But if the chip automatically reduces the current when it gets hot then there should be no problem right? 
4xNiMH starts at about 5.8 volt, but will very soon drop to around 5 volt (Depending on load).
Here is my test of some NiMH batteries:
/Eneloop%20AA%20BK-3MCC%201900mAh%20(White)-Capacity.png)
HKJ I’m not at all doubting any info you post but I have a light (pukelight v1) that uses 2 parallel strings of 4s NiMH and it runs at 5.3v fresh off the charger, dead they’re around 5.1v, why the difference?
I have multiple 4s AA lights (and the one 2s4s) running 7135 based drivers and they all work perfectly and have yet to overheat.
That sound like the charger is undercharging the batteries or maybe they rest a long time before you measure them.
I assume you measure the final voltage without any load on the batteries, that will be higher than loaded voltage.
You might not see that they overheat, it is very difficult to see if the brightness drops 30% and when you add a ammeter you add some extra resistance, i.e. the voltage to the 7135 will be lower with less risk of overheating.
The risk of overheating depends on the number of 7135 places close to each other and their contact to a heatsink (I.e. flashlight body).
what i would like to know is the difficulty in creating a single li-ion boost driver for single high Vf emitters (xp-g2, xm-l2 etc). i know the discussions with lightmalls fell through. what could have been the technical causes? large drop out, low efficiency?
Something like this: http://lygte-info.dk/review/DriverTest%201A%202.7-5.5V%20Buck-Boost%20UK.html
thanks hkj. i know buck-boosts like this are available. of course i mean i would like to know the inherent problems of a boost driver for to meet Vf at high current (>4A). i see your reviewed driver regulated well at 1A, and that boost mode was more inefficient than buck mode.
It is no problem to make a boost or buck driver, but combining to a buck/boost is a problem. There exist only a few chips that can do it.
The problem is that you basically need two switchers sharing one inductor, this requires multiple transistors. As long as all the transistors are in a chip it is fine, but a led driver does not have place for external transistors for higher power.
a mass market driver with boost should be buck+boost, and i can understand the complexities of such drivers. good explanation also! but the driver in dicussion with lightmalls would have been boost only (DD at Vin>Vf). in other words, a specialised driver for lights which can withstand DDing an emitter (i.e. 5A+).
at high current would a boost driver never reach the required efficiency to work from a single li-ion? or something else, like lack of commercial ics? i can’t even find any mass produced step-up voltage converters for this range and output.
I agree, I changed the OP accordingly. Thanks! 
Buck-Boost drivers are also less efficient than just a Buck.
If he wants to drive a XM-L2 at 4A or more he should consider having two Li-Ion cells in series to get 7.4V and using a buck converter… However it will be bulky to be able to withstand 4A.
Again, as long as the input voltage is below 6V even with fully charged cells, the driver won’t fail. But it will be wasteful and run hot.
Let’s do some maths:
Your 4*AA are at about 5V during discharge.
The LED needs about 3.3V
that means that the driver needs to lower the voltage of 1.7V!
3.3/5=66% efficiency! That’s not really good… And that means that 37% of the power is lost in the driver and heats it up.
Use a dummy cell and efficiency will go up, reducing the heat in the driver and you’ll still have the same brightness.
The general problem with a boost drivers is that it always need more input current than it delivers as output current.
Efficiency varies with design, but because the current is high it will have high ohmic losses.
i see in tests of some high discharge cells that outputs of 10A @ >3V are sustainable, so i assumed in theory a boost driver could continue to regulate on a single cell. if the loss is primarily ohmic then i presume it is linear. the size constraints make it difficult i guess.
Just a FYI/opinion:
Advantages of the direct drive drivers are also that there is no tint shift between turbo and low. Or at least very very little. True constant current drivers show a huge (to me, it might be individual) tint shift so the color of things look "off" in lower modes. I especially notice this with high CRI emitters and very cool white emitters.
And the draining of the battery does change the tint a little but it is almost unnoticeable over time. What is not unnoticeable is the rapid decline in light. At least rapid enough for me to notice it and get another battery ready to load.
And I disagree with the first line of the conclusions. The one about 1 li-ion and 1 emitter.
For the above reasons.
A linear driver can use PWM to make low modes. In that case there is no tint shift.
It’s the linear drivers that are truly constant current, even in low modes that cause a tint shift. That’s the disadvantage of constant current driver, the advantages are a better efficiency and no flickering.
A driver like this one is linear, which is good because the brightness will not vary when the battery discharges. But it also uses PWM to make lower modes. So no tint shift.
In my opinion this is better than a direct drive drivers because the brigtness stays constant. I don’t see any advantage to the cheap direct drive drivers that have a 0.2ohms resistor to limit current to safe levels…
I have to admit that I may be biased because I personally don’t care about tint shift, but I really care about the improved efficiency of true constant current drivers. I may add a sentence about tint shift and PWM to clarify things out.
I don’t think I ever posted in this thread to thank you lagman, but I do point people over here every little while. One of these days I’ll make some suggestions of my own for sprucing things up, but I think you’ve done a good job!