The AM Forum

What are the pro's and con's of a diode in series with the modulation transformer to the PA final? In mobile rigs one simple solid state diode was used. In larger home stations, I have seen an 866 or 872 in series with the mod transformer to the PA.

I know about the three diodes arrangements so I'm not looking for something that's better. I just want to know how well this worked in it's simple form?

Did anyone try it. I did in an old tube CB years ago and it seemed to work fine. I guess the idea was to keep out the negative cycle? Someone said it would more effectively modulate the carrier.

Hi Terry,

Recently I've added a high level 3-diode ultra modulation (negative cycle loading) circuit to my 4-1000A  modulated by a pair of 4-1000A's.  I ran some tests and found it did indeed help contain splatter from overmodulation.  In the past I thought it caused splatter on a few of my rigs, but have since decided it is a good safety net when used sparingly.

I know some guys use it to hard limit negative peaks to increase positive peaks as a result - however, I use it simply as protection against modulation transformer damage from 100% negative peaks and to protect against occasional splatter as a result.  In this job it works FB.

I have mine set to work at about 98% negative, so it produces a very thin line at the center line (on the scope) when working. My audio is set for it to work maybe 5% of the time or less, during normal speech.

As for a single diode - I don't like that idea since it creates no matched load for the mod transfomer. The modulation transformer is at risk, especially if the diode is leaned on for bigger audio. Maybe a small rig will survive, but I wouldn't be too agressive using it in a big KW mawl. In contrast, the 3-diode  circuit has a rather nice loading effect as produced by the loading resistor.  The resistor is selected by calculating the normal modulator load into the RF final.

Because of phase shift from the average modulation-transformer AM rig, negative peak limiting should be done in the high level area, thus the 3-diode circuit. Low level limiting is not as effective - unless the rig has exceptional linearity like a low level balanced modulator, SDR or class E rig - all transformerless.  My 4X1 limiter decision was based on this concept.


T

Wasn't that what Lafayette called "Range Boost" on their 11M tube rigs back in the day?

Phil

The single diode in series with the modulation transformer may actually produce more splatter than would simple overmodulation. The problem occurs during the part of the audio cycle when the instantaneous modulated DC voltage is rapidly decreasing towards zero. The series diode effectively disconnects the modulator from the final as the rf bypass caps in the final discharge at a slower rate than the waveform out of the modulator descends. Then, as the instantaneous audio voltage from the modulator begins its upward swing as the audio cycle continues to progress, the diode conducts once again, and the descending voltage from the discharging capacitors clashes head-on with the rising voltage from the modulator. The result is a sharp spike at the negative peak, rather than the normal flattened negative peak due to the PA being cut off for a split second while the plate is driven negative during overmodulation.

This produces splatter, even though the plate of the final is never actually driven negative at the bottom of the modulation peak.

I have sometimes used a single diode from the modulated B+ to ground (reverse biased) with a resistor in series equal to the modulator load impedance at about ~5% of the modulator rms power.

Why would I do that? Well, for one this keeps the modulator loaded at all times, even during periods of overmodulating when the modulated B+ swings negative with respect to ground. This helps keep modulating transformers alive, and reduces any increase in distortion from the modulator due to the small portion of the  modulation cycle that would otherwise go unloaded. Sure the scope baseline will still go to zero, but my experience with this single diode-resistor combo is that splatter is much reduced.

What I present does not prevent over modulation, but it does protect the modulation transformer, and it does reduce splatter.

Jim
WD5JKO

Your correct, I want to thank everyone who replied. I would use the three diode scheme but it helps to be on the air when doing it.



I thought this was interesting but it would seem to me that you would really need to hammer the audio for the diode & resistor to work. Didn't someone use a similar scheme with a mill-amp meter in series with the diode. when the meter started to waggle, your are hitting the base line??

I like to work 'em hot too. Keeps me alert and creates a unique experience.



I thought this was interesting but it would seem to me that you would really need to hammer the audio for the diode & resistor to work. Didn't someone use a similar scheme with a mill-amp meter in series with the diode. when the meter started to waggle, your are hitting the base line??

Patrick,  Yes the single diode-resistor idea is only there to transfer the modulator load from the RF tube to the resistor whenever the Mod B+ goes below zero; a mere fraction of an AC speech cycle.

So what causes splatter when we overmodulate? In my opinion splatter comes from two main issues:

1.) Modulator, if using high plate resistance tubes, no NFB, and being partially unloaded or an instant during the modulation cycle. The peak modulator AC out put might try to go 2-3x or more higher than when loaded properly. This will generate the 'click-click' we hear 5-10 Khz away from the carrier when the modulation exceeds these levels.
2.) The sudden loss of conduction of the RF tube when the peak mod B+ drops to/below zero volts. The RF tube acts much like the single series diode concept that started this thread.

So guys, does #1 dominate over #2 above, or is there another explanation? My series diode-resistor concept tries to address #1 above. I've never been told I'm wide when doing this even though the scope is 'white-lining' at the baseline frequently.

The series diode-resistor combo seeded expansion of the concept as explained at the following thread:

see reply 8:
http://amfone.net/Amforum/index.php?topic=22341.0

"super modulation with progressive negative cycle unloading with carrier boost". 

Here we started a progressive negative cycle attenuation that can be set to whatever ratio and thresholds you want. This could be transparent to -90% mod, then kick in a divide by x ratio for the negative peak for -90%, then 2X at -95%, 4X at -97%, etc. Then low pass filter to minimize diode switching transients. The circuit shown will show a 200% + sine wave at -100% sine wave without any hard clipping. Back off the audio to 90% +, and the negative modulation is also -90%. The asymmetry produced when driving hard is primarily 2nd harmonic distortion which many folks like to hear..

None of the other circuits I've seen do this. Sure the 3 diode circuit most prefer keeps the mod iron loaded, and addresses both #1 and #2 reasons for splatter above. That said, the audio waveform is hard clipped, and the distortion will therefore have a lot of odd order distortion products.

Jim
WD5JKO 



 

Jim,

Your progressive circuit attached below looks very interesting. I like the use of the progressive negative peak limiting as the positive modulation is increased.  I was thinking that the three voltages could be created using three simple zener diode circuits off the same supply.  

One must be sure they have the modulator headroom to make use of the 150% - 200% modulation. That's not an easy task for the average plate modulated rig.

I'd be curious how the splatter filter looks on a simulator. Hopefully it's not too big of a load on the mod xfmr.


I may try a modified version of this circuit later with my 4X1 rig - once the new E rig is working.  


T

Looks like this cutoff is around 12Khz.
(I assumed termination Rs at 8K)

Brian,

   The splatter really wasn't too bad with HV fast recovery diodes in there. As for the filter, all I did was play with the C-L-C values while listening to my SP600 which was tuned away from center using narrow IF bandwidth looking for splatter. I did not model it. That choke used has the iron core removed to lower the inductance. I do not recall now how I measured that inductance since I built this rig in the 1980's. So maybe the actual inductance was larger.

    If I were to try this circuit with a tube modulator at high power, I would add the series diode-resistor combo discussed earlier in this thread at the left side of the circuit, and return the network to the highest bias supply voltage. This will insure the mod transformer always has a good load upon it. Maybe we could call this the '5 diode ultra modulation circuit'.  ;D

    For kicks I went into the shed and took a snapshot of the modified Viking I that I did this to. see attachment. I wish I had room to set up a bigger station.

Jim
WD5JKO

In the "page 3" circuit, how are the resistor values determined?

  Good question! I did this in 1986... But it went something like this:

* What is the maximum % upward modulation you can achieve? In my case it was over 200%..
* When do you wish to start attenuating the negative modulation? a 150v threshold on a 750v supply for    the first diode says we kick in at 80% downward modulation. For 100v - 87%, for 50v - 94%.
* Figure the parallel equiv of the three resistors (each in series with a diode) versus the upstream resistor, and make that resistor ratio fill out but not quite clip the downward modulation peak. The more bias supplies, and resistor-diode networks the greater the detail you can preserve in the compressed downward peak.

Here is an example of a low level AM rig using this concept running 150% upward modulation and about 95% downward modulation with sine wave drive:
http://pages.prodigy.net/jcandela/CE20AQRO/j.jpg.JPG

The circuit, and whole story is at:
http://pages.prodigy.net/jcandela/CE20AQRO/
The PNCL circuit is a little different being low level, and the upstream diode/resistor is the 12AT7 plate resistance. This rig sounds quite nice in the station monitor even when driven hard.

Jim
WD5JKO

   Yes you are right! I've made a tweak to Heising modulation overcoming a longstanding problem (look for my Gonset G50 posts), and on this post I've introduced an alternate method to ultra modulation citing two real world examples that I've built. Maybe I should have listened to you Brian, it would have been easier to just accept that the state of the art with AM was reached decades ago. Sorry, but I just don't accept that idea.

Jim
WD5JKO

  Brian, The reason I corresponded with Hoisy, W4CJL was because he wanted to know 'what the heck I was doing'. In a long 80m roundtable one noisy evening, my lowly Viking I stood out from the rest. If you can boost the audio on AM by 3 db without splatter, and still sound clear, then that is huge. At that time I also had carrier boost in addition to the ultra mod circuit. That kept the diode detectors in receivers happier. I am not aware of this ever being done before, or since.

Jim
WD5JKO

Jim I noticed a screen resistor for the final. Would a choke in series with the screen supply work as well in cases where the plate volts are much higher than screen, like a 4-1000?

Patrick,  You made a good point. I know that the screen grid dissipation was high. With both the carrier boost + the heavy audio, the screen would turn red visible to the eye with a sustained audio note.

 The carrier boost was adding 300-500v to the B+ depending on speech level. I did that because 200% mod + to many receivers came through raspy mostly due to the distortion curve of a rcvr diode detector. So with just carrier control (AC mod transformer disconnected), the carrier would go from 100w to 200w by whistling into the microphone (B+ jumps to about 1200v or so).  

  I doubt it would be easy to do this with a tube modulator. The big Crown M600 I used had zero source impedance, and essentially infinite power (1340 watts RMS at 4 ohm load). So adding on two mod transformers (1 for carrier boost, other for modulation) was easy..just hook it up and go! I had a variac in there too for the carrier boost transformer.

   Anyway back to the screen supply for the 4D32. Aside from a g2 dissipation issue, the upward modulation was perfectly linear up to 200% and beyond (clipping was ~ 250% as I recall). Ironically the modulation transformer came from a Harmon Kardon Citation V tube amp (NO DC through secondary). This amp was rated at only 40 watts RMS, but this rating went down to 7 Hz. The iron was big and heavy, and I didn't modulate much power below 100 hz. I forget the transformer math, but doesn't the iron area designed for 40w at 7Hz allow for 400w at 70 hz? I know it goes something like that. So for 200% mod with 150w DC input, I needed 4X that of 100%, or 4 * 75W = 300w RMS audio. The carrier control sure complicates that.

    I hope to resurrect the old Viking I someday. Storage has resulted in front panel rust, and frozen knob shafts. The Crown M600 is still here. Powering that brute up will either cause smoke, or a successful power up.

Here is a slightly off topic, but interesting take on screen voltage:
http://oestex.com/tubes/screens.htm


Jim
WD5JKO

As used in the 3-diode high level Ultra Mod circuit:

Has anyone done actual tests on standard power supply diodes vs: high speed diodes for AUDIO use?

I'm using those 6A 1KV standard Silicon Valley diodes for all three legs of the 3-diode circuit in my 4X1 rig. (20 per leg)  I used them cuz I had a bag of them here left over from another project.

I wonder what the difference would be at the higher audio frequencies, like 7kc. I know most feel the slower diodes may produce more splatter, but has anyone run a test or can recommend one to evaluate both type diodes out-of-circuit?

Maybe the difference isn't worth the trouble of my switching over to the high-speed ones.

T

Hi Brian,

That's a good point about the diodes blowing out.

I've had a tube PDM rig which used damper diodes to snub the 130kc residual spikes coming out of the filter. I tried using regular diodes and they immediately heated up and shorted. The high-speed ones ran fine.

In addition, I have blown one leg of my 3-diode circuit once already in the 4X1 rig. (Regular diodes) I added more to the leg (now 20) and it's hung in there so far.

So, it sounds logical that if they don't hang in there, the slow response will generate heat and pop them. Or maybe make them more vulnerable to spikes.  Maybe I will think about faster diodes, BUT if they hang in there, I wonder if they are close enuff in performance to the high speed ones, splatter-wise. That's the question.

Maybe someone will suggest an audio sweep test of some kind. I wud think a resistor in series with a diode with a sig gen across the pair wud show the relative drop at various freqs.

T

I heard Dirk WA2CYT complain that his diodes were not conducting fast enough and someone was sending him faster diodes. So JJ the next time you hear him on you could ask "Notice I didn't say might ask"  him about his adventurous with the new and old diodes. That is all. Over.

Has anyone tried vacuum tube diodes for this purpose -- like in three diode limiters or just the single diode approach.

Tubes like 866, 3B28, or 836's.

I'm not sure how MV or Xe diodes would work at higher audio frequencies, but 836's (or bigger vacuum recitifiers) would.

I've used larger vacuum rectifier tubes  as PDM damper diodes. (non-mercury vapor) They showed some color but appeared to work well. In contrast, the standard silicon diodes crapped out immediately.  I don't have an opinion on how fast they are, just that they hung in there.  I've heard that mercury vapor are not as good at high-freqs, but unconfirmed.

T

W8VYZ reported using 836 in his ultra-mod circuit ... he said ss rectifiers were not fast enuff ... maybe schottkys or freds

It could be, but an 836 has a 2.5 V filament.

So, if it was used in a plate circuit running at 2KV, and the entire 2.5 VAC was added as ripple, it would be only 0.125% ripple.

Most likely the plate power supply had more ripple than that to begin with.

Plus if you ran the 836 filament off a tansformer with a CT and took the plate voltage off the CT, you wouldn't expect even that much ripple to be added to the B+.

That's what you would do in an audio stage using directly heated triodes like 2a3's which operate at lots lower signal levels without a ripple problem.

Well, in this case the 2.5 VAC is the filament voltage for the 836 and that doesn't vary with current through the rectifier. So the ripple introduced by the 836 would be constant under load, it seems to me.

But, on  your side of the argument, that is 2.5 VAC RMS and so it's really P-P of 2.8 times that much. I think ripple precentages are usually calculated as P-P.

besides high voltage schottkys, we sell what we call "tandem" rectifiers. These are two bipolar rectifiers in series on one die, and the purpose is more efficient switching due to decreased total capacitance. If you don't want any recovery time, SiC is best, but the common voltages are not as high as we would prefer, maybe only 600-1000V per device. And they are costly.

   Brian,

   To evaluate these diodes you need to use them in a transmitter. For example, take the 3 diode circuit that you like so much, and in your transmitter, run the audio up to 120% positive with a sine wave where the negative cycle will be hard clipped. Then look at the RF spectrum with a spectrum analyzer. A SDR-IQ SDR receiver would be useful here. Look at the sideband distribution of audio harmonics. Then make changes from normal Trr speed diodes to high speed diodes, and compare under identical conditions. You just might say Yikes!, and install a pi low pass filter after that 3 diode circuit. ;D

Jim
WD5JKO

Brian, you are correct in that human speech is not like a sine wave. I will also say that we don't talk in square waves either, except those that remove one sideband and the carrier, i.e. 'square jaw'.  ;)

I just mentioned a test condition using an audio generator, a transmitter, and a spectrum analyzer such that real repeatable data can be obtained, and then comparisons can be made easily when make circuit changes. The waveform used can be anything you want as long as it can be repeated exactly over and over.

Here is something good to look at written by fellow member, John Lyles:
http://lists.contesting.com/pipermail/amps/1999-July/011928.html

I was wondering if you have the book, "Motorola Silicon Rectifier Manual", published 1980? This book has 14 chapters of theory using diodes in all sorts of applications. They go into both the forward recovery and reverse recovery characteristics and show test circuits, scope plots, etc. They also talk about using standard diodes at high frequency. Do you have this book? If not, would you be interested in putting this on your wonderful amforever.com web site? I could lend it to you.

update: one on Ebay, $18.99 buy it now: 350273765073

Jim
WD5JKO

   Brian,

    Why not take LT Spice, and take a stab at modeling the 3 diode circuit? For the RF PA tube, try replacing it with a diode in series with a resistor. Since the tube will likely be a tetrode where both screen & plate get modulated. As we know doubling the plate voltage doubles the plate current at 100% + modulation..So the RF PA should simplify to a diode in series with a resistor. Just wondering out loud...

   Heck the modern engineer doesn't prototype anything! They just Cad the heck out of it, over and over, and over.... ;D

  Oh, still have the cover on pool here. The trees are still spewing buckets of reproductive mung over my pool cover and grass. I need to wait at least two more weeks..

http://www.troublefreepool.com/pool-school/
here is where science clashes with pool store marketing..BBB all the way here!!

Jim
WD5JKO

8-X  

From RCA TT-3 (not found in TT-4):

866A:
10000 PIV @cond Hg temp 25-60 deg C for the 866 is given at supply frequency=<150Hz
5000 PIV @cond Hg temp 25-70 deg C for the 866 is given at supply frequency=<1000Hz

872:
10000 PIV @cond Hg temp 25-60 deg C for the 866 is given at supply frequency=<150Hz
5000 PIV @cond Hg temp 25-70 deg C for the 866 is given at supply frequency=<150Hz

Since mercury temperature is a factor, the info (and by extension turn-on/off time) is not complete, but the Hg temp seems more a factor affecting PIV than the supply freq. is. So why have they qualified the 866 data with frequency notes, and why does the 872A have no such high frequency note?

The point is that although the MV tubes seem rather slow "by the book" for audio and the manuals don't push them as audio frequency rectifiers, others have, including the maker of this:  http://amfone.net/Amforum/index.php?topic=20179.0 and the ARRL in the handbook. so, from studying that unit's performance the question could be approximately answered.

866 Switching time has to be short at an audio frequency of 3KHz (167us per half wave) to cover the bandwidth without adding distortion, so how short? And at what real peak inverse voltage?

If the 866's failed to turn off fast enough the result would be a short across the mod xfmr secondary. Harmles in some cases, but with the big stuff could be problematic. The application also uses low voltages (~1KV) relative to 866 ratings. The 816 is also proffered in the article for low power applications.

Some LT-Spice circuits, classic 3 diode and 3 diode with soft clip + LPF. See attachments, first the 3 diode mod plot, FFT, and schematic.

Here we have a B+ of 750v @ 200ma, 150w DC input. RL = 3750 ohms
Audio pushed to 140% positive, (1.4 * 750) + 750 = 1800V peak
Keep Alive supply set to 200V

The 3 diode soft clip + LPF will be in the next post..

Jim
WD5JKO

Here I discovered my 4 diode circuit behaved more like a two diode circuit, so I removed two of them. Then added another to keep mod tranny loaded for tube modulator. The LPF remains.

Here we have a B+ of 750v @ 200ma, 150w DC input. RL = 3750 ohms
Audio pushed to 140% positive, (1.4 * 750) + 750 = 1800V peak
Keep Alive supply set to 200V

see attachments

Jim
WD5JKO

Very good work, Jim!

It's impressive how you obtained the soft limiting vs: the hard sqaurewave of before.  

Could you explain the significance of the frequency vs: db curve? I see the effects of the L.P filter on the high freq wispy stuff, but what causes the average roll-off slope in both circuits?  Do you feel that high-freq noise w/o the LP filter is a cause of additional splatter?

I assume that is 0.15 uh for the coil in the LP filter.

I want to try that circuit and replace my simple 3-diode circuit that I'm using now. I will order some high-speed diodes at the same time.  Can anyone recommend a fast diode for say, 1KV and 3A sevice?  I could settle with 1A too.  I assume the in-line diode needs to be full plate current rated, but the three shunt legs can be less current.  Maybe 3A rating for the inline and 1A for the shunt diodes will be OK.

Jim, I wonder how you calculated the three resistor values and wattage ratings?  If it's not too much trouble, could you give me the values for a 4K plate impedance of my 4X1 plate modulated rig?

Thanks, OM.

T

Tom,

   Pass along the 4-1000 RF PA B+ to go along with the 4K RL.... :) Also what is the design maximum + mod percentage? 120%? 140%

Also the value of R1 is usually too low due to the effect of the keep alive supply. See attachment. The LPF should drop down at 3 X 6 db/Octave (As I recall). Some of the charts look close to 18-20 db/Octave.

The cap values are in Farads, and coil value is in Henries.. That 0.12 is 120mh.

Getting ready for a business trip, time to start packing...

Jim
WD5JKO

OK, Jim, good on 120mh. That's like a big plate choke, no problem.  I think you meant uf for the capacitance = .002uf.


Figure about 2KV on the final. That's the low tap and I run it there most of the time.  130% positive would be fine for my voice peaks.

Thanks.

T

zero source impedance or 1-Ohm modulators are not very practical. How can one calculate the source impedance of a real tube modulator?

for example a modulator with two 6L6's running 450V @ 210mA, making 55 watts into 5500 Ohms CT, and is matched via a transformer to a 5500 Ohm load (to make it simpler). Straight from the RCA RC-25.

At some point, an abrupt change in current at the point where the degative load diode conducts is going to reflect in the modulator's voltage waveform. 

The reason for the question is that the spice simulation will need to show distortions in the modulator's voltage waveform to accompany those shown in the modulator's current waveform.

The reason for that is to play around with making a better load for the negative cycle, one that results in the least amount of distortion, which the modulator may have to try and correct for (using neg feedback etc).

same here. I think I like a rounded negative cycle better. Maybe less harsh? Looks better anyway.
B+ 1000
modulation 2000 peak
V(n012) is the modulated B+ to the RF stage

This does consume more power but the modulator current waveform is not too badly distorted.


I would like to clear up something about % modulation to avoid confusion on my part.
Is this the way it is being discussed?
100% positive mod means the peak envelope voltage seen on the scope goes to 2x the carrier level. (4x power)
200% positive mod means the peak envelope voltage seen on the scope goes to 3x carrier voltage (9x power)

Brian,

As per your schematic, I added a 1K 200w resistor across the main diode on Fabio. Hopefully this will clean up any crud generated by the slower power supply diodes I use.  It still works FB, though I have not run any tests to see a difference.  At least I won't be tempted to tear the exisiting limiter down and buy high speed diodes now... ;D

T

Hi All,

  Just waking up here now in Thousand Oaks, Ca. I hate traveling the time zones...

Anyway, it seems to me that adding R2 at a low resistance will have the unintended consequence of boosting the keep alive current demand, and also will cause the modulator current (V1) to by very high when V1 voltage drops below the keep alive supply voltage. This means the Modulator RL will drop, and drop by a bunch.

On the soft limiter I presented I had to increase R1 almost 3X because it was influenced by R2, R3. I had to play with the ratio of R2/R3 for the desired soft limit without clip, and then play with R1 until the V1 current was symmetrical, and virtually constant over the 360 degree modulation cycle.   

Maybe tonight I can focus on this again with Fabio specific thoughts.

Jim
WD5JKO

OK on all, guys -

So far the circuit in Fabio is working FB and hanging in there.

I liked the circuit posted that showed a soft, rounded limiting to the negative peaks. It required a .15H inductor and .002 caps in a LP filter.  I wonder if this soft limiting can be done without this LP filter?  I think that the hi freq attenuation will occur from the plate bypass cap and plate choke anyway, so an additional LP filter may not be needed. It will only make the hi freq curve more complex.

I'll stick with the 1K across the main diodes for now. The mod load is about 4K which wud mean 400 ohms, but I'll stay at 1K until the modeling shows otherwise.

Jim, with your new circuit would your negative lead of the mod transformer/diode connection be equivalent to the heising choke negative lead on my rig?


BTW, I only have about 55 volts as the keep-alive voltage. It produces a very thin line at about 98% negative. I know it's working cuz the line gets thinnner when out of circuit. I don't see any reason to limit at 90% or even 95% prematurely, do you?  It's just limiting the depth of modulation for no reason. Taken to an extreme, what if we set the limiting at 50% negative?   No reason I can see not to get as close to 99% as possible, unless I'm missing something here..  There does not seem to be any momentum to spill over past 99% with this circuit, no matter how hard it is hit.

T

please include spice directive for diodes in the simulation
I want to see what you are using for Trr
My guess the problem is unmatched reverse recovery times from diode to diode.

Yo Tom, when you get it all figured out and satisfied with the results, let us know your findings. I'm going to hopefully have the Maytag, belching out fire before winter sets in again.

Tom,
Remember the forward voltage drop of the diodes reduces the keep alive voltage.

OK, Brian -

Yes, I could make up that circuit. I already have an extra string of diodes mounted on the Plexiglas and have an extra 400K 200w resistor available.

First I will wait to see what Jim comes up with too, before proceeding with the changes.

Thanks for the effort.

T

Thousand Oaks are there any left?

Brian,

I built up your latest 4-diode circuit with the 40K resistor. It appears to work fine with no problems.  I see the same negative peak limiting as with the original 3-diode circuit, but have not run any detailed tests to know if there are any improvements.

W9AD did do a spectrum look at my signal tonight and felt Fabio looked clean. I ran my audio from 1.5 kc to 7kc wide and he said the bandwidth tracked well with no splatter. I leaned into the limiter at times, so it appears to do the job.  Again, I had no side splatter crud before with the 3-diode circuit, but the new circuit is probably a better bet for the slower power supply diodes I use here.

Tnx for your efforts.

Terry: Well, there ya go, OM.  Unless the circuit gets changed, this is the final one for me. Back to the class E project... ;D


T

Is Brian's 4-diode simulation there for a 4-1000 PA running 2KV at carrier? Just trying to figure out what I am studying.

extra parts don't appear to do squat

The waveform slope hitting the clamp voltage is not modified so the level of crud generated will be the same.

why not just put balance resistors and caps across each diode. What is the purpose of the simulation?

Adjust the simulation to show it because I don't see what you mean

it does not help I see no value in a diode in series with the B+

I tried a bunch of 3 diode simulation and bottom line any time you conduct / clamp you generate harmonics. The higher harmonics can be pulled down slightly by putting a resistor in series with the bias supply to modify the slope of the waveform going through the switch point but the change wasn't that great. The simulation second harmonic was down 50 dB when not clipping and only down 20 dB when clipping in my best configuration. So the only thing you could do is add a brick wall filter out at 5 KHz or so to limit wideband crud and live with the harmonic energy in band. So best to avoid hitting the limiter with low level processing. Adding a bunch of extra parts didn't change the performance.

Frank,

Let's say we use the circuit strictly as a limiter to prevent hitting -100% negative and as a safety net for mod transformer damage.

I wonder how the side splatter crud compares when using this limiter vs: simply hitting -100% negative without it?  

I ran some tests yesterday into a dummy load while listening 7kc off freq to my voice audio. As expected, when I hit the limiter hard, I cud hear some slight crud generated. Though I didn't try it with the limiter completely out of the circuit yet.  It does seem transparent when not leaned upon.

Based on this test alone, I plan to run my audio with no leaning on the limiter at all - just used as a safety net, especially for damaging negative 100% abuse to the mod xfmr.

BTW, I also have a low level limiter inline that effectively keeps the audio below -100% neg.  (Frank, I just set it up since your visit the other day)

T

Low Z bias source 1KHz 65 ref, -20dB second harmonic, -40 dB at 12 KHz
High Z bias 1 Kohm 1KHz 65ref, -23 dB second Harmonic,-42 dB at 12 KHz
zero bias V           1 KHz 65 ref, -25 dB second harmonic, -44dB at 12 KHz

so you do get a slight advantage with a high Z bias source. I used 1 K in series with the bias and ran the voltage up a bit. Still there is less crud as the bias voltage is reduced.

I have some test results with Fabio: (single 4-1000A plated modulated by 4-1000A's)

As Frank suggested, I added a 1200 ohm resistor in series with the positive "keep alive" bias supply. With a HV probe on the limiter output I could see the effects of the high level modulator audio waveform having smoother negative peaks instead of clipping hard.

In addition, I added the 40K resistor across the 4th diode that Brian proposed.  So the circuit now looks exactly like Brian's last circuit post using 4K and 40K resistors, but has the 1200 ohm resistor included as described above. I tried different resistor values, but 1200 ohms worked out best for me. Too high and there was too little limiting - too low and the limiting was hard.

With Fabio running into a dummy load and using normal voice modulation, I could see the negative peak limiting on the RF waveform to be at about 97%. It was about 94% without the 1200 ohm resistor.

The good news is the side crud products. I first tried the rig with the limiter turned OFF. I  modulated at positive 140% and hit -100% negative.  My receiver showed S9+40 carrier on frequency. Up 8kc I could see S9 crud on the S-meter as I hit -100% negative. Below -100% modulation level there was nothing.

I then engaged the limiter and did the same test. There was NO crud at all when I hit the limiter hard (-97% negative) and the positive peaks were at 140%.  

So, I have proven to myself that the limiter circuit, as described, is definately better than having none at all when hitting the baseline at -100% negative. In fact, when leaning on it lightly for higher positive peaks, I could hear almost no change in side crud.   The only limitation was when I ran out of audio headroom and the positive peaks started to flatten - this caused just as much crud as hitting -100% negative, as expected.

I would say the limiter is a worthwhile addition to a high level plate modulated rig, in addition to some low level limiting.

T


This was my test set-up:

That is but one way. There are many others.

When the clipping starts near 95% negative, does the bias supply draw more current as the % of positive modulation increases -  or does current stay constant at this turn-on point no matter how far the audio level goes past the turn-on point?

T

Yes, more or less.  I was thinking that the ideal situation would be if there was some "sag" to the bias supply caused by the series resistor voltage drop as the current increased, then there would be a tendency for a more rounded negative peak.   Just like low level limiting, the softer peak will generate less crud and harmonic frequencies when pressed.

I'll have to get in there with my HV probe again and look at the high level audio waveform coming out of the limiter with and without the resistor. I thought it softened things, but I need to recheck this to be sure.

T 

Hi All,

I'm back looking at this thread after traveling to W6 land, and working a 65+ hour week in 6 days. Still here too, but i got Sunday off.

I see great progress was obtained with the modeling, Fabio test results, and some differences in preferred approach. I'm trying hard to see what that diode in series with the B+ does, other than take us back to the beginning of this thread.  ;)

Also, take the spice simulation and add the pi LPF like my earlier post (.002uf/.12H/.002uf) and then run up the frequency to 3 Khz, and then 5 Khz. The edges of the clip point get progressively more rounded which I think is a good thing since the upper audio frequencies in our voice will not create as much crud as the lower frequencies.

You guys are advancing the state of the art with 'Advanced Modulation'.

Jim
WD5JKO

Brian,

The resistors R2, R4 form a voltage divider when diodes D1, D3 conduct and when D2 is reverse biased (when waveform is below the blue line). For R4 at 1.2K to do much, then R2 needs to me much lower than 60K. Also when below the blue line, resistor R1 needs to be higher than RL because R1 is effectively in parallel with (R2 + R4).

If you don't see it this way, look at the AC current from the audio source. It will not be symmetrical until you increase R1 above RL.

Jim
WD5JKO

Tom,

I use a single diode keep alive circuit.  You do not need to use three diodes.  I've mentioned this many times on 75M.  I use a keep alive voltage of about 10% of the plate supply voltage.  The difference is that the keep alive supply has a series of extra RC filters.  3K with about 10ufd to ground followed by another 3K with about 3ufd to ground followed with another 3K with about 0.5ufd to ground.  This high impedence supply allows the voltage to sag on neg peaks, thus rounding the edges of the clipping.   I believe it is the hard clip that causes the crud you talk about.  The output from the supply runs through a string of high speed diodes to increase the PIV.  Each diode has a 1/2meg resistor accross it without any caps.  The cathode end of the string is connected directly to the class C end of the mod xfrm.  There is no need to use the other two diodes.  Any load you think the mod iron needs to see is presented by the power supply. 

I use a 100ma fuse between the supply and the anode end of the diode string, in case the diodes fail.  Also, I have a LED in series, which will flicker when I hit the neg peaks hard.  I leave the keep alive supply on at all times.  On standby (no plate voltage),  the LED will remain on, as the PA is drawing a little current from the supply.

My xtrm, as you know, is only 50W with a plate voltage of about 650V, but I think this circuit will also work with ur big rig.  You may have to play around with the RC values.

You can place a voltmeter at the output of the supply (before the diodes) to measure the keep alive voltage.  On standby the PA will drag the voltage down to some resting point.  This will be the lowest voltage the supply will reach when transmitting.

Fred

Fred said:

"The difference is that the keep alive supply has a series of extra RC filters.  3K with about 10ufd to ground followed by another 3K with about 3ufd to ground followed with another 3K with about 0.5ufd to ground.  This high impedence supply allows the voltage to sag on neg peaks, thus rounding the edges of the clipping. "


Interesting approach, Fred.   Maybe one of the guys here will add that to the model to see. I'd be curious.

So far I've changed my circuit three times and I'm happy with present results. It appears to work as I desire at this point, but I'm always open to further changes as new ideas come about.

I'm going to replace the 1.2K and try a 40K resistor for R4 and take some scope readings of the HV waveform. 

Jim, your comments are always appreciated to keep us all thinking and trying different approaches.


T

OK, Brian, I will check out the before and after.  Though, I cannot detect any distortion with the rig as-is. 

I raised my keep alive voltage to 138V last week. So hopefully it will be enuff when I go with a 40K resistor, but if not, I can go higher.

Here's a shot of the prototype limiter board I've been working with.  There's been lots of changes on it and it shows... ;)

T

Yo Tom,

I looked at ur limiter circuitry.  Watch out that u don't hurt yourself.  I looked at all those posted curcuits.   You don't need most all of that stuff.  If u look closely at it,  the 6K load is not in any way across the sec of the mod iron,  u have a 60K resistor in series back to the bias supply.  Where is that 6K resistor presenting a load to the mod iron?  Its not. The forth diode in series with the plate supply to the B+ end of the mod iron does what?  Nothing that I can see,  of course we now should bridge it with another resistor,  1.2K or maybe 40K ?????????  U also don't need the other two diodes or any of the resistors.  I would but two or three 2K resistors in series with the bias supply.  Add some caps to ground after each added 2K,  connect it to the one diode string that connects to the class C end of the mod iron.  Remove the rest.  The 2K resistors don't have to be high wattage as there is not much power dissipated in them.  There is only a little current drawn on neg peaks.

W5hro,

I've looked at all the pics and graphs.  I see the difference with and without R4.  I don't use the 3 diode limiter, as I've mentioned.  I use a single diode setup as I've posted.  It seems to work fine in my xmtr.  IMO there is no need for the two other diodes.  While I see the difference with R4,  IMO u don't need D4 or R4, same for D2 and D3. (hope I didn't mix up D numbers, I'll check)

Fred

and....if our audio is processed prior to application and presented correctly we wouldn't
need a three diode limiter...sheezzzzz... :-\

mo junk.


73
Jack.

Ah,... you said you modulator is, I said when your audio is presented correctly.

Presented correctly...you create your overdrive on purpose to achieve some
Psycho Acoustic effect..or is it "Presence" . oh strapping yea....

To wit your responsible for your emission bad construction practice
to leave your machine dependent on a silicon scheme....go get'em tiger  :)

When one has Plenty of Good Clean Signal we don't need over driven anything...

Alas Less Audio problem more signal quality....Listen to Don's advice....

73

Jack.

 

 

Brian,

I experimented with R3, the bias resistor.  Using just 138 volts of bias, I found that the best fit was 4K.  I expanded the scope vertical to get a close-up of the negative pinch-off line. As I hit negative ~94% with voice peaks, the line was 1/4" thick. But as the modulation increased, the line grew proressively thinner down to 1/16" thick as expected. It never hit -100% as a fine line. So we are now seeing a soft negative peak limiter as hoped.

I can see that if I started with say, 225 volts of bias, I could use a larger R3 and create a larger dynamic range. However, the present soft limit has plenty of range as-is to handle my negative peaks. In fact I start to flat top at 140% positive just as the negative peaks gets close to -99%. Perfect.

So I think the additional R3 was a good add-on that will now make the 3-diode or 4-diode limiter into a soft limiter rather than a clipper.  I like that.   The side crud test on a receiver still shows nothing up 8kc when hitting the limiter hard. Without the limiter the crud is S9, NG.

Fred:  The large power resistors are used for when the diode gets back biased. There can be full HV (3KV on high tap) across certain diodes and resistors at that point. These big 200 watt wire wounds can handle substantial breakdown voltages, whereas the small 5w types would flash over. A small rig would be more forgiving as long at the small resistor voltage ratings didn't get exceeded.  I've been changing things around so much that I used all 100 or 200 watters just to be sure.

Jack: The reason for a high level limiter is because plate modulated rigs using a mod transformer have phase shift thru the audio system. This makes maintaining voice asymmetry very difficult. Thus, a low level limiter can have poor effect by the time it reaches the final.  With the limiter at the final, limiting and phase can be well controlled.  When using a class E rig, low level ricebox balanced modulator or SDR rig, low level limiting works fine cuz of minimal phase shift.  A plate modulated transformer rig is old technology and has its limits and problems - but is cool to ride, just like a Harley.

T

Make more fire in the wire an Less audio Junk....

73

Jack.

2010 Resolution:  Build less - Operate more     ???

DZT I agree.
All this extra bullpucky does nothing. A diode in series with the B+ supply is USELESS. A resistor across a useless component is also useless.
All I see in simulation is stay with the simple 3 diode circuit. Soften the blow with a series resistor on the keep alive supply to round the edges a bit. All the other changes are BS.

Send me an email to work in the morning Tom and I will send you my Spice file so you can see for yourself.
Glad you added a NPL, actually Friday night I was thinking of a way to add a NPL to SS the driver

fourth diode doesn't do anything but drop your B+ supply a few volts.

LOL. The simulations don't do anything. They are inert and inanimate. Your interpretation of the data is at issue.

your simulation doesn't prove squat

Frank,

Where were you thinking of adding the NPL in the SS MOSFET driver?

T

Brian,

Sorry that some of the posters are breaking ur chops ;D.  Ur doing good with the computer modeling.

As I mentioned here and on the air many times, I don't agree with the whole concept of using three diodes to effect a keep alive circuit.  For at least five years I've been using a single diode setup and it works perfect.

Like many of us hams I'm from the old school.  We have been designing and building circuits and xmtrs long before anyone ever heard of a PC.

There is no need for D2 or D3 as noted in ur circuit.  The fact that u have a 60K resistor in series with the bias supply effectively disconnects the so-called 6K load resistor from the mod iron.  Look at ur own circuit and tell me how the 6K resistor along with D1 is doing anything.  Where is it across the sec of the mod iron with a 60K resistor in series with the bias supply?  The 6K load is completed from ground thru the internal resistance of the bias supply thru the 6K resistor thru the sec back to ground thru the filter caps in the plate supply.  If u add 60K in series with this path the so-called load is now in excess of 66K.

This should somewhat prove that there no need to use this separate load resistor,  therefore no need for using the steering diode D2 and following this D3 can be eliminated.

This leaves one diode, the only one u need.  A single diode string in series with the bias supply.  The addition of added resistance in this path gives u the high impedance supply u need so the bias voltage drags down when the PA pulls current from the supply on neg peaks.

The overall impedance of the bias supply and added series resistance creates this so-called load on the sec of the mod iron.  I use three 3K resistors with progressively smaller caps to ground to create a time constant in the bias voltage.  This creates the needed soft clip.

There is no need for D2 or D3  since any current flowing from the bias supply is going only one way and that is to the PA.  Any current flowing out of the bias supply in not going to flow backwards thru the sec of the mod iron when there is a 2KV voltage on the other end of the winding.

Nodel analysis tells us that all the current that flows into a junction will flow out of the junction.  Current flows into the junction of the mod iron sec and the line going to the PA.  With a 200V bias supply,  I see a 2KV voltage on the other end of the sec.  This tell me that the current is going only one way, to the PA.

Conclusion,  there is no need for D2 or D3.  And to repeat,  I see no need at all for D4 or R4.

Hope some of this helps.

Fred

Fred,
The blocking diode isolates the mod transformer when it swings below zero. A stronger bias in your case allosw one to delete this diode. Keeping with the proven 3 diode design that has been around since dirt the only thing you can do is soften the slope of the modulation waveform as it transitions to the clamp voltage. You do the same thing in your case. Adding all kinds of useless parts just makes the circuit bigger.

Fred, stop your smear tactics. We must all completely agree with Brian, no matter what he says. Don't you know he is all knowing and much smarter than any of us. His big signal over the years proves this.

another good thread comes to an abrupt end. sad

You'll have that... 8)

73

Jack.

Tom,

 I would never argue with you that which you know best, but also you know I believe in the 80 to 90% rule.

And that applies with some of these problems here.

73

Jack.

Tom,
Put your values in the simple simulation I sent you and tell us if they match performance..

I know the diode isolates the mod iron as u state.  I'm still not convinced that that is necessary.  Remember,  so long as the bias voltage at the class C end of the winding remains positive, its not negative.  It can't be both at the same time.

The problem with Brian's circuit is that the load resistor R1 is connected to the wrong side of the 60K resistor.  R1 should be connected at the bias supply.  The value of 60K is way to high.  If the bias voltage is about 200V (about 10% of the plate voltage) it would only take about 3ma of current to drop the entire 200 volts.  That would not create much of a soft clip.  Tom's 4-1000 would pull way more current than that even with only 40-50 volts on the plate

My 6146 on standby, (no plate voltage, no drive) pulls about 5-8ma with only 25 volts on the plate coming from the bias supply.  On transmit, the bias voltage comes back up to about the 80 volts.  As I hit the neg peaks the bias voltage drops down, but it never hits zero no matter how hard I hit the neg peaks.

I guess that there is nothing wrong with staying with the 3 diode setup, as u say it has been around for a long time.  But I still think that the single diode along with a properly designed bias supply is all that is needed.

Fred

Fred,
I think you are right as long as the bias is strong enough to over come the negative swing of the modulation waveform. I like simple myself.  The extra diodes just isolate the transformer and loads it. This makes the bias supply smaller. I only pointed out that the wavefoem could be rounded off eliminating some of the HF emissions created by the slope going through the switch point.

I agree about the bias voltage having the current to remain positive.  I think what happens is that the negative going peak simple runs out of headroom when it reaches the lowest positive voltage the bias supply will allow.  This being the case, the two other diodes are not needed.

Keep in mind that I use extra RC filters after the basic bias supply.  Three 3K resistors in series with progressively smaller caps to ground.  Last cap being about 0.5ufd.  These caps supply sudden surges of current.  Sort of a time constant of voltage decay.

The total internal impedance of the bias supply offers the load to the mod iron while at the same time allows the bias voltage to sag on neg peaks allowing a soft clip.  It seems to work and it looks good on the scope.

I may run another test by putting a DC scope on the bias voltage and check to see if the bias voltage drops to zero or go neg on neg peaks.  Something I never did.

Fred

Only if you let it.

It is sad when terms like smear tactics get thrown around when someone dares to disagree on a technical point. Very sad.

The earth is flat
the earth is the center of the universe

I ran some tests on the bias voltage with a DC scope.  On heavy overmodulation I could see where there were neg peaks pulling the bias voltage negative.  Even with this, I still feel that there no need to use a three diode circuit.

I made some changes to the bias supply.  I reduced the overall resistance of the RC filters in the bias supply.  I changed the three 3K resistors down to about 850 ohms each (just bridged 1200ohms across each).  I saw where I didn't have a cap on the last resistor, so a .68ufd to ground was added.  The first two resistors have 20ufd and 12ufd to ground.  I also added a .022ufd cap across the LED, thought that might be a good idea since it seemed that it was only higher freq. peaks that were going into the negative.  Maybe the LED isn't switching on fast enough.  Not really sure.

Anyway,  after making the changes which reduced the overall internal impedance of the bias supply,  I ran the same test again.  The results were greatly improved.  No peaks where reaching anywhere near the negative, which is good.  Even with some pretty heavy overmodulation the bias voltage was pulled down but remained in the positive.  The wave-envelope pattern looked really good with this setup.  Nice rounded neg peak modulation without hitting the baseline.

Seems that designing the bias supply with the correct internal impedance is important.  I use a single diode in series with the bias supply to the junction of the mod iron and the line running to the PA.  A proper bias supply is probably just as important when using the more common three diode circuit.  My single diode is a string of high speed diodes with 1/2meg resistors across each one (no caps).

When one considers the added problems of added two more strings of diodes for the three diode keep alive circuit, I still think that the single diode circuit is all that is needed.  Remember, the internal impedance of the bias supply is presenting the so-called load to the mod iron.  So, a separate load resistor in not needed.

Fred

Brian,

I know there are different ways of preventing overmodulation.  I came up with my circuit about 6 years ago, long before I ever heard of the 3 diode circuit.  In fact, Timtron described it to me over the air.

True, my rig is only a 6146, but the 6550's can put out a lot of audio.  I always get good audio reports and never any distortion reports.

My first thought was that so long as the bias voltage remains positive, you can't be hitting the baseline.  But to comfirmed this, today I put a DC scope on the bias supply.  I was surprised when I saw some negative peaks dropping below the zero line on the scope.  At first I thought maybe there was something wrong with my test setup, but really couldn't see anything that would be showing me false waveforms.  That's when I decided to rework the bias supply and reduce the internal Z.  After that, the negative peaks were remaining above the zero line with room to spare.  To do the test I was hitting the audio really heavy, way overmodulating.  Under normal transmitting conditions I wouldn't be hitting it anywhere near that heavy.

Looking at the wave envelope pattern, the negative peaks look real good, they're rounded and they never hit the base line.  This is with hitting the audio fairly heavy.  Positive modulation is hitting 150% and yet the negative peaks are not baselining.  What more could one ask for with AM.

I thing there is no reason why this single diode circuit wouldn't work with higher power xmtrs.

I did look over ur other post on this subject and the waveforms you posted.  I see why you're using the forth diode.  I guess you're using a modeling program,  something that I'm not able to do hear.  I see there are issues with the switching timing between the 3 diode.  You show how the forth diode and resistor someway corrects for this.  Good reason to give my single diode circuit a try in ur next xmtr 8).

What's the worst that could happen, you blow up ur mod iron,  they're cheap enough to replace ;D

Fred