Philips 543A problem detector stage

[frsimen]

None of those tests revealed anything as being obviously wrong, other than the sensitivity.

I think the next thing to do is to establish where the gain is being lost. For this test, connect your oscilloscope across C39 (Trader). Use AC coupling for this measurement. The idea is to apply a signal to various points through the IF stages to check the gain of each of the stages. This way, a problem in the audio amplifier will not upset the results.

Start by applying a signal to the anode of the diode of V3. Increase the output until you see a reasonable level of audio at showing on the oscilloscope. I expect you will need to set the output level to near maximum.

Transfer the signal generator to the grid of V3 and adjust the output level to give the same audio level as you when feeding the diode. The signal needed for this should be substantially lower in level, perhaps 40dB less.

Transfer the signal generator to the grid of V2 (keep to the 468kHz) and adjust for the same output level. Note the signal level here. It should be at least 20dB lower than at the grid of V3.

At each stage, check the appearance of the sine wave that you see on the oscilloscope. It should look good for all the tests, if it doesn't, make a note as that may give a clue as to where the trouble is.

A test of the AGC action might be worthwhile here too. Try increasing the applied signal at V2 in 10dB steps and note how the output changes across C39. In particular note whether the distortion increases or decreases at the higher output levels.

Couple the signal generator via the 0.1uF capacitor as before.

Hopefully these tests will reveal the area of trouble.

Paula

[Christoffrad]

Thanks Paula
I think you mean 470kHz not 468
The distortion has always related to low detected audio level but I will take note when I do your tests tomorrow (wed)

Chris

[PJL]

What sensitivity do we believe is normal?
What happens to the distortion when you operate the tone control?
As the problem only seems to occur with broadcast input, does your signal generator support external modulation? If it does maybe we can try a music source at various modulation levels.

[Christoffrad]

Here are the results of the tests suggested by Paula.... Signal generator set for 470kHz 30% mod with 400Hz tone, measurement of detected audio across C76 Philips (= C39 Trader) with scope:
1) Signal applied at V3 anode - 2V emf for 280mV p-p audio at C76 (C39)
2) Signal applied at V3 grid - 22mV emf for 300mV p-p audio
V3 stage therefore providing 20.8dB gain
3) Signal applied at V2 grid - 4.6mV emf for 300mV p-p audio
V2 stage therefore providing 13.6dB gain

4) AGC test generator feeding V2 grid - measurements
200uV emf gives 2mV audio
700uV gives 30mV
2mV gives 100mV
5mV gives 400mV
10mV gives 750mV
20mV gives 1.2V
50mV gives 2V
100mV gives 2.7V
200mV gives 3.3V

There is no noticeable distortion (observed on the scope) at any input level other than very slight at 200uV emf input.
Audio is not measurable at levels below 200uV input.
Audio distortion is a bit more noticeable at input levels below 1mV when mod is increased to 50% and 2mV for 80% mod

The audio distortion on broadcast signals is of an intermodulation nature where higher audio frequencies seem to be modulated by bass ones. Broadcast audio is always low. Operating the tone control does not change the character of the distortion other than to filter it as you would expect.
I suspect the distortion is a secondary effect from not hitting the detector hard enough.

Yes my signal generator does provide for external modulation.

It has been said by others on this thread that a sensitivity of 10 to 20uV should be possible.

C

[G6Tanuki]

I wonder if the distortion could be connected with stray capacity-coupling within the valve? (or associated circuitry).

There's high signal voltage on the detector and if any of this gets coupled to the grid of the triode you can find the grid/cathode forming a 'phantom diode' and demodulating the signal as well as the demodulation in the intended diode! The recovered audio from this phantom diode will have an ambiguous phase relationship with the genuine audio, so when the two are combined it's anyone's guess what the resultant audio will sound like.

Some valves - such as the 6AT6 - have internal construction which makes this effect almost inevitable! Putting a low-value [100pF] capacitor between the triode grid and the cathode can help prevent the development of phantom demodulation.

[Radio1950]

Injecting a test signal into a BC receiver mixer is not an exact process, especially with the brute force method that is normally used, as it is a great disturbance to the circuit.

Assuming a 50 ohm sig gen and say a 0.01 uF coupling capacitor, this is loading a relatively high input impedance of the mixer with about 85 ohms at 470 KHz.

So a lot of disturbance is taking place to a stage with input signals and standing bias, and with derived AGC for higher signal levels.

It will difficult to actually state a typical mixer input level due to different mixer circuitry from receiver to receiver.

A test with another identical receiver and test setup is required for absolute comparison.

In manufacturer's alignment and test procedures, they will normally say "inject a test signal and align the IF or receiver", not specifying a level, and implying that the repairer will adjust the sig gen level until a "useable" audio output is derived with almost no AGC, or to get a prescribed audio level with AGC disabled.

So what happens normally in practice, is that repairers inject a relatively high level, and then reduce it to that suitable for alignment, and reducing it further as each stage is IF and RF peaked.

The test result for some broadcast receivers that I have worked on is that 20-30 uV should be achievable for 6-10 dB S/N, but at the antenna input.

This is the test figure which should be aimed for primarily, not an absolute mixer sensitivity.

Nevertheless, I have found that this equates to a "mixer sensitivity" of approx 50-200 uV for 6-10 dB S/N, when injected at the mixer grid for some BC receivers.

The difference is due to the brute force injection disturbance of the circuit.

Communication receivers on the other hand can sometimes have a more elegant test injection, as the circuit is designed to allow it, and the mixer sensitivity can be a defined and specified test.


I suggest that you align the front end, taking sensitivity checks before and after, if you haven't already done so.
You might find something in the process.
It wont matter if you get bad test figures; it is just part of an overall process.
And when you do this, check the peaking of the IFTs with a very slight tweak.

.

[PJL]

This set makes extensive use of ceramic capacitors for Anode and Screen feed decoupling (C23, C38, C40, C41) so they will not have been changed. How about sticking a scope on these capacitors to check they are decoupling properly.

It might also be a good idea to check the anode and screen current by measuring the voltages across the feed resistors (R6, R14, R15, R16).

Both tests can be performed whilst receiving a distorted station.

[Christoffrad]

For comparison here are the results of the same tests carried out on a Ferranti 255 which is of similar vintage and uses the same mixer valve (but uses a EF85 for V3 and tripple diode triode for detector.
Measurements as previously (30% mod at 470kHz unterminated generator voltages in emf):
V3 anode 1.4v emf for 280mV p-p audio (filtered) Compared with 2V for the Philips 543A
V3 grid 10mV emf for 300mV Compared with 22mV for the Philips
V2 grid 500uV for 300mV Compared with 4.6mV for the Philips

Sensitivity for 10dB S/N (approx)
110uV @ 30% mod
70uV @ 50% mod
Sensitivity for 20dB S/N
230uV @ 30% mod
150uV @ 50% mod

I'm still unhappy about the way the secondary of the first IFT tunes. All the other cores tune very sharply but this core can be turned about 1.5 turns between two cliff edges with virtually no dip in audio output between them.

C

[frsimen]

I carried out the same set of tests on my Grundig. The diode tests gave results which are comparable with yours, 2V emf in gives 275mV out with 30% modulation. In my case, there was a lot of noise which made accurate measurements difficult.

The input to the IF amplifier valve, an EF89 in the Grundig, needed 4.7mV emf for 60mV peak to peak output, showing a similar order of gain to what you measured.
The input to the ECH81 needed 32µV for 60mV out.

The AGC test gave the following result:
Input Output
32 µV 60mV
100µV 280mV
320µV 920mV
1mV 2V
3.2mV 3.1V
10mV 4V
32mV 5.2V
Distortion was audible for the remaining measurements
100mV 6.8V
320mV 9.3V

Looking at your results, the problem seems to be in the ECH81 stage or the following IF transformer, as the stage gain is very low. My measured gain is higher than I was expecting for the ECH81 stage. As the IF transformer has already been noted as not tuning as sharply as the others, perhaps a second look there would be fruitful.

Paula

[Christoffrad]

Yes Paula you are getting 22.5dB gain on the second stage vs 20.8 on my Philips
but you are getting 43.3dB on the first stage compared to 13.6 on my Philips

BTW were you feeding directly from an unterminated generator as I was (via a C)?
I suppose unterminated the output impedance of the generator is about 100ohms.

In the days when I designed filters it was all done at 50ohms so it was possible to evaluate the filter in isolation. I wish there was a way to do this with the IFT. Do you know what impedance it expects primary and secondary. Perhaps it would be possible, having removed the IFT, to drive via a potential divider and terminate via another (and simply calculate the divider loss) in order to plot the characteristic.
Or perhaps I should take the IFT out again to examine it and just pull it apart with a bit more force!

Maybe I should first check the decoupling C's suggested by PJL although I thought I'd already done that.

It looks like the AGC is kicking in at the right sort of level of detected audio.

Chris

[frsimen]

Yes, I was using my Marconi TF2016 with the output unterminated, so it would look like a 50 ohm source, I think. I used a 22nF coupling capacitor, but I don't think that will have made much of a difference.

Clearly there is something wrong in that area, checking the decoupling would do no harm but I suspect the problem is in the IF transformer. Quite what is wrong in the IF transformer is a bit of a puzzle still.

The Primary of the IF transformer will have been designed to work with the relatively high Ra of the ECH81 and the secondary to work into the high input impedance of the EBF80.

Paula

[PJL]

We have already tried taking the 10.7 IFT and switch out of circuit.

Measure the resistance from V2 grid to V3 grid, V3 grid to the detector diode and V3 grid to chassis.
Measure the resistance from V2 anode to C34.
Check V2 and V3 cathode to chassis connections.
Measure the screen and anode current and check the decoupling.
You have already tried replacing V2 and V3?

[Mr 1936]

Hi Christoffrad

I have been revisiting my valve IF transformer theory, as it's not something I need every day

The secondary will be 200 pF (including strays), and 570 uH to resonate at 470 kHz. Ignoring the IF amp grid loading (several megohms) the loaded Q will equal the unloaded Q. For a litz wound inductor I would estimate this to be about 100, which is equivalent to a shunt loss of 170 K.

The primary will be 115 pF (including strays), and 1000 uH to resonate at 470 kHz. With the anode loading of 700 K, the loaded Q will be a somewhat less than the unloaded Q. The unloaded Q might be a bit higher than 100, but probably no more than 150, let's say 120, which is equivalent to a shunt loss of 350 K. Add in the effect of the anode and the shunt loss becomes 230 K so the loaded Q becomes 80.

I suspect the unequal primary and secondary values were an attempt by the designer to (partially) equalise the loaded Q's, given that the primary is affected by the mixer.

Several things spring to mind:

The loaded Q's of primary and secondary aren't that different, so the core peaking behaviour during alignment should be similar.

The shunt losses are equivalent to very high resistances, so any dielectric or other losses in the resonating capacitors could have a disastrous effect. Even a ropey valveholder might do the same.

To get decent Q depends on Litz wire. If some of the strands aren't making contact it will degrade the Q of the coil. It's just possible to detect this by DC resistance measurement if you have a good low ohms meter, especially if you have another (hopefully good) IFT to compare. Clearly, if all the strands are OC then there's no DC continuity, but you aren't in that situation. It's almost always at the ends of the wire where faults are found.

If probing with a scope (and allowing for the detuning effect of its probe capacitance) I would expect to see a roughly comparable IF level either side of the first IFT, i.e. at the mixer anode and the IF amplifier grid.

[Radio1950]

Ref your post #142 your sig gen Zo will be unaffected by any (normal) external factors.

Normal to use a sig gen in terminated dBm mode in 50/75 for coms receiver work eg feeding into ant sockets, but use PD unterminated for work on the internal stages of a receiver (because you cannot match it anyway).

My references to expected IF sensitivity in post #166 are in PD; multiply by two for EMF which you have been using.

When working around Hi Z circuits eg mixer grids, ensure your coupling capacitor has a leakage of 25 megohms or more, as the Hi R AGC line, and V2 grid bias, will otherwise be affected via a possible back DC path through your sig gen attenuator.
I would use a 0.01uF 630V poly only, not a ceramic here
The capacitance value is not as important as the leakage.


Have another look at post #147 for "funny" and dry solder joints, if you open the IFT.
Have another look at post #149 to temporarily bypass wiring not used in AM.
This is what I was suggesting in my post #137.
You can do this live if you are very careful; hold an insulated shorting wire across FM IFTs, listening for a change.


If you have another junk valve receiver, you might try another IFT in anode of V2, after tuning it approx to 470KHz. Unfortunately, this may mean reducing the capacity of the internal tuning cap. Use another IFT1 not an IFT2, as the internal construction might be different to match Ra/Rg and Ra/Det.

The unusual tuning "feel" you experience with IFT1 has to be a pointer, as they are normally sharp in tuning, Pri and Sec.
.

[Radio1950]

I have made an academic math error in post #174.
Should not affect things too much, if at all, and I do not want to confuse the issue and the flow of the thread.
More later, or if required.
Apologies.

[Christoffrad]

Today I removed the IFT and made a check of it's tuning as a stand alone RF transformer, feeding the generator to the primary via 220k and connecting the scope set to X10.
(Thanks to Paula for her PM'd suggestions)

Under these conditions the Secondary with the replaced capacitors connected externally now peaks very nicely and sharply although the primary core needs to be fully extended and doesn't reach a peak. perhaps it needs a bit of parallel C to resonante at 470kHz under these particular conditions.
When it was in the set the primary resonated well so probably that can be ignored here. The important thing seems to be that the Secondary that was in question can be made to resonate externally to the radio so this suggests to me that there is something that kills it's Q when replaced into the set.

Addendum:
If I now load the secondary with 180k in accordance with the comment 173 by Mr1936 the secondary still resonates sharply and the primary also resonates sharply now too.

So the interesting comments about how the Q can be killed by stray loading is very interesting and looks like my next area of investigation

Responding to PJL: yes I have changed both V2 and V3 but only with other used valves so there can still be doubt.

Thank you all for your latest ideas and suggestions.

Chris

[frsimen]

I'm a bit puzzled by the behaviour of the primary tuned circuit. There seems to be something slightly odd about the way things are working when tested out of circuit.

I wonder of there is a poor connection in that IF transformer which is giving the varied performance.

There may also be a leakage resistance somewhere which is damping the Q of the secondary winding. When you did the tests where the 10.7MHz IF transformer was disconnected, was the switch still in circuit between the 470kHz IF transformer secondary and the valve?

If it was, another test would be to wire the secondary directly to the EBF80's pin 2, leaving the 10.7MHz transformer's secondary and the switch disconnected. You could usefully bypass the 10.7MHz transformer on the primary too, leaving only the 470kHz IF transformer in circuit.

Paula

[Christoffrad]

Hi Paula, Yes I disconnected the link between the AM and FM coils and wired from the IFT secondary to the wire going to the grid of V3. But the wiring was still connected to one end of the switch (just not depending upon it). But with the benefit of recent comments I think I need to remove any connection and make a clean direct connection to V3 grid (and only V3grid!) from the IFT secondary to eliminate leakage via any other path, as a process of determining where the excess loading is coming from.

C

[PJL]

Yes, leaving the wire to V3 grid connected would have broken that test as we were trying to eliminate any loading from the wiring/switch.

Were you using the scope on the secondary to monitor the primary adjustment? A 10pF probe has an impedance of 33K at 470KHz so that loading alone will mess up the primary when driven from 220K.

[Christoffrad]

No I was monitoring detected and filtered audio