Doherty type ham band transmitter

[N0BST]

I've been toying with the idea of building a higher power AM rig, 250-350W bracket.  I'd really like a tube design so PDM is pretty much out.  I know it can be done with tubes but the thought of dealing with 4-6KV doesn't appeal to me.  I thought about conventional plate modulation, but good mod iron is getting hard to find and expensive when I do run across it.  Plus my other considerations are size and weight as I'd like it to be at least somewhat portable.  The only modulation methods which don't use big iron are grid (can't modulate enough), Ampliphase (too complicated) and the Doherty amp.  Has anyone designed a Doherty for the ham bands and what were the drawbacks?

I was thinking of using 4CX250Bs for the PA tubes.  They're nice and compact, will create the exact amount of power I need, plus a Doherty would use only two vs. three or four in a conventional plate modulated rig.  I've got a manual for a Western Electric 443A-1 kilowatt broadcast transmitter so I have at least a passing acquaintance with how it's done, though I must confess I don't have the raw engineering smarts to pull it off myself.

Scott Todd

[W7TFO]

I've seen a few ham TX built on that principal over the years.  The only drawback I see is it is a monobander, and moving off the sweet spot more than a few KHz causes the phase relationship to go off and the final product suffers.

A ham TX will by necessity have quite a few 'tweaker' controls in the 90 degree networks.

It is one of the few transmitters tuned with an XY scope rather than plate E & I meters.

Keep in mind it is just a linear amp, with low-level grid modulation ahead, but a linear with around 60% efficiency...far better than class B topology, and from power line to output on par with high-level plate modulation rigs.

73DG

[WQ9E]

Dennis nailed the biggest problem of it being a very frequency sensitive design.  It is an interesting concept but unless you plan to lock it on 3880/3885 it isn't very practical given the amount of work, components, and of course money to build what is pretty much a one trick pony.

If you don't want to deal with a big modulation transformer a very practical route is to build a linear amplifier designed with tubes, power supply, and cooling system to perform reliably while running legal limit AM.  If it will stand up to AM you will also have an amp which will work fine for SSB/CW/digital modes and providing for rapid and easy frequency change won't be a problem.  A well built amplifier coupled with a Ranger, AF-67/68 or similar will provide a solid and good sounding AM signal.

Several years ago I built an amplifier mostly for contest use but also with the intent to easily run the full limit on AM without strain.  It is built into a rack cabinet using three 4CX800 tubes with plenty of cooling air and power is supplied by a large Dahl transformer.  The final tank uses a 5KW rated roller inductor with a vacuum tune capacitor.  A set of G3SEK tetrode boards greatly simplify the control and protection system while also providing regulated screen and control grid voltage.

I have several high power AM rigs (Johnson Desk KW, BC-610, Johnson 500, along with a Gates BC-250GY which is awaiting restoration) so at this point the amp doesn't get used much for AM.  I do use it on 160 AM a bit and with its simple passive grid input 25 watts of RF will drive it to the legal limit allowing "entry level" AM rigs to easily drive it to full power.

[kb3ouk]

With the 4CX250Bs, you could build a PDM rig and not have to run an extremely high voltage. Use 2 tubes in parallel for the finals with 1000 volts on them. That would give you about 290 watts of output. You would have to run at least 2000 volts for the power supply voltage, 2500 volts would give you some headroom. That's gonna be your next closest alternative to plate modulation.

[WBear2GCR]

so, you want to avoid the heavy mod iron/heising choke? Is that the idea?

"Portable"??

At 300 watts I think ur stretching the meaning of the term as we think of it today.
According to the Miltary the BC-610 and the T-368 were "portable" transmitters.

             _-_-

[steve_qix]

I've played around with the phasing techniques used in the Doherty amps back in the 70's.  Very tricky at these frequencies - even on the broadcast band they're tough to tune up.

The PWM is certainly the most practical non-modulation transformer idea if you want efficiency.

Here's another idea that needs some experimenters - outphasing modulation.  This works and isn't that hard to do.  An outphasing transmitter I designed and built in high school won me a scholarship to college 

Anyway, how it works is this:  You have two identical phase modulated RF paths operating from the same oscillator.  You have two drivers, two RF amplifiers, and two PI networks looking into a single loading capacitor.  You have RF phase shifting networks to adjust the phase shift between the RF amplifiers.  This is easy - it's just a tuned circuit a little out of resonance.  It will either be capacitive or inductive, and you will get a leading or lagging phase shift.  Everything starts off 180 degrees out of phase, and you move each RF path 22.5 degrees (one leading, one lagging) to get your 135 degree phase difference for your carrier.

So, the RF amplifiers are operated 135 degrees out of phase with each other at carrier.  At 100% modulation, they are 90 degrees out of phase, and of course at 100% negative modulation they are 180 degrees out of phase.  This works because the sine function is fairly linear over the phase shifts described.  You could improve the system by using a "remodulator".  A remodulator is a highly linear (both in phase and amplitude) detector feeding demodulated audio back into the audio stages ahead of the two phase modulators.

The nice thing about this is that nothing is super-critical as long as the phase modulators are linear and everything is pretty much the same on both halves of the system.  The phase differences don't have to be exact at all, and you can move around the band with little effort.

Just a thought!

[AB2EZ]

Steve

Neat!

This is the approach that is sometimes used to "externally modulate" the output of a laser in an optical communication link. The approach is used in systems that employ lasers that cannot be modulated by simply turning the laser on and off (direct modulation).

In that case, the laser output is split into two equal parts using a "beam splitter"; and one or both parts are phase modulated (or polarization modulated). The two parts (beams) are then recombined using a "beam combiner" of one type or another.

Example: a helium-neon laser operating at a wavelength of 0.00000063 meters (a frequency of 476,000 GHz) on-off modulated at a data rate of 10 billion bits per second... for a free space communication link between two satellites.

This approach is also used in sensors, where the parameter being sensed modulates the phase of the optical signal in one of the two paths... to produce amplitude modulation when the optical signals are recombined. For example, to link a towed array of pressure sensors to an array of receivers in a hydrophone system.

The modulator is called a Mach Zender interferometer.

Both the beam splitter and the beam combiner are the optical frequency version of a 3dB RF directional coupler

Stu

[w4bfs]

on k9qi website he has described a Doherty-Sainton 300W  tx using 4e27a / 803 pentodes ... he eliminates one of the touchy 90 deg phase shifts by using a Johnson counter at low level and buffering the outputs .... I emailed him with no answer

Steve ...in the low level out phasing technique how many of the required phase shifts could be done at low level by direct synthesis ... I shared a patent in 1980 on Walsh function work with k4lkq (sk) ... John

[Steve - K4HX]

There was a design for one in QST sometime in the 50s.

[W3GMS]

I've played around with the phasing techniques used in the Doherty amps back in the 70's.  Very tricky at these frequencies - even on the broadcast band they're tough to tune up.
The PWM is certainly the most practical non-modulation transformer idea if you want efficiency.
Here's another idea that needs some experimenters - outphasing modulation.  This works and isn't that hard to do.  An outphasing transmitter I designed and built in high school won me a scholarship to college 
Anyway, how it works is this:  You have two identical phase modulated RF paths operating from the same oscillator.  You have two drivers, two RF amplifiers, and two PI networks looking into a single loading capacitor.  You have RF phase shifting networks to adjust the phase shift between the RF amplifiers.  This is easy - it's just a tuned circuit a little out of resonance.  It will either be capacitive or inductive, and you will get a leading or lagging phase shift.  Everything starts off 180 degrees out of phase, and you move each RF path 22.5 degrees (one leading, one lagging) to get your 135 degree phase difference for your carrier.
So, the RF amplifiers are operated 135 degrees out of phase with each other at carrier.  At 100% modulation, they are 90 degrees out of phase, and of course at 100% negative modulation they are 180 degrees out of phase.  This works because the sine function is fairly linear over the phase shifts described.  You could improve the system by using a "remodulator".  A remodulator is a highly linear (both in phase and amplitude) detector feeding demodulated audio back into the audio stages ahead of the two phase modulators.
The nice thing about this is that nothing is super-critical as long as the phase modulators are linear and everything is pretty much the same on both halves of the system.  The phase differences don't have to be exact at all, and you can move around the band with little effort.
Just a thought!

Steve,

With the scheme you built starting with Phase Modulation, did you take into account the natural per-emphasis that goes along with Phase Modulation?  So if you did, I am assuming the some form of de-emphasis is inserted in the audio path.  Very interesting concept and I enjoyed reading your explanation.   Very impressive, especially for a H.S. kid!!

Joe, W3GMS   

[AB2EZ]

Steve

Thinking about this some more:

You combine the two RF signals by using "two pi networks looking into a single loading capacitor". I think there is an additional passive combining network that you didn't mention in your post... but in any event:

Using some simplified analysis, it appears (to me) that:

1. The electrical input power to each of the two amplifiers is essentially constant whether the transmitter's net RF output power is at carrier level, or at a 100% positive modulation peak (4x the carrier power level), or at a 100% negative modulation peak, or anything in between.

2. On 100% negative modulation peaks (zero RF voltage at the antenna port), each amplifier's output pi network will be producing zero net RF output power, averaged over each RF cycle. This may be because the impedance of the load on the plate of each amplifier's output tube (at the fundamental frequency) is zero or because it is purely reactive. If the impedance of the load on the plate of each amplifier's output tube, at the fundamental frequency, is purely reactive... then the load will be absorbing power from the amplifier for half of each cycle of the fundamental RF frequency, and returning power to the amplifier in the other half of each cycle. On 100% negative modulation peaks of the transmitter's output, each amplifier will be operating at essentially 0% efficiency; even though the electrical input power is the same as it is at carrier or at 100% positive modulation peaks, or anything in between.

3, At carrier, each amplifier will be operating at 25% of the efficiency it achieves on 100% modulation peaks....because, again, it is operating with about the same electrical input power it operates with on 100% positive modulation peaks of the transmitter's RF output.

4. Therefore, the amplifiers will dissipate significantly more power than they would with screen modulation.

[Note that with screen modulation, the RF output power at carrier is 25% of the RF output power on 100% modulation peaks... but the electrical input power is only 50% of the electrical input power on 100% modulation peaks. Therefore, with screen modulation, the amplifier operates with about 50% (not 25%) of the efficiency that it achieves on 100% modulation peaks]

If the combining were done with a 3db directional coupler, with one of the coupler's output ports going to the transmitter's antenna output port, and the other coupler output port going to a dummy load... then on 100 percent negative peaks, all of the RF power being produced by the two amplifiers would flow into the dummy load. Each amplifier will operate at maximum efficiency with any modulation level. At carrier, 75% of the total RF power being produced by the amplifiers will be absorbed by the dummy load.

The total electrical input power and the total RF output power would be the same, but the inefficiency associated with the modulation would heat up the dummy load instead of the output tubes.

Stu

I've played around with the phasing techniques used in the Doherty amps back in the 70's.  Very tricky at these frequencies - even on the broadcast band they're tough to tune up.

The PWM is certainly the most practical non-modulation transformer idea if you want efficiency.

Here's another idea that needs some experimenters - outphasing modulation.  This works and isn't that hard to do.  An outphasing transmitter I designed and built in high school won me a scholarship to college  

Anyway, how it works is this:  You have two identical phase modulated RF paths operating from the same oscillator.  You have two drivers, two RF amplifiers, and two PI networks looking into a single loading capacitor.  You have RF phase shifting networks to adjust the phase shift between the RF amplifiers.  This is easy - it's just a tuned circuit a little out of resonance.  It will either be capacitive or inductive, and you will get a leading or lagging phase shift.  Everything starts off 180 degrees out of phase, and you move each RF path 22.5 degrees (one leading, one lagging) to get your 135 degree phase difference for your carrier.

So, the RF amplifiers are operated 135 degrees out of phase with each other at carrier.  At 100% modulation, they are 90 degrees out of phase, and of course at 100% negative modulation they are 180 degrees out of phase.  This works because the sine function is fairly linear over the phase shifts described.  You could improve the system by using a "remodulator".  A remodulator is a highly linear (both in phase and amplitude) detector feeding demodulated audio back into the audio stages ahead of the two phase modulators.

The nice thing about this is that nothing is super-critical as long as the phase modulators are linear and everything is pretty much the same on both halves of the system.  The phase differences don't have to be exact at all, and you can move around the band with little effort.

Just a thought!

[AB2EZ]

Joe

The effect that you refer to as "natural pre-emphasis" will not be associated with this modulation method. The phase is being modulated; and the modulation of the phase is being converted to amplitude modulation by a phase-sensitive combining process.

The pre-emphasis / de-emphasis concept arises in communication links that use FM. Since phase is the time integral of frequency, the phase of the transmitted signal in an FM system varies by a larger amount when the audio modulating signal is at a low audio frequency v. a high audio frequency. At the receiver of an FM link, the additive noise affects the fidelity with which one can recover the phase of the incoming FM signal. Therefore, the larger the phase deviation (due to the FM modulation)... the higher the SNR in the recovered audio. Since lower frequency components of the modulating audio signal produce a larger phase deviation, the lower frequency components of the modulating audio signal will be recovered (at the receiver) with a higher SNR. To fix this, one can pre-emphasize the higher audio frequencies of the modulating audio signal at the input of the FM link to enhance the SNR at the receiver output for those pre-emphasized frequencies. One can then deemphasize the recovered output from the FM receiver to flatten the end-to-end frequency response. The net result of the  pre-emphasis done at the input to the transmitter's FM modulator, and the complementary de-emphasis done at the output of the FM receiver's demodulator is to produce the same SNR for all audio frequencies passing through the link.

I agree that this is very impressive (actually amazing) work for a high school student.

Stu

I've played around with the phasing techniques used in the Doherty amps back in the 70's.  Very tricky at these frequencies - even on the broadcast band they're tough to tune up.
The PWM is certainly the most practical non-modulation transformer idea if you want efficiency.ed
Here's another idea that needs some experimenters - outphasing modulation.  This works and isn't that hard to do.  An outphasing transmitter I designed and built in high school won me a scholarship to college  
Anyway, how it works is this:  You have two identical phase modulated RF paths operating from the same oscillator.  You have two drivers, two RF amplifiers, and two PI networks looking into a single loading capacitor.  You have RF phase shifting networks to adjust the phase shift between the RF amplifiers.  This is easy - it's just a tuned circuit a little out of resonance.  It will either be capacitive or inductive, and you will get a leading or lagging phase shift.  Everything starts off 180 degrees out of phase, and you move each RF path 22.5 degrees (one leading, one lagging) to get your 135 degree phase difference for your carrier.
So, the RF amplifiers are operated 135 degrees out of phase with each other at carrier.  At 100% modulation, they are 90 degrees out of phase, and of course at 100% negative modulation they are 180 degrees out of phase.  This works because the sine function is fairly linear over the phase shifts described.  You could improve the system by using a "remodulator".  A remodulator is a highly linear (both in phase and amplitude) detector feeding demodulated audio back into the audio stages ahead of the two phase modulators.
The nice thing about this is that nothing is super-critical as long as the phase modulators are linear and everything is pretty much the same on both halves of the system.  The phase differences don't have to be exact at all, and you can move around the band with little effort.
Just a thought!

Steve,

With the scheme you built starting with Phase Modulation, did you take into account the natural per-emphasis that goes along with Phase Modulation?  So if you did, I am assuming the some form of de-emphasis is inserted in the audio path.  Very interesting concept and I enjoyed reading your explanation.   Very impressive, especially for a H.S. kid!!

Joe, W3GMS  

[W3GMS]

Joe

The effect that you refer to as "natural pre-emphasis" will not be associated with this modulation method. The phase is being modulated; and the modulation of the phase is being converted to amplitude modulation by a phase-sensitive combining process.

The pre-emphasis / de-emphasis concept arises in communication links that use FM. Since phase is the time integral of frequency, the phase of the transmitted signal in an FM system varies by a larger amount when the audio modulating signal is at a low audio frequency v. a high audio frequency. At the receiver of an FM link, the additive noise affects the fidelity with which one can recover the phase of the incoming FM signal. Therefore, the larger the phase deviation (due to the FM modulation)... the higher the SNR in the recovered audio. Since lower frequency components of the modulating audio signal produce a larger phase deviation, the lower frequency components of the modulating audio signal will be recovered (at the receiver) with a higher SNR. To fix this, one can pre-emphasize the higher audio frequencies of the modulating audio signal at the input of the FM link to enhance the SNR at the receiver output for those pre-emphasized frequencies. One can then deemphasize the recovered output from the FM receiver to flatten the end-to-end frequency response.

I agree that this is very impressive (actually amazing) work for a high school student.

Stu

I've played around with the phasing techniques used in the Doherty amps back in the 70's.  Very tricky at these frequencies - even on the broadcast band they're tough to tune up.
The PWM is certainly the most practical non-modulation transformer idea if you want efficiency.ed
Here's another idea that needs some experimenters - outphasing modulation.  This works and isn't that hard to do.  An outphasing transmitter I designed and built in high school won me a scholarship to college  
Anyway, how it works is this:  You have two identical phase modulated RF paths operating from the same oscillator.  You have two drivers, two RF amplifiers, and two PI networks looking into a single loading capacitor.  You have RF phase shifting networks to adjust the phase shift between the RF amplifiers.  This is easy - it's just a tuned circuit a little out of resonance.  It will either be capacitive or inductive, and you will get a leading or lagging phase shift.  Everything starts off 180 degrees out of phase, and you move each RF path 22.5 degrees (one leading, one lagging) to get your 135 degree phase difference for your carrier.
So, the RF amplifiers are operated 135 degrees out of phase with each other at carrier.  At 100% modulation, they are 90 degrees out of phase, and of course at 100% negative modulation they are 180 degrees out of phase.  This works because the sine function is fairly linear over the phase shifts described.  You could improve the system by using a "remodulator".  A remodulator is a highly linear (both in phase and amplitude) detector feeding demodulated audio back into the audio stages ahead of the two phase modulators.
The nice thing about this is that nothing is super-critical as long as the phase modulators are linear and everything is pretty much the same on both halves of the system.  The phase differences don't have to be exact at all, and you can move around the band with little effort.
Just a thought!

Steve,

With the scheme you built starting with Phase Modulation, did you take into account the natural per-emphasis that goes along with Phase Modulation?  So if you did, I am assuming the some form of de-emphasis is inserted in the audio path.  Very interesting concept and I enjoyed reading your explanation.   Very impressive, especially for a H.S. kid!!

Joe, W3GMS  

Yep, your right Stu.  I was approaching this from a FM vs Phase modulated topologies for FM type of transmissions.  I have designed both and quickly became familiar with the no pre-emphasis required for modulation schemes that use phase modulation topology.  I always preferred the direct FM varactor schemes which do require pre-emphasis and the corresponding de-emphasis to get back to "flat audio".  Of coarse we all know the reasons why that is done for FM transmissions in relationship to the SNR ratio. 

Joe, GMS

[W4EWH]

Steve,

With the scheme you built starting with Phase Modulation, did you take into account the natural pre-emphasis that goes along with Phase Modulation?  So if you did, I am assuming the some form of de-emphasis is inserted in the audio path.  Very interesting concept and I enjoyed reading your explanation.   Very impressive, especially for a H.S. kid!!

Here's an interesting question: the FCC specifies a 75 microsecond pre-emphasis for FM transmission in the 88-108 broadcast band. Since phase modulation produces a 75 microsecond preemphasis by it's nature, I wonder: did the FCC adopt the rule because Major Armstrong was already using phase modulation?

73,

Bill, W1AC

[steve_qix]

Hi Stu, I'm trying to remember... and I don't remember exactly the output network, but I don't remember a coupler of any kind either.  It was fairly simple.

Somewhere, I still have my original (from 1973) notes on the transmitter, and the write up I did for the science fair.

A couple of people here saw the transmitter - noteably, Chuck K1KW (then WA1EKV) and Tim WA1HLR.

I found a very basic description in some old book I got from the library, and immediately set out to build one.

The transmitter lived for about 3 years after the science fair, but ultimately (and unfortunately) I recycled the parts and the chassis into something else that is most likely still on the air today!

There are some variations on the theme that I would like to try someday.  Notably, to combine the RF amplifiers across a transformer primary of some kind.  So, in this case the RF amplifiers are operated IN PHASE with each other (resulting in no output), and each RF amplifier is then phase shifted 22.5 degrees (in opposite directions).  This will give a 45 degree phase difference across the primary of the RF output transformer or parallel tuned network.  At 100% positive modulation, the difference will be 90 degrees, and of course at 100% negative modulation, the difference will be 0 degrees.

If you used a DSP pre-distorter, you could get a wider phase shift swing, or operate closer to 90 degrees out of phase at carrier (maybe 60 degrees or so) by pre-distorting the audio to compensate for the sine wave non-linearity that comes about as one approaches 90 degrees.

[AB2EZ]

Steve

Okay...

Fond memories of past projects!

Stu

[steve_qix]

Stu, any thoughts about combining across a balanced load?  In phase = 0 output (100% neg), 90 degrees out = 100% positive, 45 degrees = carrier.

Regards, Steve

[AB2EZ]

Steve

With two amplifiers, each with its own output pi network, feeding into opposite sides of a balanced load:

The efficiency will still be what I see as the biggest problem.

The amplitude of the modulated RF output signal will be (as you point out):

A = b x sin (c/2)

Where b is a constant, and c is the phase difference between the two pi network outputs (in radians)

With c = pi/4 (i.e. 45 degrees) at carrier, sin(c/2) = sin(pi/8) = 0.383

With c = pi/2 (i.e. 90 degrees) on "100%" positive modulation peaks, sin(c/2) = 0.707

So, as you point out, the modulation will be roughly linear. I.e. 2 x 0.383 = 0.766

But the output power at carrier will be only 0.383 x 0.383 = 0.146 x the sum of the maximum output powers of the two amplifiers. Meanwhile, each amplifier will be consuming the same electrical input power that it consumes at maximum output.

Therefore, the efficiency is very low.

If you know that you are never going to operate with c greater than pi/2, then you can optimize the loading to obtain an RF output at carrier that is 0.25 x the sum of the maximum output powers of the two amplifiers... but you will still have about half the efficiency, at carrier, that you could get with screen modulation

Stu

[pa0ast]

May be an  interesting simple diagram..

[Tim WA1HnyLR]

Hmmmmm I say,
I have been following this with great interest. In the 90s while working for Armstrong Transmitter I worked on converting an  early Continental  50 Kw Doherty transmitter to short wave. The transmitter used a pair of 6697 medium Mu triodes. I did not understand how the Doherty system functions . I tried to get my head around it. I was suffering from mental constipation. I stared at it for a long time. I fired up the 811 rig. After the smoke cleared it all made sense. It is a game of impedance matching two tubes in linear class B mode with a common power supply. The Carrier tube and the peak tube. The carrier tube is loaded lightly to get best plate efficiency , but has no headroom for positive peak modulation. The Peak tube is working into an impedance that will yield full 100% positive peak modulation. The peak tube is biased at the edge of cutoff with the normal level of excitation being applied to the carrier tube. The other major part of the equation is coming up with the impedance matching to make it all happen. This is accomplished by tuned circuits between the two tubes. In the process, the phase is shifted. The rest of the equation returning both the RF waveforms from the carrier and peak tube to be in the same phase at the outpoot. There are other variants of this system whicH I will address at another time. I have commitments and have to leave my post now,,,,,, To be continued, Tim WA1HnyLR

[N6YW]

Dennis nailed the biggest problem of it being a very frequency sensitive design.  It is an interesting concept but unless you plan to lock it on 3880/3885 it isn't very practical given the amount of work, components, and of course money to build what is pretty much a one trick pony.

If you don't want to deal with a big modulation transformer a very practical route is to build a linear amplifier designed with tubes, power supply, and cooling system to perform reliably while running legal limit AM.  If it will stand up to AM you will also have an amp which will work fine for SSB/CW/digital modes and providing for rapid and easy frequency change won't be a problem.  A well built amplifier coupled with a Ranger, AF-67/68 or similar will provide a solid and good sounding AM signal.

Several years ago I built an amplifier mostly for contest use but also with the intent to easily run the full limit on AM without strain.  It is built into a rack cabinet using three 4CX800 tubes with plenty of cooling air and power is supplied by a large Dahl transformer.  The final tank uses a 5KW rated roller inductor with a vacuum tune capacitor.  A set of G3SEK tetrode boards greatly simplify the control and protection system while also providing regulated screen and control grid voltage.

I have several high power AM rigs (Johnson Desk KW, BC-610, Johnson 500, along with a Gates BC-250GY which is awaiting restoration) so at this point the amp doesn't get used much for AM.  I do use it on 160 AM a bit and with its simple passive grid input 25 watts of RF will drive it to the legal limit allowing "entry level" AM rigs to easily drive it to full power.

Rodger
Excellent advice and it relates to almost what I have here. For those of you who have subscribed to Electric Radio for a long time may remember Cliff Kurtz N6ZU. He built 263 amplifiers of various types. I own his last amp, "The Final Final" which incorporates a 4-1000A running 5KV to the plate, has band switching with tuned input and uses heavy duty components including a vacuum variable load capacitor.
When the ART-13 isn't enough muscle to get through, I use the Flex 5000A running about 13 watts of carrier into the amp. That gives me 325 watts of carrier from the linear and I easily modulate around 1200 watts with excellent audio. In this scenario, old meeting new isn't a bad thing at all although some will argue it's not a "pure" way of doing AM. The guy on the other end surely doesn't know this because of the audio it produces, which IS processed outside of the Flex rig.
This might go against your wishes of dealing with very high DC voltage and I don't blame you. However, there is a lot to be said for using the 4-1000A tube. It's powerful and can take quite a beating but it can also produce a tremendous amount of power for a one lung affair. I have been quite happy with it.
Just throwing it out there...

[k9qi]

I realize this is an old post.  Everyone has probably moved on.  First, I saw that one of my Doherty transmitter approaches was referenced; not sure why I didn't receive the email but all my email forwards to my work account and our filters tend to be quite heavy these days.  Sorry for not responding to the email; if I had seen it, I certainly would have.

I have built several Doherty amplifiers of "both" types.  Type 1 is the classic Doherty design wherein a 2:1 ratio of driving voltage is applied to the grids of a pair of tubes running in a linear mode.  All Western Electric transmitters, running Class B, plus the Continental Electronics 5 and 10kW transmitters all worked with this basic approach.  Neither the carrier or peak tube are driven into conduction in the Continental AB1 approach (where tetrodes are used).  They are not terribly efficient compared to plate modulated rigs but, overall efficiency if high levels of average modulation are used is roughly the same as a plate modulated rig.  The benefit of the classic Doherty over plate modulated rig are fewer tubes and no iron.  Back in the day, electricity was cheap and tubes cost a lot.  The Type 2 "Doherty" is really a Sainton design; different from all the earlier Doherty variants including the Weldon, Terman and others in that both the carrier and peak tube run Class C; the control grid is driven well into conduction.  Each control grid is biased identically; the screen is the control element.  There can be a modulation transformer but it is quite small; its only purpose is to feed both screens in series yet isolate them at DC potential so that one can be set to normal (screen saturated) operating point and the second (the peak tube) to a state near cutoff.  In the case of a tube such as the 4-1000A, this means about ~500V on the carrier tube screen and about -75V on the peak tube screen. Audio then drives the peak tube to about 500V at 100% modulation.  The screen tube voltage is varied to minimize distortion; it is a crude form of matching mu of the tubes.  Since, at 100% peak modulation (or higher  if you wish), the tubes are not yet matched since the carrier tube will have just over 1000V on it's screen and the peak tube will be roughly 500V or so.  To further reduce distortion and null out carrier shift, a small amount of modulation is applied to the grid of the peak tube to further raise its operating point so that it matches that of the carrier tube.  Fortunately, the curve fitting of this is quite good; performance is maintained at all modulation levels. 

As mentioned, I have built both types.  I built only 1 Type 1 true Doherty.  It works fine but is a pain to tune up and operates at one frequency (+/- 20kHz on 75m) and at one power level (250w).  The Type 2 Sainton Doherty is a bit more forgiving.  Rather than using a 90 degree network between the carrier and peak tubes, I used a Johnson Counter which takes a signal at 4x carrier and divides by 4 in such a way to give 0 and 90 degree reference carriers.  I then amplify these and drive the grid with these signals.  regardless of the frequency, the quadrature reference is always maintained.  I used both low Q networks and toroids to step up the voltage to drive the grids.  My designs were at three power levels:  About 8 watts using 6CL6 tubes, about 200W using 4-65A tubes and about 300W using conduction cooled versions of the 4CX250 (8560); the latter using an old VHF Micor cabinet as the basis for the transmitter.  They all run on 75m.  I also did a pentode design using 5-125A/4E27 tubes and a second using 803 pentodes.  These are a modified sainton design where the supresor is modulated; not the screen.  Beam tubes cannot be used in this manner; they must be true pentodes.  The 803's are bullet proof but the amp only makes about 150W.  The 803's are great in that the screen is heavy duty; you have no worries about excessive screen dissipation when the suppressor is biased to cutoff on the peak tube (about -110V).  The 4E27 is a little more fragile; I used a screen dropping resistor to limit the dissipation at a slight expense in distortion.  All designs had below 5% THD between 25 and 95% modulation with very small amounts of feedback (roughly 8 - 10dB). 

My most recent design is based upon a pair of 4-1000A tubes.  I want a full 375 watts of carrier with lots of positive modulation capability.  The 4-400 and 4-1000 are great choices as they have plenty of filament emission (necessary for this application) and can take a real beating with regard to the screen.  They are also easy to drive; no more than 5w is needed for each tube.  My screen modulator will be a cathode follower.  You want a low mu tube for best distortion performance as the screens draw quite a bit of peak current; the impedance varies considerably on the cathode follower stage.  But, low mu tubes that can handle 1000V on the plates are difficult to come by; therefore, I am using a pair of 4-65A's triode configured.  I thought about using older school triodes such as a 35T but I had a bunch of 4-65's so I started using them.  I hope to finish the transmitter this winter; hard to say if I will have the time. 

Tuning up these rigs does require a scope.  That is the best way to tune them.  If you step your impedance down in multiple steps (from the plate to the antenna), you achieve the necessary harmonic reduction and you can use the final leg as a matching "tuner" to get the impedance right at the intermediate point.  Then, you simply adjust the carrier tube for resonance and the peak tube to 90 degrees phase shift (between the peak and carrier tubes) as indicated on the scope. 

Plate modulation is a lot easier and a lot more forgiving.  Again, this approach saves the big iron.  For me, I did it because I worked with Joe Sainton.  He was my mentor on these transmitter designs.  I've also built one ampliphase approach and a PDM approach; all using tube PA and modulators (in the case of the PDM).  I think the Sainton or even Class I Doherty are easier than the others.  The PDM is easy on paper but impedance matching is key to making the series LPF filter work and not have an incredible arcing monster on your hands.  (And don't' forget the damper diode...).  The ampliphase is pretty easy in theory but linear phase modulation is what you start with, then you convert it through feedback and some predistortion (RCA called it the regulator stage - basically grid modulation that took care of a good portion of the downward modulation cycle) to an arcsine function.  That provides the correct mapping of phase to amplitude to get low distortion envelope modulation.  I cheated; I developed an arcsine phase modulator  - actually a pair of them out of phase - and drove my PA with this.  My transmitter was only 100W but it was a handful.  It worked at one frequency.  I was always impressed with the RCA engineers that were able to build HF / SW transmitters that were re-tunable using the outphasing method of ampliphase. 

Sorry for the long soliloquy on subject; it is fun experimentation and they will all work on the ham bands.  But they are all more difficult than a plate modulated rig.  Perhaps a true Doherty (Type 1), tuned for 3885kHz, would be a good project to take the output of a 25W  AM rig and crank it up to a quarter kw or so.  It would be a specific use amp, though....

73,

greg - k9qi

[Opcom]

Thank you, that post was very interesting. Do you have schematics for these high efficiency amplifiers and the Counter?  I can readily see the 4fc being divides via tubes, very nice.. However, you get a constant amplitude wave, and therefore require the screen or suppressor modulation? A picture's worth 1000 words!

I want to experiment with PDM more than Doherty style, but I'm no expert on it and have enough projects to do first.

One thing that comes to mind is to adapt the usual tube regulated supply to such a use (obviously QRP here as I don't like explosions) - but an intermediate step is to convert it to a class A series mod, that is, to find a place to inject audio against the feedback voltage and modulate its DC output to follow the audio waveform. That part is easy. I have converted one tube type lab regulated supply to a class A series modulator for AM before and it worked well, within experimental limits.

Taking it another step farther by adapting the 'convert your linear supply to a switcher' scheme seen with solid state regulators from time to time - why not? Or I could just call it the 'lazy hack's PDM'.

The only two things I had to do, with the linear-regulator-to-self-excited-switching-regulator scheme on solid state supplies, was to make sure there was a lot of gain in the error amp, and make sure there was more than enough base drive at the bipolar pass transistor to totally saturate it when ON. Seems it might be easy with tubes.

Any more info and details you have from experience on PDM would be gratefully read and studied.

[John K5PRO]

Yes, Greg, very interesting. Post schematics if you can. Happy black Friday I guess.
John

[w4bfs]

hi greg et al .... I have been studying some of these direct waveform synthesis methods for some time now ... Taylor modulation seems to be a simplified method of Doherty but I am not able to find any performance criteria concerning distortion .... the sdr used as a spectrum analyzer is making these methods more interesting ....73 ...John