toptone: Detection and AGC in a 5 tube AC/DC radio
I have measurement results for Detection and AGC performance in a late production Japanese AA5 radio model SM100 by Topline.
The tube lineup is the classic 12BE6 pentagrid converter, 12BA6 IF pentode, 12AV6 dual detector diode and audio preamp triode, the 50C5 beam tetrode and the 35W4 power rectifier. The only noteworthy variation from the cannonical design, is that the 12BA6 IF pentode was loaded with a single tank circuit, rather than the usual 1:1 455kHz loosely coupled transformer.
I included measurements for AGC behaviour, detector performance, and stage gain for the original radio and several variants with improved detector gain and AGC compression.
In one of the circuit variations, I got the detector gain boosted to over 5, and AGC delay improved compression for signals in my urban home area.
The original AGC action compressed 40dB at the input into a 10.7dB range at the detector output. I added a 3V AGC delay and reduced the output further, to a 6.9dB range. The greatest benefit of the AGC delay was a combination of increased gain for weak signals and a more effective moderation of the strongest signals. The largest gain increase was 12dB.
Detector measurements and improvements with positive feedback:
The detector improvement came as the result of a musing about what whould happen if audio feedback from the voice coil were applied to the cathode of the 12AV6 dual diode & triode.
See pages 5,6,7,9 of AGC-detector-schematics
The common cathode shared by the triode and diodes meant that detection would be affected. As it turns out, Negative voltage feedback to the triode cathode works as Positive feedback for the 12AV6 detector diodes sharing the same cathode. This positive feedback is very helpful to sharpen the soft forward characteristic of vaccum tube diodes.
The sharpening of the diode characteristic happens because the audio at the cathode works as a variable diode bias that follows the audio envelope. The total voltage drive presented to the diodes is increased, as a result: IF voltage drives the detector anode and audio drives the detector cathode in opposing directions.
The amount of Positive feedback applied to the detecting diodes is set by the volume control between the detector diodes and the 12AV6 triode grid.
At full volume, which may be needed for the weakest signals, the positive feedback more than doubled detector output. This was observed in the attached measurements, and also resulted in much better audio fidelity for low level audio signals.
As was the original goal, the negative feedback for the triode section also improved audio tone at any volume level. The amount of negative feedback for the triode is fixed because it does not go through the volume control.
While pursuing possible improvements for AGC action, I thought to double detected audio and AGC voltage output by replacing the 100k filter series resistor at the detector output, with a 1N34A diode. This transformed AGC detection from peak envelope detection to peak-to-peak envelope detection, thus doubling detected output for all IF signals driving the detector above about 50mV.
See page 3 of AGC-detector-schematics for the first implementation of p-p detection with the 1N34A.
As expected, very low level signals are more difficult to detect with the two diode drops from the original thermionic diode in series with the 1N34A, than with just one diode drop. But once audio feedback was applied as described above, the low level limitation was eliminated.
Guyla Kiss in Holand has used p-p detection in some of his AM tube radio designs.
The combination of voltage doubling from p-p detection, and Positive feedback to the thermionic diode cathode resulted in a 5-fold increase in detected audio output. Listening tests confirmed improved low level signal fidelity.
See page 5 of AGC-detector-schematics with p-p detection for AGC and Audio.
The intended improvement from p-p detection in signal compression was negligible.
The 5 fold increase in p-p detected output with PFB is useful to make up for the gain loss from audio feedback around the audio amp.
Full Gm/gain from output Pentode:
While tinkering with the radio, I also tried to increase forward audio gain, while reducing component count. I did this by taking the required negative bias for the 50C5 beam tetrode from the local oscillator section of the 12BE6 pentagrid converter.
See page 4 of AGC-detector-schematics
This particular oscillator variant used a Hartley circuit with a coil tap for the cathode and thightly coupled top of the tank to the grid. This thight coupling that does not rely on a low capacity coupling to the grid for oscillation, as a result, it yields a very flat grid leak self bias voltage around -12V over the entire tuning range. Oscillators with very light capacitive feedback to the grid tend to have more oscillation amplitude variation over the tuning range. The negative self bias is approximatelly half of the p-p oscillation amplitude at the grid.
I saw about 1V of variation in self bias over the tuning range. This negative bias trick could have been used since the dawn of the local oscillator in superhet sets, but I have never seen it applied anywhere.
The increase in gain comes from the hard grounded cathode of the 50C5. Shorting out the 150Ohm unbypassed cathode resistor adds about 6dB with the full realization of the available 7.5mS transconductance of the 50C5.
In a private communication with Ed Lyon (Radio Age editor), he recalled seeing this kind of bias as a cost saving measure in some Japanese radios that were contemporary with the Topline SM100 radio that is the subject of these experiments. Ed warned of the potential for overheating in the radio, should the local oscillator stop running. I made extra measurements for current and power draw with the 50C5 driven with 7.5V bias and with 0V bias. See the schematics on pages 8 and 9 at AGC-detector-Measurements. These schematics show measured currents and power dissipation as a result of loosing bias on the 50C5. In this particular radio, the added dissipation of the 50C5 at zero bias did not cause dangerous rise in internal temperature.
This method of pentode bias from the LO would have been preferable to the method that was common when electrolytic capacitors were expensive and Power tetrode gain was not very high. This other method employed a resistor in the B- path to create a negative voltage. Often this voltage had a lot of hum.
Another place where this method of bias would have been very convenient is with battery radios, because the power tetrode had a grounded directly heated filament. These radios used the resistor in the B- leg, and it had to be bypassed by an electrolytic cap.
But, keep in mind that using the LO as a source of bias requires that the oscillator amplitude, and thus the DC bias it generates to be fairly constant over the tuning range.
AGC delay is a misleading term that means the elimination of AGC voltage for any signals below a certain level. With AGC suppressed for the weakest signals, the radio runs at full gain when it is most needed. A 3V delay means that only signals that exceed 3V in amplitude at the detector will start to reduce RF and IF gain.
Pages 6 and 7 of AGC-detector-schematics show the first experimental version of adding a 3V delay, and the final version, that employs a cathode resistor to obtain the 3V delay.
As I mentioned the introduction above, the 3V delay improved the original AGC compression from 60dB/21dB to 60dB/14.6dB. This improvement also includes the voltage doubling from p-p detection in the AGC voltage. The added 12dB gain for signals that previously only developed -1V of AGC voltage, and the more effective compression of the strongest signals made a noticeable difference in tuning the radio across the dial.
The last page of AGC-detector-Measurements shows the gain of each stage (12BE6 converter and 12BA6 IF) as a function of AGC voltage. Note that AGC is most effective in the -7V to -2V range. For larger signals, the gain drop is sacrificed to maintain low distortion. Also note that a grid bias below -1V causes a reduction in stage gain due to slight grid conduction that loads the high Q tank circuits at the grids.
The Philips796A from 1936:
After I finished my experiments, I came across a very similar detector design in the Philips 796A from 1936. This radio includes the AGC delay resistor at the cathode of the triode, as well as feedback from the voice coil to the same Triode cathode. This was a completely fortuitous, but fun coincidence. So this radio should have higher sensitivity, better AGC action and cleaner detection than other contemporary 5 tube superhet designs.
This Philips 796A includes a tuning eye indicator that is driven by the top of the volume control which has no AGC delay, this means that the tuning eye is active for signals below the AGC delay voltage.
One drawback of introducing AGC delay to my radio is that I lost signal indication for signals below the 3V AGC delay. The IN-13 monitors the current of the IF stage, so it is dependent on the delayed AGC signal.
A 10MByte video clip showing operation of the radio with the IN-13, but may be too large to post at RM, can be see on my personal page .
Because of your nice picture with the neatly done tuning indicator with a bargraph IN-13 I was looking at your model. Maybe my eyes are not good enough but I think according to the picture with the model plate (thank you) the model should be named "Topline 5M-100" instead of SM100 (5 instead of S). I have changed it but would like to ask you if this is all right for you.
I also like your well done schematics which you have uploaded to the model. The photo "Tune-a-lite" would be nice to put to the "tube" and then link it, not a foreign link. Since it is a russian tube we write it (internally as tube model with small letters as in-13). We seek to be selfreliant and therefore prefere links to pages on RMorg. The second reason is that we want to prevent that in the future (many years) we are a linkage grave. We would not have this vast information if we would have just linked ...
Your different schematics with explanations would also be most welcomed on the model ;-)
Something for the future: A larger article is always better than two shorter ones. Therefore I would suggest in such a case to just add a second post or more to the same article. Google will like it much more ...
The name problem Toptone / Topline we will have to solve one day. Now the model can be found both ways. Perhaps the maker ist really "Optical & Radio MFG Co Ltd" which used to manufacture in 1959 models like Toptone AR-65 (Mascot, Remington) but was also known as TOPCON. Maybe Topline is just a part of the model name maybe it is the company.
At least at a certain time Toshiba had 40 % of Topcon's shares. Tokyo Optical made also the AR-610 which also has been sold by ACME as "Tops all" Ch-610 and by Haltone as AR-610. Toptone is also a Korean speaker manufacturer with factory in China.