Sure. The amp-second product of the return current won’t be any more than what it was on the cycle that charged the primary inductance in the first place. That’s limited by what the three valves on each side can sink. And I wasn’t seeing concerningly high cathode currents. What clued me in to the fact there was no load was how low those cathode currents actually were. They’d spike a little at clipping, but only to tens of mA. Hell, most of that was probably screen current since there wasn’t squat for plate load.
There is a long thread here that discusses the problem of damage when a tube amp is operated without a load.
Tube output protection
I have done extensive experimentation trying to find a perfect solution to this issue. So far nothing is completely foolproof, and sonically benign. My best attempt involved some TVS diodes, and discharge tubes, and some RC snubbers. These were on both the primary and secondary side of the OPT.
Nothing could prevent dead parts if the amp is continuously operated into hard clipping into an open. I was looking at sensing the light from the GDT's and using that to limit drive to prevent melting the GDT when my 41 year engineering career ended and I had to pack up and move 1200 miles on three weeks notice.
The diodes were a common solution in some guitar amps many years ago, then they mostly disappeared. I did successfully blow some diodes in my testing. When a diode blows, it fails to a short, placing the power supply directly across half the OPT's primary. I experimented with resistors in series with the diodes.
These experiments led me to discover that the diode blows from reverse breakdown as apposed to forward current as I thought. Diodes in a guitar amp overdriven 10+ db into clipping are sonically obvious, and they get HOT! it's not from forward current, it's from acting like a 1KV zener. In an open circuit the OPT behaves more like an automobile ignition coil that a transformer (that leakage inductance thing). 2 to 3 KV spikes can be seen in a cranked guitar amp on a 430 volt B+ supply WITH a speaker at resonance. A better OPT may not bee this bad though.
Attempting the fuse the DC supply is a futile effort without special 600 volt DC rated fuses. Ordinary glass fuses EXPLODE in a BIG way when run on DC and faced with a dead short. The resulting surge / arc can fry an OPT, or magnetize it. An AC fuse in the power transformer circuit can stop the current when the diode blows, but the OPT still gets to eat the stored energy in the filter caps. Those big A$$ caps in the power supply are NOT your friend here.
I managed to make a big magnet out of an OPT when I ignored the voltage rating on a feedback resistor connected to the plate of the output tube. The costing on the resistor failed and an arc ensued from the plate terminal to the ground plating on the PCB. The power supply had 1000 uF charged to 650 volts, which burned a hole in the PCB and magnetized the OPT before the power supply's current limiting could react.
I managed to demagnetize the OPT so that in no longer attracts a screwdriver, but it doesn't sound the same and the measured THD is about 1% higher than a new one.
Worse case is an amp running at full crank when the load goes open. I managed to set an identical OPT on FIRE about 20 years ago when trying to squeeze over 170 watts from a pair of 36LW6 tubes on 750 volts. I was north of 160 watts when my cheap series - parallel combo of Radio Shack non-inductive load resistors on a heat sink blew open. I promptly bought a pair of 8 ohm 500 watt resistors.....Even I can't blow them!
Coincidentally both of these experiments involved 36LW6 tubes on big voltage and some guitar amp quality OPT's that I have lots of. I bought 200 of them to make guitar amps over 20 years ago, and have only blown two. I still have a few left and don't mind offering one up to science experiments.
I still have a 500 WPC tube amp on my bucket list, and I have seen 300 WPC flow from ONE PAIR of 36LW6 tubes, but they were rather mad, red faced kinda mad. These little $16 "80VA" OPT's did successfully pass 253 watts at 1KHz into clipping. Real Plitrons (the ones I want to protect) went over 300 watts. Posts #22 and 25 here:
If UNSET and the RCA50W Had a Baby
I plan to do more testing with the small OPT's before actually building "the big one." I aim to run the amp into clipping, then disconnect the load.
Back in the 70's I saw an Ampeg SVT 300 watt bass guitar amp burst into flames when the speaker went open at full crank at an outdoor concert. In a strange twist of fate I would wind up fixing that amp about a year later. That vintage SVT ran 6 X 6550's. The older version ran 6146 transmitter tubes. They blew up all by themselves due to Ampeg's violating the screen grid ratings on the 6146. The large music store had declared the dead amp "beyond repair." The OPT was fried, as was the power transformer, some small parts, two tube sockets, and some wiring. It even had a dead 12AX7, with an open heater.
Tube output protection
I have done extensive experimentation trying to find a perfect solution to this issue. So far nothing is completely foolproof, and sonically benign. My best attempt involved some TVS diodes, and discharge tubes, and some RC snubbers. These were on both the primary and secondary side of the OPT.
Nothing could prevent dead parts if the amp is continuously operated into hard clipping into an open. I was looking at sensing the light from the GDT's and using that to limit drive to prevent melting the GDT when my 41 year engineering career ended and I had to pack up and move 1200 miles on three weeks notice.
The diodes were a common solution in some guitar amps many years ago, then they mostly disappeared. I did successfully blow some diodes in my testing. When a diode blows, it fails to a short, placing the power supply directly across half the OPT's primary. I experimented with resistors in series with the diodes.
These experiments led me to discover that the diode blows from reverse breakdown as apposed to forward current as I thought. Diodes in a guitar amp overdriven 10+ db into clipping are sonically obvious, and they get HOT! it's not from forward current, it's from acting like a 1KV zener. In an open circuit the OPT behaves more like an automobile ignition coil that a transformer (that leakage inductance thing). 2 to 3 KV spikes can be seen in a cranked guitar amp on a 430 volt B+ supply WITH a speaker at resonance. A better OPT may not bee this bad though.
Attempting the fuse the DC supply is a futile effort without special 600 volt DC rated fuses. Ordinary glass fuses EXPLODE in a BIG way when run on DC and faced with a dead short. The resulting surge / arc can fry an OPT, or magnetize it. An AC fuse in the power transformer circuit can stop the current when the diode blows, but the OPT still gets to eat the stored energy in the filter caps. Those big A$$ caps in the power supply are NOT your friend here.
I managed to make a big magnet out of an OPT when I ignored the voltage rating on a feedback resistor connected to the plate of the output tube. The costing on the resistor failed and an arc ensued from the plate terminal to the ground plating on the PCB. The power supply had 1000 uF charged to 650 volts, which burned a hole in the PCB and magnetized the OPT before the power supply's current limiting could react.
I managed to demagnetize the OPT so that in no longer attracts a screwdriver, but it doesn't sound the same and the measured THD is about 1% higher than a new one.
Worse case is an amp running at full crank when the load goes open. I managed to set an identical OPT on FIRE about 20 years ago when trying to squeeze over 170 watts from a pair of 36LW6 tubes on 750 volts. I was north of 160 watts when my cheap series - parallel combo of Radio Shack non-inductive load resistors on a heat sink blew open. I promptly bought a pair of 8 ohm 500 watt resistors.....Even I can't blow them!
Coincidentally both of these experiments involved 36LW6 tubes on big voltage and some guitar amp quality OPT's that I have lots of. I bought 200 of them to make guitar amps over 20 years ago, and have only blown two. I still have a few left and don't mind offering one up to science experiments.
I still have a 500 WPC tube amp on my bucket list, and I have seen 300 WPC flow from ONE PAIR of 36LW6 tubes, but they were rather mad, red faced kinda mad. These little $16 "80VA" OPT's did successfully pass 253 watts at 1KHz into clipping. Real Plitrons (the ones I want to protect) went over 300 watts. Posts #22 and 25 here:
If UNSET and the RCA50W Had a Baby
I plan to do more testing with the small OPT's before actually building "the big one." I aim to run the amp into clipping, then disconnect the load.
Back in the 70's I saw an Ampeg SVT 300 watt bass guitar amp burst into flames when the speaker went open at full crank at an outdoor concert. In a strange twist of fate I would wind up fixing that amp about a year later. That vintage SVT ran 6 X 6550's. The older version ran 6146 transmitter tubes. They blew up all by themselves due to Ampeg's violating the screen grid ratings on the 6146. The large music store had declared the dead amp "beyond repair." The OPT was fried, as was the power transformer, some small parts, two tube sockets, and some wiring. It even had a dead 12AX7, with an open heater.
The only thing I haven’t been able to figure out yet is why (ie, the mechanism) the power did not go up as expected with the higher voltage HT trafo. With the 430 volt, I get 70 volts peak (all on the 8R tap where the feedback is taken) with light loading playing music into a dummy load. That is with just under 600 volts DC. Over 300 watts music power. With a 250 Hz sine wave, the supply drops to 537 volts, and I measure 58 volts peak (41 RMS) to give me the 210 watts. That is all one would ever expect t from two 400 VA power trafos. 88% regulation. Par for the course for my typical 1 kW solid state beast. Given this, one would believe that there is enough screen voltage to support the peak current required for 300 watts.
Now switch to 475 volt trafos. This gave me 698 volts with no load, and 660 under music conditions with a decent peak/average ratio at the onset of clipping. Peak voltage only reached 73 volts. A measly 10% increase in power, which could easily be attributed to the plate voltage not being pulled low enough. But what had me stumped was the continuous sine wave test. I only got 63 volts peak, or 248 watts at 8 ohm at clip. The power supply was dropping to 605 volts. At slightly *under* 600 with the lower voltage trafos, I was getting MORE peak voltage (ie more efficient operation). If the peak currents were being limited by the screen supply, shouldn’t the limit be the same?
The screens are being fed by a separate 310 volt REGULATED supply. There is a dropping resistor in the drain, but would drop out of regulation at half an amp of screen current (67 volts across the resistor). I did load test the screen supply - to make sure it could withstand a momentary short. This was the reason the screen was fed from the lower rail rather than the HT - the smaller safety resistor required lest me have higher peak screen current before limiting. Could it really be drawing that much on the peaks under normal conditions? If that’s the case I *really* don’t want to be running the higher voltage anyway. But I do eventually want to understand it. There are bigger, higher voltage amps on the horizon. They may just end up being “science experiments” - but the more I understand from *this* the greater chance of success.
Now switch to 475 volt trafos. This gave me 698 volts with no load, and 660 under music conditions with a decent peak/average ratio at the onset of clipping. Peak voltage only reached 73 volts. A measly 10% increase in power, which could easily be attributed to the plate voltage not being pulled low enough. But what had me stumped was the continuous sine wave test. I only got 63 volts peak, or 248 watts at 8 ohm at clip. The power supply was dropping to 605 volts. At slightly *under* 600 with the lower voltage trafos, I was getting MORE peak voltage (ie more efficient operation). If the peak currents were being limited by the screen supply, shouldn’t the limit be the same?
The screens are being fed by a separate 310 volt REGULATED supply. There is a dropping resistor in the drain, but would drop out of regulation at half an amp of screen current (67 volts across the resistor). I did load test the screen supply - to make sure it could withstand a momentary short. This was the reason the screen was fed from the lower rail rather than the HT - the smaller safety resistor required lest me have higher peak screen current before limiting. Could it really be drawing that much on the peaks under normal conditions? If that’s the case I *really* don’t want to be running the higher voltage anyway. But I do eventually want to understand it. There are bigger, higher voltage amps on the horizon. They may just end up being “science experiments” - but the more I understand from *this* the greater chance of success.
Two quick tests come to mind.
Verify that the driver is not the limiting factor, and it can drive the grid positive if that is part of the plan.
Make sure that the bias on the output tubes hasn't shifted to far negative when the bigger transformer is used.
I test all my big amp experiments on a variable 0 to 650 volt 1.7 amp supply on B+ and a smaller supply on the screens before deciding on what power supply will be needed for the complete amp.
Big amps are often limited by the idle dissipation in the output tubes. Applying more B+ requires turning the idle current down to keep the dissipation down. This shifts the tube's overall operating point. The 125 WPC mods I made on Pete Millett's Engineers amp were tuned to keep idle dissipation to 18 watts on 24 watt tubes. Any more than 600 volts on the B+ resulted in too much idle dissipation. The tubes would go slightly over 24 watts at about 80 WPC, but nobody listens to 80 watt sine waves continuously...for long.
They are also limited by tube dissipation somewhere in the power range, varying from max dissipation at full output for a true class B amp, to max dissipation at zero output for a true class A amp.
This is usually governed by the plate efficiency in the output tube. For every tube, plate voltage, screen voltage, load impedance, and bias point there is a plate efficiency. For an amp designer starting from square one, all of these are design knobs. Often OPT load is either one of a few choices, or a given (how much power can I make with these OPT's).
At this point I start with a driver board that can drive the grids out of anything, and knobs on everything else. I then spend some making a spreadsheet to find the optimum operating point for a given tube. This is where I find that some tubes can make stupid amounts of power, and do it forever, but only under the right operating conditions. Turning up the voltage can lower the operating efficiency so that much of that added voltage turns to heat in the plate.
Playing the grid bias and the screen voltage against each other to keep the idle current in check can move that optimum point so that you MAY be able to get more power at a higher voltage. Keep an eye on the screen grid dissipation, it is often the limiting factor on some tubes, particularly TV sweep tubes.
Here is a page from a chart I made testing the venerable 50C5 tube found in old radios making a watt or two in class A. I typically make a separate page for every 20 volt step in B+ and screen voltage over the range of interest for a given tube.
Here a pair are running in AB2. I tried 3300, 5000, 6600, and 8000 ohm OPT's at plate voltages from 200 to 400, and screen voltages from 90 to 150 volts. This is the 340 plate volt / 150 G2 volt page. Plate and screen current as well as THD and power output are measured at power levels from zero to the point where some tube parameter goes out of spec. I arbitrarily chose a 10 watt plate dissipation spec, since the tubes don't red plate until about 14 watts. The G2 spec is from the data sheet, and must be respected in these (and most other ) tubes. Two different brands of tubes were tested. Yes, a pair of tiny 7 pin tubes can make 25 watts without damage. The screen grid goes out of spec at 30 watts, and glows brightly when driven to hard clipping at about 40 watts.
I could get over 20 watts at any B+ voltage from 320 to 400 volts, but 340 volts is an easy B+ voltage to make from an isolation transformer, so I optimized for that voltage. I chose to drop the screen voltage to 130 with a mosfet and incorporate current limiting into the screen supply do prevent damage in guitar amp applications. The design will make 1.4% THD at 20 watts, and it survived an overnight test at 20 watts output. I'm planning on using it for a guitar amp.
In this chart, the plate efficiency keeps going up as the drive is increased. This is usually the case if the amp is allowed to clip. If you set an arbitrary THD point (I use 3%), and look at efficiency VS plate voltage there will point where efficiency drops off, or a limiting factor comes into play. Here the screen grid dissipation is the limit. It is different for each brand of tubes, but "screen limit occurs near 3% THD" is a common trait for sweep tubes.
Verify that the driver is not the limiting factor, and it can drive the grid positive if that is part of the plan.
Make sure that the bias on the output tubes hasn't shifted to far negative when the bigger transformer is used.
I test all my big amp experiments on a variable 0 to 650 volt 1.7 amp supply on B+ and a smaller supply on the screens before deciding on what power supply will be needed for the complete amp.
Big amps are often limited by the idle dissipation in the output tubes. Applying more B+ requires turning the idle current down to keep the dissipation down. This shifts the tube's overall operating point. The 125 WPC mods I made on Pete Millett's Engineers amp were tuned to keep idle dissipation to 18 watts on 24 watt tubes. Any more than 600 volts on the B+ resulted in too much idle dissipation. The tubes would go slightly over 24 watts at about 80 WPC, but nobody listens to 80 watt sine waves continuously...for long.
They are also limited by tube dissipation somewhere in the power range, varying from max dissipation at full output for a true class B amp, to max dissipation at zero output for a true class A amp.
This is usually governed by the plate efficiency in the output tube. For every tube, plate voltage, screen voltage, load impedance, and bias point there is a plate efficiency. For an amp designer starting from square one, all of these are design knobs. Often OPT load is either one of a few choices, or a given (how much power can I make with these OPT's).
At this point I start with a driver board that can drive the grids out of anything, and knobs on everything else. I then spend some making a spreadsheet to find the optimum operating point for a given tube. This is where I find that some tubes can make stupid amounts of power, and do it forever, but only under the right operating conditions. Turning up the voltage can lower the operating efficiency so that much of that added voltage turns to heat in the plate.
Playing the grid bias and the screen voltage against each other to keep the idle current in check can move that optimum point so that you MAY be able to get more power at a higher voltage. Keep an eye on the screen grid dissipation, it is often the limiting factor on some tubes, particularly TV sweep tubes.
Here is a page from a chart I made testing the venerable 50C5 tube found in old radios making a watt or two in class A. I typically make a separate page for every 20 volt step in B+ and screen voltage over the range of interest for a given tube.
Here a pair are running in AB2. I tried 3300, 5000, 6600, and 8000 ohm OPT's at plate voltages from 200 to 400, and screen voltages from 90 to 150 volts. This is the 340 plate volt / 150 G2 volt page. Plate and screen current as well as THD and power output are measured at power levels from zero to the point where some tube parameter goes out of spec. I arbitrarily chose a 10 watt plate dissipation spec, since the tubes don't red plate until about 14 watts. The G2 spec is from the data sheet, and must be respected in these (and most other ) tubes. Two different brands of tubes were tested. Yes, a pair of tiny 7 pin tubes can make 25 watts without damage. The screen grid goes out of spec at 30 watts, and glows brightly when driven to hard clipping at about 40 watts.
I could get over 20 watts at any B+ voltage from 320 to 400 volts, but 340 volts is an easy B+ voltage to make from an isolation transformer, so I optimized for that voltage. I chose to drop the screen voltage to 130 with a mosfet and incorporate current limiting into the screen supply do prevent damage in guitar amp applications. The design will make 1.4% THD at 20 watts, and it survived an overnight test at 20 watts output. I'm planning on using it for a guitar amp.
In this chart, the plate efficiency keeps going up as the drive is increased. This is usually the case if the amp is allowed to clip. If you set an arbitrary THD point (I use 3%), and look at efficiency VS plate voltage there will point where efficiency drops off, or a limiting factor comes into play. Here the screen grid dissipation is the limit. It is different for each brand of tubes, but "screen limit occurs near 3% THD" is a common trait for sweep tubes.
The driver develops over 150 volts peak to peak. At that point, it gets a bit asymmetrical. It stays visually symmetrical at 100 volts p-p,which is all it should require. I’m running the idle bias down at 20mA per tube, and vg1 tends to run around -42V give or take. The required setting doesn’t seem to move much regardless of which power trafos were used, because the entire front end along with the screens are separately powered. This made light bulb limiting of the output stage easy during development, where the entire front end is operated normally just plugged straight in. I was able to get the full 300 watts peak power out of it (with the lower voltage supply) with the bias set as low as 2mA per tube. It didn’t even sound too bad that way, but it is better up at normal bias. It just doesn’t seem *logical* that it can’t get 300 watts if the HT can be maintained at 600 volts, if it can for single cycles before the original supply drops under load. The peak drive is the same in either case.
I suppose at some point I can drop the 475s back in the old prototype, and put in a set of trasher tubes and a smaller cheaper OPT with the same impedance and keep the frequency in the midband. Then mess with it some more, perhaps running the screen higher, or use set of sweeps that I can run the screens down at 100 volts. Don’t want to blow up a good working amplifier.
I suppose at some point I can drop the 475s back in the old prototype, and put in a set of trasher tubes and a smaller cheaper OPT with the same impedance and keep the frequency in the midband. Then mess with it some more, perhaps running the screen higher, or use set of sweeps that I can run the screens down at 100 volts. Don’t want to blow up a good working amplifier.
Hey George, just so you think of it, when you intend to disconnect the load whilst under power, make sure your contactors are OK with break voltage and break current. Had some fun making into large currents( the record so far is just shy of 9000A, maybe add another battery in ||? ). And get a bigger shunt...LOL
cheers,
Douglas
cheers,
Douglas
Wouldn’t the total continuous current be limited by the fusing current of the primary winding? Dead short and it *will* open up. How much can 20 or 22 gauge wire take before giving up the ghost? I seriously doubt that’s more than your *beaker panel* can take. Aren’t most of them have interruption ratings if 10 or 20kA?
Every flaming amp or flaming OPT incident that I have seen that was caused by a load going open started as an arc due to extreme voltage, either inside the OPT, or from pin 3 (plate) to pin 2 (heater) on the output tube socket, or inside the base of the tube itself. Once this arc starts it's fed by the DC B+ voltage. In many cases the primary side fuse, the rectifier tube, or both fail and stop the fireworks. Often the OPT, tube or tube socket are the only casualties.
In the case of the Ampeg SVT the primary side fuse was rather large, 10 amp maybe, I don't remember, but it did not blow. This amp was operating on stage at an outdoor venue where the line voltage could have come from a long extension cord, with enough resistance to keep the fuse from blowing. I was in the crowd out in front of the stage when I heard the rattling sound of a voice coil meltdown coming from the bass guitar speaker. A few minutes later there was no bass guitar, and a lot of yelling on stage with smoke coming from the bass amp. About a year later I talked to the bass player. He had used someone else's speaker cabinet that day because his was tied up in a recording session. It couldn't eat the 400 watts an SVT could make running in clipping.
For my 100 watt level testing I'll simply yank the banana cable out of the jacks at full crank. I don't know if I'll ever have the nerve to do a full power test on my 400 watt Plitron OPT's. They are no longer made, and therefore not replaceable if something bad happened.
In the case of the Ampeg SVT the primary side fuse was rather large, 10 amp maybe, I don't remember, but it did not blow. This amp was operating on stage at an outdoor venue where the line voltage could have come from a long extension cord, with enough resistance to keep the fuse from blowing. I was in the crowd out in front of the stage when I heard the rattling sound of a voice coil meltdown coming from the bass guitar speaker. A few minutes later there was no bass guitar, and a lot of yelling on stage with smoke coming from the bass amp. About a year later I talked to the bass player. He had used someone else's speaker cabinet that day because his was tied up in a recording session. It couldn't eat the 400 watts an SVT could make running in clipping.
For my 100 watt level testing I'll simply yank the banana cable out of the jacks at full crank. I don't know if I'll ever have the nerve to do a full power test on my 400 watt Plitron OPT's. They are no longer made, and therefore not replaceable if something bad happened.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.