Improved 3-Diode Negative Peak Limiter

Discussion in 'Technical' started by W5HRO, Apr 8, 2014.

Thread Status:
Not open for further replies.
  1. W5HRO

    W5HRO Administrator

    Attached below is a greatly improved 3-diode negative peak limiter with results as to why it is improved.


    After much experimentation with my transmitter I discovered that the current flowing through diode D3 in the circuit hits a brick wall as the circuit AC conducts. This was observed as random transient voltage spikes on the modulation envelope during modulation. This was further verified by running additional tests with a spice simulator and it shows exactly what happens during 150% modulation peaks. Without the addition of R2 in the circuit there is no way to prevent the spikes from occurring. The values of both R1 and R2 must also be 2 times the RFPA plate load ohms because they are essentially in parallel with each other during the negative half of the cycle. This is also more than likely why other hams through the years have reported the diodes often being blown and it's probably because of the transient spikes. It's also likely to cause excess splatter on the bands when it happens as well.

  2. KF7JAF

    KF7JAF Member

    so how much is the bias voltage on the diodes? Could a battery be used?

  3. W5HRO

    W5HRO Administrator

    That depends, but you need enough current to hold the new secondary baseline solid.

    The whole concept is to create and raise a new secondary baseline so it won't start to turn the RF amplifier tube(s) off during modulation. Some people like to raise it up to get the highest possible positive modulation peaks, but most tubes can't really handle it well. If you raise it up too high then it will start to flattop/clip as it goes too far positive.

    As a general rule limiting your negative peak to about 90% max is about right so you would adjust the bias supply voltage to achieve no more than that. In my case my modulator was built for almost 200% so I can only run it up a little. My diode bias supply (keep alive supply) runs about +100V to +150V max.

    Below is what mine looks like on the simulator set for 100% modulation with the bias supply set to only +100V. It just barely starts to clip the negative peak. Again, the supply needs enough current to hold the new secondary baseline solid though. If it starts to cave then it wont work very well.

  4. W5HRO

    W5HRO Administrator

    After receiving some questions via PM I have attached the LTspice file for this improved 3-Diode limiter circuit, but you will need the LTspice program to run it found here: LTspiceIV for Windows or LTspice IV for Mac OS X 10.7+

    In addition, you will need to go into the LTspice program directory and add a line to add the diode.

    With the Windows version go to C:\Program Files\LTC\LTspiceIV\lib\cmp\standard.dio and open the standard.dio file and add the line in blue below then save it.

    .model 20KV D(tt=0 Iave=10 Vpk=20000 mfg=W5HRO type=Silicon)

    This is a perfect diode with no delays. Without R2 in the circuit the current in D3 will be a square wave.

    Happy Simulating...

    Attached Files:

  5. KM1H

    KM1H Guest

    In amp PS diode strings I run a .0047/.005 across the full string and another to ground at the output to bypass any spikes around the diodes and to ground. With noisy power lines a lot of stuff gets stretched going thru the transformer.

  6. W5HRO

    W5HRO Administrator

    What's really happening with the original 3-diode limiter circuit with only the 1 resistor is because of the RF tube's load (RFPA Load) and not with the diodes being used.

    Attached below is what happens with the current in D3 in the original circuit and then what happens if you remove the RF tube load completely. Part of it is due the timing of the overall circuit, but the tube's load is what causes it. Adding the second resistor and making both resistors equal to the tube's RFPA load ohms together solves the problem entirely.

    Anyway, similars thing happens when you load down the output of a halfwave supply that wont happen with a fullwave supply. Same principle…

    With the RF tube load...


    Without the RF tube load...

  7. W5HRO

    W5HRO Administrator

    The attachments below should explain exactly what is happening for anyone still having trouble seeing it. The example is a have-wave output circuit with a negative cycle clip diode with the main diode DC forward biased.


    Well, whomever the BC design engineer was that came up with the original 3-diode limiter circuit design back in the day when they were using it with 866A and 872A tube diodes didn't factor in what the RFPA load would add to the equation when used in the actual transmitter. Basically they goofed, but at the same time they didn't have computer spice simulator programs back then where you could look at everything like the currents and see what was happening through each component. Unless you were a real wizard in math like Stephen Hawking and could think in multiple dimensions then mistakes like what I have found would often be made. Don't use these 3-diode limiters without using the 2 resistors. Only using one resistor can cause splatter on the bands and also risk blowing the solid state diodes when they are used instead of tube diodes.
  8. W5HRO

    W5HRO Administrator

    I've recently tested this new improved limiter design in my transmitter with the two resistors instead of only the one and it works great now. The spikes and everything I saw before are completely gone and I think the modulated bandwidth (splatter) has also been reduced now as well. I'm not even using a soft limit resistor is series with the bias supply anymore.

    I'm currently working on the transmitter's new exciter deck which will have the limiter's bias supply and I'm debating on what type of metering to use. It had a 150V volt meter to monitor the limiter's supply, but I may decide do something else before it's over. If I come up with a new idea on how to add some kind of a limiter adjustment monitor/indicator then I will post it here when everything is finished.

    Anyway, don't use these 3-diode limiters without using the new two resistor design because it definitely solves the problem.

    Old 1-Resistor Limiter Design

    New 2-Resistor Limiter Design
  9. KA4KOE

    KA4KOE Guest

    Question 4U

    Can this widget take the place of a Volumax or Optimod? I worry about frying the mod iron in my Gates but getting this bugger on-air is gonna require SOME exterior stuff. If I can build one of these devices then I can possibly delay acquiring Volumax/like gear.


  10. W5HRO

    W5HRO Administrator

    Those old CBS units are more like audio compressors for maintaining the same level of audio whether the input is low or high. It can help limit the peaks and on the negative side, but it wont work near as well as the 3-diode negative peak limiter will. This limiter WILL protect the modulation transformer, but it wont compress anything. Even if you have a Volumax you should still install it.

    What happens is when the negative going modulation peak reaches 100% it starts to turn the RF amplifier tubes off and if it goes down below that it will slowly damage the modulation transformer. The limiter and it's bias supply (keep alive supply) create a new secondary baseline so the tube will never start to get turned off therefore protecting the transformer. It also increases the positive modulating peaks when that happens.

    To be honest this 3-diode limiter AND a Volumax combination would be awesome, but the Volumax really isn't necessary. If you do build one use the new 2 resistor design because the problem I found was in fact real.
  11. KA4KOE

    KA4KOE Guest

    I've priced Volumax's and unless I find one in a Goodwill somewhere one is gonna go for at least 2 or 3 bills. What TX are you working with at the moment?
  12. W5HRO

    W5HRO Administrator

  13. AB3HT

    AB3HT Member

    This 3 diode limiter looks like a nice addition to my t-368b.
    Any good ideas on where to find 'keep alive' bias voltage in the t3? Tap an existing circuit? Lash up something new?
    73, ED
  14. W5HRO

    W5HRO Administrator

    You are probably going to need to add one unless you change out the big HV 80K bleed resistor (R58 I think) on the +2,400V supply and convert it to smaller values in series to get a tap close to the bottom to give your around +240V max. The bias supply should be no more than 10% of the HV plate supply.
  15. WD5JKO

    WD5JKO Member

    I have never seen the insides of a 368, or even put a hand on one.

    That said, I have heard that these rigs are tough to work on when energized. Maybe that is good since it is safer that way. Therefore debugging it like a Johnson Valiant on the side is not possible.

    If the bleeder is 80K on the 2400V B+ supply, then that means the bleeder current is 30ma. If you were to open the ground side of that bleeder, and insert a 9000 ohm 10 watt resistor to ground then the bleeder current drops to about 27 ma. This current through that 9K will drop 243 volts...that be 10% of the total. Put a bypass Capacitor across it (maybe 1 uf @ 450V), and that will be your keep alive supply. A 9K 10W resistor might be tough to find. Perhaps two 18K 5Watters in parallel would do.

    I don't know if this is feasible to make a keep alive supply, but I just wanted to throw this idea out there so we can bat it around.

  16. W5HRO

    W5HRO Administrator

    More like a 40uF to 100uF 450V electrolytic cap. Even 30mA is going to be a bit light, but it may still work. A 1uF won't do hardly anything during modulation and it will cave. The circuit will consume more current that what you think and will start pushing the new baseline down if there is not enough. Keeping the diodes FWD biased in one thing, but keeping the new baseline solid is another.
  17. WD5JKO

    WD5JKO Member

    I don't know if the 368 keys the B+ or not with PTT. If it does, then we want a short R-C time constant upon PTT key-up. So 80K and 100uf equates to an 8 second time constant, or about 1/2 minute to fully charge 5 time constants. If the HV is always on, then 100uf would be ideal.

  18. KA4KOE

    KA4KOE Guest

    Definitely gonna build one of these when I get to that point for the Gates. What is the voltage rating of your cap?

  19. W5HRO

    W5HRO Administrator

    That wont matter and a slow rise wont hurt anything and will help where it really does matter. When the modulation starts and the negative peak hits the new baseline the bias voltage is going to dip and not be very stable with a bleeder divider tap setup like that anyway. It will be poorly regulated so a larger cap is needed to slow it down. At least 40uF. You would just need to play around with the value to find the best compromise, but a 50uF is usually ideal. That's what is on my schematic and what I'm using with the supply in my transmitter and that supply turns on with the PTT.

    To be honest and I don't know how well it would work, but adding maybe like one of the big 200V to 250V metal zener setups to ground from the end of the 80K resistor might work. If so then you could use like a 20uF cap across that and have decent regulation.

    If you have never built one of these big limiters for use with a high power transmitter with a 2,000V to 4,000V plate supply then you wouldn't know what to expect.
  20. W5HRO

    W5HRO Administrator

    I took a good look at the B model's supply and it's not going to be that easy. You would have to get in there and disconnect the whole overload relay protection circuit and ground the center tap on the 3B28 filament transformer instead. That's probably way more trouble than what's its worth :icon_thumbdown:

    I would just add a small separate supply in there somewhere where you could also control the voltage to set limiter's clipping level. That would be best anyway.

    Pages from T-368_TM 11-809-35.jpg
  21. W5HRO

    W5HRO Administrator

    It was actually good this limiter bias supply subject came up because I ran a couple of quick simulations looking at the current being pulled out of it during modulation. If I remember correctly the transformer I'm using with mine is either 130mA or 150mA, I forget which and I have a 50uF cap on the output.

    A transmitter with a 2000V plate supply and a 200V limiter supply (keep alive supply) will limit the negative peak to around 90 - 95%. Below is the current being pulled out of the limiter's supply for such a transmitter at 100% modulation and then at 150% modulation with a 1kc tone. The simulations were ran using the same circuit already posted in this topic.

    Bias supply at 1kc with 100% modulation

    Bias supply at 1kc with 150% modulation

    As you can see at 100% modulation the peak current being pulled out of the supply is about -65mA. At 150% its about -230mA. This is why the cap needs to be fairly large. During normal talking with different voice tones and peaks the average current being pulled out of the supply will be less, but it will still be fairly substantial and certain tones or words can pull it down for much longer durations.
  22. W5HRO

    W5HRO Administrator

    I looked at my limiter supply last night and it turns out my transformer is 250mA and not 130mA or 150mA. I was thinking it was smaller for some reason :icon_wtf:

    Anyway, I drew up my actual circuit below for reference. This is what's in my transmitter now and with the exact part numbers. I must have picked the bigger transformer and just forget what size it was, but it also jives with the simulations. I remember I tried my old 4-400CG screen and grid supply transformers and the voltage just caved down too far on modulation peaks so I had to order the bigger Hammond, but I was thinking it was only 150mA max. As it turns out the larger of those old transformers I tried was 130mA. What I also have is a 40uF cap on the supply output, but then a 10uF cap on the rheostat's wiper. That's where I get the 50uF.


    With a 2000V transmitter capable of only around 100% modulation before adding a limiter a 100mA transformer would be large enough. However, if it can do 150% modulation before adding a limiter then it needs to be at least 225mA to 250mA like mine. A 4000V transmitter capable of 150% modulation before adding a limiter would need to be about 300mA for sure.

    The ideal regulation situation is to have the voltage at 100% modulation be a 10th of your plate voltage and then be a 20th during full load like at 150%. For a 2000V plate supply that would be 200V at minimum load and then 100V at full load. That's close to what my supply does, but it's really hard to achieve with the selection of components available today.
  23. W5HRO

    W5HRO Administrator

    I looked under my limiter supply chassis and saw I had put another 10uF cap on the input side of the choke. I'm sure I did that to get the voltage up to full output. The 40uF and 10uF on the output side are made up of a metal can cap mounted vertical on top of the chassis. It's a 20-20-10uF cap and the two 20uF sections are tied together to make the 40uF.

    Another thing is the 20K rheostat has always been way too big and the adjustment is so tight. It just barley adjusts off of the end before the output falls off the face of the earth. I just measured it and it's set to less than 500-ohms. I should have known it would do that, but I was trying to use just the one 50W rheostat for the bleeder and everything. Eventually when I get the chance I'm going to change out that rheostat to a 500-ohm one and just use a 20K bleeder in series with it.

    Anyway, I've updated the circuit in my last post to reflect the above. My supply only works from about 150V down to 100V and I have it set to where it never drops below 100V under max load. If I go above 100V during max load (modulation peaks) it flattops because the positive peaks are way too high for my RF amp tube to handle.
  24. WD5JKO

    WD5JKO Member


    Something looks peculiar with that pot adjustment. If the B+ from the supply is 225 volts, then the bleeder current is around 10ma. That would mean the pot end to end might vary the wiper voltage 5 volts. I think your original circuit had a larger adjustment range. Might leave the pot out, and just vary the input capacitor value to adjust the output voltage. No cap (choke input rms *.9 if bleeder current meets critical inductance) would be about 160 volts, and a big cap (rms * 1.414) would give about 250 volts.

  25. W5HRO

    W5HRO Administrator

    No, not under full load. At 150% modulation the max current peaks being pulled out of the supply are -230mA with a 200V output and -200mA with a 100V output. Granted the peaks are short unless you go awhaaaaa… into the microphone.

    Also, don't forget the 5U4 drops the voltage down a bit too, but never as much as the data sheets show if you use the right tube. The brands differ a bit and normally if you use the right one it will only drop the voltage down around 20V to 25V total instead of 50V unless you use a 5R4. A 5R4 will drop it down that far and then some. Some of the 5U4 sheets do say a 50V drop per plate, but I've never seen a 5U4 do that before. I think those ratings were based on the old peak-to-peak values of transformers and not the RMS values. With the Hammond transformer I'm using those values listed are all RMS values. I think that's where the discrepancy could be. My transformer is outputting 180Vrms to the input of the 5U4. Then after going through that and passing through the choke I'm left with about 150Vdc loaded.

    Anyway, the 500-ohm pot wouldn't adjust the voltage very far like you are saying with no load or at minimal load, but it will at full load. It's just like throwing a current limiting resistor in series with the output. It would be better to use a Variac on the primary side, but I don't want to go that route. I just want something on the front panel to adjust where I can set it quick and the rheostat works for me. It just needs to be 500-ohms max with my setup instead of the 20K. Think of the rheostat working more like a current limiter instead of like a voltage adjustment. I just set the supply for 100V during peak modulation and whatever is rises up to when it's less doesn't matter and the whole thing works sort of like a soft limit supply.

    I think part of the problem is people don't realize when you modulate a lot of current will actually be pulled out of the limiter's supply. It's more current than you would ever imagine. Think about it, how much current/voltage is your modulator inducing onto your plate supply from primary to secondary? I don't know about anyone else, but the current on my 810's peak up to 450mA on the primary side with 330mA on the secondary side from the RF plate supply load. So, how much is half of that or the difference on the secondary side for the negative half of the modulated cycle?

    Anyway, since the current is intermittent you can get away with having a supply with about the same max current as what the modulator will induce into or pull out of the limiter. If it was continuous then it would need to be at least 100mA to 200mA greater than that.

    So basically the idea of using a tap from a bleeder string like in the T-368 would have never worked. With only 25mA to 30mA available it would cave.
Thread Status:
Not open for further replies.