HI-FI AM transmitter

ID: 236679
HI-FI AM transmitter 
15.Nov.10 02:46

Joe Sousa (USA)
Articles: 664
Count of Thanks: 35
Joe Sousa

Fellow Radiophiles,

I have distilled some of what I have learned in this forum into a new HI-FI AM transmitter design. This design takes lessons from earlier AM transmitter designs and from what I learned about neutralization.

One thread that was particularly instructive for me dealt with the AM modulator in the Granco ARC60 FM-to-AM converter for car use. I learned here about problems with spurious FM modulation, and distortion in the modulation process.

For completeness sake, I repeat here a list of AM transmitter threads that I am aware of:

Dynatron tetrode transmitter

1 Watt AM Transmitter (solid state)

Home made AM transmitter with dual gate mosfet

A survey of AM modulator/transmitters

Another survey of AM modulators/transmitters

Am tube modulator this is one of the most extensive AM transmitter threads in German (I read it with the Google translator).

Simplest diode modulator

Granco-ARC60 AM Modulator/transmitter runs off 12.6VDC.

 High Performance modulator with Resistance stabilized oscillator


Another relevant thread explains neutralization of grid to plate capacitance (Miller capacitance) in the analysis of the Grundig 5040W/3D. The topic of neutralization came up at different points in the 5 part analysis with particulary enlightening explanations and analysis by Hans Knoll. One particularly thourough analysis of neutralization in IF amplifiers was contributed to the Grundig 5040W/3D thread by Andreas Steinmetz.(I read this thread with the Google translator)

The operation of Dual Control Pentodes is also directly relevant to this design.

Single Pentode Transmitter concept

The main characteristics can be summarized as follows:
-Linear Class A operation for the oscillator for low harmonic content
-Class A operation also means that the oscillation period will not be lengthened by the class C current pulses. The period will be determined by L and C.
-Spurious FM modulation from Miller capacitance gain variations will be eliminated with proper neutralization of parasitic feedback capacitances.
-Linear modulation with two quadrant multiplication for 100% modulation without distortion and suppressed carrier modulation is thus possible too.
-No tuned circuits affect the modulation bandwidth. The modulation path is limited only by transit time and stray parasitics. The band of interest is AM, but perhaps this could be used for Video modulation at VHF.

Circuit blocks

1- Cathode follower oscillator with undistorted sine waves at the tank voltage and at the cathode current.
The idea here is to control the class A oscillation amplitude with the DC level of the grid bias, instead of grid-leak derived duty cycle control as is usually done with local oscillator class C designs in nearly all radios.
The sine wave at the top of the tank is 6x larger than at the cathode. The DC envelope of this oscillation at the top of the tank is detected by the 1N34 diode to extract a DC AGC voltage bias voltage for the control grid G1. This bias voltage affects transconductance to control oscillation amplitude.
The DC AGC voltage will stabilize at a point where the linear class A oscillation path has a net AC loop gain of 1. The input resistance at resonance at the coil tap was measured as 600Ω. The cathode drives a sine wave at the tap that is stepped up by 6x to the top of the tank and then stepped back down 0.33x by C1 and C2. The net step up from the tap to the grid is 2x. An attenuation of 0.5x is needed from the grid to the cathode for unity gain around the AC oscillation path.The AGC voltage presented to the control grid will stabilize at a point where the cathode resistance is also 600Ω to obtain the needed attenuation of 0.5x. My final choice of the pentode was the 6AS6. The plate and screen transconductances from G1 are each 4mS with 150V at the screen and plate. The triode transconductance is therefor 8mS at 150V, and the lowest possible internal cathode resistance is 1/8mS=125Ω.  When the control grid G1 bias drops to -3V, the internal cathode resistance increases to 600Ω and the unity loop gain condition is satisfied.
The peak AC amplitude at the top of the tank is 3.2V leaving a total swing of 1V p-p at the control grid G1 with respect to the cathode. The cutoff voltage for the control grid G1 with 150V at the plate and screen is about -7V. This means that the grid operating under linearly in class A with only 1Vp-p at a bias of -3V.
The first tube types I tried for this transmitter were remote cutoff RF pentodes. The plate voltage would have to run at a low value for modulation by G3. See how the EBF89 remote cutoff pentode compares to the dual control 6AS6 pentode. However, the expected gain control advantage of the EBF89 did not materialize. The reason for this is that the EBF89 was designed for gain control over 3 decades with a control grid G1 bias range of 30V and G2=125V. One compromise that is made for this control range is that the tube gain is more nonlinear at low G1 bias voltages than the gain of an equivalent sharp control cutoff pentode, like the 6AS6. The 6AS6 is more linear with 1V p-p swing and -3V G1 bias than the EBF89 at the same bias level. The EBF89 can be very linear with 1Vp-p swing at G1, but with a much more negative G1 bias level, like -20V, where there is little transconductance left for high performance oscillation.
Permeability tuning at the tapped inductor was chosen over capacitance tuning to keep the tank circuit capacitance at a maximum 300pF. This desensitizes the oscillator to changes in parasitic capacitance due to warmup, and to the amplified effect of these capacitances as the tube gain changes (Miller capacitance gain).

2- Suppressor grid G3 modulation with complementary modulation push-pull outputs at the plate and screen grid. The RF phase at the grid and plate match, but the relative amplitude between plate and screen is controlled by the suppressor grid G3. The  6AS6 dual control pentode was chosen because it was designed for easy control of the plate and screen currents by the suppressor grid voltage G3. Other pentodes can be made to work but with different operating voltages at the plate and screen grid G2. Doug Coulter suggested the 6AS6, and it became the ideal choice for single supply operation.

While suppressor grid G3 controls the RF amplitude at the plate and screen, it has negligible effect on the amplitude of the oscillation at the control grid G1 and cathode.

The most important aspect of this modulation approach is that G3 operates in the middle of it's linear range to effect a low distortion two-quadrant multiplication of constant AC cathode current because the output is taken as the difference between plate and screen currents. A zero carrier output occurs when the suppressor grid G3 voltage apportions the plate and screen currents to drive the output transformer with equal amplitude. The AC RF phase of the plate and screen currents is always equal and the same as the cathode current. When the transformer is driven equally by the plate and screen, the net RF output at the transformer secondary is zero. This is how the carrier can be completely suppressed without distortion.

The carrier suppression bias level at suppressor grid G3 will vary from tube to tube and this level is adjustable with an external control to cover this variation. G3 bias is best set more positive than the null point, where the G3 transfer function is most linear. A variation of this bias changes the carrier amplitude, and thus the modulation percentage, for a fixed audio input. The 2Vp-p output from a CD player is enough to get 100% modulation. If the null point is at -6V for a particular tube, G3 is then biased at -5V for 100% modulation. But the output RF voltage could be doubled if the G3 bias were moved to -4V and the modulation input increased to 4Vp-p.

Modern AM radios cope well with high modulation nearing 100%, but pre-war sets were designed for lower modulation levels. 50% modulation sounds nice on my regenerative grid-leak detectors.

(Perhaps having a fixed opposing RF added to a high powered plate modulated class C transmitter was how old tube transmitters could achieve 100% Plate modulation without cutoff distortion. Any historic info on this would be appreciated.)

One way to envisage the two quadrant multiplication of G3xG1, which is output in complementary fashion at screen grid G2 and the Plate, is to think of G3 as one input to a differential pair, with the complementary outputs at screen grid G2 and the Plate. The tail current is provided by the cathode under the control of control grid G1. The sum of plate and screen currents is very independent from the voltage at G3; the sum of the plate and screen currents depends on the control grid G1, which is the other input to this two quadrant multiplier. The Audio modulation signal appears as a differential component at the plate and screen, while the RF oscillation at the cathode appears as a common mode component. The net modulated signal is the differential component of the product of these two signals.


3- Frequency Stability.

There are at least three distinct problems facing frequency stability. One is sensitivity to output loading. Another problem is the real time sensitivity of frequency to the modulating audio signal. This is spurious FM modulation. Yet, another problem is overall frequency stability as a function of tube aging, supply voltage and temperature.

Insensitivity to output loading was achieved with the high fixed 1500pF capacitance presented to the control grid with the 1/3x scaling to the top of the tank circuit. Another measure of insensitivity is given by relegating the oscillator function to the grid and cathode. The plate and screen are not directly involved in the oscillator operation.

Frequency insensitivity to audio modulating signals can be achieved with careful Neutralization, or balancing of Miller capacitance from G2 to G1 and from P to G1 seeks to keep total Miller capacitance feedback gain independent of modulation to minimize or eliminate FM modulation. This is a key problem with single tube transmitters, as Jacob Roschy found from direct experience; see Granco AM modulator.

This trim is presented here only as a suggestion, because the transformer turns ratio yielded a negligible 100Hz of spurious FM modulation. When I tried this neutralization adjustment with 1pF from the plate to the control grid, the spurious FM modulation dropped to 30Hz for a sweep of G3 from 0V to -10V. The un-neutralized insensitivity was somewhat fortuitous because it depends on the layout of the components. I have no Neutralization trim in my final construction because it was not needed. Note that the 10k trim pot replaces the 10k output load. It makes sense that the needed neutralization capacitance would be from plate to G1 because the parasitic from G2 to G1 is necessarily larger than the parasitic capacitance from the plate to G1.

The 10K output load bounds the output amplitude to about 1Vp-p if the suppressor grid G3 is biased between 0V and the suppressed carrier bias level. This relatively modest output swing helps to keep the frequency insensitive to antenna loading and audio modulation.

The last frequency stability problem is long term insensitivity to temperature, tube gain and supply voltage. The large fixed capacitance presented to the grid also helps here, but the key aspect of the design that insures insensitivity to these environmental parameters is the AGC regulation of the class A oscillator loop gain of 1. This high gain AGC loop forces the cathode resistance to 600Ω as described above. This means that the tube may loose gain over time, or the supply voltage may vary, or the tube may warm up slowly over time, but it's gain remains fixed by the AGC loop. These environmental inputs simply result in a variation of the AGC bias level and total RF output. Keeping the tube running at a fixed gain eliminates Miller gain variations because the Miller gain is now fixed by the AGC loop.

The gain of the AGC loop in this oscillator is a direct function of the linearity of the AC gain. In other words, with an ideal linear AC loop gain, the AGC gain loop gain is infinite. This is a direct result of the requirement for stable oscillation amplitude the loop gain of linear system be exactly one. Even a small amount of non-linearity greatly reduces this AGC loop gain because now it is possible for parts of the wave to operate at a loop gain higher than 1, while the rest of the wave operates at a gain that is lower than one. This real time variation of gain, which is a way of describing the cause of distortion, changes the frequency. Under these conditions a variation in temperature, supply voltage, or tube gain also results in a variation in frequency, in addition to the variation in amplitude. (This problem was at one time found by Dr. Bernard (Barney) Oliver of HP in the original founding product at HP, the Wien bridge sine wave RC oscillator. It was found that when the oscillators were too linear, with a distortion well below 0.01%, the light bulb stabilization AGC loop would become unstable from excessive loop gain. Bringing the distortion up to 0.01% was enough for AGC loop stability. I first learned of this in a riveting presentation by analog guru and staff scientist at Linear Technology, Jim Williams. Oliver published a formal paper on the subject)

One merit of the  High Performance modulator with Resistance stabilized oscillator posted by Prof. Dietmar Rudolph is that the real time variation of the AC loop gain is kept low with the careful selection of a gain setting resistor. The real time stabilization of the oscillation amplitude of this loop still relies on a necessary level of distortion because there is no other way to change loop gain with a bias level. A conventional Class C oscillator also gets all of it's amplitude regulation from very wild real time variations of AC loop gain, but the use of a high Q tuned circuit, with less than a 1% loss per cycle (Q=100) can still achieve adequate stability despite the wild changes in real time gain because most of the time, the tube is cut off in class C and the period is defined by L and C.

The constructed transmitter

The power supply is based on a dual primary 120VAC mains transformer with two 6.3VAC secondaries (Signal Transformer DPC-12-2000). The original purpose of the dual primary configuration was for operation of the 6.3VAC secondaries from a 120VAC or 240VAC mains power. My local power is 120VAC, thus leaving a spare winding for the high voltage. The total available power using only one of the 120VAC windings as the primary, instead of both windings in parallel, reduces the maximum deliverable power because of higher Ohmic losses, but the total power required of this circuit is much less than what the transformer was designed to deliver. So much so, that I had to include a 5.1Ω resistor to drop the lightly loaded secondary to deliver 6.3VAC at 150mA to the 6AS6. R8,R9,C19,C10 eliminated hum injection and cathode RF leakage. I used the spare 6.3VAC winding to generate the variable 0 to -10V suppressor grid G3 bias, but the heater winding might have been used instead. I always include an appropriately sized fuse in all my constructions. This 64mA fuse is just slightly higher than the 50mA draw of the transmitter.

The transmitter was constructed in a Campbell's Tomato Soup can with built-in power supply. The paper label is an original 2010 edition, and not a fine art reproduction from 1968.

I opened the can lid with a side-cutting can opener that keeps the lid reclosable. I ate the soup before continuing. Four 4-40 screws close the lid.
The control knob on the right sets the bias voltage at suppressor grid G3 to control the output carrier amplitude. A little red 12V incandescent bulb just above the knob serves as knob marker and indicates that the transmitter is "ON THE AIR". The standard mini phone jack on the right connects directly to a CD player or MP3 player. A sub-mini jack is needed for the smallest MP3 players, and RCA or DIN plugs are best to connect to home stereo gear. The power switch is directly above the white power cord. The removable 3 foot rod antenna is held on the left with RCA jacks. The upper jack delivers the modulated RF, and the lower jack is just a support. 
The long multi-turn set screw in front of the tube moves the ferrite core in the main tuning coil to set the operating frequency.  The coil was mounted in a hole that was drilled where he pull tab was originally attached.
The RF side of the transmitter is entirely mounted on the lid, while the power supply is mounted at the bottom of the can, with the individual caps, diodes, and resistors mounted directly on the transformer terminals.
Note the hand-wound output RF transformer with it's Litz wire reaching to the output RCA jack. The construction of this transformer is not very critical.
During development, I used a toroidal 1:1 transformer that is normally found as the mains filter in a PC power power supply. You could use one of these too if you add an extra winding for the output and double the number of turns for the plate side of the two existing windings.
The main tuning coil in the foreground was originally a stock coil with a single winding. I added the enameled wire layer over the original coil for the lower half of the tapped coil. The coil terminals are wired to the tube socket terminals with stiff wires made from paper clips to eliminate frequency drift that would be caused by wobble in the tuning coil.
The can is wired to the circuit ground and includes a 1500pF cap to AC-couple the can to the ground lug in the polarized power plug. Grounding the can was necessary to eliminate hum pickup.
The choice of antenna is dictated by the type of AM radio that will receive the signal. Radios with loop antennas will work well with a transmitting loop as shown below, while radios with external wire antennas will work better with the Electric field rod shown below. These choices are elaborated further in a post exploring personal AM transmitter antennas. Jacob Roshy's post on AM and MW home transmitters also elaborates on antenna choices.
The loaded Q=10kΩ/400Ω @720khz=25 for this 89uH loop has a negligible effect on transmitted bandwidth 720kHz/25=28kHz, which results in a 15kHz detected audio bandwidth. The 10k is the transmitter impedance, and 400Ω is the reactance at 720kHz.
When choosing a transmission frequency, look for an empty channel. However this is not as obvious as it seems. If you hear a whistle in the radio when you tune your Tomato-Soup personal AM transmitter, that is a sign that there is a weak carrier in an adjacent channel that could not be detected until the carrier of your transmitter made detection of the weak carrier possible. If you hear a whistle, move the transmitter to another frequency. This will also help assure unwanted interference of your transmission to radios in your neighbor's house. He may hear your transmission weakly, but not on top of his program.

Measured results

This first set of photos shows the RF output for three bias levels (-9.5V, -5.9V, -0.5V) at the suppressor grid G3. The scope sync is derived from the unmodulated cathode voltage. The middle photo shows the 720kHz carrier completely suppressed. The two scope photos for bias levels on either side of the null point shows the phase reversal that is expected from the two quadrant multiplication between G3 and G1. The first three scope photos were taken with the output under a wide band 10kΩ load.



Note that when the suppressor grid G3 modulation input is biased to get zero RF carrier at the balanced null point, there is some residual second harmonic that is at least 20dB down from the full carrier. Loading the 10k secondary impedance with a tuned 89uH transmission loop reduces this second harmonic to the point of undetectability as seen in the photo on the left.

Scope photos of modulation in real time with 1kHz and 10kHz audio signals. Time plots and trapezoidal modulation XY plots demonstrate high linearity and the capability of suppressed carrier modulation.

The photo on the left was taken with all scope knobs in the calibrated position to represent the absolute levels being measured at the RF output and the audio input.
The first 4 photos show a 1kHz audio signal and the last photo shows a 10kHz audio signal.
These last two scope photos show the static relationship between plate and screen grid G2 currents as a function of suppressor grid G3.

 Parametric sweeps


Pentode to pentode variations

The following plots summarize operational results for 14 dual control pentodes of different brands and types, 11 of which were the 6SA6, one was the 6HZ6 and two were the 6GY6/6GX6. The pentodes are serialized from #25 to #38.

I conducted this evaluation of design robustness because of the well founded concern from Felix Schaffhauser over parametric variations among tubes.

This survey resulted in the adjustment of the turns ratio of the push-pull output transformer load for Plate and Screen from 1:1 center tap to 2:1. The reason for this shift is that it was not possible to balance the plate and screen outputs of all tubes with a centered tap, but it was with a 2:1 turns ratio favoring the plate.
This reduced screen swing by 4, and increase plate swing by 4. This had a very fortuitous effect on Miller capacitance, and spurious FM modulation was reduced from 1kHz to 250Hz as G3 varies from 0V to plate cutoff.

If I increased the supply voltage to the plate, while operating with a 1:1 center tapped transformer primary, I would have increased the plate output, as an alternative to using the 2:1 primary tap favoring the plate. But separate supplies and windings for plate and screen are inconvenient.



Final thoughts, going further and acknowledgments.

When G3 is set for a null RF carrier output, the AM detected sound is like the "Donald Duck" sound you might hear in the SW bands, coming from SSB ham operators.

Adjusting G3 with the knob is easily done by ear, by first listening for the RF null point on the knob with the characteristic "Donald Duck" sound, then easing back the bias to 50% of that setting. If more modulation is desired, the bias can be moved closer to the carrier null point just before clipping can be heard. G3 works as a control for the null carrier output and as a modulation control. The further from the null point, the lower the modulation percentage for a given audio input.

The output can drive a loop for reception on radios with loop antennas, or the 3-foot metal rod for radios with wire antennas that are sensitive to the electric field. The 3 foot rod and loop antenna work well together because the loop antenna tunes out some of the output parasitic capacitance.

The final listening tests sound absolutely sweet. It has made me want to check the alignment of a couple of my AM radios that show very little treble response. Perhaps I will cheat and add some over-coupling capacitance from primary to secondary of the last IF transformer to peak the treble response, while moving the AGC detector to the primary side to retain monotonic AGC response to the carrier.
Additional tube types that can be tried are sharp cutoff heptodes like the 6BY6, beam deflecting modulators like the 6ME8, 6JH8, 6AR8, and Compactron triple triodes like the 6AC10. The Beam Deflectors are ideally suited for this topology of single tube transmitter. The only down-side is that they would be used exactly as intended, and all the fun of subverting pentode operation would be gone. A much more worthy use of a beam deflecting tube like the 6JH8, would be Robert Weaver's single tube direct conversion receiver (homodyne or synchronous detection) or his one tube double reflex superhet receiver.
There is a list of suitable dual control pentodes at the end of the post on dual control pentodes.
Gate beam tubes like the 6BN6 are related to dual control pentodes, but their low gm1 makes then unsuitable to work as stable oscillators in this topology.
The list of dual control pentodes includes a number of GE Compactrons, like the 6AL11, which include a high transconductance dual control pentode and an audio power pentode. A higher power version of this transmitter could be built with one of these Compactrons to broadcast 1W with the dual control pentode wired as discussed in this post, and the power pentode serving as a linear class A output amplifier.

I learned a lot from experimentation with Professor Pentode, however, much of what I know about neutralization and spurious FM modulation in AM transmitters was learned in the RMORG Forum.
I had this concept in my head for a while. It was our own Professor Dietmar Rudolph who sparked the realization of this transmitter when we discussed the operation of dual control pentodes like the 6DT6 used for FM sound detection in American TV sets of the 1950's and 1960's.
The development of this transmitter was first shared via email with a number of very supportive tube enthusiasts. One of whom, was Barrie Gilbert. The roaring fire that Barrie lit under my butt with his ardent enthusiasm helped too!

This transmitter is dedicated to all of you.

Comments are welcomed, in particular, about any results from similar circuits or your built version of this circuit. It would be great to hear about a high power (1W) version of this transmitter with one of the GE Compactrons.

Best regards,



To thank the Author because you find the post helpful or well done.

Excellent article 
19.Nov.10 10:12
401 from 46707

Felix Schaffhauser (CH)
Articles: 189
Count of Thanks: 28
Felix Schaffhauser

A big "thanks" to Joe Sousa for this excellent article about this HI-FI AM transmitter. After a lot of investigations, calculations and practical constructions Joe has come up with a very good concept of such an AM modulator. It is good to see, that with just one multigrid tube (here a Pentode, although a little bit special) a practical "Heimsenderlein" can be realized. I am sure, many of us will be happy to use his thoughts and recommendations for constructing a similar piece of  hardware and appreciate the many considerations given in order to come up with own solutions.

I personally think, that there are so many nice tubes around that it will make sense to apply Joe's ideas to them and experiment with it and hopefully share the results with the community.

I hope that we can enjoy many such articles of Joe in our forum in the future.

Felix Schaffhauser

To thank the Author because you find the post helpful or well done.

21.Nov.10 09:57
523 from 46707

Bob Isaac (USA)
Articles: 108
Count of Thanks: 57
Bob Isaac

Wow! This is a really intersting and thorough article. You da man, Joe!

To thank the Author because you find the post helpful or well done.