selector 2 = 30.
We can untune the Selector 1 to decrease volume; at perfect tuning, unfortunately, there is reaction (too much gain or regenerative operation between the coils of the LO and the antenna?)

Conclusion: first test is good, I am going to build some another lampes to see if I can improve performances.
It is also possible that the reactions were due to the characteristics of the lamps, higher than original ones.

B) Now that the operation is validated, I can do some checking ...

Firstly, given the low value of the IF (42kHz) the frequency image is not far on the dial, in other words, we must be able to receive France Bleu, for 2 different settings o Selector 2.
Indeed, the 2° setting is for Selector 2 = 24.
By adjusting Selector 2 to 27, we tune on the incident frequency, in other words, there is a violent reaction with antenna!

C) adjustment, to improve the sound quality.

By adjusting the following manner:
-Selector 1 = 25 (instead of 30)
-Selector 2 = 30
-Battery = 1.5
-Volume = 6
This gives a sound not so bad.

D) intensities.

The measures were:
- IA (heating) = 0.23 A
- IB45 (detection) = 1.3 mA
- IB90 (+90 General) = 12.5 mA

E) testing the 5 lamps mode.
I removed the AF # 2, and putting in place of the AF # 1 (-E). The HS is connected to the sockets on the front panel.
The result is not very good: clashes, level really too low.
I get the impression that because the rheostat "Battery" is set at minimum, its resistance in series with all filaments (5 ohm) helps the reaction to occur ...

Pondering a little, I thought that the problems of reception perhaps due to the fact that the ersatz lamps should have too much gain.
So I tried to replace the lamps ... I had already done during the first test, but you never know ...

For example, why is the best reception corresponds to rheostat "battery" setting at the minimum? I.e. for a heating voltage very low (about 2.2 V)?
So I plugged the lamp 99-A in LO slot again, test which gave nothing.
By turning clockwise the rheostats, the reception emerged, but very bad (reaction).

And there's the chance intervenes: whereas until now, every time I swapped lamps, I put the power supply off, this time, I did not do it!

I removed the AF # 1 for changing it and then, surprise: Lamp is absent, but I heat receiving, low, but clear, with no reaction !
In other words, the gain is so high that even with the AF#1 lamp is absent, the signal can pass (capacity leakage or induction leakage between 2 AF transformers ?).

So I decided to look at the signals on the lamps grids.
Here's what I seen (on 864kHz receipt, on day, loop antenna in the basement, Battery = 3, Volume = 5)

The lamps were:
99-D 99-A 99-D 99-D 99-D 99-E (except that for each measure corresponding lamp is removed)

A) grid of the IF # 2




We have already 600mV peak after the reflex-mixer stage!

B) detector grid




Near 3V peak, more than enough to properly detect. Importantly, with this level, a single lamp might be sufficient, and there is still 3!

C) AF # 1 grid




There are some 20V peaks !
It's too much, this lamp must be completely saturated.

D) grid of the final AF # 2




peaks over 200V !
Needless to say that the entire AF stage is saturated and this explains the reaction: the power supply can not work well.
***********************************

With these results, there is simple test to do, if the gain is too strong, it is sufficient to decrease it .... by removing a lamp!

Just replace the AF # 1 lamp with a 10nF condenser 10nF between its grid and anode, so, no amplification for this stage.
The result is quite satisfactory: the sound is good, there is no reaction, we can settle the 2 rheostats with flexibility and progressiveness, in short, the set works better.

It should be noted that in order to achieve this function, you have to put the 99-A in LO stage and no other lamp is suitable for this stage! (nor the 99-E)
That means that I must study another lamp specifically for this function (in fact, the 99-A is wired with a resistance between grid and anode 2, but not the 99-E)

However, the 99-B may replace the 99-D at the three remaining slots (RF, IF#2, Det)

Similarly, the 99-E is better in final AF#2 than others that work anyway (but with more distortion)

Here is the radio set with lamps:

The idea is to try to find a substitute that could be suitable for all slots, including LO stage.
I tried a 3D6 according to the following drawing:

  Substitute with 3D6

This substitute is suitable everywhere except in LO stage! Note that the gain is lower than 99-D, for example, which means that if I put 6 ersatz like this one on the radio, I would resolve perhaps the problem of gain.
But it means that I need to build at least 4 subsitutes !

Why is the 99-A only suitable as oscillator?
That needs further investigation ...
So I installed the oscilloscope on coils of the oscillator (the hot wire connected to the varcap), and I observed the signal.

First, when the reception is blocked, it is not because LO is blocked, but at this moment, the signal is polluted by a 42 kHz modulation, in other words, IF stage reacts !
This reaction does not depend on the LO lamp, but on this at RF-IF#1, and does it makes sense: This lamp has 2 functions: RF amplifier and IF amplifier.
In short, the receipt is correct when this lamp hasn't too much gain.
So I installed a 99-F at that position, and with this condition, I can put into any lamp at LO stage (99-D 99-A 99-F)
However, as soon as it's a 99-D at RF-IF1, everything goes wrong. This version has too much gain. I think I will give it up.

So here is the used lamps set:
99-F 99-F 99-A 99-B 10nF cond. 99-E
For best reception, -C voltage has been set at -10V.
This is logical, versions -F use an 3D6 which works perfectly with this bias, which is not at all the case for 2SH27L used in the version -D.

I even put a 99-B at AF#1 instead of the cond, and it remains some saturation, but not as much. I think that if we put in versions -F at detection and AF#1, it should work.

Well, I still have to make two other versions -F substitutes.

Here are the 4 manufactured ersatz "99-":




With these 4 lamps installed at RF, LO, IF and Det, we have good receipt, although the heating voltage decreases: indeed, the power supply can't give more intensity, it's time to build a stronger power supply !

Moreover, we must build this power supply which will be dedicated to the job and that I can permanently installed in the compartiement batteries.
Here the set with its lamps; at left side the four "99-F" and right side the "99-A" and "99-E":

During testing the power unit, cracks have sometimes occurred.
I first attributed them to a poor reception, but as testing continued, they increased in intensity.
Here is a description of the symptoms:
- The receipt is now impossible without interferences covering completely sound.
- When you unplug detector lamp, you can hear in HS a noise loud enough , which seems random, with bass tonality.
- When you remove the lamp BF#1, the noise disappears.

I first thought that AF transformer T10 between the 2 lamps was out.
- When you bypass the primary winding of the transformer
- When you measure the voltage drop at the terminals of the 2.2 k of the power supply +45, detector lamp removed, you get: 0.27 V.

Conclusion: the capacitor C2 leaks. You can even evaluate his leakage, the open circuit voltage is 83V. The current is: 0.27 / 2200 = 123 µA. So the leakage resistance of the cond. is: 83/123µA = 676kohm!
We must change this capacitor, because if it's been in short-circuit, there is danger for the primary transformer T10, which will be directly connected between +45 and the ground (however, the internal resistance of power supply would limit the current, but do not risk is better ...)

So, I have to disassemble Catacomb again!

Well, This Catacomb is removed, shield removed too. I decide to change C2 and C4, which are side by side. I used 150pF/250V. I had to lengthen one leg to connect the cond on each side.

Here capacitors wired:







Reassembling can start ... Duration of the repair: approximately 1:30!


Tests: the failure has been repaired, the noise no longer manifested if we remove the detector lamp. A voltage measure on 2.2k shows there is no current output when the lamp is absent, proof that the new cond does not leaking.
The reception is again correct, its level is stronger than with the original condensator.