MOSFET Dissipation (for Power Scaling / VVR)

[Gregg]

These issues have been discussed already in this thread:

http://music-electronics-forum.com/t32007/

Cathode biasing is cathode biasing and can't be a substituted for the fixed biasing.
As seen from the schematics a clever idea by Marshall in their AFD100/YJM100 models to the current powerscaling designs was to introduce a variable NFB pot section which pretty much takes care of some already known PI issues. As a result the crossover distortion is not so drastic and abrupt and can be easily avoided by tweaking the master control or applying other approaches (provided that optimal operating point at lower voltages is already set).
If somebody can offer a simple kit for fixed bias amps that would provide close to perfect linear tracking of HV and bias voltage + feedback it will be a clear winner.

[g1]

If somebody can offer a simple kit for fixed bias amps that would provide close to perfect linear tracking of HV and bias voltage + feedback it will be a clear winner.

Real "Power Scaling (tm)" kits are so far off from that?

[Gregg]

Some people here are very anxious not to reveal any details concerning the kits of this person however some competition wouldn't hurt I guess.

[mhuss]

Do your have a suggested part for the optoisolator?

Also, is that a depletion mode device?. If not, shouldn't the opto diodes be pulling "up" instead of "down?"

[R.G.]

Do your have a suggested part for the optoisolator?
Also, is that a depletion mode device?. If not, shouldn't the opto diodes be pulling "up" instead of "down?"

It's right there in the pdf...
that I just replaced. Thanks for catching that. In defense of the reversed PV diodes, I can only say that they're drawn that way in the data sheet and monkey see, monkey do while drawing. oops.

The PV is somewhat unique, for its fast quoted "on" time; the MOSFET is enhancement mode, and has a low gate charge that ought to work well with the PV - if they both adhere to their datasheets!

[mhuss]

I'm a little dense sometimes, but I still am not getting the opto connection. I guess the gate floats 'on' when open circuited?

By the way, the Vishay data sheet -- darn, it's SMD only -- shows the diodes connected the way you had them originally ("Single MOSFET Driver Application"), but the way I understand photodiodes to work (normally reverse-biased), you have it correct now. :-)

[R.G.]

I'm a little dense sometimes, but I still am not getting the opto connection. I guess the gate floats 'on' when open circuited?

Well, it was supposed to be easy until I futzed up the diagram. Photovoltaic effect is when incoming photons knock electrons out the "wrong" way on the diode, as you knew before my erroneous diagram confused you. The stack of diodes on the "voltaic" side has to be cathode to gate, anode to source.

How it works is that light from the LED hits the diodes, the cathodes go positive and shove charge into the MOSFET gate. When the light quits, they no longer charge the gate, so the MOSFET would stay on and drift sloooooowly off. Most LED/PV MOS drivers tell you to put some kind of turn-off circuit to turn it off faster; this one has a built-in active turn-off circuit to suck charge out of the gate when there's no light.

So LED-> on, MOSFET -> on. LED - off, MOSFET -> off. What's different about this pair of devices is the sub-100uS times for both of those.

By the way, the Vishay data sheet -- darn, it's SMD only -- shows the diodes connected the way you had them originally ("Single MOSFET Driver Application"), but the way I understand photodiodes to work (normally reverse-biased), you have it correct now. :-)

Yeah, I can only plead that I looked at their datasheet instead of correcting it.

[J M Fahey]

I'm interested.
I never liked linear power solutions because they're intrinsically wasteful. This seems much more elegant, the tricky bit will be managing the layout and noise...

In general that's right, but a Tube Amp is so incredibly inefficient and wasteful even under normal conditions that I guess linear voltage regulation loses a lot of the stigma.

And that wthout delving into fine points such as sagging supplies, tube rectifiers, tubes overbiased "because they sound better" and other un-green delights.

Just sayin'

[R.G.]

In general that's right, but a Tube Amp is so incredibly inefficient and wasteful even under normal conditions that I guess linear voltage regulation loses a lot of the stigma.
And that wthout delving into fine points such as sagging supplies, tube rectifiers, tubes overbiased "because they sound better" and other un-green delights.
Just sayin'

True enough, I guess. A hippopotamus on a diet is a terrible thing.

On the other hand, if you're trying to regulate down a lot of power in a tube amp and have to dump the heat out of your devices, a less-fat if not svelte hippo starts looking better. I don't mind so very much that the existing tube amp could poach an egg, if not literally fry it, but a MOSFET I put in shouldn't get overheated unless there's no other simple way. Giant heat sinks are so un-organic with tube amps.

And the one nice thing about this particular form of not-doing-linear-regulators is that it uses exactly the same rectifiers, whether tube or not, gives the same sag +/- a gnat because it's still the rectifiers (tube or not) and filter caps doing the sag.

It just takes a little electron gymnastics.

[nickb]

Sorry to necro a thread but maybe this will be of interest.

I always thought the phase control idea was jolly good idea and would be fun to try. I had a couple of iterations.

The first used a CMOS LMC555 driving a HV MOSFET inverter which then drove the main MOSFET switch. I took this approach as I just couldn't find a source of the photovoltaic optocouplers in small quantities. The downside of this method is the voltage drop across the FET is the same as the gate-source voltage required to turn it on. Just 3 to 4 volts so a huge improvement over the linear method. This, having no bias control, ended up a cathode biased amp. The timer current consumption was so low I was able to use a dropper resister off the HV supply and this made for simplicity in wiring ( rectified HV in, controlled volts out).

Version two used a PIC microcontroller to generate the timing and the bias. Again for wiring simplicity I used the rectified bias winding to provide the power for the micro. A simple power law of V^1.4 worked well for the bias voltage. There's a quite a bit of complexity in the code to ensure than things are controlled smoothly and that the pot had a nice feel. Also I use a three quadrant triac rather than a FET with a fairly sensitive gate. This meant the the power dissipation was very low and this no heatsink was needed at all. I took care in the HT wiring to keep the loop area small to avoid the noise from the fast current rise time being a problem. This ended up in Fender Twin 135W. This was a candidate as the original fibre board had become too conductive to be saved and so was virtually a rebuild.

Thanks RG for the idea

[mhuss]

Hey Nick,

I'd love to hear more about your PIC/Triac solution. Seems like overkill on the surface, but the parts are cheap and it avoids all the extra heat dissipation.

[nickb]

Try this.

This is a (untested) revision on from the one in the image above. The earlier one used a dropper resistor and a zener for the 5V supply. I also simplified the bias PWM circuit and added more protection around the FET since should the micro misbehave the FET could, in theory, remain on and self destruct.


PSPC.pdf


The pot is arranged so that the highest 5% of travel selects full power. In this mode the triac a sent a burst of pulses until half way through the cycle ( i.e. the crest ) to ensure the holding current is sufficient. Out of that range a single 10uS trigger pulse is linearly varied over 60% to 91% of the time of the cycle using the pot. The pot position is heavily low pass filtered prevent the bias getting too far out of step.

The bias control signal is pulse width modulated at 4kHz. The (high) output voltage is sensed to drive the bias calculation.

The power line frequency is auto sensed during power up to be 50 or 60Hz.

[soundmasterg]

Has anyone done a reliable 50 watt or larger amp using PS/VVR? Maybe using a CPU chiller?

I think VVR was based off an earlier version of London Power's Power Scaling circuit and it can only do around 50 watts. London Power's later Power Scaling designs can do up to 500 watts if I remember correctly. Its all in his books, including which MOSFETs to use. I've put it into several amps and regularly gig with one I made for myself that is probably 60-65 watts or so RMS. (It was 55 watts with no visible distortion on the scope, but the RMS power should be measured with 5% distortion, which I haven't done yet.) It is a gutted Bogen chassis with my own circuit and uses four 7868's at around 470V with power scaling on the power amp but nothing before that. It has a master volume at the end of the preamp to control how hard the phase inverter and power amp get hit. I forget right now which MOSFETs I used, but they have no issues at all with heat. I bolted them to an aluminum block and bolted that to the chassis. I used an actual power scaling kit from London Power for this amp and another I did for a customer that was about 25 watts. I was going to upload pics but for some reason it isn't working right now when I try to do so.

Greg

[nickb]

Greg - are the 500W versions the same basic idea ( control plate voltage with a linear dropper) or something different?

[SoulFetish]

Sorry to necro a thread but maybe this will be of interest.

I always thought the phase control idea was jolly good idea and would be fun to try. I had a couple of iterations.

The first used a CMOS LMC555 driving a HV MOSFET inverter which then drove the main MOSFET switch. I took this approach as I just couldn't find a source of the photovoltaic optocouplers in small quantities. The downside of this method is the voltage drop across the FET is the same as the gate-source voltage required to turn it on. Just 3 to 4 volts so a huge improvement over the linear method. This, having no bias control, ended up a cathode biased amp. The timer current consumption was so low I was able to use a dropper resister off the HV supply and this made for simplicity in wiring ( rectified HV in, controlled volts out).

Version two used a PIC microcontroller to generate the timing and the bias. Again for wiring simplicity I used the rectified bias winding to provide the power for the micro. A simple power law of V^1.4 worked well for the bias voltage. There's a quite a bit of complexity in the code to ensure than things are controlled smoothly and that the pot had a nice feel. Also I use a three quadrant triac rather than a FET with a fairly sensitive gate. This meant the the power dissipation was very low and this no heatsink was needed at all. I took care in the HT wiring to keep the loop area small to avoid the noise from the fast current rise time being a problem. This ended up in Fender Twin 135W. This was a candidate as the original fibre board had become too conductive to be saved and so was virtually a rebuild.

Thanks RG for the idea

Homeboy, did you custom design the main board and power supply board to be made for this project?
That looks dynamite, man. Pretty inspiring work.
I haven't dug into this threat too much yet, but is this based on the concept of sampling the incoming AC to track the phase and set the ON/OFF duration of a switching FET to limit the peak voltage over each half cycle? Is there a name for that?
The first time I came across that was reading a later iteration of RG's Mosfet Follies. I remember immediately thinking, "damn... that is a great idea. This is how it should have been done all along", and if I was ever going to build one again, I would do it this way.