Falling voltage compensated Soft Starter

[JimmyM]

I've hit upon an idea that will kind of bridge the gap between the existing JM-SST and a true regulated design.
It only really works when the battery voltage is high enough to pop a bulb when the pack is fresh and will be OK when the pack voltage has dropped a bit. Basically it raises duty cycle as pack voltage drops. It doesn't monitor output voltage. I started thinking about this while building nitnapz' Mag458. The pack voltage would pop the 458 when fresh so I had to dial back the duty cycle, but the lamp voltage dropped out of the "optimum" range as the pack settled down into the area where it would spend most of its discharge time.
Below is a solution for the above mentioned Mag458. Pack voltage would initially start at 26.1V, but the Bulb would see 19.95V. As the pack voltage drops (pink line) the RMS voltage dips a little (down to ~19.28V), then rises again.
From a Vbatt of 22V to the point where the FET is at 100% duty (19.98V) the RMS output is pretty steady (19.87-19.98V)

This isn't a regulator though.
Think of it as a soft-starter "plus".
It would require a little tuning by me or the user via 2 onboard pots. The design is scalable from a 5761 to one of petrev's 1100W monsters. 6V to 40V. Sorry Hyper-Blitz guys. The absolute maximum voltage is 42. After that the smoke comes out.
I still have to do some breadboarding to work out the kinks.

Thoughts?

 

[LuxLuthor]

That's a really great step forward, Jimmy....and a very impressive rms out curve in the face of those input voltages. Is the setup the same for the variety of bulbs and voltage ranges you mentioned?

 

[JimmyM]

The curve is similar regardless of voltage. Things get bent out of shape if you try to run a 6V bulb on a 15V pack. The lower the duty cycle, the less linear the output curve is. Using a 13 cell pack (1.45V down to 0.9 V/cell)
Gives quite a "V". Starts out at 6.25 but drops to 4.35 before starting to climb again to ~5.7V Not very good.
There are 2 adjustments.
1) Adjusts the location of the bottom of the "V".
2) Adjusts the output voltage at max Vin.
Perhaps I'll put together graphs for a few common packs with a normally just out of reach bulbs. One I used before was the 5761 on 2 Li-Ion cells. Hot off the charger (8.4V) the bulb would start at 7.07V (Vbatt=8.4V), fall to 7.0V (Vbatt=8.3V), then slowly rise to a maximum of 7.1V (Vbatt=7.1V). After that the FET was fully on and the Vbulb dropped with Vbatt.
This is just a "take the edge off" kind of thing. Of course... it will soft-start.
Can you think of a few bulbs that normally pop on a fresh pack but run well if the pack has the "edge taken off"?
1185 on 3xLi-Ion cells.
64458 on 18 cells.
Etc.

 

[Bimmerboy]

Subscribed!

IIRC, the 1185 can handle 11.1V with a soft start, but someone definitely correct me if that's wrong.

However, I'd like to use 3 X Emoli for my Mag85, which sag less than Li-Co (the regular type Li-ion), so a soft-start "plus" may be a great thing in that scenario.

 

[JimmyM]

Subscribed!

IIRC, the 1185 can handle 11.1V with a soft start, but someone definitely correct me if that's wrong.

However, I'd like to use 3 X Emoli for my Mag85, which sag less than Li-Co (the regular type Li-ion), so a soft-start "plus" may be a great thing in that scenario.

If I remember correctly, The eMolis start off at 4.2V, have a nominal voltage of 3.7 and are done at 3V. Correct?
So, some adjusting of the "pots" in my spreadsheet works out to...
Pack voltage.....Bulb Voltage
12.6V.............. 11.1V
12.53V............. 11.04V
11.1V...............11.1V
11.1-9V.............pack voltage (FET@100%)

Looks like this is a good solution for the 1185 on 3 x eMoli. You won't see that 0.06V dip at 12.53V

 

[Bimmerboy]

If I remember correctly, The eMolis start off at 4.2V, have a nominal voltage of 3.7 and are done at 3V. Correct?

According to Silverfox's Li-ion shootout testing, the 26700 Emoli's stayed above 3.75V for a little more than half their runtime at both 3A, and 5A loads, and held above 3.6V for practically the entire run. The drop off to 3V at the end is quite sharp.

Long story short, that, combined with your spreadsheet figures, looks like an absolutely perfect 1185 solution! I'm psyched!

Many thanks for your continued efforts, Jimmy. :thumbsup:

I know this is still in the conceptual stage, but Paypal is now set on a hair trigger, ready when you are.

 

[JimmyM]

I'm putting together a parts list for Digikey. A few approaches, tracking regulator and non-tracking regulator, LDO, blah blah blah.

 

[JimmyM]

A little wrangling with a shematic and some math shows the design would be good from 6V to 50V. Of course the "L14K and bigger" crowd would probably like to know if it will work with their 600+ watt lights. My best answer at the moment is... probably. But dual FETs would be better. That can be done too.
Of course there will be some options regarding the FET. But the IRF2804 should suffice for most applications 9V-40V.
If you're going to push that 40V mark, the IRFS3206 is the FET for you.
I've got an order to place with Digikey... gotta go.
This is what the curve looks like for petrev's PK1000 on 10XA123s

 

[monkeyboy]

This would be good for 64623 running on 13 nimh cells and 64625 on 11 or 12 cells.

I don't think 1185 on 3 x li-ion needs this (or soft start) with smaller li-ion cells so long as the batteries are not charged too high, but it would certainly improve bulb life. I guess it depends on the cells though. 17670 probably has enough voltage drop to be OK without the driver but 3 x li-ion D or emoli probably doesn't.

1331 on 3 x li-ion might need this. I've read that some people have experienced instaflash and others not.

 

[JimmyM]

Actually, this removes the whole worry about which cells hold voltage better, which ones sag, how "fresh" off the charger they are, etc, etc.
If you have an 1185 on 3 Li-Ions, it will instaflash if you have big cells, or charge over 4.1V, etc, etc. This circuit handles all that.
If you have a set up that will pop the bulb if the cells are too fresh, or if you do resistance mods, or you use brand X versus Brand Y cells. This fixes that.
If your mod won't pop the bulb under any circumstances, then you don't need any kind of soft starter.
Below is the plot for an 1185 on any pack under 12.6V Max.

It starts at 11.1V max, then falls to 11.01V at a pack voltage of 12.49V.
However, if your pack holds voltage really well (3.75V/Cell), the bulb voltage will self adjust to 11.098V at a pack voltage of 11.25V.
I'll be doing some testing this weekend.

 

[Raoul_Duke]

This is realy cool...

I bought a couple of JM SST in anticipation of using the 64458 on 7 emoli, but it turns out that is perfect for not flashing that particular lamp.

They are still in the packaging you sent them in....I thought about trying to fit on in my 7C mag...but never got round to that.

It still all adds to the usefull parts bit in the box, I'm guessing they might have a use one day.

So the difference between this, and the first is that this will protect from over voltage, untill the cells droop below the set pint.

The first SST could be set to not flash the lamp, but as the pack sagged, the offset remained the same, and also sagged.

I'm not sure exactly what a regulator does different to this at the end of the day...but it seems like this provides a regulated output untill the battery pack poops out......works for me.

Do you think its possible to fit the pots to be adjusted via a a drilled hole in the kiu socket, and then adjust on the bench something like this so it could be tuned to each set up, including resisatnce of the host etc.

I would like to run the 64633 on 5 26700 emoli ( in a 7C cell mag if you could shrink the board down a little PLEASE.)

 

[JimmyM]

So the difference between this, and the first is that this will protect from over voltage, untill the cells droop below the set pint.

The first SST could be set to not flash the lamp, but as the pack sagged, the offset remained the same, and also sagged.

That's right. This would increase it's duty cycle as the pack voltage falls.

I'm not sure exactly what a regulator does different to this at the end of the day...but it seems like this provides a regulated output untill the battery pack poops out......works for me.

A regulator has a feedback loop that adjusts itself based on conditions at its output. This does not.

Do you think its possible to fit the pots to be adjusted via a a drilled hole in the kiu socket, and then adjust on the bench something like this so it could be tuned to each set up, including resisatnce of the host etc.

I do plan on having the 2 adjustment pots available on the top of the board for adjustment while installed. It will be a little tricky to do, basically you have to install a fresh pack and put a small load on the output, then set the 1st pot for the correct voltage (measured at the load by RMS meter), then allow the batteries to discharge until the pack is at the same voltage you want the maximum for the bulb to be, then turn the 2nd pot until voltage just stops increasing.

I would like to run the 64633 on 5 26700 emoli ( in a 7C cell mag if you could shrink the board down a little PLEASE.)

That will be a real trick since the FET won't even fit under the C cell KIU socket.

 

[JimmyM]

This is realy cool...

I bought a couple of JM SST in anticipation of using the 64458 on 7 emoli, but it turns out that is perfect for not flashing that particular lamp.

The 458 runs direct on 7 eMolis? They must be sagging like crazy. They would have to sag almost immediately to 3V each to keep the voltage to the bulb below 21V. Do you use an NTC or anything? How is your switch modded? Does it get warm?

 

[Raoul_Duke]

The 458 runs direct on 7 eMolis? They must be sagging like crazy. They would have to sag almost immediately to 3V each to keep the voltage to the bulb below 21V. Do you use an NTC or anything? How is your switch modded? Does it get warm?

Whoops....Typo. I meant 64657 ( 24V 250w) flashes at ~29.4.

( I have run the 64458 on 5 emoli perfectly...but TBH i dont see the point of that verses the 4 x emoli on a 64623.

Same lux readings on luxs charts, ( and ~ same temperatures on the reflectors ~ ( OK prehaps it is a bit whiter...but not a great deal to my eyes...prehaps I should revisit?)



I got 24V at the pins on start up over a judco switch. Pretty bright for a maglite. Prehaps a bit hot too

 

[JimmyM]

Whoops....Typo. I meant 64657 ( 24V 250w) flashes at ~29.4.

( I have run the 64458 on 5 emoli perfectly...but TBH i dont see the point of that verses the 4 x emoli on a 64623.

Same lux readings on luxs charts, ( and ~ same temperatures on the reflectors ~ ( OK prehaps it is a bit whiter...but not a great deal to my eyes...prehaps I should revisit?)



I got 24V at the pins on start up over a judco switch. Pretty bright for a maglite. Prehaps a bit hot too

OK. That makes more sense. 24V bulb. Gotcha.
The 64458 can do 6 eMolis on this set up and not exceed 20.5V. The benefit would be almost constant brightness with increased run time. During the run, it would only drop off to 20.1V at the "dip" before rising again to 20.5V. After that it would fall with battery voltage since the FET would be at 100% duty.
I really can't wait to get into the shop.

 

[Raoul_Duke]

That's right. This would increase it's duty cycle as the pack voltage falls.

A regulator has a feedback loop that adjusts itself based on conditions at its output. This does not.

I do plan on having the 2 adjustment pots available on the top of the board for adjustment while installed. It will be a little tricky to do, basically you have to install a fresh pack and put a small load on the output, then set the 1st pot for the correct voltage (measured at the load by RMS meter), then allow the batteries to discharge until the pack is at the same voltage you want the maximum for the bulb to be, then turn the 2nd pot until voltage just stops increasing.

when I said like """this""" i actually meant like this... the Kiu socket, in this pic, holding the lamp, is then plugged into an already installed kiu inside the host, giving more access to the pots.

(Blant rip of one of Petrev's images...Opologies in advance) ........... :nana:
Might make it a bit easier

That will be a real trick since the FET won't even fit under the C cell KIU socket.

Bum

How about If I remove the stock switch, and flip the Kiu upside dow and use a smaller switch and..... ( Prehaps I will let you make the "D"'s first, and we might look at this later. :thinking :green:

...

 

[Raoul_Duke]

OK. That makes more sense. 24V bulb. Gotcha.
The 64458 can do 6 eMolis on this set up and not exceed 20.5V. The benefit would be almost constant brightness with increased run time. During the run, it would only drop off to 20.1V at the "dip" before rising again to 20.5V. After that it would fall with battery voltage since the FET would be at 100% duty.
I really can't wait to get into the shop.

LOL...You talk of a dip of 0.4V like its a big deal....I'm used to watching my V lamps dropping faster than this per second ( well more actually, like its counting down 0.1V like seconds) for the constant run of the light.

I can cope with watching it dip 0.4 or so V for a minute and then return.

Outstanding work DUDE!!!!

What king of soft start are we talking here time wise. I quite like the hotwired lights to light up instantly...that the fun with hotwires over hid etc.

 

[JimmyM]

...

The design would have to be a double sided board and the bulb tower would have to sit on top of that somehow. It would most likely be a custom build for that baby. Then again, I did manage to fit a FET switch inside of a stock C body switch for the ultimate low resistance ROP-HI Lithium Edition.

 

[JimmyM]

LOL...You talk of a dip of 0.4V like its a big deal....I'm used to watching my V lamps dropping faster than this per second ( well more actually, like its counting down 0.1V like seconds) for the constant run of the light.

I can cope with watching it dip 0.4 or so V for a minute and then return.

Outstanding work DUDE!!!!

What kind of soft start are we talking here time wise. I quite like the hotwired lights to light up instantly...that the fun with hotwires over hid etc.

Typically (in the JM-SST design) there is a pause of about a tenth (or less) of a second then it ramps quickly to about 90% of setpoint in about 1/4 second, after that it reaches full voltage in another 1/4 second. I make the soft-start gentle enough so as not to pop bulbs that are on the ragged edge. Like a 64458 on 21 volts. I'll have to fiddle with soft start speeds.

 

[JimmyM]

Just finished testing a breadboarded design. Looks very promising.
Set for 11.1V, and swept input from 12.6 to 11V. The ouptut never dropped below 11.05V Until the input voltage dropped to the point the FET was at 100%.
I'll be fiddling with adjustments, but initial component values seem dead on.
Next step will be to try reducing component count by redesigning a reference voltage section using a zener. After that I'll get a schematic into software and see if it can fit on a 30mm round board with a monster FET.