Super Senior Go Boom!

[WB5IRI]

Was talking to W9GAS on 7290 when there was a pop! Snap! Crackle! and flames shooting through the fan outlet on the front panel of the Super Senior. SWR was 1:1, carrier set at 275 watts, 100% modulation according to my REA monitor.
Sent the unit back to Bruce, K7DYY, to see what the heck happened. Pics show the damage.
Doug

[WB5IRI]

And a final pic.

[K1JJ]

It looks like a fairly easy repair.

I wonder... does that D design use transorbs for protection across the gates and  drains/sources?  I don't see any in the pictures across those 11N40s.

When I ran a  24 pill 11N90 E-rig, I believe they did a great job protecting the rig from my dark side tendencies.

They can be tacked in very easily.

T

[w1vtp]

You're not the only person having problems with the K7DYY transmitter.  Bruce should get with some additional safety / overload engineering or that transmitter will become the "Chernobyl Senior Transmitter."

Good luck, Doug.  Bruce is very responsive with his repairs but my friend Peter W1ZZZ just lost his for the 7th time.

I think some reengineering is in order

Al

[Opcom]

A senior moment for sure.. but I thought the bugs were worked out a long time ago.

[w4bfs]

I have no idea specifically what is happening to these rigs but I can't help but wondering if it is electromigration that could be the cause of these failures .... if a test could be run at partial power and it indicates increased longevity due to less current density in power devices, this could help  find out its problem

google electromigration, Black's equation for more information

[WB5IRI]

From everything I read, I thought the bugs were worked out, too. Could have been a fluke. We will see. Maybe some more-robust devices?

I don't have the schematic with me right now, but I will check it for transsorbs later. You would put them across the drains/gates? Never used them before, so need some advice on their use. I note that Digikey has them.

Doug

[K1JJ]

Hi Doug,

Here's a class E schematic from Steve/QIX's forum that shows a good example of where the transorbs are placed.  (Look for the diode-type devices)

http://classe.monkeypuppet.com/viewtopic.php?f=2&t=4

Notice that the 11N90 drain to source uses a unidirectional transorb, (540V  D7, etc)  while the gate to source uses a bi-directional transorb (18V  D5, etc).

I'm not sure why the class D design does not use 11N90s instead of 11N40s since the 90s are 900V devices rather than 400V.  The price is close, like about $3 each.   This would increase the existing D-S voltage survival by a factor of ~ X2.
Not sure if the output match would be the same, but could be adjusted if need be.

Anyway, maybe some of the other solid state guys can comment on why these changes would or would not be a good thing for a user that had a hot soldering iron in his hand looking for better longevity.  

(Gawd, I just love to modify stuff)

T

  

[W3RSW]

May simply be a bad batch with the continued failing repair parts coming from the same original OEM order.

I ran into this big time trying to get a faulty engine sensor repaired. Dealer replaced the sensor, the connecting wires, the computer and finally the sensor again. Each time the car would run about a hundred miles and fail again, mileage weather dependent; ran farther in cold weather.  I traded back to dealer the car in the summer knowing it would survive in around-town trips only.  Original sensor lasted about 80k miles.  Everyone was pulling hair I can tell you.  Always wondered what next owner and dealer franchise would think when finding tape wound wire trunk and documented repair record.  The next sensor from hopefully different source would work.

[WD5JKO]

May simply be a bad batch with the continued failing repair parts coming from the same original OEM order.

   It can be quite telling to remove a greased transistor from the heatsink, and examine how the grease was squished. If the two surfaces are flat and shiny, it takes very little grease to do the job. In fact, if the grease squishes out much, then there is too much grease. Too much grease might be worse than none. More is not better. The TO-220 package when tightened down by the screw hole in the tab will tilt, raising the portion needing good thermal coupling to the heatsink. Tightening the screw more only worsens the problem.

    Ran into parts quality with the commercial 3KW class E RF Generators my company services. First it was Thermalloy that changed their heat sink grease in order to be RoHs complaint (was P/N 250 now 250g). The stuff continually migrates from under transistor till the device overheats. Back to Dow Corning 340 grease. Then the power transistors used have a postage stamp sized footprint, and some of them have a domed bottom instead of being flat. The domes are not always centered either. Some were blowing in a way that the internal gate resistance (Rg) would jump from less then an ohm to 100K or more. These would still work on DC tests, but with AC drive the R-C low pass (Rg and Ciss) would kick in below 100 hz!

   Desperate for answers I call Thermalloy and IXYS. Thermalloy admits to no change going to Rohs compliance (lead free), but I can clearly see the particles used before (often aluminum oxide) are gone now. The folks at IXYS on the other hand listened, wanted failed units returned, and has sampled me new transistors. Without actually telling me, they know that their Asia foundry has big problems. The samples have domed bottoms too.  

Jim
Wd5JKO

[K1JJ]

Some more tests to do:

The damaged FET heat sink surfaces should be inspected closely with a magnifying glass for metal burrs. I have seen this cause arc-overs.

Also, carefully check all MOSFETS in operation to insure they are all close to the same temperature.  Key the rig briefly and touch all the FETS one at a time to insure that one is not running away from the others in heat.  An IR gun can be used too, though I have found a touchy-feely works just as well.  After a long transmission, they should all be close to the same temp.  If not, it pays to pull the hotter one out, re-clean the surfaces and even use some fine sandpaper to insure the surface is smooth, and paste or mount  it back up with Sil-pads.  

Some FETs are simply manufactured of different specs and may need to be replaced to get an even balance, even though they work OK.  (The weakest FETS may show the beginnings of thermal runaway)

Be sure the FET screws are all torqued down evenly.

When I ran 24 pills, it was a job to initially get all FETS to be of equal temperature, especially when they were run very hard so that small inconsistencies showed up between them..

T

[John K5PRO]

Besides fabrication variabilities or need for transient protection, single event upsets are also problems for power semis, esp MOSFET and IGBT. We have tested for them using neutron beam at LANSCE, and had 1100 volt biased IGBTS go boom in short time. Lower voltage gives long time between failures. This is a good reason to use much larger Vdss rated parts if you can live with the slower speed from the capacity.

http://www.smpstech.com/power-mosfet-single-event-burnout.htm

[WD5JKO]

The damaged FET heat sink surfaces should be inspected closely with a magnifying glass for metal burrs. I have seen this cause arc-overs.

   Tom,

   If you look carefully at the pictures, the FET body and tab appears to be plastic covered. The part number is FQPF1140C. The "F" in the P/N signifies plastic mount such that an insulator is not required when bolting to a heatsink. This will greatly ease assembly, and also help with cutting down drain to heat sink capacitance. The downside though is the thermal impedance from junction to case jumps from .93 to 2.86 degrees C / watt. For normal class D operation, the FET's run cool anyway, but during a fault, or during a shutdown sequence, that increased thermal impedance might be critical as to whether the FET survives the event or not. A lot of trade-offs in a design like this.

http://www.smpstech.com/power-mosfet-single-event-burnout.htm

  John, That link you provided seems to have a wealth of good yet hard to find information. Makes me wonder about all the semiconductors my company uses in the ion implantation systems we make. We only go up to a few Mev, but for sure opening the doors, and jumpering interlocks can result in a Geiger counter to pin the meter on the bottom 2-3 ranges.

Jim
Wd5JKO

[steve_qix]

Having done extensive solid state design, I find that in order for things to be highly reliable, you need to have:

100% voltage headroom ABOVE the _WORSE CASE_ predictable scenario.

300% current headroom above the worse case predictable scenario _AND_ normal operation within the DC rating of the device all the time.  This means:

With MOSFET such as the 11N90, with an 11A dc rating and a 44 amp peak current rating (for something like 100us) - at carrier operating at 1.1A DC, the peaks there will be about 3.5A (because class E/D do not operate at 100% duty cycle - more like 40% or a bit less).

Then you modulate - 200% positive modulation is 3x the carrier current, so now the peaks are around 11A.  Still WELL under the peak rating of the device.

Keep in mind, all of this is NORMAL operation.  Worse case scenario is an overload at full modulation.  Now this is outside of the DC rating, but well within the peak rating.  If the overload system is very fast, everything will be just fine.

Now you can see where 24 MOSFETs came from for a 1kW transmitter.  You could probably do the same power with 12 or even 10 MOSFETs, but there would be no headroom whatsoever.

Marginal designs using solid state will fail.  The more marginal, the more failures will happen.

Fast acting overload shutdown circuits are complex, but absolutely necessary.

Just some thoughts....and some reflections from experience often gained the hard way.  There were many failures along the way 

[W3GMS]

Paul-W2JTD continues to have problems with his DYY 80/40M rig.  Can't count the number of times its been back. Bruce says "check SWR".  Paul is a long time Ham and definitely knows how to run his gear.  His SWR is as flat as a dummy load, so that's definitely not the problem. 

I have never had one on the bench to look at various parameters so I can't say what the issues is.  I would be very interested in looking at any cross conduction along with voltage margin on the components.

The 160/80 meter ones seem to be solid. 

Joe-W3GMS       

[n1ps]

I understand the 80/160 units also are prone to failure.  As earlier noted there are no D-S or G-S protection devices shown on the published schematics.  No indication of any shut down circuits.  No meters to watch drain currents. I dont understand not using the higher rated 11n90s.    Guys are shipping these units back at their own expense and in some cases several times as noted.   Chernobyl Senior indeed.  Its too bad because there are few if any commercial E/D transmitters on the ham market for HF.  I hope he is able to fix his problems.  My suggestion is get some engineering help as noted by Al VTP.  Call Steve?

p

[WB5IRI]

From Bruce, K7DYY:

"I've repaired the transmitter. The FET failure is consistent with mechanical damage, probably during assembly. The source lead may have fractured the case.
I will do burn in and further testing tomorrow and email then."

Stuff happens. We'll see. If another FET dies, what would be involved in replacing them all with the beefier FETs? Surely not just a drop-in replacement, I am assuming. And the transorbs? Where, how many, what type? You are dealing with a guy making the transition from thermionic emission here, decades too late!

Doug

[WD5JKO]

Stuff happens. We'll see. If another FET dies, what would be involved in replacing them all with the beefier FETs? Surely not just a drop-in replacement, I am assuming. And the transorbs? Where, how many, what type? You are dealing with a guy making the transition from thermionic emission here, decades too late!

    One thing to consider is that transorbs often have considerable device capacitance, sometimes in the range of 1-10nf (.001 to .01uf). On 7290Khz, one RF cycle is 137ns long, or a half cycle 68.5 ns long. The act of adding a transorb will increase the FET gate rise time, and increase the FET gate fall time. The pulse width will be increased beyone 1/2 cycle, and the time in the linear region will be increased as will the switching losses. Transorbs that may work on 160m, and 80m will be a bigger problem on 40m. Looking at the circuit, there are two 160 ohm (161) resistors in parallel on each FET between gate and source. The gate side may be OK, but the Drain side is another thing more open to debate.

    The FET output topology is Class D with two bridge circuits that are series combined with an isolating transformer. A bridge circuit divides the HV B+ by two for the two devices per bridge that are OFF (open), hence the lower FET Volt Drain Source voltage rating. The Class D output is a Low Pass Filter, instead of a high Q resonant circuit as in Class E.

    The Class E circuit that many are more familiar with usually do not use a bridge circuit. With Class E, the output matching circuit is resonant, and when operating properly, the peak FET drain voltage is just under 4X the HV B+ value (3.8X as I recall), hence the higher FET voltage rating necessary.

    One thing I worry about with the DYY design is how the bridge circuit drive minimizes overlapping conduction; where one pair in the bridge is turning off while the other pair is turning on. This overlap in conduction can be a big deal, and should be avoided as much as possible.
    
Jim
Wd5JKO

[steve_qix]

I understand the 80/160 units also are prone to failure.  As earlier noted there are no D-S or G-S protection devices shown on the published schematics.  No indication of any shut down circuits.  No meters to watch drain currents. I dont understand not using the higher rated 11n90s.    Guys are shipping these units back at their own expense and in some cases several times as noted.   Chernobyl Senior indeed.  Its too bad because there are few if any commercial E/D transmitters on the ham market for HF.  I hope he is able to fix his problems.  My suggestion is get some engineering help as noted by Al VTP.  Call Steve?

p

The gate input capacitance is CONSIDERABLY higher with the 11n90s.  So, with these, the driver circuits would need to be beefier as well.  Doing a conservative design cascades through everything, and the whole transmitter becomes bigger physically, and more expensive.  But more reliable which in my book supersedes cost or size.

[K1JJ]

Stuff happens. We'll see. If another FET dies, what would be involved in replacing them all with the beefier FETs? Surely not just a drop-in replacement, I am assuming. And the transorbs? Where, how many, what type? You are dealing with a guy making the transition from thermionic emission here, decades too late!

Doug

Hi Doug,

Hopefully the rig will run well after this repair.

BUT, if the rig blows up again, based on the last two posts from Jim and Steve, I would keep the 11N40s in there for now and focus more on the transorbs and a possible outboard overload/shutdown circuit.   The question is if the transorbs will work on 40M. I think the class E designs used them OK on 40M, but I never tried them above 75M myself.  

Steve/ QIX:  Did you ever come up with a simplified overload/shut-down-board version using the Hall effect circuit or are we still using the original board?  I wonder if that could be implemented into the class D design...

T

[steve_qix]

The newest revision of the overload/shutdown board is simpler (much simpler to set up and adjust) than the original board.  But, it still takes a reasonable number of components to make this circuit.  The board does continuous samples of current and voltage, and uses an analog comparison of the two values.  If the current rises faster than the voltage, that is a an overload.

[John K5PRO]

Transorbs are essentially back to back (or single) avalanche diodes. Would it make sense to use what high power thermonic tube RF amplifiers use, gas discharge tubes? These have about 1 pF of shunt capacitance and are very reliable. I use a $10 Littlefuse CG31.5 leaded model to protect the control grid and power supply for a recent workplace HPA design. For screen grid I chose a higher energy 2600 volt "UBD" part from High Energy Devices LLC, which costs $460. This is the old EG&G/Perkins Elmer spark gap line. With a tetrode costing $330,000 and power supplies at $10,000, the cost of protection is relatively low. Photo attached shows the UBD in place in the bias filter box alongside a HV feedthru cap and current transformer to view the pulsed screen current on scope.

I checked and Littelfuse has them down to about 100 volts rating, for telecom applications, in surface mount even!
http://www.littelfuse.com/products/gas-discharge-tubes/squared-gdt.aspx

[steve_qix]

I think the gas tubes, used on their own, are too slow.  I have used them in concert with TransZorbs because they can absorb a LOT of energy.  TransZorbs are essentially instantaneous in the turn-on time.  But, they are limited in the amount of energy they can sink.  The combination is generally very effective.

I'm not at all sure if TransZorbs will help.  A totum pole arrangement (I believe that is what they're using) generally protects the devices from drain-to-anything overvoltages.  However, since the gates are connected via transformers (at least some of them are), it is possible to have gate spikes.

Over Current is another story.  Any condition where both devices in a totum pole can conduct at the same time will cause device destruction.

I started out with totum pole arrangements back in the '80s and early '90s.  Had the same sorts of issues.  Yes, they work at low frequencies, but the higher you get, the harder it is.
 

[John K5PRO]

Ok I see. With big tubes the speed of gas tubes isn't a limitation so much, as there is large mass of metal that can take an arc for so long (typical limit is 5- 10 Joules of energy) before damaging the tube badly. We crowbar the anode (plate) voltage power supply in < 10 uS and that is enough, with a series R of a few ohms at most, to minimize damage. Try that with silicon, and I suppose you will have a molten puddle or blown bond wires.