bench testing of incandecent soft start ckt

[wquiles]

Andrew,

Thanks much for the information about DesignWorks Lite. They "do" offer it for Windows, and I have 30 days to try - only $40 to register, and it is very easy to use . Here is my redrawn 'mark 2' without the pots, as I recall the actual values used (have not been to the bench yet today, so these values are from memory and might be off a little!):

I think I will try the additional FET/transistor this coming weekend for the discharge path. Even as it is today, it works great with the charged battery pack, so I want to do some more tests while the pack discharges. Theory is great, but nothing beats actual bench testing "and" being able to see/measure using the scope

Will

 

[andrewwynn]

Oh.. in the case of using LiONs to power SF lamps.. i would be prepared to include like 0.075 to 0.01 ohm power resistor (high powered sense resistors) in series with the whole shabang to keep the initial voltage low enough to protect the lamp.. might not be necessary from the 10.8V solution but from the 7.2V solution it is.

I just noticed.. that with R3 there.. you could put a zener in parallel witht he C1 and it would make for a maximum voltage on the gate and consistant operation with any battery voltage. I love how this is coming together, hope some people will clone it for their own use.

-awr

 

[wquiles]

You mean like this below? If I pick a Zenner with a voltage around 3 Volts (which is above the Vgs required to turn it on, which I think it was 2.8V or so), that could work well

 

[andrewwynn]

exactly what i'm talking about.. but maybe a little higher value... thre reason being that the Rdson is lower with higher Vgs.

Z1 might make D1 obsolete

doncha love having DWL?

-awr

 

[wquiles]

I picked up a 3.0V Zenner today at Fry's - I hope to try it tonight if I can get my kids to sleep early

Yes, DWL simply rocks, and it was so EASY to use and setup. The very large library, easy way to create new simbols, and easy routing makes it a bargain for $40

Will

 

[wquiles]

Andrew,

Well, I really have no clue what is going on :mecry:

The zener is not working like I expected at all , and in fact never seems to get to the 3V it should :thinking:. From what I have measured, it seems to clamp at about 1.2 volts or so :thinking: :thinking: :thinking:

Instead of using the 3V zener (NTE5065A) across C1 and R3, I decided to try it just accross C1, and guess what, even though the zener is not behaving like it should, the whole circuits works perfect now

Here is the new circuit:

Here is the circuit with the zener disconnected (to show the "before" view of things). Note how between each switch closure the Vc3 keeps increasing since there is not enough time between presses to discharge it properly and since there is nothing preventing the cap's voltage from rising towards Vbat:

Here is the "mistery" circuit above with the zener connected across C1 (this is the "after" view of things). Note now how the circuit is just perfect, and how there is never a Vbat spike and how the Vc3 is now capped (somehow) by the misterious 3V zener who is not working according to what I expected.

The awesome thing about this circuit is that now, I get a soft start (I can actually see the bulb getting brighter until is full ON, so it is not just delayed start) after each switch press, not just the first one like before.

I also tested not connecting D1, but it does help to speed the discharge of C1, so I kept it in there. I tried playing with various values of R2 and C1, but these values matched the strange (to me) behavior of the zener.

I tried different values of R3, but the circuit (with the zener) is very particular to these values as when I tried 20K the whole balance of the zener was ruined since the starting voltage was now higher and the C1 effectively gave the gate of Q1 a higher voltage quicker than before, so now I loose the soft-start for switch presses #2 and so on. With a lower value (10K) then (because of the zener), there is never enough voltage at the gate to fully turn Q1 ON.

So, do you have some clues/suggestions for me?

Will

 

[andrewwynn]

What you ended up with was my design.. i looked too quick at your first mark3 design... with the zener to ground, it might work with a 5V zener.

When making small ckts.. big caps = sucky.. that's why i originally did a design with 4.7uF for the cap.. did you work on the ckt with a smaller cap? you should be able to just use exactly proportional increase in R2 but can leave R1/R3 the same..

What are you kidding? we have a circuit!

R1/R3 set up a voltage bridge that is non-capacitive.. they simply have to be a ratio that puts a voltage just below the gate threshold voltage.. the only time you'd hvae to tweak R3 is if you change the voltage of the bat... or a different FET that had a different threshold voltage.

It would be a 'cleaner' design to have yet another zener in parallel with R3 or even in replacment of R3.. but better put at least a small (200ohm maybe) resistance in series.. because when the ckt is turned off C1 will dump a bunch of charge through very quickly.

with the relatively small fractional value from Vbat to Vbase (bottom of C1) based on R3/(R1+R3).. it won't swing enough during typical voltage swings, but you have to make sure that with fresh batteries C1 base doesn't have enough voltage to turn on the ckt immediately.

I am thinking your 1.2V measurement was across C1? when you had the zener across C1, R3 ? that would make sense if your base voltage is 1.8V which is right about what i'd expect...

It helps greatly to have the zener only across C1.. what happens is when it's working.. you get 1.8V at the base and that puts 4.8V on the gate.. not a bad gate voltage.. that's usually going to get you the fast majority of your Rdson taget (i.e. some FETs will be 4mohm at 4Vgs and 3.3mohm at 10Vgs).. it's a moot point as long as you get to that first target value. Anyhow.. in this same example.. as soon as you turn off the ckt, you will have 3V on the cap, but now the base of the cap immediately drops to ground which puts the gate at 3V, which is right at the threshold of being on, but the diode will quickly discharge that 3V through only 50kohm.. that actually could be lower for a faster discharge.. but it seems that could be a moot point, looks like the turn on and turn off is working great even for 'blinking'.

I would work on putting in a lower cap and higher R2.. the rest should work as-is with that solution.. the RC calculator says 431 msec to 23% which is my rough guess of 1V increase from the start before FET sweeps from 'just starting' to 'full on'.. if you use 4.7uF cap, the calculator says 350kohm is the proper resistance.

with the bigger R2, it would necessitate D1 sticking around, but would make an even bigger difference in the discharge vs charge rate.

I really like how the D2 comes into play and does exactly what i planned for it.. the best part is that it means it works with a wide range of Vbat without the ckt changing performance much.. R3 could be exchanged for a different zener that would have to match the needs of the particular FET.. come to think of it.. probably can't replace R3.. would have to have it parallel to R3.. that way if Vbat is higher, it clamps the 'base' voltage, but keeping R3 is needed to do the quick 'drop' on the Vgs.

Wow, it was really neat to see those o-scope traces.. thanks for doin' all the legwork for me. very exciting project.

-awr

 

[wquiles]

Andrew,

You make multiple excellent points/comments, so I will try to address them one by one

R1/R3 set up a voltage bridge that is non-capacitive.. they simply have to be a ratio that puts a voltage just below the gate threshold voltage.. the only time you'd hvae to tweak R3 is if you change the voltage of the bat... or a different FET that had a different threshold voltage.

It would be a 'cleaner' design to have yet another zener in parallel with R3 or even in replacment of R3.. but better put at least a small (200ohm maybe) resistance in series.. because when the ckt is turned off C1 will dump a bunch of charge through very quickly.
-awr

That makes sense. That is why changing R3 (once I had a nice value) makes such a big difference in circuit performance. I might try the additional zener with R3 later, once I figure out why the current 3V zener is not working as expected.

I am thinking your 1.2V measurement was across C1? when you had the zener across C1, R3 ? that would make sense if your base voltage is 1.8V which is right about what i'd expect...

It helps greatly to have the zener only across C1.. what happens is when it's working.. you get 1.8V at the base and that puts 4.8V on the gate.. not a bad gate voltage.. that's usually going to get you the fast majority of your Rdson taget (i.e. some FETs will be 4mohm at 4Vgs and 3.3mohm at 10Vgs).. it's a moot point as long as you get to that first target value. Anyhow.. in this same example.. as soon as you turn off the ckt, you will have 3V on the cap, but now the base of the cap immediately drops to ground which puts the gate at 3V, which is right at the threshold of being on, but the diode will quickly discharge that 3V through only 50kohm.. that actually could be lower for a faster discharge.. but it seems that could be a moot point, looks like the turn on and turn off is working great even for 'blinking'.

-awr

Actually, that is not how the circuit is working. That is why I am still confused. The measured steady-state voltage across C1 once the circuit is ON is right at 1.20V. The measured steady-state Vgs, (or across C1 and R3 - which is the same voltage) is exactly 2.90V, which is what I already had measured to be the minimum voltage necessary to have Q1 turn ON - these two measurements clearly confirm that although the 3V zener is helping and make the circuit work perfect, the 3V zener does not "trigger" at 3V as expected/proyectec by theory. When I had the 3V zener in paralell with R3, the lamp would never come ON . I know I should be happy that the circuit is working perfect "as is", but I "need" to know why the zener is not working as expected

I would work on putting in a lower cap and higher R2.. the rest should work as-is with that solution.. the RC calculator says 431 msec to 23% which is my rough guess of 1V increase from the start before FET sweeps from 'just starting' to 'full on'.. if you use 4.7uF cap, the calculator says 350kohm is the proper resistance.

with the bigger R2, it would necessitate D1 sticking around, but would make an even bigger difference in the discharge vs charge rate.
-awr

I will try to use different values for R2 and C1 today - in theory it should work just the same way

I really like how the D2 comes into play and does exactly what i planned for it.. the best part is that it means it works with a wide range of Vbat without the ckt changing performance much..
-awr

Yes, assuming we can explain why the zener works when it shouldn't

Wow, it was really neat to see those o-scope traces.. thanks for doin' all the legwork for me. very exciting project.
-awr

You are welcomed. I would also "love" to work with you on the regulated circuit and/or LDO if you can use the help.

This has been a very simple circuit, but most interesting to test on the bench. Once I understand the weird zener behavior, then I will move to the 3xprotected 17670 battery pack with the MN60/61, since ultimately that is the "goal" of my project for my SF M6 .


One more thing - For testing with the 3x17670 cells, I will NOT be doing the "initial" bench work with the MN60 nor MN61 - I want to "debug" the circuit with a much, much cheaper bulb that is close enough for the early development work. What other "similar" bulb, even if it uses/requires less power, can I use? I have even considered using three of the 3xcell M*g bulbs in series to simulate a realistic load of about 1Amp with the right voltage - what do you think?

Will

 

[andrewwynn]

the zener is not working exactly as expected i think because your Vbat is not high enough to make the zener conduct until you turn off the switch.

If C1 is 1.20V and Vgs is 2.9 that means that the Vbase (of C1) is 1.7V.. I think i need some voltage reference from each of the important points.. also crank up the msec/div and see if you are just reading funny voltage readings because of an oscilation or such.

I'm not sure what the reading on the oscope is.. Vgs or Vc1.. it would be good to see both of those.. also let it run longer 'til it reaches steady state maybe just do two or three cycles.

I use a 12V, 20W lamp for my testing or an 1185.. running the 1185 at 1/2 voltage it still draws over 2A and just glows nice... yuou can pick up a 20W 12V bulb from just about anywhere for like $3-5.. the likes of home depot.. look in the lighting dept for a 'quartz halogen' bipin bulb.

keep it coming.. hey i do have another LDO solution in the works very similar to what you are working on.. send me an email.. flashlight at rouse.com

-awr

 

[wquiles]

Andrew,

I took some additional steady state measurements last night with my Fluke 87:

Vbat = 4.97V (this is under load, of course)
Vr1 = 3.30V
Vr2 = 2.07V
Vr3 = 1.66V
Vgs = 2.85V
Vds = 0.058V
Vbulb = 4.7V

The circuit just plain works awesome. I will keep draining the battery as I want to see what happens when the battery is depleted. By depleted, I mean something about 1.1Vx4 (with no load) is my guess at the moment as a good place to stop. Sound right to you?

I showed the working unit on my bench to my neightbor next door (a little bit of a flashaholic as well, but not as "committed" as I am!) and he was "oohhh, aahhh" over how cool the circuit works

Email sent to you as well

Will

 

[wquiles]

Another data point from tonight. I put the 4xAA back into my 1D modamag host, driving the WA1319 (about 250B Lumens or so) long enough that I detected it "spent", and when I was done I got a Vnoload of about 4.75V. Guess what? At this "low" voltage the circuit no longer turns the M*g 3D bulb ON - barely lits at all!. What this means is that by careful picking of R3, you can have a built-in low voltage "detector" since the light will dim from the FET not being turn ON all of the way, and, this low point can be set above the point where the cells would be damaged by overdischarging - very nice

If I changed R3 from 16K to 20K I was able to once again get the 3D bulb to lit, so somewhere around 15-17K seems to be the sweet spot. I am recharging the battery pack and will try again tomorrow with it - if I recall the no-load voltage was like 5.6 to 5.7 volts, which would of course be the worst case for the circuit in terms of inrush current :naughty:

Of course, now what I have this working, I have to try with the actual WA1319 to fine-tune the values to the actual bulb, and then again in the future once I move to the 3x17670 pack

Will

 

[wquiles]

Andrew,

Today I tried using a smaller C1 and R2. For some weird reason (probably becuase we artificially set the voltage with R1 and R3) when I changed C1 to 4.7uF, the value that worked OK was much smaller than I expected - somewhere around 70K for R2 was about the same delay as before :thinking:. Since I also wanted to speed up the ON time a bit, I ended up with a value of 60K for R2, which works well as shown in these shots here - horizontal is 400mS/div:

I then also proceeded to work with the actual bulb for this battery pack, the WA1319. Since the load on the battery is greater, I needed a larger value on R3 to get enough voltage on Vgs to turn on the FET fully (around 2.9V). I got to a value of 19K which works like this - horizontal is 400mS/div:

Look at the circuit without the soft start (removing R3). Look at the almost two volts negative spike (blue trace - never mind the yellow trace since it is floating here) - horizontal is 100mS/div:

And here with the circuit fully operational - horizontal is 100mS/div:

Since for R3=19K I had to use a 50K-POT to get, I tried a "standard" value of 20K. It does not work as well, but almost good enough - horizontal is 100mS/div.

and here is the same circuit, also with R3=20K, but with multiple ON-OFF cycles. Note how the battery dips a little bit more than before, but probably still "acceptable" - horizontal is 400mS/div.

Next thing to try is to let the battery discharge so that I can fine-tune R3

Will

 

[wquiles]

Here is an image of the latest circuit (Rev 5):

Will

 

[wquiles]

Andrew,

Some more tests today with the battery pack getting discharged. Due to the larger load of the WA1319 (compared to the 3xcell M*g bulb) I need to increase R3 to around 24K to have enough voltage left to be able to still barelly keep the bulb operational when the pack reaches a loaded voltage of about 4.1-4.2 volts (4.7v unloaded). This of course means that although the initial voltage spike is still going to be contained (mainly determined by the R2*C1 delay), additional, very quick ON-OFF cycles will suffer a little since C1 will still have some charge in it. At this point I am re-charging the pack to test with the 24K and a fully loaded setup (worst case again). I will plot some screens once the pack is ready again

Additionally, since I will be building this prototype in my 1D setup, I went ahead and setup the switch to be in-line with R2 (as you initially had it) and direct connect the bulb to the battery "+". I did change this on my bench setup and was able to instantly gain some brightness since (as we all know), that "normal" switch consumes some power compared to my super efficient FET switch

Will

 

[NewBie]

Have you made any current spike measurements lately (red trace on your older circuits-yes I know you can infer them from the battery voltage)?

Another thing to consider, is that 0.1 ohm sense resistor does serve to help limit the in-rush current to the bulb, altering the reality of the final circuit.

After moving the power on switch to R2, did you go back and look at the battery sag, since you removed the power switch resistance from the main current path?

Andrew, there was a thread awhile back where I listed a whole shebang of op-amp/comparator combos, such as the LM392 (which are under a dime each). I think I was talking to Al at the time. (in case you want to go down that path in the future).

Good work fellas on your circuit.

Question, does this circuit work as is, when a person changes to different bulbs, or do you need a different circuit for each bulb (part value changes)?

Say like when a person uses a PR-6 and then swaps to a 1181, 1185, 1329, 100W H4, or a 64623?

What happens if someone uses a different set of cells, with different internal resistances?

As far as the zener, consider at 1V across it, there could be up to 50uA of current flowing. Try sticking it in series with your meter on uA scale to a power supply, and turning up the power supply voltage to the voltage you were reading across it, and checking out the leakage current.

With only 81uA flowing through R1 and R3, the leakage in both C1 and D2 can have an effect on the circuit's operating points, especially if temperature changes. D1 only has leakage currents in the nA range, and you can probably not worry about that.

Typical DC leakage for tantalums can easily reach 2uA or more.

Some information on leakage currents and temperature, when using tantalum capacitors can be found here (a 40C change can alter the leakage current by a factor of two):
http://www.avx.com/docs/Catalogs/techsuml.pdf

 

[wquiles]

Newbie,

Thanks so much for jumping in here - I always appretiate your feedback

In all of the recent posts I have eliminated the 0.1 power series resistor. I only had it initially while investigating how many amps I was getting on a cold bulb (post #14 above). Once I started fine-tunning the R/C values, I removed it to make sure I had a more realistic circuit. I have since not used the 0.1 ohm recently, once I knew what to look for on the Vbat spikes, but you are right, I should do a couple more measurements with a freshly charged pack again.

As of right now, the circuit only works efficiently for the battery and bulb combination. I had to change R3 for the bulb change (3xD M*G to WA1319) and I also had to fine-tune R3 to account for the initial and depleted battery status. By no means this is perfect or trully versatile, but it sorta-works well enough for now

After all of the teaking involved, I would predict that the following values will have to be adjusted with each new battery pack/bulb combo:
- R3: A larger value is needed if the load on the pack is greater since the ration of R1/R3 dictates the starting point for C1 to charge. Unfortunately, too large of a value on R3 will cause a bigger negative spike on the battery, and too low a value will make the circuit dim before the battery is really depleted. The key is to maintain about 2.9V on Vgs of the FET - this or a higher voltage makes the FET conduct very well, specially since this FET has a 3mOms at Vgs=4.5V (in my tests this voltage, due to the zener, goes to about 3.1V). Because these are all competing forces, the circuit will only work "well" on a narow range. Also note that with a completely new battery pack, the ratio of R1/R3 would have to be changed as to achieve a Vgs of about 2.80V with a depleted battery pack.
- R2/C1: Might need to be adjusted as with a larger battery voltage the currents will get proportionaly higher and affect the effective charge times. I have not done any tests on this particular area, so I am not 100% sure - just an educated guess at the moment.
- For large currents (I am only testing at about 2 amps), the metal tab (thermal path) on the FET should be potted to the metal body since when the FET is not fully ON, the current will heat the FET some


Now, not that I am done with this fun project, but after all of this time/work on this purely analog soft start circuit, I think that it "might" be far more effective/efficient to have a small 8-pin PIC microcontroller (with a built-in A/D) use PWM to monitor and then control the duty cycle to achieve the "appropriate" startup voltage and regulation. The firmware can then monitor the battery voltage and/or voltage at the bulb and adjust the PWM to the FET as necessary, and also perform the low battery condition and shutdown sequence (even gaving some warning pulses, or something similar). That might be the next evolution of this work, at least for me

Will

 

[NewBie]

Hi Will.

Go back and read my post, I put more information in there.

Also note, ceramics typically have lower leakage currents...

With your uC A/D and PWM deal, pay particular attention to the conversion and setup time, you may find that you will want to do this in an analog fashion to get the speed you need to prevent surges from hitting the bulb, since at time zero, you will in fact be hitting the bulb with 20-100 Amps, depending on the bulb type. You will be hitting it with full current, each time you switch the FET, which will reduce over time, as the filament heats up. In this case, you are still slamming the filament, but you are spreading it out over more time, in smaller chunks, so it isn't stressed as bad. Not as easy on the filament as the linear softstart solution.

The next step is to go to a L-C switcher, where you can actually control the in-rush current. One thing that will help is a driver that can command the mosfets to 100% duty cycle- they do exist.

Yes, it gets complicated in a hurry, once you go beyond the basic standard linear softstart.

If you need to monitor the high side current off the cell, you can always use a:
http://focus.ti.com/docs/prod/folders/print/ina193.html

The advantage of this part, is you have a gain of 20, which helps alot when measuring currents, and you can use much lower values for sense resistors. Don't be tempted to the higher gains, as you will have a lower bandwidth, and you may not see the full surge (slow response). In this case, you could lower your 0.1 to 0.005 ohms. Remember to kelvin connect it to the sense resistor, so you don't introduce the error of the copper board resistance.

Otherwise, go with a low side sense resistor, it takes less parts (just hook up scope to the resistor), and isn't bandwidth limited at all (the probe and scope may be...)

 

[andrewwynn]

nice on figuring out the low-volt hack.. i don't think it's important to 'eliminate' the spike, just reduction is fine.. and i personally don't often 'blink' my lights.. i'm more concerned about when it's been sitting over night and i turn it on.

newbie makes a good point about the sense resistor it will already be helping to limit in-rush.. i usually just measure mV drop across a wire somewhere to measure current.. for example in my bench supply i have a 40mohm resistance.. it's perfect for getting an accurate current measurement... just divide the mV drop by 40.

I actually ordered some of those opamp/comp combos they are cool.. i had a whole ckt designed around one of them, but re-designed it around one of the micrel LDO drivers.. i found a cheaper source for them.. they are 'double' the price i'd expect to pay for the opamp/comparator so i will probably build that ckt i described since it's already designed.. i'll be a buck 20 less to build.

to ans. the question about load.. this particular ckt is fine-tuned to the voltage it outputs.. you would need to tweak the base resistor

newbie has a very good point about the leakage through the zener.. i meant to mention that is likely a cause of biasing the voltage levels and perhaps why things don't read as you expected.

The beauty of this ckt is it is a buildable by anybody ckt.. you can wire discrete components right to each other in a bundle and fit into a light host. (of course heat-sink the FET to the body of the light somehow)..

I have the 'next generation' design lined up for full regulation in a simpler design than the one in my hotdriver.

For this simple soft starter.. one more mod that would maybe make it work better.. either using an active component to take place of the diode for helping discharge.. that will get you another 1/2V quicker drop.. and the other thought.. rather than clamping C1 with the zener.. clamp that whole ckt branch.... with a larger value zener.. actually i'm realizing that means you could eliminate R1, and replace with the appropriate zener.. so the 60k/4.7uF always gets the same voltage regardless of Vbat.. i guess that'd screw up the D1 discharge.. so we might have to work in that active discharge ckt.

It'll be a moot point for me n you.. the regulation ckt is not much more difficult to wire up and the regulator is $2.50, so for about $3.00 (or 7 if you want a really nice trimpot).. you can build a full regulator.

I'm working on re-designing the requlator to have the FET below the lamp like in this design.. it's tricky, but i think i solved the main problem (since the ground of the regulator is the drain on the FET it's regulationg!).. it needs a 'hack' to jump start.. once running it should stay running.. feels a little bit like perpetual motion.. hoping it won't oscillate.

I still have the 'ludicrously simple' ckt to build.. maybe Will can breadboard one of those... it uses only discrete components so just about anybody could hand-wire one on one of those neat prototyping boards avail. at RS.

here post in the thread with the final design.. can be built completely from discrete components. I think this is the ckt you want to play with next. You can scroll up on that thread to read the prequil... i can't remember what the value of the D1 is.. pay attention to the fact the ckt is 'upside down'.. Mr Al convereted it to nch fet for me and it was easier than re-drawing it.

Oh i could probably send you the DWL file i'm curious if they are cross-platform compatible anyhow.

Oh.. btw.. per usual i made a quick reference to this thread and that one..

http://softstart.rouse.com leads to this thread
http://ldo-simple.rouse.com leads to that other one.
http://hotdriver.rouse.com leads to the hotdriver

http://cpf.rouse.com leads to my flashlight collection of links

-awr

 

[wquiles]

OK guys, I have what I think will be my final scope screen shots for this circuit . In all of them, the blue trace is Vbat and the yellow trace is Vgs, and both are on a 1Volt/div.

These two were created with R3=24K and a fully charged battery pack (Vnoload=5.65Volts):
(this one at 100mS/div)

(this one at 400mS/div)

These two were created with R3=22K and a fully charged battery pack (Vnoload=5.65Volts):
(this one at 100mS/div)

(this one at 400mS/div)

I tried putting the 0.1Ohm resistor in series with the bulb, and although it did not seem to affect the circuit much, I was never able to get reasonable measurements of the peak current. I ended up measuring the steady-state current with my Fluke M87 with no resistor in series, of about 1.97Amps (exactly where it should be), and with the resistor in series, by measuring the voltage right accross the 0.1Ohm I got 1.95Amps, which again is perfect. Sorry I was not able to give you nice screen shots of the pulsed current during the initial ON cycle

Will

 

[wquiles]

After moving the power on switch to R2, did you go back and look at the battery sag, since you removed the power switch resistance from the main current path?

Yes, it was a tad larger, but not my much. All of the screen shots on post#39 have the switch only in the path of the FET circuit.

As far as the zener, consider at 1V across it, there could be up to 50uA of current flowing. Try sticking it in series with your meter on uA scale to a power supply, and turning up the power supply voltage to the voltage you were reading across it, and checking out the leakage current.

With only 81uA flowing through R1 and R3, the leakage in both C1 and D2 can have an effect on the circuit's operating points, especially if temperature changes. D1 only has leakage currents in the nA range, and you can probably not worry about that.

Typical DC leakage for tantalums can easily reach 2uA or more.

Some information on leakage currents and temperature, when using tantalum capacitors can be found here (a 40C change can alter the leakage current by a factor of two):
http://www.avx.com/docs/Catalogs/techsuml.pdf

Thanks. That actually helps in understanding why the circuit in the bench behaved a little bit different from what I expected

With your uC A/D and PWM deal, pay particular attention to the conversion and setup time, you may find that you will want to do this in an analog fashion to get the speed you need to prevent surges from hitting the bulb, since at time zero, you will in fact be hitting the bulb with 20-100 Amps, depending on the bulb type. You will be hitting it with full current, each time you switch the FET, which will reduce over time, as the filament heats up. In this case, you are still slamming the filament, but you are spreading it out over more time, in smaller chunks, so it isn't stressed as bad. Not as easy on the filament as the linear softstart solution.

Thanks so much. I suspected it was not going to be that easy!. Maybe a future project then

The next step is to go to a L-C switcher, where you can actually control the in-rush current. One thing that will help is a driver that can command the mosfets to 100% duty cycle- they do exist.

Yes, it gets complicated in a hurry, once you go beyond the basic standard linear softstart.

I think I will stick with the simpler circuits first, before I go to the L-C switcher, where how you route wires can and does make a difference. I don't think I am ready for that yet

If you need to monitor the high side current off the cell, you can always use a:
http://focus.ti.com/docs/prod/folders/print/ina193.html

The advantage of this part, is you have a gain of 20, which helps alot when measuring currents, and you can use much lower values for sense resistors. Don't be tempted to the higher gains, as you will have a lower bandwidth, and you may not see the full surge (slow response). In this case, you could lower your 0.1 to 0.005 ohms. Remember to kelvin connect it to the sense resistor, so you don't introduce the error of the copper board resistance.

Otherwise, go with a low side sense resistor, it takes less parts (just hook up scope to the resistor), and isn't bandwidth limited at all (the probe and scope may be...)

I will read into this. It might come handy in a future circuit - thanks!

nice on figuring out the low-volt hack.. i don't think it's important to 'eliminate' the spike, just reduction is fine.. and i personally don't often 'blink' my lights.. i'm more concerned about when it's been sitting over night and i turn it on.

Then, we are done. The current circuit does that perfectly

For this simple soft starter.. one more mod that would maybe make it work better.. either using an active component to take place of the diode for helping discharge.. that will get you another 1/2V quicker drop.. and the other thought.. rather than clamping C1 with the zener.. clamp that whole ckt branch.... with a larger value zener.. actually i'm realizing that means you could eliminate R1, and replace with the appropriate zener.. so the 60k/4.7uF always gets the same voltage regardless of Vbat.. i guess that'd screw up the D1 discharge.. so we might have to work in that active discharge ckt.

It'll be a moot point for me n you.. the regulation ckt is not much more difficult to wire up and the regulator is $2.50, so for about $3.00 (or 7 if you want a really nice trimpot).. you can build a full regulator.

I'm working on re-designing the requlator to have the FET below the lamp like in this design.. it's tricky, but i think i solved the main problem (since the ground of the regulator is the drain on the FET it's regulationg!).. it needs a 'hack' to jump start.. once running it should stay running.. feels a little bit like perpetual motion.. hoping it won't oscillate.

I still have the 'ludicrously simple' ckt to build.. maybe Will can breadboard one of those... it uses only discrete components so just about anybody could hand-wire one on one of those neat prototyping boards avail. at RS.

here post in the thread with the final design.. can be built completely from discrete components. I think this is the ckt you want to play with next. You can scroll up on that thread to read the prequil... i can't remember what the value of the D1 is.. pay attention to the fact the ckt is 'upside down'.. Mr Al convereted it to nch fet for me and it was easier than re-drawing it.

Oh i could probably send you the DWL file i'm curious if they are cross-platform compatible anyhow.

Yes, please send me the Mac version of the final DWL drawing for the simple LDO design. I will start working on it today or tomorrow, since I am still on vacation and have "some time" available