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

[Alan B]

That shouldn't be a problem. You just set the voltage you want, and let the regulator work out the rest. It really doesn't have anything to do with the bulb itself. If you set it for 11.1V (for example), you can run any bulb you want on 11.1V (assuming it doesn't blow on 11.1V). The pack can be just about anything over 11.1V and under 30V. I don't have testing results on small filaments with large battery-to-bulb voltage differentials. Small filaments have less thermal mass, they MAY be more affected by a large voltage differential. I think I'll tap Lux for a little testing.

If some of the parts (such as the FET) are rated at 30V maximum, that means that 30V should NEVER be exceeded. One of the things we have seen is that a lot of older regulators die after awhile (we have not seen that with this new design, but we don't have the same level of experience with it yet either). It is important to derate the operating voltage enough that these maximums are never exceeded. As we now know, there is a slight inductive kick apparently from the filament that spikes the voltage up as the FET goes off. There is still some work to be done in the spike area, but in any case we need some headroom for this and other voltage excursions to occur and still stay below maximum ratings.

Note also that using an FET driver may make this spike go even higher, but it also becomes much shorter and harder to see, so we might think it is getting better when it is in fact getting more dangerous for the FET. We have not seen this happen, but is that because it is not happening, or because it is harder to measure? The amount of energy to be dissipated is constant (the inductive energy in the filament), so dissipating it in a shorter time due to faster switching requires a higher voltage...

Based on the unknowns I would recommend a substantial safety margin on the component voltage ratings. For example, derate 30V parts to something on the order of 20V. This will make the system far more reliable, which is a good quality in a flashlight.

-- Alan

 

[Alan B]

Jimmy, while your enjoying your new year, once again I have been trying to think clearly in order for you zero in on our needs.

M@g85: ordinarly we would use FM's 9xAA battery holder to keep things from blowing up, that with additional runtime from FM's 12xAA holder would really make a very, very nice flashlight with significant brightness with our choice of average runtime or extended battery options.......since we have discussed Emoli's before, 3x would be normal, and 4x would offer the average flashaholic a stunning performer with great runtime too.

Basically copy cat the above with the 623.

Bob E.

That shouldn't be a problem. You just set the voltage you want, and let the regulator work out the rest. It really doesn't have anything to do with the bulb itself. If you set it for 11.1V (for example), you can run any bulb you want on 11.1V (assuming it doesn't blow on 11.1V). The pack can be just about anything over 11.1V and under 30V.

.

The regulator will compensate for the pack voltage, as Jim indicates.

However the protection voltage should be adjusted for the battery pack. Protecting the batteries is useful and important, especially so folks can use things like unprotected Li-Ions. The protection circuits in the cells are problematic for these high current applications, so using the protection in the regulator makes a lot of sense.

-- Alan

 

[JimmyM]

However the protection voltage should be adjusted for the battery pack. Protecting the batteries is useful and important, especially so folks can use things like unprotected Li-Ions. The protection circuits in the cells are problematic for these high current applications, so using the protection in the regulator makes a lot of sense.

-- Alan

Oh, yes. I forget about onboard voltage protection.

I've yet to test the DPAK 40V FET at high power. My order just shipped from Digikey, so I'll be testing that one as my preferred FET.
I've been zooming in on the rising and falling edges during testing. While testing the 30V FET I have now, and the 40V FET I'm expectiing, I'll capture some waveforms while driving things at real world voltage and currents. I think I have an extra IRLU7843 for testing. Maybe I'll test one driving an 64623 at 30V In/14V out, and see what happens.

 

[donn_]

Any plans (down the road) for one of these which will fit a C-Mag tube?

 

[JimmyM]

Perhaps, but mounting would be difficult. The KIU bipin mount for C cell mags leaves nowhere to mount any electronics. It would definitely have to have components on both sides of the board. But how would it be mounted/wired within the light?

 

[donn_]

I wonder if there's a way to modify AW's C-Mag softstarter.

 

[Alan B]

I wonder if there's a way to modify AW's C-Mag softstarter.

Probably make more sense to start from scratch and design for a full switch replacement, which is what AW did. Rather costly to start with a $75 part.

We already have a D mag design that works that way (full switch replacement), and it uses all fullsize parts. Shrinking it to C size would probably not be too hard.

I just received a set of D Li-Ion cells. Nearly 5AH (measured about 4500 mAH as I recall). Now those are going to work nicely in a regulated 3D M@g85. I think I'll stick with D myself.

-- Alan

 

[JimmyM]

I wonder if there's a way to modify AW's C-Mag softstarter.

I suppose it's possible. But, per Alan's comments above, it would make more sense to design a new switch body. To use the existing switch body would be quite difficult.

 

[donn_]

I agree. I hadn't factored in the initial cost of the AW part.

The reason I'm hot on Cs is the fact I prefer A123 cells to Emoli or IMRs. They have most of the benefits, and other than lower capacity, no downsides. And they can be recharged so much faster, the difference is night and day.

I can easily run them in D tubes, but I also prefer the form factor of the D tube.

 

[JimmyM]

Maybe I'll test one driving an 64623 at 30V In/14V out, and see what happens.

I'm testing a 64623 on 29.3V input with a 12V setpoint.
There is an inductive kick at shutoff. It shoots up to 39-40V. Schottkys across the FET and bulb have no effect. As we've seen during development. The FET is getting warm, but no overly so. I can easily hold it between my fingers indefinitely. I think it's the fact that it is carrying quite a bit of current while ON.
From Lux's tests...
64623 on 12V = 8.8A Therefore -> 1.364 Ohms.
29.3V/1.364Ohms = 21.48 Amps while in the ON state.
It's been running for 10 minutes so far. I'll let it run for another hour or so and see what happens.
Note: This is using an FET driver and 5V gate drive. Not 12V.

 

[Alan B]

I'm testing a 64623 on 29.3V input with a 12V setpoint.
There is an inductive kick at shutoff. It shoots up to 39-40V. Schottkys across the FET and bulb have no effect. As we've seen during development. The FET is getting warm, but no overly so. I can easily hold it between my fingers indefinitely. I think it's the fact that it is carrying quite a bit of current while ON.
From Lux's tests...
64623 on 12V = 8.8A Therefore -> 1.364 Ohms.
29.3V/1.364Ohms = 21.48 Amps while in the ON state.
It's been running for 10 minutes so far. I'll let it run for another hour or so and see what happens.
Note: This is using an FET driver and 5V gate drive. Not 12V.

Great test. Would be interesting to know how it does without the driver. That voltage kick is endangering the 40V FET. Probably not acceptable the way it is, no margin. Might be ok on 24V or 25V, but there may be more peak voltage than the scope is seeing.

The voltage kick should be L* di/dt. So it should be worse with the driver since dt is smaller. L and di are fixed. All you can control is dt. Need to slow it down a little to reduce the voltage kick.

-- Alan

 

[JimmyM]

Great test. Would be interesting to know how it does without the driver. That voltage kick is endangering the 40V FET. Probably not acceptable the way it is, no margin. Might be ok on 24V or 25V, but there may be more peak voltage than the scope is seeing.

The voltage kick should be L* di/dt. So it should be worse with the driver since dt is smaller. L and di are fixed. All you can control is dt. Need to slow it down a little to reduce the voltage kick.

-- Alan

2 hours later and the 30V FET isn't letting up. It's still just quite warm. That's all. No problems. Hmmmm. I'll see how the 40V FET works out when it gets here. But the 30V FET seems to be holding on just fine.

Why wouldn't a schottkey have any effect? It should shunt that kick back to the bulb positive.

EDIT check that "no effect" part.
Below is a cap of the scope. 10V/div, 1uS/div.
Green = WITH BAT42W Schottky diode (5nS reverse recovery)
Red = No Diode.



Also, I switched to a 64625 and the trace was EXACTLY the same.

 

[Mr Happy]

Below is a cap of the scope.

I don't see a pic...?

 

[JimmyM]

Here's another cap. Both with schottkys.
10V/div, 1uS/div
Green = 64625
Red = Automotive type 1157 dual filament (both filaments driven)

 

[wquiles]

2 hours later and the 30V FET isn't letting up. It's still just quite warm. That's all. No problems. Hmmmm. I'll see how the 40V FET works out when it gets here. But the 30V FET seems to be holding on just fine.

Why wouldn't a schottkey have any effect? It should shunt that kick back to the bulb positive.

EDIT check that "no effect" part.
Below is a cap of the scope. 10V/div, 1uS/div.
Green = WITH BAT42W Schottky diode (5nS reverse recovery)
Red = No Diode.



Also, I switched to a 64625 and the trace was EXACTLY the same.

If you recall my early scope shots on th email PhD thread also showed the schottkey had no effect for me either (at least nothing significant) 🙁

Will

 

[JimmyM]

If you recall my early scope shots on th email PhD thread also showed the schottkey had no effect for me either (at least nothing significant) 🙁

Will

I remember. It didn't change things for me either. I just can't believe it has no effect. I HAD to try it again.

 

[JimmyM]

Alan, The 40V FET I found has a logic level gate and has 3.8 mOhm Rds(on) at 5V. I thought it was a standard gate. This could allow me to reduce component count on the board.
I've also fould a fixed 30mA/5V linear regulator that has a Max Vin = 45V. Unfortunately, Digikey doesn't stock it, so I'm looking into samples from the Mfg and lot pricing.

 

[Alan B]

That provides good voltage margins. 5V gate is good, especially for those high current 6V bulbs with low voltage packs.

The voltage spike is not important as long as it does not exceed ratings of devices. Adding parts to control it is probably not desirable in these tight and cost sensitive applications.

Did you try driving the high power bulb without the FET driver? Is the driver really required?

-- Alan

 

[JimmyM]

Did you try driving the high power bulb without the FET driver? Is the driver really required?

-- Alan

I haven't tried it yet. Maybe I'll try this afternoon.
I'll get scope caps of the turn-on / turn-off. I just can't quantify it, but I really want to use the FET diver. I suppose I'm going to have to prove it to myself that it isn't needed. I just feel like I need to isolate the AVR from the gate drive duties.

 

[Alan B]

I haven't tried it yet. Maybe I'll try this afternoon.
I'll get scope caps of the turn-on / turn-off. I just can't quantify it, but I really want to use the FET diver. I suppose I'm going to have to prove it to myself that it isn't needed. I just feel like I need to isolate the AVR from the gate drive duties.

It is just an engineering tradeoff, it would be good to know what the cost/benefit is. As long as we don't exceed the per pin current rating of the micro it should not hurt the micro to drive the FET, and as long as the FET dissipation and voltage ratings are not exceeded, it should not hurt the FET. I would expect with the really high current bulbs there might be a need for it to keep the FET power dissipation down, but it would be good to figure out where the extra drive is required. These FETs are so incredibly high gain and low on resistance they can do amazing things when driven directly with the Micro. This performance was not available some years ago.

-- Alan