Thanks, Bocka. I just got back for a look.
From Figure 1, page 3 of the device spec, I read that the device will hold 33A below 1V. A quick spin of Ohm's law gives 30mOhm, which matches the notation of "typical" for the chart, less than the guaranteed 40mOhm.
The normalized numbers read as you note, and even up at elevated temperatures, it's under 2x the 25C rating. That's under 80 mOHm.
You have a good eye for datasheets. The data is there, it's just sometimes well hidden and you have to **read** to find it.
80 mOhm does indeed cause some power dissipation, but you're dead right that the sub-fractional ohm MOSFETs are not the problem - the output devices are going to be glowing in the dark if I'm getting 400W off the MOSFETs 🙂
The thought occurred to me - ballast resistors. Substantially every power amp uses a ballast resistor per output transistor. Those things are up in the 0.22 ohm class, about 2.5-3 times the max, high temperature resistance of the MOSFET. They don't need 400W of heatsinking, at least in most amps.
anatech, I think you misunderstood me.
Let's agreed that the intrinsic failure rate of a MOSFET is lower than the IFR of a relay, given that each is run within its design maximums. I think that's a fairly non-controversial view. How they fail is less important than how often, as we are discussing the use of both devices as last ditch protective devices when something else has failed. The point is indeed to protect expensive electromechanical devices - the speakers - in a situation where we posit that part of the power amp has become non-functional. I probably did not make myself clear.
The MOSFETs are not in series with the amp output. They're sitting in the power supply rails feeding the power amp, pre-poer-supply-rejection and pre-feedback. While I will agree that there will possibly be some second order nonlinearity in the MOSFET rds, it is (a) small as the whole drop across the MOSFET is a tiny fraction of the amp output, and the non-linearities are small compared to that and (b) rejected by the amp's power supply rejection ratio, which is typically quite large.
I didn't intend to put MOSFETs in series with the speakers. Yes, you could hear that.
I mentioned cost just because some of the posters brought it up. I think that either a relay or MOSFET is cheap protection compared to the speakers dying, and compared to repairing the amp. On this point we're in violent agreement.
As to failure modes - it's silly to design a protective component for other than the "normal" operation of the device being a fault in the rest of the amp. Competent design of a MOSFET shutdown on a power rail to protect speakers should take into account shorted outputs, shorted output devices, inductive crossover components and speakers on the outputs, too-low speaker impedance, and all the other flotsam and jetsam that gets hooked up by ignorance and inattention to the outputs of amps.
As you note, MOSFETs are not terribly expensive compared to amp repairs, and you have already bought amp repairs if the MOSFETs get activated to shut down. So design in some fault tolerance. Use more of them.
Going back to my example device, $28.00 gets me ten of them, five in each power rail, for a combined DC current rating - not pulsed - of 310A per rail. The pulsed rating is 220A per device, total of 1100A. For my off-the-cuff 90V power rail, the current is limited to 90V divided by the resistance of whatever is connected.
Since the scenario is that the output device is shorted, the speaker is still good. Its DC resistance will limit fault current to 90A (90V/1ohm, assuming DC resistance of a 2 ohm speaker). We believe that the normal output device protection will catch output shorts - that has to be provided for that anyway.
At shutdown, all ten are trying to close down the fault current from the power supply to the output line before the speaker can oveheat. There will be conditions when the MOSFETs will not do the job, but a lot where they work just fine.
It is always tricky to guess what will fail when there is enough current flowing to mechanically move wires and make contacts weld.
Anyway, I've spent a lot more time on this than it calls for. I have proven to myself by operation on my own amps that it is possible to design MOSFET switches to switch off the DC power to an amp under some if not all fault conditions. How well this works depends, as always on the skills of the designer and the particulars of the situation.
It's something any of you can use or not, as you like. Enjoy.
From Figure 1, page 3 of the device spec, I read that the device will hold 33A below 1V. A quick spin of Ohm's law gives 30mOhm, which matches the notation of "typical" for the chart, less than the guaranteed 40mOhm.
The normalized numbers read as you note, and even up at elevated temperatures, it's under 2x the 25C rating. That's under 80 mOHm.
You have a good eye for datasheets. The data is there, it's just sometimes well hidden and you have to **read** to find it.
80 mOhm does indeed cause some power dissipation, but you're dead right that the sub-fractional ohm MOSFETs are not the problem - the output devices are going to be glowing in the dark if I'm getting 400W off the MOSFETs 🙂
The thought occurred to me - ballast resistors. Substantially every power amp uses a ballast resistor per output transistor. Those things are up in the 0.22 ohm class, about 2.5-3 times the max, high temperature resistance of the MOSFET. They don't need 400W of heatsinking, at least in most amps.
anatech, I think you misunderstood me.
Let's agreed that the intrinsic failure rate of a MOSFET is lower than the IFR of a relay, given that each is run within its design maximums. I think that's a fairly non-controversial view. How they fail is less important than how often, as we are discussing the use of both devices as last ditch protective devices when something else has failed. The point is indeed to protect expensive electromechanical devices - the speakers - in a situation where we posit that part of the power amp has become non-functional. I probably did not make myself clear.
The MOSFETs are not in series with the amp output. They're sitting in the power supply rails feeding the power amp, pre-poer-supply-rejection and pre-feedback. While I will agree that there will possibly be some second order nonlinearity in the MOSFET rds, it is (a) small as the whole drop across the MOSFET is a tiny fraction of the amp output, and the non-linearities are small compared to that and (b) rejected by the amp's power supply rejection ratio, which is typically quite large.
I didn't intend to put MOSFETs in series with the speakers. Yes, you could hear that.
I mentioned cost just because some of the posters brought it up. I think that either a relay or MOSFET is cheap protection compared to the speakers dying, and compared to repairing the amp. On this point we're in violent agreement.
As to failure modes - it's silly to design a protective component for other than the "normal" operation of the device being a fault in the rest of the amp. Competent design of a MOSFET shutdown on a power rail to protect speakers should take into account shorted outputs, shorted output devices, inductive crossover components and speakers on the outputs, too-low speaker impedance, and all the other flotsam and jetsam that gets hooked up by ignorance and inattention to the outputs of amps.
As you note, MOSFETs are not terribly expensive compared to amp repairs, and you have already bought amp repairs if the MOSFETs get activated to shut down. So design in some fault tolerance. Use more of them.
Going back to my example device, $28.00 gets me ten of them, five in each power rail, for a combined DC current rating - not pulsed - of 310A per rail. The pulsed rating is 220A per device, total of 1100A. For my off-the-cuff 90V power rail, the current is limited to 90V divided by the resistance of whatever is connected.
Since the scenario is that the output device is shorted, the speaker is still good. Its DC resistance will limit fault current to 90A (90V/1ohm, assuming DC resistance of a 2 ohm speaker). We believe that the normal output device protection will catch output shorts - that has to be provided for that anyway.
At shutdown, all ten are trying to close down the fault current from the power supply to the output line before the speaker can oveheat. There will be conditions when the MOSFETs will not do the job, but a lot where they work just fine.
It is always tricky to guess what will fail when there is enough current flowing to mechanically move wires and make contacts weld.
Anyway, I've spent a lot more time on this than it calls for. I have proven to myself by operation on my own amps that it is possible to design MOSFET switches to switch off the DC power to an amp under some if not all fault conditions. How well this works depends, as always on the skills of the designer and the particulars of the situation.
It's something any of you can use or not, as you like. Enjoy.
R.G. said:
Yeah...but output stage ballast resistors DON'T need to pass DC...only AC....and only half waves at that....
Your MOSFET is required to interrupt as well as pass nominally pure DC...(ripple ignored)...which is where the problems with dissipation come in...
Actually, it's not even that bad. 50A is what you'd see as momentary peaks with brutal loads or under shorted output transistor conditions. The speaker is presumably at least one ohm - we're not trying to protect against a shorted output.
So under actual operating conditions, you get your choice of sine wave current split two ways, positive and negative rail MOSFETs, or the much lower current of the 10 to 20db down music that goes through the amp. That's your static dissipation number. The fault current of 50A and up is a short term pulse. It lasts long enough for the MOSFET to turn off.
As I noted in my last post, it just can't be worse than the ballast resistors, which are much larger resistances.
If the MOSFETs are dissipating hundreds of watts, the OUTPUT TRANSISTORS AND BALLAST RESISTORS are emitting thermal X-rays 🙂
In non-fault conditions, the ballast resistors, output transistors, and MOSFETs are in series. That kinda makes their currents the same, right? There's a MOSFET switch in each rail.
The MOSFETs each pass a DC -voltage- that the output tranisstors let half waves of -current- through from.
(by the way, if it's not apparent, I'm talking about Class AB amplifiers, not single ended Class A amps. Those have other problems I won't address.)
Nevertheless, with adequate heatsinking, and sufficiently low Rds(on), this is probably a viable solution......certainly superior to EM relays in the signal path.... 
R.G. said:
Resistors tend to have a much greater tolerance to overload than semiconductors.....
But i reckon you'll find that a gross DC fault tends to blow out your ballast resistors pretty rapidly....
...especially if the fault is accompanied, or caused by a prolonged short-circuit at the output....
That's not the point. As I noted in my earlier post, you can fairly simply design the MOSFET switches so they are not operating in overload for the case of a shorted output transistor.
And while, yes, resistors have a better tolerance for overloads, you don't have to have overloads for the MOSFET switches, either current or power, if you care to do the design to fit the circumstances. The devil is always in the details. Broad brushes are not suited to design work.
My comment was in response to your statement that:
This is clearly not the case. I used the statement about them being in series with the ballasts and output transistors to illustrate that no, they pass exactly the same current, same conditions as the ballast resistors, not some fixed, unvarying full fault-condition current.
The MOSFETs also "don't need to pass DC... only AC... and only half waves at that...". The same current flows in both devices - until the output transistor blows. At that point, it is possible, assuming good, careful design with attention to details, to shut off the MOSFET, or MOSFETs.
By the way, I completely agree with you - switches in the power supply lines are vastly preferable to EM relays in the audio path. That's what I was trying to get away from.
Hi RG
What sort of heat-sink provision have you made for your MOSFETs?
Can you provide photographs..please?
cheers.
What sort of heat-sink provision have you made for your MOSFETs?
Can you provide photographs..please?
cheers.
I'd have to tear an amp down - I just mounted drilled and tapped holes in the main heat sink and mounted them there with sil-pads. That's one reason I was a little baffled at the power dissipation questions - the sinks don't get noticeably hotter than they did before, and the sinks are kind of marginal in these amps anyway.
These particular amps are old Thomas Vox Super Beatle guitar amps. The output transistors in these things are marginal, and so they're very prone to dying, and taking four speakers with them when they do. It's a chore to take the amp chassis out of the box, not as simple as just pulling off a cover plate.
Of course, as I said, this is a smallish amp compared to what you might want to use it on. The power supply is only +/-32Vdc. Let me be clear - I have not put MOSFET switches on kilowatt monsters. But I can't find a reason why it should not work. I can easily see that if you were shaving pennies and nickels from an amp's cost in production you might be tempted to use not quite enough MOSFET current and power capability to do the job, but that's not true for most DIYers.
The real questions to deal with for this technique is the control circuit - determining when to shut the MOSFETs down - and the gate drive. Big power MOSFETs are most reasonably available as N-channel devices, and so you need a100% duty cycle high side driver on the + power rail. I frankly expected a lot of discussion of that stuff. My control circuit is amateurish, just a single slope RC integrator and then a latch to turn on/off the MOSFETs. I think a much better job could be done.
These particular amps are old Thomas Vox Super Beatle guitar amps. The output transistors in these things are marginal, and so they're very prone to dying, and taking four speakers with them when they do. It's a chore to take the amp chassis out of the box, not as simple as just pulling off a cover plate.
Of course, as I said, this is a smallish amp compared to what you might want to use it on. The power supply is only +/-32Vdc. Let me be clear - I have not put MOSFET switches on kilowatt monsters. But I can't find a reason why it should not work. I can easily see that if you were shaving pennies and nickels from an amp's cost in production you might be tempted to use not quite enough MOSFET current and power capability to do the job, but that's not true for most DIYers.
The real questions to deal with for this technique is the control circuit - determining when to shut the MOSFETs down - and the gate drive. Big power MOSFETs are most reasonably available as N-channel devices, and so you need a100% duty cycle high side driver on the + power rail. I frankly expected a lot of discussion of that stuff. My control circuit is amateurish, just a single slope RC integrator and then a latch to turn on/off the MOSFETs. I think a much better job could be done.
Hi R.G.
As long as it works.
I still think a crowbar with a triac is the answer. Heck, if it fails, it fails short. Just trigger it for DC offsets.
-Chris
As long as it works.
I still think a crowbar with a triac is the answer. Heck, if it fails, it fails short. Just trigger it for DC offsets.
-Chris
R.G. said:
Yes indeed RG...the drivers would require isolated supplies....probably developed from the source of each MOSFET...the drain is out of question...for obvious reasons...
....a charge pump is a possibility....this is getting expensive....voltage doubler perhaps?
What is your solution?
I used an IR dual isolated gate driver (IR2011). I even went ahead and designed a small PCB to accomodate the driver and an N-channel TO247 MOSFET for each rail. My intention was to power the drivers from it's own little supply.
I went no further with the idea when I thought about the problem of how to implement latching protection and dealing with o/p stage startup glitches. And the fact that I have not actually had a problem with relays in the output, which works out smaller and cheaper and also avoids startup pops/clicks.
I went no further with the idea when I thought about the problem of how to implement latching protection and dealing with o/p stage startup glitches. And the fact that I have not actually had a problem with relays in the output, which works out smaller and cheaper and also avoids startup pops/clicks.
richie00boy said:
Yes...but if you use n-channel mosfets exclusively, you'll need two floating supplies: one for the negative rail, and the other for the positive rail...
Remember, you're switching DC, so stuff like bootstraps, isolation transformers...etc...are out of the question...viz: Your gate drive has to be sustained indefinitely..
My solution?
How about isolated photovoltaic drivers?
http://www.irf.com/product-info/datasheets/data/pvin.pdf
That's by far the simplest solution. Toshiba makes a similar device. There are a few tricks to make turn offs faster in the IRF app notes.
Frankly, it's just not that hard to take a 555 or a 7660 charge pump to make a floating voltage 10V over the power rail either. Last I checked, 555's and the appurtinent R's, C's, diodes, etc. were pretty cheap. Mouser sells NE555D's for US$0.32, resistors cost US$0.02 each and electro caps are maybe $0.10 each. I guess it depends on what your definition of "expensive" is.
On sequencing: the idea is to either not do sequencing, or to use the implied ability to do fast logic for sequencing to make the + and - rails work at the same time. I personally would not sequence the drivers and output stage. I would - as I said first - just use the power switches to switch off the DC power to the whole mess. As I envisioned it, it's a protection device, to be used in emergencies to protect things. There is certainly an opening there if you're sufficiently clever to make it do other tricks, but it doesn't make much sense to tar the basic protection use with the taint of not doing other things easily and well.
As I've said before, it's not a panacea. There are probably places where it won't work; I have not tested it exhaustively; I am not an attorney; your mileage may vary; it's not necessarily the best thing since sliced bread, but it does work in the cases where I've tested it, and it seems like a useful way to avoid putting a relay into the audio path, which is something some people want to do; you should always test it in an inconspicuous location before using it everywhere; read all the warning labels before proceeding.
How about isolated photovoltaic drivers?
http://www.irf.com/product-info/datasheets/data/pvin.pdf
That's by far the simplest solution. Toshiba makes a similar device. There are a few tricks to make turn offs faster in the IRF app notes.
Frankly, it's just not that hard to take a 555 or a 7660 charge pump to make a floating voltage 10V over the power rail either. Last I checked, 555's and the appurtinent R's, C's, diodes, etc. were pretty cheap. Mouser sells NE555D's for US$0.32, resistors cost US$0.02 each and electro caps are maybe $0.10 each. I guess it depends on what your definition of "expensive" is.
On sequencing: the idea is to either not do sequencing, or to use the implied ability to do fast logic for sequencing to make the + and - rails work at the same time. I personally would not sequence the drivers and output stage. I would - as I said first - just use the power switches to switch off the DC power to the whole mess. As I envisioned it, it's a protection device, to be used in emergencies to protect things. There is certainly an opening there if you're sufficiently clever to make it do other tricks, but it doesn't make much sense to tar the basic protection use with the taint of not doing other things easily and well.
As I've said before, it's not a panacea. There are probably places where it won't work; I have not tested it exhaustively; I am not an attorney; your mileage may vary; it's not necessarily the best thing since sliced bread, but it does work in the cases where I've tested it, and it seems like a useful way to avoid putting a relay into the audio path, which is something some people want to do; you should always test it in an inconspicuous location before using it everywhere; read all the warning labels before proceeding.
Thanks for that RG....although somewhat slow, your driver certainly saves on the extra circuitry and dedicated supplies required to implement two separate charge pumps...
I confess, i had no idea such a simple galvanically-isolated driver existed...
i.e: one that did not require a dedicated, floating supply to pull its output high in so-called high-side applications....
Further info:
http://www.irf.com/technical-info/appnotes/an-1017.pdf
http://www.irf.com/technical-info/appnotes/an-1068.pdf
Cheers.
I confess, i had no idea such a simple galvanically-isolated driver existed...
i.e: one that did not require a dedicated, floating supply to pull its output high in so-called high-side applications....
Further info:
http://www.irf.com/technical-info/appnotes/an-1017.pdf
http://www.irf.com/technical-info/appnotes/an-1068.pdf
Cheers.
Gentlemen, is problem for you to let custom made transformer with separate windings ? Or to take for it some small transformer with dual secondary windings, which cost like one " Big Mac " 😉 .
Huge problem me old mate....transformers pass AC...not DC...
As your n-channel mosfets are configured for 'normally-off' operation, you need to keep their gates at least +10V DC (preferably +12V) with respect to their sources constantly...i.e: for as long as you want your supply rails connected to your amplifier....
This means your transformer is simply out of the question...
As your n-channel mosfets are configured for 'normally-off' operation, you need to keep their gates at least +10V DC (preferably +12V) with respect to their sources constantly...i.e: for as long as you want your supply rails connected to your amplifier....
This means your transformer is simply out of the question...
mikeks said:
Hi Mikeks:
I guess (at least understood that) Upupa was talking about *supply* transformers (i.e. floating low power supplies).
That works, but only as long as they are not subject to swinging as in the case of the gate drive for the upper (if n-channel) device.
Rodolfo
- Status
- Not open for further replies.