Hotwire PWM Regulator for Mag D body, JM-PhD-D1

[JimmyM]

Test conditions: 20V in, 12VRMS out, 64625 bulb.

This is the OFF transition (voltage measured across the FET)
10V/div, 5uS/div
Green = FET Driver with 510 Ohm resistor in series with gate
Pink = AVR direct drive with 68 Ohm resistor in series with gate.



This is the ON transition (voltage measured across the FET)
10V/div, 5uS/div
Green = FET Driver with 510 Ohm resistor in series with gate
Pink = AVR direct drive with 68 Ohm resistor in series with gate.


I know the FET Driver has a 510 Ohm resistor in series, but the AVR with the 68 Ohm resistor (calculated in Dev thread) shows a nice drive (slew rate-wise) , but the voltage spike at turn off holds at the "kick voltage" for longer.
I'm going to turn things up a bit and see what happens.

 

[JimmyM]

Here's the scope...
Conditions: 29.4Vin, 12VRMS out, 64625 bulb. AVR direct drive.

10V/div, 5uS/div

The kick is still there, as before. But the 30V FET I'm using is getting a 41.6V Vds whack every time it's shut off. But it's not complaining.
Also, it's been running like this for the last 15 minutes with no heat problems. I.d say it's a little warmer than before, but nothing serious so far. I hope not, it's my last IRLU7843.

I could probably increase the AVR-to-gate series resistor to 100 Ohms to further reduce the demands on the AVR's internal FET and still get good gate performance.

 

[JimmyM]

When the FET driver and AVR both habve a 100 Ohm resistor in series, the waves are precicely the same. Stored wave forms for each lay right on top on one another.

 

[Mr Happy]

While I'm following this thread, let me throw in a dumb question:

Is is possible that the voltage spike being attributed to the bulb filament is also partly or mostly from the inductance of the wires from the circuit board to the bulb? And if that happened to be the case, could the spike be reduced by twisting and tying the wires together so that their inductance in each direction canceled itself out? In addition the twisting might introduce a bit of capacitance that would also help to balance out any remaining filament inductance?

Forgive me if this question is too dumb for words, but sometimes it doesn't hurt to ask the crazy questions...

 

[JimmyM]

While I'm following this thread, let me throw in a dumb question:

Is is possible that the voltage spike being attributed to the bulb filament is also partly or mostly from the inductance of the wires from the circuit board to the bulb? And if that happened to be the case, could the spike be reduced by twisting and tying the wires together so that their inductance in each direction canceled itself out? In addition the twisting might introduce a bit of capacitance that would also help to balance out any remaining filament inductance?

Forgive me if this question is too dumb for words, but sometimes it doesn't hurt to ask the crazy questions...

That's not actually a bad thought.
I forst separated the power leads as much as I could.
Then separated them.
Below are the results.
Green = wires separated
Pink = wires together.
White = Shortened wires together.
It's tough to tell the difference between the pink and white. But the white one seems to be the shortest duration kick.

 

[Mr Happy]

Interesting.

Does the flat top of the spike look to you as if it has been clipped? From the shape of it, it almost seems as if it should extend upwards by another two or three divisions to a sharp peak, but has been clipped after a rise of about 40 V as if by a voltage breakdown somewhere?

Is it just my imagination, or is the ringing more prominent in the white trace?

If the spike is caused by inductance, maybe it could be reduced in height by adding a small capacitor in parallel with the bulb? The capacitor might absorb the excess voltage and lower the rise speed at the cost of a bit more ringing afterwards?

 

[JimmyM]

Interesting.

Does the flat top of the spike look to you as if it has been clipped? From the shape of it, it almost seems as if it should extend upwards by another two or three divisions to a sharp peak, but has been clipped after a rise of about 40 V as if by a voltage breakdown somewhere?

Is it just my imagination, or is the ringing more prominent in the white trace?

If the spike is caused by inductance, maybe it could be reduced in height by adding a small capacitor in parallel with the bulb? The capacitor might absorb the excess voltage and lower the rise speed at the cost of a bit more ringing afterwards?

Very interesting thought regarding the clipping.
The below scope cap shows the difference between 29.4V, 20.0V, 15V, and 12V input.

Green = 29.4V
Pink = 20.0V
Orange = 15V
Red = 12V



Alan, Will... Thoughts?

 

[Mr Happy]

Wow. That does look like clipping.

I hesitate to suggest something that might blow up your FET, but I wonder what would happen if you connected a capacitor of about 1 uF across the source and drain pins of the transistor? There would be an inrush current when the FET turned on, but when it turned off the cap might absorb the inductive current and reduce the height of the voltage peak? In effect, doing what the Schottky diode somehow was failing to do. The capacitor rating is just a rough guess based on the slope of the dV/dt rise curve on the voltage spike. It might need some experimenting to pick a good value.

Incidentally, regarding the diode, I had a thought about that. When the FET is switched on, the current is flowing in a forward direction. Therefore any diode you place in the circuit has to be reverse biased to the normal current flow, or it will short out. However, when the FET switches off, the inductance in the circuit will try to make the current continue to flow in the same direction. Therefore any diode that was reverse biased before will still be reverse biased. This leads me to think that instead of a Schottky diode, you may want a Zener diode instead? The Zener should have a voltage above the normal operating voltage of the supply, but below the maximum rating of the FET. That way the Zener will break down before the FET does and bypass the current safely.

I'm not sure whether a capacitor alone, a capacitor plus Zener, or a Zener alone would work best in this situation...?

(As before, I'm not an expert in this, so my thoughts should be taken with a measure of due dilligence...)

 

[Alan B]

Wow. That does look like clipping.

I hesitate to suggest something that might blow up your FET, but I wonder what would happen if you connected a capacitor of about 1 uF across the source and drain pins of the transistor? There would be an inrush current when the FET turned on, but when it turned off the cap might absorb the inductive current and reduce the height of the voltage peak? In effect, doing what the Schottky diode somehow was failing to do. The capacitor rating is just a rough guess based on the slope of the dV/dt rise curve on the voltage spike. It might need some experimenting to pick a good value.

Incidentally, regarding the diode, I had a thought about that. When the FET is switched on, the current is flowing in a forward direction. Therefore any diode you place in the circuit has to be reverse biased to the normal current flow, or it will short out. However, when the FET switches off, the inductance in the circuit will try to make the current continue to flow in the same direction. Therefore any diode that was reverse biased before will still be reverse biased. This leads me to think that instead of a Schottky diode, you may want a Zener diode instead? The Zener should have a voltage above the normal operating voltage of the supply, but below the maximum rating of the FET. That way the Zener will break down before the FET does and bypass the current safely.

I'm not sure whether a capacitor alone, a capacitor plus Zener, or a Zener alone would work best in this situation...?

(As before, I'm not an expert in this, so my thoughts should be taken with a measure of due dilligence...)

First of all, good work, and good comments!

I have not had time to really dig into this, but:

1) The FET inherent avalanche mode is probably doing the clipping. At these low stored energies there may be no danger to the FET, but it would take more research to be sure. The fact that it doesn't fail right away is good but not sufficient.

2) I have had the same thought about the diode polarity, but have not gone carefully through it to be sure. It would seem to be similar to the relay coil driver protection diodes that are commonly used, but the lack of effectiveness makes one wonder.

3) Be very careful about putting capacitance across the bulb. This causes huge current spikes in the FET. The normal way to approach this is to use a snubber, which generally consists of a resistor in series with a capacitor. This will cause the transient to go away faster as energy is absorbed in the R.

4) Using a zener is also possible. If the FET's avalanche specs are not exceeded, there may be no reason to add a part. Further analysis of the FET spec sheets may shed some light on the capacity of the inherent avalanche capability of the device.

5) The inductance of the leads and circuit board are too low. The bulb's coil on coil construction will dominate the inductance of the circuit. I plan to measure that soon, I have an LC meter that is very accurate. Unfortunately it is still in kit form and I need to assemble it...

-- Alan

 

[JimmyM]

First of all, good work, and good comments!

I have not had time to really dig into this, but:

1) The FET inherent avalanche mode is probably doing the clipping. At these low stored energies there may be no danger to the FET, but it would take more research to be sure. The fact that it doesn't fail right away is good but not sufficient.

It ran for 3+ hours at 29.4V in, 12VRMS out with a 64625. Ut was warm, but that's it. I'll have to look into avalanche ratings, etc.

2) I have had the same thought about the diode polarity, but have not gone carefully through it to be sure. It would seem to be similar to the relay coil driver protection diodes that are commonly used, but the lack of effectiveness makes one wonder.

I had the diode connected such that current would flow from the FET drain to the Bat+ across the bulb. Otherwize, it would conduct everytime the FET turned on as Mr Happy said.

3) Be very careful about putting capacitance across the bulb. This causes huge current spikes in the FET. The normal way to approach this is to use a snubber, which generally consists of a resistor in series with a capacitor. This will cause the transient to go away faster as energy is absorbed in the R.

Yeah, the capacitor idea just doesn't make sense. It would pull more current from the battery when the fet was on and dump it back through the bulb when it was off. It would increase the RMS voltage to the bulb.

4) Using a zener is also possible. If the FET's avalanche specs are not exceeded, there may be no reason to add a part. Further analysis of the FET spec sheets may shed some light on the capacity of the inherent avalanche capability of the device.

I'll be doing some research regarding this.

5) The inductance of the leads and circuit board are too low. The bulb's coil on coil construction will dominate the inductance of the circuit. I plan to measure that soon, I have an LC meter that is very accurate. Unfortunately it is still in kit form and I need to assemble it...

-- Alan

Agreed, the bulb's coiled filament is the biggest contributor.

 

[JimmyM]

Below I've calculated the FET amps based on different bulbs and a maximum input voltage.
You can see that even though the bulb itself may not be a super powerful bulb, it can still pull a lot of amps while the FET is on.

Bulb . Bulb DC voltage ..Bulb DC amps .. Supply voltage .. Peak FET Amps
5761 7.2....................5.5.................40.....................30.56
64275 7.2...................6.6.................40....................36.67
64623 15.5.................10.3................40....................26.58
64625 13.2.................9.8..................40....................29.70
64458 20...................10.4.................40....................20.80
1185 11.2..................3.4..................40....................12.14
ROP-HI 7.2.................4.4..................40....................24.44
64430 9.....................7.3..................40....................32.44
62138 13.2.................9.8..................40....................29.70
64655 26...................12.2................40.....................18.77
64657 28 ..................12.3................40.....................17.57

So there may be a limitation on the Vrms to Pack voltage ratio. I'll have to do more testing to make sure. But you can see that running a 64275 on a 40V pack will put a heavy load on the FET.

 

[Roland]

Nice to hear about regulated mag D switches.

Simple option 1) sounds good
The internal circuit is completely dead when off and it will only leak as much current as the MOSFET itself will allow. That's in the area of a few nano or micro amps. However, you're limited to just ON or OFF operation.

Easy v-bulb adjustment for different bulbs is nice and even necessary.

Is a 5761 (7,2 volts) on a supply voltage of 7,3 volts or 8,4 volts also a problem?

 

[JimmyM]

Nice to hear about regulated mag D switches.

Simple option 1) sounds good
The internal circuit is completely dead when off and it will only leak as much current as the MOSFET itself will allow. That's in the area of a few nano or micro amps. However, you're limited to just ON or OFF operation.

Easy v-bulb adjustment for different bulbs is nice and even necessary.

Is a 5761 (7,2 volts) on a supply voltage of 7,3 volts or 8,4 volts also a problem?

This no problem. Just a turn of the on board pot. I'll have development hardware soon and should be able to start the software development.

 

[SafetyBob]

Jimmy, a raging fire of epic proportions is starting at Paypal awaiting the "send money" button directed to your bank account!!

I cannot wait to slam a couple of these into the flashlights I will have waiting for them.....which reminds me I need to get some more big Emoli batteries.

As a result there will be strange sightings of flashing lights pointing towards the sky from Oklahoma when you finally send them to me you know.....

Anyhow, glad to see we are all but done except for fine tuning. I just can't say how much I look forward to these being done. The answer to about 90% of our problems solved by one piece of equipment!!

Bob E.

 

[LuxLuthor]

Below I've calculated the FET amps based on different bulbs and a maximum input voltage.
You can see that even though the bulb itself may not be a super powerful bulb, it can still pull a lot of amps while the FET is on.

Bulb . Bulb DC voltage ..Bulb DC amps .. Supply voltage .. Peak FET Amps
5761 7.2....................5.5.................40.....................30.56
64275 7.2...................6.6.................40....................36.67
64623 15.5.................10.3................40....................26.58
64625 13.2.................9.8..................40....................29.70
64458 20...................10.4.................40....................20.80
1185 11.2..................3.4..................40....................12.14
ROP-HI 7.2.................4.4..................40....................24.44
64430 9.....................7.3..................40....................32.44
62138 13.2.................9.8..................40....................29.70
64655 26...................12.2................40.....................18.77
64657 28 ..................12.3................40.....................17.57

So there may be a limitation on the Vrms to Pack voltage ratio. I'll have to do more testing to make sure. But you can see that running a 64275 on a 40V pack will put a heavy load on the FET.

Jimmy, that is very interesting, and surprising the Peak FET readings! I totally agree with SafetyBob.

 

[Mr Happy]

Doesn't it seem rather exceptional though, to run a 7.2 V bulb on a 40 V pack? Isn't it more reasonable to match the pack voltage so it is only a few volts more than the target bulb voltage?

 

[JimmyM]

Doesn't it seem rather exceptional though, to run a 7.2 V bulb on a 40 V pack? Isn't it more reasonable to match the pack voltage so it is only a few volts more than the target bulb voltage?

I think it would be much more common for folks to run a few extra cells for added runtime. A 7.2V bulb on a 12V pack that they can use for anything up to 12V is probably a really common application. Along with things like an 18V pack for the 12V 100W bulbs like 64623, 64625, and 62138.

 

[Roland]

Jimmy, that is very interesting, and surprising the Peak FET readings! I totally agree with SafetyBob.

I figured out that the Peak FET readings can be predicted exactly according to:
Peak FET Amps = Bulb DC voltage * Bulb DC amps / Supply voltage

 

[JimmyM]

I figured out that the Peak FET readings can be predicted exactly according to:
Peak FET Amps = Bulb DC voltage * Bulb DC amps / Supply voltage

That's right. Straight forward Ohm's Law. Use DC voltage and amps to figure the filament resistance then divide the supply voltage by it. Your equation is the reduced version. Quite straight forward.

 

[SafetyBob]

Mr Happy, you are exactly right.....it is exceptional to run a 12VDC bulb with a 40VDC battery pack. It is runtime we want, and lots of it as long as everything else holds up.

This will change world order....maybe not with a 623 because you will be able to fry everything (mostly your hand) after say 30 minutes, but think about that wonderful 1185 or 1111 if you like. Super bright, super white bulb with runtime we thougth was only possible with and led flashlight.

Seriously though, most of the 100watt plus lights are for showing off, that is until you could actually use them practically for 30+ minutes. An hour?

Let me predict our next problem. How to keep these cool with an integral cooler. Such as 4 big emoli batteries in a 5 D with a tail cap with small slots milled in it to allow cooling air to go through the body pushed by a small 5VDC fan embedded in the tailcap....channels milled in the plastic used to keep the batteries from moving about and finally slots milled in the head to let all that new found heat out. Who needs water proof when it will boil upon contact?

Can you tell I have been thinking about keeping a high powered M@g cool for awhile? I don't know if that good insulation that Lux uses will be good enough to keep the heat away or not.

Bob E.