BLF A6 FET+7135 Light Troubleshooting and Mod thread

BTW, these drivers can be run faster to speed up the PWM… The problem is, the more pulses per second, the more time the waveform spends rising and falling. This makes the light less efficient and more voltage-sensitive. Comfychair and I both measured this a while back, me with some improvised tools, comfy with an actual oscilloscope, and we got roughly the same answer: At the time scale of a millionth of a second or faster, the leading and trailing edge of each pulse become pretty significant — sometimes enough to outweigh the rest of the pulse.

For example, a FET-only driver can’t get a proper moon mode at 19kHz because each pulse (at fast PWM=0) is only half a nanosecond, and the actual output varies from ~3.0 lm to ~0.001 lm depending on the battery voltage. It never reaches a full “on” state, so the question becomes how high the pulse can go in that time. A single 7135 chip fares much better, because it can use longer pulses to achieve the same total volume. It’s still voltage-sensitive, but the output only varies from like 0.6 lm to 0.4 lm.

At the other end of the spectrum, slow PWM makes the output very stable, but it also introduces audible sounds and visible strobing.

So, we generally aim for the slowest speed which will be neither visible nor audible, meaning about 20 kHz.

On the tiny25/45/85 series, this will be 25kHz or 31kHz, but I haven’t tried it enough yet to figure out where the sweet spot is.

PWM on moon mode is only visible with close up shot . Thanks Toykeeper for nice explanation :slight_smile:

TK, I’m not convinced that running them higher will completely solve the problem. It seems like the high frequency is somehow causing certain parts to vibrate at a lower frequency. My S8 is normally silent, but if I loosen the tailcap the tiniest amount a faint tone can be heard. If my dog whistle app is to be believed, that tone is around 13-15khz. The problem seems to manifest itself much more in FET drivers than linear drivers. Adding more solder to connection points like switch tabs usually helps.

Awesome tips pilotdog68!
Could almost start a thread on eliminating flashlight PWM harmonics.
Great info like this often gets buried in the vast knowledge of BLF threads way too fast.

There you go! I was hoping the great minds of blf would show up here! :party:

Moon runs at half the usual PWM speed, about 9 kHz instead of 19 kHz. It uses a slower setting to improve the stability and reduce voltage sensitivity. And at such a low output level, the 9 kHz tone it makes should be fairly hard to hear.

FET-based modes are louder because there’s a lot more current pulsing. 350mA vs 5000mA, roughly, so it’s like turning the volume up by a factor of 14.

I’m not convinced it will help either. Changing the speed doesn’t guarantee that parts won’t resonate any more. But a different speed should at least change the harmonics and might make the waveforms stop aligning with the hardware’s sweet spots. Then again, it could also make things worse. Depends on the individual host, I suspect.

It’s probably more effective to do as you said, physically alter the connections a bit, and thus change the hardware’s resonant sweet spots.

Technically since we’re playing with unknown resonances of hosts/sprimgs/joints, lowering the pwm a tiny bit might help just as much as raising it.

Oh well.

Just tried the black fabric thing with an A6. It actually works! :smiley:
I can hear it and according to this phone app it is at 13.5 kHz:

Which is also what I measured in the past on an oscilloscope with other firmware running fast PWM with the AtTiny13 at 4.8 MHz.
I know according to the Atmel datasheet it should be 4.8 MHz/256 = 18.75 kHz, but I think we are missing something somewhere. Phase-correct PWM gave 7.2 kHz v.s the 9.4 kHz expected.

Related:
The MCU output signal going to the 7135s on a 105C is not pretty with high frequency PWM. These chips were never designed to switch at 20 kHz. This is why PWM < 5/255 is not stable with high frequency PWM. The signal needs that 5/255 time just to stabilize. You also need to add about 9 to the PWM level to get the expected current with fast PWM, e.g. 73 vs. 64 for 25% current.

So I tried a 9.6 MHz/8 MCU clock with phase correct PWM. This gives a PWM clock of 2.35 kHz according to the datasheet. I measured 2.22 kHz, so much closer to the calculated value than with the 13.5 vs. 18.75 kHz above. I can see the PWM when flicking a business card, and it is slightly audible, but it is not bad. Moon mode can go very low and is much more stable with dropping voltage. I like the trace much better on the oscilloscope, and 64/255 actually gives 25% current. Will have to try it in a light to see if it is acceptable.

I have a somewhat off-topic question: Why are the threads at the tail end of the battery tube anodized on this light? I know that it allows lockout, but why would you need/want lockout capability for a light with a mechanical switch?

I’m wondering if the unanodized version of this light, when it’s available, will have less problems :)…

Jim

About buried stuff: I can find the output and currents of the BLF-A6 firmware (like above in the OP) but can not find back how in each mode the current is cleverly divided between FET and 7135, and what frequencies are used. Where can I find that again, or can someone repost it?

I personally haven’t been locking mine out but I can see since I am using the pocket clip for edc how you could switch it on trying to put it away.

I think the anodized threads fall along the lines of trying to please as many people as possible.

Two more reasons to anodise the threads: 1) it is easier in the manufacturing process, you can do the anodising at the end instead of halfway before machining the threads and extra risking damage of the ano, 2) anodised threads are way more wear-resistant than bare aluminium.

The best compromise is the threads on the head side not anodised for maximum electrical contact, and on the tail side anodised against wear from swapping batteries, and possibility of lock-out.

Looks like I'm having some common 7135 issues. I'm trying to get at the driver and I get this brass retainer ring off. I've tried turning it with quite a bit off force and no joy. Should I try gentle heat or is there some other trick?

Try a drop of penetrating oil (“nano-oil” works well for me — ’oogle it for more, it’s debated among watchmakers, gunsmiths, and other users of moving parts)
and/or a drop of tuner cleaner/electronics cleaner/lubricant (I have an aerosol can from somewhere)

Don’t pour the oil in and don’t spray the lube in directly.

Just take a drop, smallest possible amount, put it in on the end of a toothpick, and draw the tip all the way around the circle wetting it where it can work its way into the threads.
Then put it so gravity works for you and leave it overnight.

Wipe out the threads with something that won’t leave more stuff in the threads where they’re rough — microfiber cloth maybe.
Look hard for any little ‘boulders’ of swarf sitting and jamming the thread where the ring needs to turn.

Wiggle the ring gently with the needlenose pliers.
Wipe and maybe oil/lube one more drop, no more.

Why all this? Guessing/speculating these were assembled without being cleaned and/or in a dirty environment, so it’s possible any sort of stuff is down in the threads in small sized bits.
Anything from metal filings to, well, anything. So you may be trying to dissolve something, or dislodge something.

This is basically how I got the stuck head off my light.

I’m not intending to argue (but I guess I am), but it seems like the problems that some (a lot?) are having with lights not working because of slightly different tube dimensions because the threads are anodized are way worse than the positives mentioned? The button is not stick out of the rear, so it seems like it’d take some effort to turn it on accidentally when in pocket especially with the clip?

In my case, I lucked out and my light seems to work ok, but if the end of the tube was just slightly too short and prevented the contact to the retaining ring, then… no light…

I can’t find it either. TK’s firmware links may tell. But she will know for sure what. I think 1-4 use the 7135+ pwm, then just pwm 5 6, fet dd for 7.

Thank for that, makes sense. I’ll put a few drops on and leave it a few days for the oil the penetrate.

Anything volatile is going to evaporate and deposit further inside the light.
LED emitters are known susceptible to contamination from anything volatile — that’s why there’s sealant over the emitter, but it’s not a perfect protection.
Anything — temperature control gunk, O-rings, solder/flux, cruft, penetrating oil — is going to volatilize to some extent.
The reflectors are really not easy to clean without scratching; if your light’s bezel/head is removable (with all the precautions about pressing on the glass while rotating the bezel, being careful not to dedome the emitter) then you might take the reflector away first and leave the LED open to the air (sit the thing on an open grid/mesh)

One drop of whatever, with that drop led around the groove with a toothpick, and let it work its way in.

Don’t mind me, I’m just nitpickery about this stuff.

Pics added! function check, tools, spring bypass, sanded star. Member vinte77 found an awesome pocket clip of you’re into that sort of thing :wink: check in the opening list for a link.