Reactor's sparking when carrier goes off

[AMLOVER]

Last night reactor flashed when I unkeyed even it is isolated from the sassis, this morning I checked and was black between the wirings, I hope not dead...
I am using modified Heising with a modulator 30:1, 3.6K to 4 ohm using a solid state amp and having grounded one side of the modulator. The other side is connected with reactor and final due to a 1.5uf/10kv cap.
I unkey in the same moment with one order to 3 different solid state relays, the drive, the screen and add in the cathode a 50K resistor.
When I key no problem at all, everything goes right.
My questions are:
a) is my sequence right or I should have some certain delay between drive  and screen?
b) do I use the right Heising configuration by grounding one side of the iron as soon as I use a 4ohm af amp with no high voltage in the primary of the iron or I should ground one side of the modulator through the cap and connect with the reactor the other one?
c) I put off the solid state amp first and then I unkey the transmitter but I hear a loud noise. What is the right way to put off the af amp without this "boop"?
d) Is it a good idea for avoiding sparking to put mod and reactor irons in oil suitable for transformers or to keep them out as they are?

Thanks in advance for any help
Stefanos
 

[K1JJ]

Stefanos,

I had the same problems with my 4-1000A modulated by a pair. 

The thread below is about 9 pages long and covered the subject quite well. There are sequencing suggestions offered. It cured my spike problems.

See if this helps, if not, then axe away.

http://amfone.net/Amforum/index.php?topic=22073.0


T

[flintstone mop]

TOM VU is da man!
He knows dis stuff

[AMLOVER]

Tom,

Thanks for the thread, I read it 3 times.
The results are :
a) I take the solid state amp far from the rf amp, in my case different room 30' away and I'll carry the 4 ohm energy by 2x30' RG213 to the rf amp mod department using the inners for af and the braids for rf isolation.
b) I must drop the fixed bias to as less as possible, now is -210v and goes to -390v with 2.5k resistor. I'll leave this for last because I feel very safe with high fixed bias as far I keep HV always on the tube.
c) I'll add a 100uf in paralel with the 50K cathode resistor, that's the easiest part, I'll do first.
And finally I'll keep on the drive for 1/2 sec after screen is off and cathode is grounded through the 50k.
I have not a T/R relay because I listen from an external receiver and the transmitter is monobander.
I must add that I unkey screen from the ac side so I notice the screen volts coming down in 1/2 second from unkeying whatever this means...

If there is smthing more that I lost please let me know. After the changes I'll let you know the results and post some photos of the "Lady in red".

If somebody can help me with the right Heising configuration for my case will be highly appreciated.

Stefanos

[Bill, KD0HG]

What happens across a larger inductor when the voltage and current across it goes away?
Zapp!

[WA1GFZ]

E=L dI/dt

[K1JJ]

Stefanos,

With the items you listed, it looks like you have a good start and the knowledge to pull it off. Sometimes the sequencing takes some time to get right. You might even buy a 4-step sequencing board from Jay/W1VD. That's what I needed to do for my class E AM rig, but was able to get the 4X1 rig working with just a few R/C delays in the ant and keying relays.

Since that thread, I switched over to a homebrew solid state FET audio driver designed by WA1GFZ. The 100 watt 8 ohm amp with the backwards driver transformer did finally work well when placed far from the 4X1 rig. However, this FET driver sits right next to the rig and is far superior with less phase shift. You might build one up later.

http://amfone.net/Amforum/index.php?topic=23632.0



BTW, about your Heising question... will this simple Heising schematic approach help?

http://www.amwindow.org/tech/htm/813/813.htm

Another recommended approach is to tie the Heising cap to ground. There is a good thread on this subject about three months ago. I can find it if you can't.

T

[AMLOVER]

Bill, you are very right. Energy stored in a big inductor is unpredictable sometimes.
The story is that this happened suddenly one night when was working without problems for more than 6 months.
It is possible that the reactor could handle the enrgy till a point after that it sparked.

Tom, thanks for the recomendations but in my case the af amp is not a driver for a tube af amp, it is the final af amp. In few words it is a Phonic 5100 the same like Crown XLS 5000 a huge amp with 5000w bridged 4 ohm output...That's why I thought to ground one end of the iron. There isn't any high voltage in the primary but just the pp af voltage 360v-400v. The schematic of the 813's is considering that the pri and sec of the iron have the same voltage, in my case this configuration would give 5000v to the sec of the iron and big difference with the pri...I thought that this could be dangerous or not?
I attach my configuration to become more clear.
The bad news are that the reactor is off...I already opened the lamination on the bench and found lot of smoke between the upper windings and the inside of the E iron...New winding is ordered with 25kv isolation between windings and iron.
Unfortunately I can't see the difference of the 3 changes I already did in the transmitter because all even before the changes worked perfect without the reactor in line. Change a. I added a 100uf/450v in parallel with the 50K catode resistor. Change b. I added aan RC delay to the drive on off relay and Change c. I moved the heavy af amp closed to the audio chain about 10m away from transmitter and in diff room. Tomorrow I must connect with RG213 the bridged 4ohm output to the transmitter and wait for the new reactor winding. The reactor has all E and all I laminates separated with a 2mm fat isolation material like a choke, is that right or it could work better if i mix the laminations like a transformer?
I still need your opinions about the rightest Heising configuration for my special case.
Thank you,
Stefanos

[AB2EZ]

Stefanos

Hi!

I have been running a similar configuration for a couple of years now... with no problems. I'm using a 700W QSC ampliier to plate modulate a pair of GS35b's. The transformer is a 1:25 step up transformer*, the Heising reactor is a Peter Dahl 50H 300mA, and the coupling capacitor is 1.4uF. The plate voltage at carrier is 1800 Volts, and the plate current at carrier is 300mA.

*The transformer is a ferrite core toroidal power transformer which has nearly ideal characteristics. It's magnetizing inductance is much larger than 50H, so the reactor dominates. The frequency response of the modulated RF amplifier (as measured with an off-the-air monitor) is 3dB down at 20Hz, and flat out to beyond 10kHz.

I believe that because the impedance looking back through the transformer into the amplifier is very low (the amplifier looks almost like an ideal voltage source), the current that is flowing in the reactor will not have to change as rapidly when the rf tube(s) is (are) turned off or cut off on negative modulation peaks (compared to a traditional design). The current will just be rerouted into the amplifier (via the coupling capacitor and the transformer). Therefore, di/dt is not anywhere near as large as it would be if the modulator were a traditional tube design (with a relatively high output impedance).

Like Tom, I believe that the effect you are observing is being caused by a transient generated by the amplifier. This would be of particular concern with a 5000 watt amplifier.

Best regards
Stu

[AMLOVER]

Stu,
First of all many many thanks for all I have learnt years now from you and all the amers in this forum.
I used to use the traditional mod style till I met all of you and I found the right way to avoid saturation etc etc...
You are absolutely right, the trancient came for some reason from the amplifier and wounded heavily the reactor.
After your reply I thought to check the amplifier and found that the protections lamps are always on now, the amp takes too many amperes with no load and after 1 minute about, one of the two fuses on its back comes out...
For sure the amp got ac or input rfied and created the transient on the reactor, after I didn't start the amp because I focused on the reactor's sparking and so I thought that the problem is sequencing or anything else.
Today I'll bring the amp to the service, middle of next week I'll get the new reactor with only 30H inductance because of heavier isolation and I'll keep all the new changes that Tom suggested.
The only doubt is for the 10m long audio cables from the amp's bridged output to the mod iron, they seem to me long but definately may be not...
I am thinking to key the af amp by interrupting a relay on its 4ohm output, now I key it with relay in the ac line and hear an annoying "boop" in my monitor,  is it dangerous for the amp to stay on but unload even with all pre audio chain off or not?
Sorry for all boring questions but it is completely new to me how to use fragile solid state amp with heising.


Thanks,
Stefanos
 

[AB2EZ]

Stefanos

1. With respect to the cables from the bridged output of the amplifier to the transformer

The impedance at that point is very low (essentially zero), and therefore the capacitance of the cables (the capacitance of each cable's center conductor to the grounded shield) will have no effect on the audio.

2. With respect to the output relay

Do not open the amplifier's output relay. What keeps the reactor and the mod iron from arcing over when the plate current to the rf tube suddenly turns off (either because you have removed the rf drive, or because the tube has been driven into cutoff on a negative modulation peak, or because you have inserted a large resistor between the cathode and the negative side of the B+ circuit) is the relatively low impedance path through the audio amplifier that the reactor's current can take.

3. I don't think that the audio amplifier cares if it is looking into an open circuit (i.e. when the rf tube is turned off). However, it might be a good precaution to put a resistive load (for example: 20 Ohms) across the output of the amplifier, in parallel with the modulation transformer. Since the modulation resistance of the rf amplifier is around 5000 Ohms (B+/plate current at carrier), the audio amplifier sees a load (from the rf amplifier) of around 5000 Ohms/[30 x 30] = 5.5 Ohms. Putting a 20 Ohm resistor in parallel will make the amplifier deliver an extra 25% of audio power; but you have plenty of extra power in the audio amplifier.

A key question to ask... which is specific to this particular audio amplifier, is whether the overload protection circuitry of the audio amplifier is sensitive (as designed) to the sudden increase in output current that results when the reactor current is diverted from the r.f. tube(s) to the audio amplifier. In fact, you may find that the overload protection circuit is trying to respond to that by opening up the output relay (just a hypothesis). You may want to consider bypassing the output relay.

Stu

[W2VW]

Any sequencing should be backed up by good spark gaps in case drive power were to be switched off by failure or mistake.

[AB2EZ]

Stefano

I few more thoughts:

1. With respect to the 100uF capacitor being placed across the 50k Ohm resistor

It makes sense to me to open the relay (RL1) across the 50k Ohm resistor + 100uF capacitor first, because doing so will cause the plate current and the screen current to drop slowly, toward 0mA, when unkeying. With the plate current and the screen current at 0mA, it should make little difference as to what you turn off next. Note that the 100uF capacitor will slow down the turn off of the plate and screen current. If the total plate + screen current is (for example) 1A, then the 100uF capacitor will initially charge up at a rate of 1A/100uF Volts per second = 10 Volts per millisecond. Thus, it will take several 10's of milliseconds for the tube(s) to turn off. This will, of course, reduce the concerns regarding the current in the reactor having to be diverted into audio amplifier during the turnoff process.

Note: with the 100uF capacitor charged to whatever the cathode cutoff bias is... closing the relay, RL1, is going to make a "bang" unless you place some resistance in series with the relay contacts.

2. With respect to turning the audio amplifier on and off

I think it is not a good idea to turn the audio amplifier on and off by turning its AC power on and off. What I do is to just open the audio input line to my audio amplifier. Even that may be unnecessary, but it seemed prudent to do. [I.e., in case I decided to set my audio chain to produce a high-level  sine wave test signal, I didn't want that to go into the input of the audio amplifier until I was ready to transmit]. If you turn off the AC power to the audio amplifier, then there will be an open circuit across the input to the modulation transformer... and I think it is better to keep a very low impedance (the impedance looking back into the audio amplifier) across the input to the modulation transformer.

3. With respect to the Heising configuration

Since the primary of the modulation transformer is at 0 volts DC relative to ground, you should keep your present configuration. I.e. ground one side of the secondary of the modulation transformer, and couple the other side to the Heising reactor/plate using the Heising capacitor. This will keep the entire modulation transformer (primary and secondary windings) a 0 volts DC. 

Stu

[AMLOVER]

Stu,
I thought about the addition of the 100uf and found it needless for my case. The reason is that I turn off the screen supply from its ac side, this supply is a C (150uf/1200v) input one  and with 170ma on the screen it comes down to 0ma in about 100ms when I turn it off. I think that if I turn off first the audio chain, then the screen's supply from ac pri + 50k to cathode and finally the driver, it will be right and safe.
The suggestion for not turning off the amp's pri was my wish for that Christmas and thank you for that precious present...It was my nightmare that "boop" but I was very afraid to leave the amp on and a huge backward trancient to destroy it. Definately that happened but because of different reason.
So I thought to add a switch between the output of my comp/limiter and the amp's input and keep the amp undriven due to receiving period.
The reactor is with separated E from I laminations by a 2mm thick isolation material, is that better or a mixed laminations core would give more inductance and work properly,too? Now that I have openned it is a good opportunity to corect it in case it is wrong.
Ending I decided to manually open and close the 3 switches for on and off untill I feel sure and safe for the sequencing. I'll also construct 2 pair of spark gaps as W2VW suggested to protect the irons.
Give me please an idea about the wattage of the 20ohm resistor and keep an eye to the new schematic that I am oriented to follow.
Of course all your comments and suggestions will be highly appreciated.

Thanks,
 Stefanos

[AB2EZ]

Stefanos

I agree that turning off the screen voltage first, by removing the AC input to the screen supply, is an excellent strategy for avoiding a sudden change in the current through the Heising reactor.

The 20 Ohm auxilliary audio load resistor would have to dissipate about 25% of the average audio power that is going into the transmitter. As long as you don't apply a sine wave at 100% modulation (at least not for a long period of time), the average audio power will be much less than the peak audio power.
In the end, you may decide that including this resistor is not necessary. Recall that in the normal audio amplifier power-on sequence, the audio amplifier's output relay is open until the amplifier gets through its power on transient. Therefore, operating without any load is nothing unusual for a modern audio amplifier.
But, if you do include the resistor, I would suggest a resistor capable of dissipating 100 Watts of average power. You might consider using a spare 1kW peak / 100 Watt average dissipation 50 Ohm RF dummy load... since all you are trying to do is to keep some load on the audio amp when the transmitter is off.
I don't know as much as the other folks on this board about the subject of how to lay out the laminations of the reactor.

[k4kyv]

In none of the posts to this thread do I see any mention of the simple solution I have used for years, and which has been in use by AM broadcast transmitters since the advent of plate modulation in high power transmitters, dating back to the 1920s.

Use a spark gap across the modulation reactor, set just slightly wider than what will spark over on modulation peaks. The idea is to let the voltage breakdown occur harmlessly at the spark gap, rather than in the wiring of the transmitter, or worse, through the internal insulation in the modulation transformer and/or reactor.

To protect the modulation transformer, a similar gap should be installed across the primary, not secondary of the modulation transformer.  A gap across the secondary and/or across the mod reactor, without one across the mod transformer primary, may actually increase the risk of insulation break down in the mod transformer, due to the voltage generated by the dI/dt that occurs in the pulse that results when the gap sparks over. You want the dI/dt to occur before the modulation transformer, not allow it to be reflected back through the transformer to the primary from the secondary load.

I also install a gap across the power supply filter chokes.

[W2XR]

To protect the modulation transformer, a similar gap should be installed across the primary, not secondary of the modulation transformer.  A gap across the secondary and/or across the mod reactor, without one across the mod transformer primary, may actually increase the risk of insulation break down in the mod transformer, due to the voltage generated by the dI/dt that occurs in the pulse that results when the gap sparks over. You want the dI/dt to occur before the modulation transformer, not allow it to be reflected back through the transformer to the primary from the secondary load.

Don,

RCA installed spark gaps across both the primary and the secondary windings on some of their broadcast modulation xfmrs (at least the ones I happen to be personally familiar with).

I have an RCA type 900777-502 BC mod xfmr (I had two at one point; gave one away many years ago to another AM'er), and both of them had factory-installed gaps across both windings. This xfmr was used in the RCA model 250K 250 watt-class AM rig, and the xfmr was designed to handle the full 250-300 ma. unbalanced secondary current; there was no mod reactor used in this rig. This xfmr easily weighs in at over 150 lbs.

Perhaps RCA knew something here that you may have overlooked???

By the same token, I am not aware of any Gates/Harris AM BC rigs that used spark gaps on either winding of the mod xfmr.

Just my 2 cents.

73,

Bruce

[AB2EZ]

Don et. al.

One has to adjust the various rules of thumb that apply to a traditional plate modulated AM transmitter, when one is employing a plate modulated transmitter that uses an audio amplifier whose output impedance is essentially zero.

Consider the case of a classical audio amplifier that employs a pair of tetrodes in push pull.

The current that will flow through each tetrode is essentially (not exactly) independent of the plate voltage on each tetrode. Therefore, the (output) impedance looking back toward the audio amplifier... from the modulation transformer... is essentially infinite (in any event, very large compared to the modulation resistance of the rf amplifier).

As a result, if the rf tubes are cut off, the current flowing in the mod reactor (for the case of modified Heising modulation) or in the secondary winding of the modulation transformer (for a non-Heising configuration) has no place to go. Therefore, di/dt is essentially infinite... and the reactor or transformer needs a spark gap to limit the voltage and to dissipate the energy stored in the magnetic field of the reactor or transformer.

If the audio amplifier employs triodes in puish pull, then the situation is not quite as bad. The (output) impedance looking back from the modulation transformer toward the audio amplifer (adjusted for the turns ratio) will be comparable to the the modulation resistance of the rf amplifer. Thus, if the rf stage plate current is disrupted, the current in the reactor or the secondary of the modulation transformer has a path to flow into (to allow the stored energy in the magnetic field to be dissipated, and to limit di/dt to a safe value)

For the case of a modern audio amplifier, whose output impedance is essentially zero (because it is a feedback-controlled voltage source), a disruption of the plate current in the rf stage will not cause a disruption of the current flowing in the Heising reactor or in the secondary of the modulation transformer. The current will be diverted into the audio amplifier. As long as the audio amplifier does not shut down (e.g. by opening up its output relay), the problems associated with a sudden cutoff of the rf output stage plate current will not cause a large voltage to develop across the modulation transformer or (in the case of a Heising configuration) the modulation reactor.

Stu

[k4kyv]

RCA installed spark gaps across both the primary and the secondary windings on some of their broadcast modulation xfmrs (at least the ones I happen to be personally aware of).

I have an RCA type 900777-502 BC mod xfmr (I had two at one point; gave one away many years ago to another AM'er), and both of them had factory-installed gaps across both windings. This xfmr was used in the RCA model 250K 250 watt-class AM rig, and the xfmr was designed to handle the full 250-300 ma. unbalanced secondary current; there was no mod reactor used in this rig. This xfmr easily weighs in at over 150 lbs.

Perhaps RCA knew something here that you may have overlooked???

By the same token, I am not aware of any Gates/Harris AM BC rigs that used spark gaps on either winding of the mod xfmr.

Nothing wrong with placing gaps across both windings. The problem is when there is a gap across the secondary but none across the primary.

With a reactor (assuming a gap there), a gap across the mod xfmr secondary would serve no useful purpose.

On my BC1-T, I re-mounted the mod transformer on stand-off insulators, and added  spark gaps across the primary.

My HF-300 rig uses the mod xfmr from an old RCA 250-watt transmitter, although I have seen it modulate as high as 150% positive at 1 kw DC input.  That transformer has gaps across the primary, but none on the secondary.  I  have a reactor to match that transformer, but the one I use in the rig is a UTC LS-103, to which I added gaps across the terminals. I'll have to check to see if the RCA reactor has a set of gaps.

I have one of those RCA 250 watt mod xfmrs designed for use without a reactor; it's the same size as their 1 kw transfomers designed for use with a reactor. I don't use it; if someone wants it, I would be willing to swap it for something, or even sell it, but it has to be pick-up only; I refuse to try to pack the thing and then haul it to a destruction shipping company, since it weighs 75 lbs or more. One of the problems with it is that its turns ratio (1.7:1) is too much step-down @ 3:1 Z ratio.

One danger may exist in ham transmitters without gaps across the transformer, in the case of a relay across the secondary to short it out for CW or linear amplifier operation.  The relay contacts may inadvertently serve as an arc gap, with none across the primary.

Regardless of whether I used triodes or tetrodes, I would still feel safer with a mod transformer protected by gaps.

[W2PFY]

I'll carry the 4 ohm energy by 2x30' RG213

I don't know what the output voltage of your 5000 watt amplifier is but it seems to me that the wires are too small for this amount of energy delivered to the transformer. If you have 100 volts there you would be asking those wires to handle 50 amps?  I would suggest that you run at least # 6 wire to carry the energy.

[W2XR]

Don et. al.

One has to adjust the various rules of thumb that apply to a traditional plate modulated AM transmitter, when one is employing a plate modulated transmitter that uses an audio amplifier whose output impedance is essentially zero.

Consider the case of a classical audio amplifier that employs a pair of tetrodes in push pull.

The current that will flow through each tetrode is essentially (not exactly) independent of the plate voltage on each tetrode. Therefore, the (output) impedance looking back toward the audio amplifier... from the modulation transformer... is essentially infinite (in any event, very large compared to the modulation resistance of the rf amplifier).

As a result, if the rf tubes are cut off, the current flowing in the mod reactor (for the case of modified Heising modulation) or in the secondary winding of the modulation transformer (for a non-Heising configuration) has no place to go. Therefore, di/dt is essentially infinite... and the reactor or transformer needs a spark gap to limit the voltage and to dissipate the energy stored in the magnetic field of the reactor or transformer.

If the audio amplifier employs triodes in puish pull, then the situation is not quite as bad. The (output) impedance looking back from the modulation transformer toward the audio amplifer (adjusted for the turns ratio) will be comparable to the the modulation resistance of the rf amplifer. Thus, if the rf stage plate current is disrupted, the current in the reactor or the secondary of the modulation transformer has a path to flow into (to allow the stored energy in the magnetic field to be dissipated, and to limit di/dt to a safe value)

For the case of a modern audio amplifier, whose output impedance is essentially zero (because it is a feedback-controlled voltage source), a disruption of the plate current in the rf stage will not cause a disruption of the current flowing in the Heising reactor or in the secondary of the modulation transformer. The current will be diverted into the audio amplifier. As long as the audio amplifier does not shut down (e.g. by opening up its output relay), the problems associated with a sudden cutoff of the rf output stage plate current will not cause a large voltage to develop across the modulation transformer or (in the case of a Heising configuration) the modulation reactor.

Stu

Hi Stu,

Thank you for your usual detailed and insightful response.

This would probably explain in large part why Gates, who almost exclusively used triodes as modulator tubes, did not see the need to include spark gaps on the windings of the modulation xfmrs or reactors used in their plate modulated BC rigs.

To carry this one step further, and the answer is probably obvious, but would this also suggest that plate moduated rigs employing triode modulator tubes would be less likely to exhibit modulation transformer or modulation reactor failure in the event of periods of over-modulation, where the carrier is cut-off and the modulation transformer is effectively unloaded?

73,

Bruce

[Bill, KD0HG]

I am surprised that no one has mentioned negative cycle loading, using a diode stack and resistor, to help protect mod iron.

[k4kyv]

I have never worried about crapping out a modulation transformer with negative overmodulation.  At the moment of overmodulation, the plate current is minuscule, approaching zero, and at exactly 100% it is zero.  IMO you would have to drive the plate of the final hundreds of volts into the negative region before this would happen. However, it could happen with separate power supplies if you somehow lost plate voltage to the final but maintained full plate voltage to the modulator tubes.  I have accidentally done that a few times and nothing happened, but I still would not intentionally allow it to happen.

I never cared for the ultramodulation circuit or negative cycle loading.  But a variation would protect the modulation transformer during negative overmodulation.  Obtain a wirewound power resistor the same value as the modulating impedance (DC plate voltage/DC plate current), and a diode with PIV at least twice the DC plate voltage, plus substantial margin for safety. Connect the resistor from the positive plate lead coming off the modulation transformer, and ground the  resistor through the diode, with the anode grounded and the cathode going to the resistor. The diode keeps the resistor disconnected as long as the voltage to the final is positive or zero.  As soon as the final is overmodulated, the plate lead goes negative and the diode  conducts, placing the resistive load on the modulator just as the final cuts off.

[Bill, KD0HG]

Yup. Doing an ultramodulation circuit here since ER #3, 7/99.

No matter how hard I hammer the audio, I can't push it beyond 95% negative. And doing about 110%-120% positive. I absolutely cannot cut off the RF final.

I recently smoked one of the 15 yo ultramod diode stacks, which I just replaced today. Six 6 Amp, 1000 volt fast diodes in series now in place instead of the 1 amp stack. Looks great on the scope.

if you've got excess audio power, and good iron, Ultramodulation is a good way to go and helps protect your iron. Negative cycle loading is a cheap way to protect against human screw ups.

Don, you should write up an article on the technique. Actually, you already sort of did.

[k4kyv]

But I don't like to  load down the modulator over the negative cycle before 100% modulation.  That's what the classic 3-diode ultra modulation circuit does, actually attenuating the entire negative half of the audio cycle via resistive losses. The abrupt discontinuity as the audio waveform crosses the baseline point produces splatter if a low pass filter is not inserted between the modulator and final. In practical use, splatter with this circuit is reduced to a near-satisfactory level by the rudimentary filter formed by the leakage inductance of the mod transformer combined with the rf bypass and plate tuning capacitors in the final, aka "building out" the mod transformer. 

Any time you tamper with an audio waveform, by definition, you generate distortion and greater bandwidth. The ultramod circuit produces a positive carrier shift along with even harmonic distortion. Better to take advantage of the natural asymmetry of the human voice to generate enhanced positive peaks. The increased output power you see with ultramod  mostly comes from the rectified audio voltage added to the DC plate voltage, plus even harmonic distortion audio products.  This would be identical to the effects of running controlled carrier modulation while deliberately introducing even harmonic distortion elsewhere in the audio chain.

Years ago I tried the ultramod circuit.  It looked good on the rf ammeter, but signal reports told me the audio was not any louder than without it, but the additional distortion was noticeable.

The single diode and resistor circuit I described produces no distortion, and can serve as a handy overmodulation indicator when an 866 or 866-A, or other MV rectifier tube, is used as the diode.  At the point of overmodulation, you see the 866 flash blue.  I once used one for this purpose, with the 866 mounted inside a darkened box with a viewing window cut in the side.  If desired, the plate of the rectifier can be biased positive in order to get it to flash at less than 100%, for example at 95% negative.