The AM Forum

  OK, this is a post about a recurring failure I am seeing on a piece of commercial gear that I do repairs on. The power supply is part of a 13.56 Mhz 3KW RF amplifier. Since this post is for a commercial part, and part of my livelihood, please moderators feel free to delete if this content is deemed inappropriate.

  There are three power conversions in this design, AC to lumpy DC, lumpy DC to 0-200V DC, and 0-200V DC to 0-3000W RF. Overall the efficiency is about 63% at full output.

  The first conversion is where I see more failures then I think there should be. I am attaching a schematic, and a few pictures of a transient filter the OEM design has.

   The three yellow discs appear to be 20mm diameter varisters, and the little white thing appears to be a sparc gap. If I hi-pot these things, the varisters break over at about 800 volts (1 ma), and the spark gap begins to flicker illumination from within at about 300 volts.

   We are seeing the big FW Bridge die (short), and then the relay on the next turn on becomes a welded mass. Then the customer resets the breaker over and over until it too fries. The power is 3 phase 208 VAC Delta. The FW Bridge has 1200V PRV. I just cannot understand how those MOV's would do much...they don't sacrifice themselves at all...they survive!

   Was thinking of just adding some 300L40 MOV's phase to phase, but that spark gap seems to be there to trap a common mode lightning bolt.

Any thoughts?

Jim
Wd5JKO

This should be a good grey matter exercise for some high power transmitter folks who live here in AMFONE.
 
I'm not a 3 phase power type, but would some imbalance between phases be killing the rectifiers? It must be pretty ugly after the diodes pop and then the relay takes the final hit.

OR how 'bout this......Are the diodes capable of handling the peak voltages and currents?
A call to the manufacturer might help for this recurring failure of the same components. Maybe they have an upgraded replacement diode available.

Fred

If it were mine I'd add a zero crossing relay and several rectifiers in series and call it a day.

I wonder if the diode bridge is failing because there isn't enough series resistance in each leg to limit the peak current flowing through each diode when it starts to conduct on each cycle.

Stu

yup ... needum step start

What is the size and rating of the cap?
What is the peak inverse voltage the diodes are supposed to be withstanding?

"OR how 'bout this......Are the diodes capable of handling the peak voltages and currents?
A call to the manufacturer might help for this recurring failure of the same components. Maybe they have an upgraded replacement diode available."

  The AC RMS current per phase is a maximum of 15 amps at 3KW RF out. The Power factor is .9 at 750 watts, and rises to .95 at 3 KW. The thing is PFC (power factor corrected) so the bridge rectifier does not have to charge up a big capacitor..there is very little energy storage in there. The diodes are rated at 30 AMPS RMS, and have a huge transient peak current rating.

"If it were mine I'd add a zero crossing relay and several rectifiers in series and call it a day."

  Not easy to get all three phases at zero crossing at the same time unless the power is off. More diode PRV though might be necessary, or perhaps better transient suppression..the topic of this post. All the power semiconductors are water cooled; everything else cooks in hot stale air. These units consume 5 KW (AC Line) to make 3 KW RF. All this in a box the size of a shoe box.

  The machines these units go into use 14 of these amplifiers. In addition there is a 100KV 100 ma DC supply, and two electromagnets that each consume 5KW. Add in a half dozen roughing pumps, and six turbo pumps, and 4 cryo pumps. Lots of power management....whole thing takes 480V 3 phase at 100 amps / phase. All this to make a multi-mega-electron-volt particle accelerator.

"I wonder if the diode bridge is failing because there isn't enough series resistance in each leg to limit the peak current flowing through each diode when it starts to conduct on each cycle."

  Not much series resistance for sure. Adding just a little would add considerable thermal heat loss and lower efficiency.

  The OEM maker added the varistors and spark gap after many years of not having them. I don't think they do much. Beefing up the transient suppression might mean more failures from blown MOV's.

Jim
Wd5JKO

According to the schematic, the MOVs don't go directly to ground so they do very little if anything.

As other have alluded to, you either have too much surge current going through the rectifier diodes or a large reverse voltage is spiking them out.

If you have PF inductors ahead of this circuit, you could have large PRV's occurring that are not being quenched.

Phil - AC0OB

Wow, it looks like an accident waiting to happen !

There should definitely, absolutely be some sort of step start or other soft start mechanism in place. There's nothing to mitigate the in-rush current - not even a power transformer to add some loss.

A couple of questions - does it always fail on turn on?  If so, inrush current is really looking like the problem.  If it fails while running, there are some major surges coming across the line from someplace.

A step-start would not be a big deal. Another relay would be necessary and of course the series resistors that are shorted out by the step start relay, and whatever you use to establish the time delay.  I use a diode going into an R/C right across the line.  The C charges through the R, and a small relay closes when a critical voltage is reached.  This relay in turn sends power to the step-start relay and closes it.  Done !

Can provide more details if necessary, but assuming the failure occurs on start-up, a step-start will probably fix the problem.

Regards,  Steve

   With modern supplies it seems that PFC is more of a cost savings marketed as a feature. They get rid of about 90% of the energy storage causing lumpy DC, and then Taylor the feedback loop to correct for the varying input at some multiple of the power line frequency. Having 3 phase power sure has an advantage since the unfiltered ripple is manageable. So with PFC they replace capacitors with Silicon. With PFC there is less EMI ripple conducted up the AC power line.

  That rectifier diode has an NTE replacement. Like all NTE stuff there is a premium price to be paid. There is also a higher PRV rated part from Powerex..I need to see, but I recall the higher PRV resulted in less current surge ratings. There is very little room to add anything. Modern supplies like this use very high density packaging, and these days everything is approaching 100% surface mount...even at multi-kilowatt levels.

  I am still contemplating using more application specific varistors such as in the picture. For example, If I used V140LA20A's instead, but wired as in the schematic drawn, then I would have 280v RMS rated protection phase to phase. Each has a 340V clamping voltage at 100 amps, so that provides 680V protection phase to phase.

http://www.littelfuse.com/products/varistors/~/media/Electronics_Technical/Datasheets/Varistors/Littelfuse_Varistors_LA_Datasheet.pdf

  The power might be so dirty that varistors as proposed would not survive...

  So why is that gas discharge spark device in there? I see these in some RF Ham transmission line products. Do they survive multiple discharges?

Jim
Wd5JKO

  It's hard to say when the failures occur. When I see these units, they are long removed from the machine. I suspect the failures (bridge rectifier) occur from multiple causes. Down stream on the DC buss is a 20A fuse that sometimes opens, but not always. The load is an Iso-Topp H-Bridge with 100A rated Fet's. Sometimes one side of the H-Bridge fuses to a short, which blows the fuse, and sometimes the bridge. When the bridge is blown the NEXT power cycle closes the contactor into a short which welds the contacts. Then the customer resets the breaker into a dead short taking it out too.

  The contactor gets power from an auxiliary 24v supply that verifies proper AC power, and then drops the contactor. So in a sense there is a step start. Without a big capacitor bank to charge up, the turn on transient is uneventful. I monitor current on all three phases; no big surge unless its broken.

  I have a nice collection of dead parts, and some spectacular photos... :o

Jim
Wd5JKO

The load draws down a fraction of the charge on the capacitor 6x on each cycle ("lumpy" DC at the output of this supply). Therefore, there are 6 current surges (one through each pair of diodes) each cycle, corresponding to the six rechargings of the capacitor from the AC source. The peak current on each of these current surges may be much larger than the rms current on each phase. The surge current rating of the bridge may not apply to surges that pass through each diode twice per cycle.

The lower the AC source resistance, the larger the peak current on each surge will be.

Stu

Could be that the MOVs (if anywhere) should be connected on the other side of the relay.  Problem might be some back surge when the relay is opened.

Fred

Sorry. I took my 3 phase hat off yesterday when I left work :-[

hi Jim .... an interesting problem to be sure .... it's beginning to sound like a problem downstream of the lumpy supply causing a failure cascade that the 20A fuse can't catch .... a turn on surge problem doesn't seem to fit after all; however, the ESR of the capacitor sets the uncharged to going charged current amount not the capacity

it would be interesting to add an overcurrent trip ckt in the 24V contactor supply feed by sensing the negative supply lead with a low value resistor and low coil voltage relay (or suitable optocoupler) .... perhaps this could stop the cascade long enuff to id the culprit

I learned at TI a long time ago that the actual lifetime of high power/current electronics is a function of electromigration of junction materials due to current density and it is finite ....

Too bad that these units can be removed by the user and now you have no idea what exactly is going on. These units read like they are updated, cheap shortcut designs, a so called improved version, of something more reliable made a number of years ago.
Does the manufacturer have any input as to what the real problem is?
If not, then it's a cash cow for you. Just replace the defective components until the user gets tired of the constant repair bills. How many users of this device?
You are repairing these for many, from a region, or just locally?
Can you give us the make and model of this device?
As you kindly reply to the responses, the more I read that this is a technical merry-go-round........too many things going wrong and no control over what exactly the problem is.
Fred

   That is a possibility, and there might be a way to do this with the space constraints.


  Thanks. These failures are usually after years of operation, so your "metal migration" concept could fit. That said, certain failures can be immediate initiating a cascade of carnage.

  Looking at the range of three phase FW bridges out there, there are many with very stout ratings.None will fit the pre-drilled, and water cooled heatsink though. Look at the IX YS stuff, this company seems to be leading the industry on many fronts, including FET's for Cla ss E transmitters. We use two of those IX YS Fet's in this box paralleled to make 3 KW RF.

  I am attaching two pictures. It occurred to me though, that some of the bridge failures might be from the isolated mount (2500V RMS rated) punching through. The Varistor / spark gap pictured in post 1 of this thread might help protect with isolation failure, but not do much with diode PRV punch through failures. That said, most of the diode failures have no outward sign of device trauma.

Good input from all, Thanks.

Jim
Wd5JKO

I was thinking turn off transients may be what is killing the rectifiers.

Some interesting ways of mitigating:

http://www.crydom.com/en/Tech/Whitepapers/3P_HC_whitepaper.pdf

Jim

Put a 0.05 ohm 50W wire wound resistor in series with each of the three branches (to keep things balanced).

With the normal output load attached, measure the voltage waveform across each resistor, using a transformer to couple the voltage across the resistor to a scope.

[Or, even better: use a current probe (rated at 200A or more) if you have one]

You may be surprised by how large (peak amplitude) the current pulses are. On each phase, you should see 2 positive pulses and 2 negative pulses in each 60Hz cycle.

I also agree that the most stressful event in this circuit will occur when the contactor closes, and the capacitor charges up from zero coulombs... but that event will occur only during turn-on... not multiple times per 60Hz cycle.

If the designers skimped on the size of the output capacitor ("lumpy DC" output), they may have reduced the turn-on stress (avoiding the need for a step start circuit) in trade for the stress that occurs multiple times each 60Hz cycle.


Stu

I would say that the rectifiers are being destroyed by transients.  The voltage rating of 800 volts on the MOVs is too high.  With them connected in Y, they won't conduct until 1600 volts appears from phase to phase.  That is going to wipe out 1200 volt diodes nicely.  I would install a more suitable MOV in delta across the phases.  The peak voltage for 208 is something under 300 volts, so I would pick a MOV around 450 volts or so, I'm not sure what's available.

It could also be you have an insufficient heat conducting path from the rectumfiers to the cold plate.

You might also try some Bi-Directional Surge Suppressors such as the NTE 4993,5, two in parallel for each phase leg:

http://www.nteinc.com/specs/4900to4999/pdf/nte4903_99.pdf

I have seen some pretty nasty three-phase power in both 208 and 480VAC lines.

Phil - AC0OB

For each pair of input lines (phases):

If the line to line voltage exceeds the voltage across the capacitor (in either polarity), the bridge will conduct current into the capacitor.

Therefore I think that a line to line voltage surge would result in a forward current surge through the associated pair of forward biased diodes (possibly destroying one or both diodes)... rather than creating a PIV problem.

For line to line voltages, the MOVs will only come into play if the contactor is open.

The MOVs allow all three lines (phases) to share the same spark gap to ground, to protect against a lightning strike that would raise all three lines with respect to ground (I.e. common mode, not line to line)

Stu

   I was looking at one of these amplifiers yesterday. The cap is 180 uf / 400 V. Now get this, they Epoxy two 27 ohm 10W WW resistors to the water cooled center plate, and the resultant 13.5 ohms is placed in series with that capacitor. Then there is a 200K 1/2 watt across the cap. So there is the impedance to limit peak current from the rectifiers. On occasion I find these resistors shattered, and open...as if the contactor had been chattering..along with the usual cascade of carnage.


   Remember that MOV's are rated for the RMS voltage in a AC circuit, and an energy absorption in Joules.

http://www.littelfuse.com/products/varistors/~/media/Electronics_Technical/Datasheets/Varistors/Littelfuse_Varistors_LA_Datasheet.pdf


  Good point! I bet there was some sort of compliance test that needed that circuit added to be certified. I used to do Mil Spec 461 testing, and all sorts of things like this had to be done.

Jim
Wd5JKO

Jim

If there really is 13.5 ohms in series with the capacitor... then the load on this supply must be drawing occasional, short duration pulses of power from the supply.

I.e. if there were (as an example) 10A of average current flowing into the load, then the 13.5 ohm resistor would dissipate an average power of 1350 watts!

However, if the load is drawing only an infrequent, short duration pulse of current from the supply, then the capacitor would deliver the corresponding charge (coulombs = the integral over time of the pulse of current being drawn by the load) and the corresponding energy (joules = the integral over time of the pulse of power being drawn by the load).

In between these occasional, short duration pulses of power being drawn by the load... the capacitor would be recharged by the AC source delivering current (and power) through the diode bridge and through the wire wound resistor.

The power flowing into the wire wound resistor would exceed its dissipation rating for the duration of the recharging process... but the wire wound resistor can handle the these short duration dissipation overloads, provided there is enough time between capacitor rechargings for the associated heat to be dissipated by the heat sink.

If all of the above is what is actually happening... then the next thing I would examine is whether the thermal conductivity to the heat sink is insufficient to dissipate the power produced in  the series resistor when the load is drawing its maximum number of short duration pulses of energy per unit time.

Stu

   Stu,

   the power buss is bypassed by four .47 uf Poly caps, and then four 0.1uf poly caps. Part of the cascading carnage is these handle a LOT of AC RMS current, and they get HOT. After about 10 years they spill their guts out. Perhaps the resistors have to handle more AFTER the caps start popping...failure is not immediate.

  The power buss feeds a big H-Bridge PWM converter about running 50 Khz. I have been replacing the .47uf's with something more readily available, and with lower ESR. These also get hot, but the overall efficiency of the unit (line in KW with PF correction) versus RF power output increases a few percent, like 62% to 64%..remember there are three power conversions in there, so 64% overall isn't too bad.

Keep in mind these designs came out 20 years ago. Pretty Hi-Tech for that time...Clas s E 3KW RF @ 13.56 Mhz...The size of a shoe box.

Jim
Wd5JKO

Jim

I am still confused as to how there can be 13.5 ohms of resistance in series with the capacitor... unless the load that it is feeding is drawing only occasional pulses of power from this supply.

I guess that I need a more complete schematic of this power supply (showing the bridge, the 13.5 ohm resistor, the capacitor, and how the load is connected) and a better understanding of the nature of the load it is feeding.

Stu

Jim

Since you said that this power supply is one component of a large particle accelerator... is it possible that this supply is basically sitting idle (i.e. delivering no power or very low power to the bigger system) until someone "pushes a button" that causes everything to very briefly turn on to inject a single bunch of electrons into the accelerator. Then, this power supply would sit idle until the next bunch of electrons needs to be injected into the accelerator.

Stu

If the above is correct, then there could be a large transient voltage pulse produced between both sides of this supply's capacitor and the grounded side of the spark gap (common mode) every time the larger system is pulsed to inject electrons into the accelerator.

Given the path to ground through the MOV's and the spark gap ... this could result in a large pulse of current passing through half of the diodes in the bridge. If the amplitude of the current pulse is large enough... those diodes could be damaged.

If that same current flows through the 13.5 ohm series resistor, it may also cause a voltage drop across the remaining three, reverse biased diodes... that exceeds their PIV rating.

Stu

May be a late reply, but the sparkgap surge arrestor   in series with the mov's to ground is  to prevent a to high leakage current from the mov's to ground. A sparkgap has nearly no leakage if not ignited and when its ionised it has something as 0.5 Ohm. The mov's have always a rest current of something as 1 mA each.
The residial  current as in the standards ( UL, Csa, Kema , Gs etc) is less then 0.5 mA. This cannot be met by mov's alone.
Anton

  Makes sense Anton. This arrangement has its purpose in compliance issues, but seems to do little to protect that FW bridge, unless it is to keep the isolated bridge package from breaking through beyond its 2500 V RMS rating. Same for the ISO-Topp FET's in the H-Bridge switcher.


   Maybe the attached scope plots will better illustrate whet I was trying to express in words. At 2ms per division the +/- sides of the bridge rectifier are lumpy DC ranging from 90-180v, essentially providing dirty power to the high frequency switcher. Both plots are of the same thing, only the scope sweep rate changes.

   I also added a PDF showing one stage of the linear accelerator (linac). Each stage has RF phase shifters to optimize the energy multiplication, and transport efficiency. All the Rf amplifiers slave off a common Rf source that is called the master oscillator. The resonator cans are a bit larger then a beer keg, are VERY high Q...just a few Khz from tuned and the reflected power gets really high. The RF amplifiers have a built in protection circuit to limit the reflected power to no more then 250 watts. Usually when tuned there can be as high as 2750 watts forward, and < 5 watts reflected. The tuning loop is always active since things change fast as parts warm up..even with water cooling.

   The RF stuff runs at various power levels, and not all are used at a time. For some needs they all run (up to 14 of them). They are on CCS, often 24/7. The RF is on whether or not an ion beam is running through the linac. It might take a Forrest of trees to offset the carbon foot print from one of these beasts.

Jim
Wd5JKO

Jim

In the second oscilloscope photo... that shows the 2ms per division time scale:

It appears that you are using a pair of single-ended (grounded) probes (2 superimposed traces), rather than a differential probe.

If so, what is the "ground" side of each probe connected to?

I see 3 pulses in the top waveform in 1/60th of a second (16.7ms). I would expect to see 6 pulses in the top waveform in 1/60th of a second. I.e. in each 1/60th of a second (separated by 1/360th of a second)

Phase 1 - Phase 2 reaches its peak positive value
Phase 1 - Phase 3 reaches its peak positive value
Phase 2 - Phase 3 reaches its peak positive value
Phase 2 - Phase 1 reaches its peak positive value
Phase 3 - Phase 1 reaches its peak positive value
Phase 3 - Phase 2 reaches its peak positive value


The first oscilloscope photo (longer time scale) shows the voltage switching between two levels (as if the load were switching between two values). What is the time scale (time/division) of the first photo?

Stu

Stu,

   The time scale is part of the picture label, i.e 2ms and 2 us per division

The H=Bridge is very similar to this one:
http://omapalvelin.homedns.org/tesla/SSTC/pics-fullbr/bridge.gif

  The + and - power in the image go to the + and - outputs of the 3 phase diode bridge.

The capacitor in series with the transformer is four times 4.7 uf polypropylene in parallel. These get HOT but seldom fail. The transformer is a mule, about 2"X2"X3", and is water cooled. The secondary output is ground referenced...(FW CT rectifier follows) so it isolates the AC Line stuff while making 0-200VDC at around 20 amps maximum.

The capacitors in the image between the + and - rails are the ones that fail. Local to the H-Bridge, clustered between the four FET's are six (I said 4 earlier) .1 uf polypropylene, and then on a daughter PCB there are four .47 uf polypropylene all in parallel across the +/- buss.

The scope probe was ground referenced looking at the transformer primary..the side with the series capacitor, hence seeing +/- 150v (with 3 phase power line ripple on the buss).

Jim
Wd5JKO

Jim

okay... with respect to the 2nd oscilloscope trace photo...I think I get it.

The scope ground reference for these measurements (two probes in the 2nd oscilloscope picture) is equivalent to putting a local balanced "Y" load directly across the 3 phase AC source... and defining the point where the 3 legs of the "Y" come together as the measurement ground reference. This measurement ground reference would not be the same as the AC source ground to which the spark gap is connected (which could have all sorts of noise on it, relative to this measurement ground reference). If the oscilloscope has an "a-b" feature, I guess you could subtract the lower trace from the upper trace to display the voltage between "+" and "-", and to cancel out the effects of any extraneous noise on the measurement reference point relative to both "+" and "-".

With respect to the first oscilloscope photo (2us per division), I guess I don't understand how this system works. It is hard for me to see how the diode bridge output voltage (either + or - side) with respect to measurement ground can change that quickly (rise/fall time much less than 2us)

Stu

Jim

I think I now understand what 1st oscilloscope trace photo is.

There is only one probe producing the one trace that is shown in the 1st photo (whereas, there are two probes producing the two traces shown in the second photo)

The (single) probe's tip is attached to the + side the diode bridge; and the "ground" lead of the probe is attached to the "H bridge" at the point where the source of the top FET on the left side of the "H bridge" attaches to the drain of the FET below it.

It appears that the four switching inputs to the H-bridge are 100kHz (approximately) square waves (i.e. a little longer than 10us to complete one cycle).

Therefore, as the H bridge FETs switch on and off at 100kHz (approximately), the probe's ground connection point is, in effect, switching at 100kHz (approximately) between the voltage on the "-" side of the diode bridge (plus the voltage drop across the bottom FET that is conducting), and the voltage on the "+" side of the diode bridge (minus the voltage across the top FET that is conducting).  

Therefore, what the scope shows is a 100kHz (approximately) square wave whose peak-to-peak amplitude is approximately the voltage between the "+" and "-" terminals of the 3 phase bridge... with the 3 phase ripple superimposed as a fuzz (due to the overlapping of multiple traces that are all triggered to sync with the square wave).

Anyway, getting back to the failures in the 3 phase diode bridge:

It appears that the 13.5 ohm resistor is limiting the average voltage across the capacitor on the output side of the 3 phase bridge... and also setting the shape and amplitude of six charging current pulses per 60Hz cycle that flow through the diodes in the bridge into the capacitor.

However it also appears that the load (power delivered to the "H bridge") is being drawn directly from the +/- terminals of the diode bridge. I.e. the load is not across the capacitor... but, instead, across the series combination of the capacitor and the 13.5 ohm resistor.

Therefore, charging current is flowing into the capacitor via the 13.5 ohm resistor, and discharging current is flowing out of the capacitor via the same 13.5 ohm resistor... and the average power dissipated in the 13.5 ohm resistor (which depends upon the integral of the square of the current) is less than it would be if the "H-bridge" load were directly across the capacitor. 

What is still surprising to me is that the wire wound resistors and the associated heat sink (a pair of 27 ohm 25W resistors in parallel, epoxied to the heat sink) can handle the power that is being dissipated in those resistors... 

Stu

Stu,

  Both images are done with one scope probe, and use only one scope channel. Both are done under the same exact conditions within the RF amplifier, i.e RF on at 500 watts RF output. The only difference was the scope sweep rate, 2 ms /div and Line sync to see the 3 phase ripple, and 2 us / div and channel 1 sync to see the ~ 100 KHZ switch rate.

   Let me make a comparison to something else. Imagine a class AB solid state audio amplifier with complimentary NPN and PNP output transistors. There are dual supplies, + and -. The connection between the two transistors (the emitters) will drive a speaker. The +/- power supplies are referenced to a common point (ground). The power supplies have lousy filtering so the ripple at 120 Hz is about 5%. The amplifier has about 20 db NFB.

  So lets drive this amplifier with a 1 Khz sine wave until it is running 50% of whatever the watt rating is into the speaker load. A nice sine wave input becomes an amplified sine wave out. The power supply ripple makes no difference because the amplifier is running well within its dynamic range, and there will be a high power supply rejection ratio (PSRR). The scope is hooked to the output, and to the common point (ground).

  Now lets drive the amplifier until it clips badly as the output slams against the power supply rails. When this happens, the 20DB NFB goes away, and PSRR goes to zero. The clipped portion of the sine wave (1 khz) will now have 120 hz ripple present riding on the flattened peaks. When this is happening, if you adjust the scope sweep to 2 or 5 ms / division, then the scope will mostly show what appears to be two traces of 120 hz ripple...i.e the top trace will show the + supply ripple, and the bottom will show the - supply ripple. if the scope is a digital one, the 1 Khz portion might not even be visible.

  The H-Bridge switcher on each side is behaving like the audio amplifier driven into hard clipping. The ripple is common to each side of the transformer, and the series cap to one side of the primary removes any DC component, and also acts like a high pass filter. There is NO ripple in the 0-200V output that this H-Bridge feeds into.

  Back to the AC 3 phase supply feeding this amplifier. The power distribution to the test building is 480v delta 3 phase. Then it is stepped down to 208v 3 phase 'Y'. Here at the common point in the 'Y', we connect both the neutral, and ground wires at this common point. The electrician ran the 3 phase 208 service to the test room using a 4 wire cable (3 phases and ground...no neutral). Since the ground wire connects back at the transformer the same as a neutral, grounding my scope to the amplifier chassis (power line ground) is a valid way of doing this measurement of the power buss ripple.

  If I wanted to make meaningful measurements of the 100 Khz into the transformer, I'd need two scope probes, sum the channels, and invert one of them to make the measurement. I have two nice 100X compensated probes, so I may try this. I would connect a probe to each side the the transformer primary.

  So what is the rms ripple % of unfiltered 3 phase power? If I follow this thread, I take it to be about 5%.

http://forums.mikeholt.com/showthread.php?t=132994

  These RF amplifiers go back about 20 years. At one intermediate point, the OEM was putting a power diode across that 13.5 ohm resistor (two 27 ohms in parallel). I don't remember which orientation. Whenever I run across one of those diodes, if I need to dismantle that module, I will remove the diode. The 'current' production amplifiers do not use it, nor do the real old ones.

Jim
Wd5JKO

Jim

Excellent analogies and reference thread!

I learned a lot "discussing" this with you and the other posters.

I'm sorry that I can't provide much regarding the cause of the 6 diode bridge failures, other than those caused by downstream short circuits, and possibly the temperature rise of the heat sink from the power dissipated in the wire wound resistors.

Best regards
Stu

Late to this party but 2 cents..

On the single phase setup here, 120V primary peak current at full power is about 54A each half cycle (2.5A peak rectifier current)
(two supplies, one on each side of the 240V to neutral). Not too surprising but substantial, and is the reason why a 240V 50A dryer outlet was put in.

When keying the TX, the simulator says the inrush for one cycle is by simulation 230A but it's probably a lot less due to wiring resistances etc. There was no digital scope to capture the transient with at the time I measured. Anyway it translates to 10.5A through the rectifier for the inrush.

On transients when turning off the HV, which is done by a relay. The set is keyed by control of the power to the primaries of the plate transformers and final screen supply transformer. When the relay is opened there can be a transient over 15KV on the rectifier plate leads. This is why MV rectifiers are used. I am too cautious but don't like large strings of rectifiers and would not put less than a 30KV SS rectifier in place of a 673 rates 15KV. The rectifier that blew on January 18, 2009 (that took a  33A plate transformer relay (http://amfone.net/Amforum/index.php?topic=18335.msg127928#msg127928) with it) went because it was bad, not a flashover.

I know it's not the same situation but maybe the information will parallel what you are having and suggest things.

Is the diode stack water cooled? Ok. "Loss of water cooling not reported by customer?"
-For whatever reason.  
Chain of custody information, Pump, valve or flow failure, etc.

  Good question!  Yes that happens. Having 14 parallel, and individually valved 2 gallons / minute flow loops is hard to manage. On occasion the pumping system (a remote chiller) aerates the water such that bubbles build up over time, at a valve, or elbow. These can restrict water flow.

  What should happen is that the amplifier senses an over temp condition (Airpax TO-220 cased sensor), and then shuts down. Sometimes this causes the operator to condemn an amplifier as bad, and a perfectly good amplifier goes through the repair loop. Putting in a new unit will purge out the water lines, so the person believes the old amp was bad, and the new one fixed it. Sometimes the Airpax sensors are bad, and the amplifier does self destruct. The symptom is that the Thin-film RF power parts get hot enough to unsolder themselves, and cause one hell of a flame spray in the process. I have seen several of these bad sensors, and when I find a bad one, that 3 phase bridge rectifier survived.

  One test performed is to run the amp at 1500 watts with the test set water interlock defeated. If an amplifier does not trip off in 2 minutes, then the sensor is bad.

Jim
Wd5JKO

Sounds like the whole thing was designed by physicists  ;D

In some situations loss of data for a moment or two is more critical than meltdown and calling you.
Don't think your ever getting the whole story, heh, heh.

As mentioned, keep the cash cow and don't appear too bright.