IF-Filters in Early French Superheterodyne Radio Receivers |
Harald Giese
![]() Editor
D Articles: 78
Schem.: 65 Pict.: 536 05.Feb.18 14:01 Count of Thanks: 9 |
Reply
|
You aren't logged in. (Guest)
1
IF-Filters in Early French Superheterodyne Radio Receivers
|
Dietmar Rudolph
![]() Editor
D Articles: 2202
Schem.: 831 Pict.: 453 05.Mar.18 10:40 Count of Thanks: 6 |
Reply
|
You aren't logged in. (Guest)
2
The motivation for performing another series of measurements on IF filters/transformers found in french superheterodyen radios of the late 1920s originated in the fact that the resonance frequency indication obtained using an 'Exciter' were in some cases questionable. Such dubious answers were obtained on two occasions: (i) When trying to measure the resonance frequency of the primary coil of IF-Transformers. In the present context the term 'IF-transformer' will be used for an IF- filter in which only the secondary coil has a parallel capacitor to form a resonant circuit on the intermediate frequency, while the primary has no such capacitor. Examples are the GAMMA and SOLENO filters (see table in post #1) (ii) When trying to measure the resonance frequencies of a strongly coupled band-pass filter, where both, primary and secondary coil have parallel capacitors, i.e. both sides of the filter are tuned to the IF. In the case of the PATHEDYNE IF-filter, the oscillation frequency of the Exciter jumped stochastically between two values. In order to clarify these points, it was decided to run more precise measurements using a wobbulator (frequency sweep generator) and a spectrum analyzer. Two parameters were now measured for every filter type:
The Low Capacity Injector Sensor (LCIS)The frequency dependant impedance Z(f) was recorded using an active probe, which in the following discussion will be called 'LCIS'. As shown in the following schematic, this self-built active probe combines a low capacity signal injector with a low capacity sensor and a JFET amplifier. In order to minimize the influence of the output impedance of the signal generator (wobbulator) on the resonance behaviour of the DUT, its output signal was coupled into the DUT (in our case the resonance circuit) via a very small capacitor of <1pF. The RF-Signal building up acrosss the DUT is proportional to its impedance. It is sensed via another small capacitor, again < 1pF. The transistor acts as an impedance transformer allowing to record the output signal on an oscilloscope or on a spectrum analyser. Both, coupling and sensing capacitor were realized by neighbouring copper strips on the printed circuit board. The dimensions of this active probe are so small, it can be accommodated in a copper water pipe of 15mm dia (yellow tube in the lower part of the picture). Signal connections use standard coax cables and BNC connectors. The JFET operating voltage of 3 - 18V DC is provided by a standard wall plug power supply. Working with the LCIS is particularly comfortable if a spectrum analyzer with built-in signal generator is available. The measured curves can then be displayd on a logarithmic scale, leading to better resolution in the stop band of the filter. Obviously, in place of our LCIS one can use any active probe for this type of measurement. The problem is that in the interest of minimizing the influence of the measurement equipment on the DUT, the coupling and sensing should be made through very small capacitors. Putting the probe in the vicinity of the DUT to achieve very loose coupling is a bad idea, since the reproducibilty of the measurements will be poor.
The "GAMMA" IF TransformerAs a first test specimen an IF- transformer made by the french enterprise "GAMMA" was analyzed. This type of IF-filter has an aperiodic primary consisting only of coil L1 (bottom) and a tuned secondary consisting of coil L2 (top) and a parallel MICA capacitor C = 225 pF. When triodes are used for amplification, the smaller coil L1 is connected to the plate of the preceding tube, and the resonant circuit with the big coil L2 is connected to the grid circuit of the following tube. Before commencing the frequency response measurements of the filter, the inductance and coupling factor of the primary and secondary coils were measured. This obviously required removing the 225 pF resonance capacitor from the secondary coil. The test equipment used in this first stage of the investigations were:
The following inductance and resistance values were found: (the non-measured coil was always left open): L1 = 9,35mH (9,1mH) R1 = 103,5 Ω (105,3 Ω) The results obtained from 'LARU', resistance meter and LCR-T5 show satisfactory agreement. The coupling factor of the primary and secondary coil was derived by measuring the inductance of each coil in two conditions: (i) with the other coil left open, and (ii) with the other coil shorted. An in-depth description of this approach can be found in "Berechnungen in der Radiotechnik". L1 = 7mH (7,1mH) with L2 shorted The values in paranthesis were obtained using the LCR-T5. Using the curves found in the above quoted reference one can easily deduce a coupling factor of k = M/(L1L2)½ ≈ 50% , a value which can be considered sufficiently exact in the present context.
Frequency dependance of the "GAMMA" filter impedance Z(f)Measurements of Z(f) using the LCIS
The impedance curve Z(f) shows two peaks. Since the peak at 418.5 kHz has a higher amplitude, one would expect that the EXCITER will oscillate at this frequency. However, comparing this value with the result of the Exciter measurement given in the table of post #1 (467 kHz), one observes a very distinct difference. This indicates that inspite of the extremely loose capacitive coupling, the LCIS produces a higher capacitive load on the resonance circuit than the Exciter!
It seems worth mentioning that the measured Z(f) characteristic of L1 shows, that the load resistance seen by the triode is rather compex. This may serve as a hint that this type of amplifier stage might oscillate on unforeseen frequencies.
In this constellation the Z(f) characteristic shows only one maximum. The position of this maximum corresponds to the lower maximum of the previous graph. As a consequnce of the shorted secondary coil L2 the inductance of L1 dropped to a lower value. A second upper maximum is no longer observed. Obviously this occured due to coupling with the unloaded secondary coil.
Only one resonance maximum ist observed. The resonance frequency is determined by L2 and stray capacitances.
Shorting coil L1 reduces the inductance of coil L2 which results in a higher resonance frequency. Apart from that, no significant influence of coil L1 on the impedance characteristic of L2 is seen. This is in contrast to the opposite case shown in the first two graphs above, where the LCIS had been connected to L1 and the change of L2 from open to shorted had a very dramatic impact on the Z(f) curve of L1.
The impedance characteristic of L1 is now very similar to the one with coil L2 shorted. The characteristic has no second maximum. The response curve resembles the case "LCIS connected to L1 with L2 shorted (without C)"
The Spectrum Analyzer settings were identical to the previous cases. At first glance only noise is measured. However, on the far left side a somewhat higher peak may be identified. Accident?
To clarify this point, the area around the assumed peak was magnified by reducing the displayed frequency range from the previously used 30 - 600 kHz to 40 - 100 kHz, and by increasing the output signal level of the RF generator.
Now the resonance of the filter secondary circuit L2-C is clearly seen at 50.050 kHz. So the vague peak in the previous measurement was not accidental but a realistic signal, camouflaged by the background noise!
Additional Tests without LCISSome additional tests were run on the GAMMA filter to quantify the effect of tighter coupling of the signal generator and spectrum analyser to the circuit. In these tests the RF signal generator was injected into the hot end of the secondary circuit L2-C via a capacitor of 12 pF, and the spectrum analyser was connected to the same point via its standard probe in series with a capacitor of 2.7 pF. (The LCIS used coupling capacitors of <1pF)
One observes a weak resonance peak at 50.8 kHz (marker).
One observes that, compared to the previous graph which was recorded using the LCIS, the measured characteristic is rather noisy, but what is more important: None of the two curves shows the typical resonance characteristic of an LC resonant circuit. The reason for this is the capacitive couping of the RF signal generator and the probe to the hot end of the LC resonant circuit which prefer higher frequencies. This leads to a significant distortion of the typical LC resonant curve. The typical LC resonant curve will be obtained if the RF signal generator is coupled to the foot of the resonant circuit. Therefore, a fooder C = 66 nF was installed in series with the circuit resonance capacitor of 225 pF. Injection of the RF is parallel to the fooder C. The spectrum analyzer stays connected to the hot end of the circuit, again using a 10:1 probe in series with 2.7 pF. As one can see, that foot point signal injection produces a proper resonance curve.
Measurement of the frequency response curveWhile all measurements described above had the objective of showing the frequency dependant impedance of the IF filter primary and secondary circuit, the following investigations were made to show the frequency response of the complete filter unit under realistic load conditions in a radio set. For this purpose the output of the RF signal generator (Ri=50 Ω) was now connected to the hot end of L1 via a resistor of 33 kΩ, representing a realistic simulation of the internal resistance of the tube feeding the filter. The spectrum analyzer (Ri=1 MΩ) was connected to the hot end of the secondary resonant circuit (L2 & C =225pF) via its 10:1 probe with an additional series capacitor of 2.7 pF. This means that the secondary also saw a load comparable to the control grid input of the following tube.
We can now see that operated under realistic load conditions, with signal injection into L1 and signal pick-up at L2, the GAMMA filter shows a clear filter behaviour.
In retrospect one might ask oneself why we did all the measurements described above, if a realistic measurement would have given the correct answer straight away. The answer to this question is found in the first part of this report. Using the EXCITER as a simple gadget to find resonance frequencies of LC-circuits, we had found some questionable results, like e.g. the surprisingly high resonance frequency of the GAMMA filter primary. Only the detailed measurements of the frequency dependant impedance of the filter primary and secondary circuit allowed understanding the sometimes unexpected results obtained from EXCITER. The origin of such strange results can be in primary coil oscillations with its own intrinsic capacity or in primary-secondary-interaction. The problems encountered here in the context of the EXCITER measurements may also be found in modern semiconductor circuits. The difficulty results from a loose coupling of the coils of a RF transformer. A practical example is the oscillator in a transistor set. An oscillator is de facto comparable to an Exciter: both have the task of stimulating a passive LC circuit to oscillate.
The "SOLENO" IF TransformerLike the "GAMMA" IF-transformer described above, the "SOLENO" IF-transformer has an aperiodic primary coil and a secondary coil tuned to the IF frequency using a parallel capacitor. While in the "GAMMA" transformer, the secondary circuit used a MICA capacitor of 225pF to tune the IF to a fixed value of ≈ 50 kHz, the "SOLENO" has a built-in variable capacitor that allows tuning the IF between 41 and 77 kHz.
Measurements of Z(f) using the LCIS
Measurement of the primary coil shows (comparable to Gamma) a maximum far above the nominal IF frequncy. However, the coupling between primary and secondary coils is tighter than the Gamma transformer, and therefore a second maximum (59.5 kHz) shows up. The Exciter indeed oscillates with the frequency of the absolute maximum, cnf. table in post #1.
Here again a circuit resonance curve shows up. The circuit is capacitively coupled to it's "hot" end.
Measurement of the frequency response curveAgain the measurement of the frequency response curve used a source resistance of 33 kΩ for the injection from the RF generator. From the hot end of the secondary resonance circuit a 10:1 probe with 2.7 pF in series picked up the the signal for the spectrum analyzer.
Compared to the frequency response curve of the GAMMA, the SOLENO has a lower Q-factor, and is consequently less selective. This is a consequence of the fact that the GAMMA coils were cheese-wound, whereas the SOLEMA coils were wound arbitrarily.
The "INTEGRA <412>" IF Bandpass FilterThis INTEGRA type <412> is a "genuine" bandpass filter. It has tuned primary and secondary circuits.
Measurements of Z(f) using the LCIS
Although both coils are "far" from each other, the coupling between them is sufficiently strong to produce over critical coupling, which results in the appearance of two maxima in the frequency response curve.
Here again two maxima show up. Now the secondary maximum is higher.
It can be seen that the tube's (simulated) internal resistance now attenuates the primary circuit to such an extent that the second maximum disappears. Judging from todays knowledge, this is not the optimal design for a bandpass filter.
Measurement of the frequency response curveLike in the previously tested filters GAMMA and SOLENO, the measurement of the frequency response curve was run with a source resistance of 33 kΩ, and the probe connected to the hot end of the secondary circuit via an additional series capacitance of 2.7 pF.
As one might have expected, the frequency response curve differs only marginally from the previous measurement using the LCIS.
The "INTEGRA <406>" Lowpass π-Filter
Normally, a regenerative detector is followed by a lowpass filter (of RC or LC type) for blocking RF signals. In this set, however, the lowpass filter has a π structure.
Networks with a π structure are usually found in impedance matching units between a transmitter and an antenna. In regenerative receivers, however, such a structure is very rarely found. For optimum performance such a lowpass π - filter has to be terminated on both sides with its characteristic resistance, which in the case of the INTEGRA <406> amounts to approx 2.8 kΩ only (see comments in post#1 of this thread). This value is far too low for the application in the LAGIER ZZ radio set. Therefore one has to assume that this π structure was not installed as a genuine AF-lowpass filter but for RF blocking only.
Using again a source resistance of 33 kΩ, and the output only loaded with the probe, a resonance maximum at approx. 14.1 kHz is found. This result is in good agreement with the value found with the EXCITER. The frequency response curve also shows that the INTEGRA <406> π filter is not a lowpass as expected from its topology.
Using alternatively a source resistance of 8 kΩ (corresponding to the internal resistance of an REN904), and also a load resistance of 8 kΩ, one observes an increase towards higher frequencies. This would allow compensating to some extent the drop of the AF-amplitude towards higher frequencies caused by the narrow bandwidth of the IF filters in the preceding stages. Maybe this was the original purpose of introducing the INTEGRA <406> π filter between the regenerative stage and the AF amplifier. Up to now this point could not be finally clarified.
The Pathedyne IF Filter
The IF filters used in the Pathedyne SB receiver are bandpass filters. Primary and secondary coil have parallel capacitors of nominally 250 pF and 350 pF (measured values: 215 pF and 290 pF), which tune the circuits to approx. 50 kHz. The mechanical setup is surprisingly simple: Two arbitrarily wound coils on cardboard bodies with very little space between them. Therefore, the coupling factor k becomes much too large for a "genuine" IF bandpass filter, and an over critical coupling will result. It can be expected that the frequency response curve will show two maxima.
Measurements of Z(f) using the LCIS
The LCIS measurement at the primary circuit shows two distinct maxima.
The LCIS measurement at the secondary circuit shows the same peaks, however, with a slightly lower amplitude, and at somewhat shifted frequencies.
Measurement of the frequency response curveIn a first measurement campaign, the original circuit capacitors were used. As already mentioned above their capacitance values had dropped from the nominal value of 250 pF to 215 pF and from 350 pF to 290 pF, respectively.
The IF filter is over critically coupled. Both peaks are clearly visible. The maximum at 65.25 kHz is higher and belongs to the secondary circuit which is unloaded.
The maximum at 41.75 kHz belongs to the unloaded primary circuit.
A source resistance of 22 kΩ corresponds to the internal resistance of a triode feeding the bandfilter. As a consequence of the lower source resistance the primary circuit is now damped to such an extent that the second peak disappears. The over critical coupling is no longer observed.
When injecting the signal into the secondary circuit (reverse signal injection), the 22 kΩ damping causes the impression as if one had a well dimensioned bandpass filter with a flat top frequency response curve.
In a second measurement campaign, the original capacitors were removed from the filter body and replaced by high quality ceramic capacitors of 240 and 350 pF. Before mountig these new capacitors they were checked on the C- meter 'KARU'. 'KARU' is a resonance capacitance meter, with a panel instrument giving an indication of the oscillation amplitude. It was surprising to see that new ceramic capacitors produced a much higher oscillation amplitude than the old capacitors, indicating that the latter had not only lost part of their capacity but also that their quality was poor. After replacement of the filter capacitors, the 4 preceding measurements (2 with 82kΩ source resistance and 2 with 22 kΩ source resistance) were repeated to see whether there was any visible difference in the resonance curves.
As a consequence of the higher values of the new capacitors, the resonance peak of the secondary circuit dropped from 62.25 kHz to 59 kHz. The resonance curves are steeper than with the old capacitors.
The replacemant of the capacitors shifted the resonance peak of the primary from 41.75 kHz to 38 kHz. Here also the resonance curve is steeper.
Loading the IF filter with the simuated internal resistance of the triode only one peak is observed at 52.75 kHz. (With the old capacitors the peak was at 60.25 kHz).
In the case of reverse singnal injection, a clear resonance peak appears. This may be due to the higher Q-values of the ceramic capacitors.
Summarizing the results obtained for the PATHEDYNE IF - filter one can say that with sufficient damping the two pronounced humps disappear, and the more or less flat top of the frequency response curves shows a certain similarity with a bandpass filter. One should keep in mind however, that a 'proper' bandpass filter in modern understanding does not only have a flat top but steep slopes on both sides, and this is most definitely not the case with the PATHEDYNE filter. Well, one should not forget that these radio sets were built during the childhood of superheterodyne radio technology and that enginineers were still learning. We have also seen that, without sufficient damping the frequency response curve of the PATHEDYNE IF filter shows two pronounced peaks. This explains the surprising results obtained when measuring the resonance frequency of these filters using the EXCITER. The EXCITER stimulates the filter circuits without putting any significant load on them. At which of the 2 peak frequencies 41kHz or 65 kHz oscillations were stimulated appeared to be purely random. By the way: This frequency jump never occured once the EXCITER had started the oscillations. But switching it off and on again, it proved to be completely unpredictable, which of the two results one would get. The 'jump' phenomenon typically occurs when over-critically coupled resonant circuits form part of an oscillator - like is the case with the EXCITER. In vintage single stage AM transmitters the jump phenomenon posed a serious a problem. Rukop has investigated this phenomenon and published his results here: "Reiß-Diagramm" (rip-diagrams). |