These chips won’t drop dead at 2.69v. Atmel just does not guarantee reliability outside of specs. Official specs are often conservative. So it depends on if your just doing a few diy hobby drivers or not. For commercial products you should follow the specs.
One should do tests to get a safe answer. I would assume yes, since on the one hand the power consumption of the Attiny increases roughly linear with cpu clock (which means less battery capacity for the LED), on the other hand the Attiny25 may shut your light off earlier than Attiny25V.
Sorry, takes a while to sink in. Your OP doesn't mention the diode, but post #6 and #15 do. Are you basing the 0.3 volts from measurements or just the spec sheet? Why I'm asking is because I'm reading the diode spec now (least ones I bought), here: http://www.mouser.com/ds/2/115/ZLLS410-92660.pdf, and I'm thinking the chart on top of pg 3 for Vf is maybe the voltage consumed? I'm not understanding it clearly.
[quote=Halo…]
Exactly that’s my point. A flashlight shoud be reliable.
Exactly, the chart says maximum forward voltage is 300mV (up to 10 mA current, even more at higher currrents). This voltage drops at the diode and is missing at the mcu power pin.
[quote=Sharpie]
[quote=Flashy Mike]
Tell this the judge after harming somebody with a device you designed with components driven out of specification.
Just fyi, I just dropped a cell at 2.7v into a BLF KRONOS X6 that has the 25 (not 25V) MCU, and it seemed to work fine in doing what it should do - modes selections working fine, memory is enabled and is working, strobes work, and LVP kicks in with 7-8 or so noticeable drops in output, til it finally drops the last time to OFF. In theory the MCU should be getting 2.4v running Bistro at 8 Mhz, assuming the China made boards are using the same or worse Schottky.
I repeated this two more times, still working as it should. Ob boy - if we stayed within CREE specs, like 3A per XM-L2, I might not be in this hobby anymore 
.
But still, anyone who's been in the biz a while has probably seen what out-of-spec usages can do though. Protos work, pre-prods work, then go to full production, and things start failing all over the place - seen this with out-of-spec designs. I've seen it work well for years, then a new batch of parts suddenly fail.
Currently most of us modders wanting super throw or super output depend on CREE's out-of-spec operations, and when it changes in batch's, we gotta deal with it.
Just realized that a flashlight can be run faster than my first computer! 
I agree, but this driver is also used in BLF X5/X6 and Cometa, bought also by “non-flashlight-professionals”, and it will problably get used in even more future flashlights.
Another fyi... I brought the cell down to 2.55v now, and the Tiny25 base X6 light continues to work fine. I also measured the voltage drop across a Schottky diode on a 7135 based driver using the same 2.55v cell, and it does read a 0.25v drop.
So in theory, the 25 is getting about 2.3v at most, and still functioning fine, though it's safe range is 2.7v+.
> It’s not as if your life depends on a torch ?
It has, and likely will again.
You don’t live in earthquake country, then?
I do. That’s why I poke at reliability questions, I’m from the moonlight end of the community here not super-brief-flashbulb end, and why this thread caught my attention.
Read the stories from the Loma Prieta quake — which was much less than the “big one” — and you’ll find the word “flashlight” showing up quite often.
Just saying — I’d appreciate (and will ask for) attention from the folks designing drivers here on this end of the range between reliable and stunning.
Reliable wins in the crunch.
PS, Murphy's Law and the Risks of Designing \"Off Data Sheet\" | Analog Devices is really important to understand.
I’ve tried for literally decades to convince amateurs why the pros talk about “wounding” not “killing” semiconductors through careless handling.
That page does a good job — if people will read down through it.
Checked: high PWM frequency is no problem with Attiny25V at 4 Mhz, I only had to set fast PWM instead of phase PWM (what Bistro does).
Just did a quick test with Bistro on 4 MHz Attiny25V:
- set the clock prescaler to 2 for a cpu clock of 4 Mhz
- commented out the SOFT_START definition to gain some more bytes program space
- divided the BOGOMIPS definition by 2
Works (on my breadboard, haven’t checked all options yet.)
I'm think'n the important things are the delay msecs function, the WatchDog timer, and PWM's of course. I assume the high speed PWM is slower at 4 than 8. I think phase does about 15 KHz at 8. When I tried this a while back, couldn't get the msecs delay working properly I think.
PWM is exactly the same, we only have to toggle between PHASE and FAST as we had to with Attiny13A.
Delays are based on BOGOMIPS, just change this value to 1000. Haven’t checked watchdog yet.
So — is it fair to sum up and say that for the driver described as
the actual drivers made for resale never did use, or need, the “V” version of that chip?
Fair to say this is a labeling problem, not off-spec, not corner-cutting, not a surprise to the designers?
Leaving aside the arcane discussion about what might yet be improved — just sorting out whether the name should describe what’s in the thing, or not ….
Yeah, this is asking the designers who actually knew what they asked for to say something, if y’all are willing, about this little oddity.
Ooops, get'n back to the main topic - You are right if you add one "not" and remove a couple of "not"s, if I understand this all correctly.
It is not a simple labeling problem, it is off-spec, it is corner-cutting (the 25 version is probably available cheaper than the 25V).
Can't answer for the designers... Mike has been pointing out the "25V" version (as listed) is within spec for the voltage cutoff level configured, but the "25" (actually used) is at the hairy edge, or actually just out of spec. From my own testing though on my one piece of the BLF X5/X6 driver, it still worked fine even further out of spec. But again, that's tested on just one piece, one sample, under ideal room temp/humidity/etc. conditions.
With the same silicon inside and the 25V just being the “cherry picked” ones - could you explain why the operation clock of Atttiny25 is specified up to 20 Mhz and the “cherry picked” 25V only up to 10 Mhz?
Please enlighten me, i am not a chip designer.