c. Patrick jankowiak and Dennis Brady
This project evolved from a desire among several long-time CW operators to experiment with the AM mode. Operating CW on the 40M band near 7163 KHz, and often hearing the AM activity on 7290 KHz, interest increased, as several members resurrected old AM gear or modified existing equipment, and went on the air with live audio.
Dennis had built a 250-TH power amplifier for CW operation on 40M, and with 2KV on the plate and -75V on the grid, the unit idles at 50mA and provides about 275 watts at 200mA. Operation is close to class-B, and the rugged, overbuilt unit performs superbly on CW and also works well as a linear amplifier. It was proposed that we experiment with this RF amplifier and try to amplitude modulate it.
Various options for modulating the 250TH were considered. The Ancient Tomes of Terman, Editors and Engineers, Everitt, Orr, Nilson, Hornung, and Sterling dusted off and their sage wisdom regarding this matter was revisited. A clean and linearly modulated signal, as close as possible to the audio standards of "high fidelity" would be our goal.
Plate modulation of this stage as then configured could net up to 800W peak envelope power with a high percentage of modulation and a very clean signal. This would require approximately 200 watts RMS power, and present a load impedance of 10K ohms. An Altec 1570B amplifier was available which could provide this power if modified for a capacitor input power supply, raising the plate voltage on its pair of 811A's from the stock 930V at 175W to 1250V for a potential of 250W. The plate to plate load impedance would be about 10K-12K ohms, up from the 6500 ohms of the stock output transformer. The modulation transformers on hand provided 8000 ohms CT to 2800 ohms. This would be a gross mismatch, but if two were run back to back for a 1:1 match, it would have been OK. Unfortunately, the voltage rating was only 2KV on the old transformers and we didn't want to push it. It was considered to run the 250TH at a lower plate voltage such as 1000V, but the necessary modifications to the power supply were inconvenient, and 1KV is pretty low voltage for a 250-TH.
Series (pass tube) plate modulation was discussed, but the inherent inefficiency of the modulator in this circuit with its class-A pass-tube would have required us to use something like a 4-1000A (or at least a 3-500Z) and a 4500V power supply, not to mention the difficult design and complex regulating circuitry required to set and keep the operating point, consistent with a linear audio transfer curve. This method is, however, excellent for low power operation where very high fidelity and very wide frequency response is desired.
Grid bias modulation was also considered, but it was not very desirable because of the low efficiency of the RF stage and the relatively high audio distortion (nonlinearity) at high modulation levels.
Cathode modulation was considered, and there were a few drawbacks to it as well, mainly due to the difficulty of creating an overall linear transfer function, in essence, the distortion problem would remain, but not to the degree of that which would present itself in a grid modulated scheme. The final stage efficiency would only be expected to be about 50% as well.
Cathode modulation won out, as, in the absence of an appropriate plate modulation trnasformer, it was the best compromise between simplicity, distortion, efficiency, and minimal expenditure. The audio power actually required would be quite low, only 30-50 watts, and the impedance would be somewhere between 300 and 500 ohms. It was both easy and fairly safe to insert a connection, between the 250-TH's filament transformer center tap, and ground.
The Altec 1570B has a 32 ohm output which was too low to be directly connected in series with the 250-TH's filament circuit. Digging through our junkboxes (or junk buildings in this case), an 120V-to-80VCT, 6A power transformer from an old solid state amplifier was found. Connecting half the secondary winding to the 32 ohm output of the Altec 1570B, the impedance transformation would provide an output impedance of approximately 288 ohms. There was some skepticism concerning the frequency response of the transformer, but at this point we were too eager not to just try it and see what happened.
Tests and Alignment
The test setup consisted of a 50 ohm dummy load, a dual trace oscilloscope, and an audio oscillator. The audio oscillator was routed not only to the Altec's input, but also to the sync input of the scope. RF was sampled at the dummy load and connected to the scope vertical input. The RF amplifier was energized and tuned. It was discovered that under modulation, the plate and grid current would rise dramatically whenever the modulation was increased beyond about 30%.
Increasing the grid drive alleviated the problem slightly, but that was not the problem. It was determined that the modulator was driving the 250-TH cathode to a point substantially more negative than the grid during a half-cycle of audio, causing excessive grid current at audio frequencies. The tank circuit of the 250-TH looks like a short at audio frequencies, thus the high plate current. More grid bias would be needed.
The grid bias was increased to -185VDC, and the 250-TH was operated 110V beyond cutoff. Drive from the exciter was increased to maximum so that the 250-TH output was 150 watts. Modulation was applied, and the plate current meter than showed only a slight increase. Modulation percentage was approximately 50% at which time the RF waveform began to clip. Loading was reduced and the amplifier re-tuned. Grid tuning was performed on the amplifier, and the exciter output stage was re-tuned. The RF carrier output was then 80 watts, and the modulation approximately 95%. Grid current remained essentially similar. PEP was calculated to be 160 watts, and the minimum envelope power was calculated to be 8 watts. This seemed to be the best set of operating conditions which could be found that allowed the 250TH to run cool (medium red, not orange) and yet obtain good overall performance. Trying to get closer to 100% modulation without distortion resulted in a substantial increase in sensitivity of the overall system to any mistuning or misadjustment. A good balance between ease of alignment and modulation percentage was reached at 90% modulation. The plate current meter still showed the slight increase under modulation, so the audio generator was switched to a triangular waveform in oerder to examine linearity of modulation. The oscilloscope showed a nonlinearity consisting of an outward (towards increased power) curve applied equally to both the leading and trailing edge of the waveform. This was calculated to be equivalent to 3% second harmonic distortion. Speculation is that this is due to a spare power transformer being used as a modulation transformer.
The Altec amplifier was known to be extremely linear, but the new modulation transformer may present reactive load, especially at the higher frequencies used in speech, and the transformer was attached outside the 1570B's feedback loop.
With the slight plate current increase under modulation sufficiently accounted for, the oscillator was switched to sweep mode and the sweep ramp output (sawtooth) was connected to the oscilloscope horizontal input. The system was swept from 100Hz to 10KHz at 90% modulation referenced to 1KHz. The modulation frequency response from 100Hz to 6KHz was found to be flat within 1dB. This was attributed to both the large size of the modulation transformer and the capability of the Altec amplifier to deliver power far in excess of that required to modulate the stage. The overkill modulator power was thought to also contribute to the relative insensitivity to any possible impedance mismatch and the overall ease with which a high percentage of low-distortion modulation was achieved.
Once the system was tested satisfactorily, a mixer/preamplifier was connected to the input of the modulator. The audio generator was conected to one input and an RCA "FM base station" desk microphone, made by Sure Bros., to another. This allowed the instant exchange of tone and voice, for testing. The scope was used to monitor the RF output for modulation, and its sweep rate was set between 20 and 40 Hz.
The system was operated into a dummy load, and a voice check was made on a receiver in another room (some RF always leaks out!). The experimenters took turns reading some material "over the air" (to the other experimenter who was listening in the separate room), while observing the operating parameters, and adjusting the mic gain channel for as much modulation as possible without hitting 100% negative modulation.
The test receiver was a Hallicrafters SX-100 with the bandwidth set to 3KHz and then 5 KHz. Not only was the audio extremely clear, but in the 5 KHz position, the character of the received audio was similar in quality to that of a commercial AM broadcast station.
A slight hum was noticed, and it was discovered to be coming from the 250-TH's power supply. The percentage of ripple was about 5%, and was not noticed on CW. We are looking into reducing it to 1% or less, so that it will be less noticeable wnen AM is used.
On The Air
There was a CW rag chew was in progress on 20M, and since several of the participants were known to one of the experimenters, we broke in and asked if anyone would mind going to 7163 and give us a report. After a short delay to allow a couple of them to get their spare receivers tuned to our transmit frequency, an on-air test was then made, and the reports were all positive, containing comments such as "A-1 sound" and "sounds like a broadcast station". At least one participant was inspired to get his AM rig going and give us a yell next weekend.
Next Saturday, the rig was tried out on the air on the 7163 CW/AM net. The reports on our Texas Experiment were very good, and compliments came from places as far away as Missouri. Reports from QSO's said audio quality VERY good. One guy said his S-meter held steady as I spoke, which (according to him) is a good indicator of correct modulation. Not too bad for an 80 watt carrier feeding an inverted V through a transmatch.
Details as to the measurements during voice conditions and type of RF circuit are as follows:
Plate volts: 2200
Plate current: 75ma carrier only, 150ma voice peaks
Grid current: 7.5ma (never changes)
Grid bias: -185v
Voice peak audio voltage at cathode was indicated at 60V on an ac averaging meter, so probably was much higher peak voltage.
split (balanced) grid tank circuit,
PI-L output circuit.
Based on article by Bill Orr.
So what is this all about? It's not about converting CW guys to AM or hogging bandwidth, not at all. It's about experimentation, cut and try, vacuum tubes, and innovation. It's about amateur radio, and the love of the hobby. It's also about respect for doing things the way they used to be done, and respect for the people who did them that way. Real men use tubes.
Dennis Brady, W5FRS, builds much of his own ham radio gear. Preferring balanced amplifiers, and doing it the old fasioned way, it is created with the same type of parts found only in the vintage amateur and commercial gear of the 1930's and 1940's. The gear is often patterned after that shown in texts of the day, and is meticulously designed and handsomely crafted. To look at some of it, you would think you are seeing pristine museum pieces from the era.
Patrick Jankowiak, KD5OEI, has always been interested in plate-modulated transmitters, as an outgrowth of interests in RF and audio. His design and construction tends to the functional, diverse, and sometimes unusual, with some examples being tube-type PC speakers, high fidelity power amplifiers, pass-tube modulators, and on the solid state side, a 1 GHz spectrum analyzer is a high point. He earned his FCC radiotelephone operators license in 1986, and a Technician ham license in 2001.
Both gentlemen are members of the Vintage Radio and Phonograph Society and work in the Electronics and IT fields.
Frank's tube data sheets
DEFBOO.CMD technology hobbyist site
AM nets are also commonly found on 7290, 3880, 3885, 3890, and 3895 KHz. Times vary, but try Noon-3pm CST for 7290 and mornings for the 3.8 MHz frequencies.