I had a similar problem on a BLF17DD that I built a while back. It was playing up sporadically, & was working in ‘next-mode memory’.
I was trying out the updated location of the CAP that eliminates the over-current spikes to the MCU.
I ended up moving the CAP back to the original position on the board, & the problem went away.
I never pinned down what caused the problem, but I’m guessing it was to do with either the CAP it’s self, or a bad solder joint of the CAP.
You had some other issue we could never get to the bottom of, relocating the cap (C1, not the off-time cap) flat out works, the only way for it to not work is for something to be wrong somewhere.
Yep, that’s why I suggested that my CAP was either faulty in some way, or I had not soldered it properly.
C_K has mentioned he’s changed the off-time CAP, but not the CAP at C1.
I didn’t try changing the C1 cap but all the 17 boards (v1.0’s) are the new version with the new cap location (the 16x7135 is to the new style but didn’t seem like the 7135 drivers needed to have an update).
My reflow settings are 315* at 4.5-5 air settings, I hold the air on the component 15sec max if it doesn’t work I give up and remove/reapply solder and try again after it cools.
I still have the assembled driver removed from the S6, I’ll start working on it replacing C1 first to see what that does, OTC setting is stock @ 130, what direction would I move to lower the time?
FWIW I never had this issue on any of the 50+ old style BLF DD drivers I’ve built, only the new ones, I’m glad I had so many extra old style ones, I’d hate to send one of these to a customer. Also why I ordered these in soon as they came out and installed them in all my personal lights.
Solder joints are all for sure good, 1 (not to sound stuck up) but I’ve gotten really good at soldering recently and 2 I chance nothing, before I try powering anything up the first time I test every solder joint with a DMM for continuity, especially the OTC testing for short to pins 1&3 and verifying the cap it’s self isn’t bad and just a sort to GND.
Just to clarify, I wasn’t suggesting that the location of the CAP was bad, just the possibility of a bad joint/CAP.
Interested to see how the board from the S6 turns out.
That's exactly what I was thinking as soon as I read your initial post: are you using the new capacitor location? Now that we know that's the changed variable, the next step is looking for the solution, which hopefully is as simple as finding a new threshold value.
I found out early on that the off-time timing is sensitive not only to the 1uF capacitor on the pin, but also the main capacitor powering the MCU. I built a batch of boards where I inadvertently put 1uF capacitors where I normally put the 10uF...none of them worked right until I swapped the 10uF back in.
The easy solution would be to put a scope on the off-time capacitor and see what's going on timing wise....the other solution would be to start playing with the off-time threshold values to see if you can get something to work.
I'm still running the old style boards and "slow" 9.4khz PWM...I guess you could call me old fashioned! 
It could be the caps you are using. SMD caps are evil little creatures. And the smaller their physical size and the closer they are run to their working voltage spec, the more evil they become. For instance, a 10uf 6V cap might actually act like a 1 uF or smaller cap if run at 5V. What you think are 1uF caps may be acting like 0.1uF caps… I always use the biggest package/highest voltage rating caps that I can fit on the board.
Also, note that capacitor values can change over time. Some dielectric types are particularly bad… avoid the cheaper/more unstable dieelectrics if a stable capacitance value is important. For instance X7R caps can be spec’d at +80/-20% values as shipped from the factory… and can get worse from there…
Which is the correct cap voltage for off - time memory?
Just measure it zere. Think what Texas is trying to say is, get the biggest quality cap you can find and stick it in there. The voltage doesn’t have to match, long as it’s higher than what your putting in.
I didn’t get to this today unfortunately, I have the old S6 driver sitting out on the bench waiting, even got a 10-pin ribbon cable with flying leads soldered on for programming over and over trying to mess with the values (what direction should I start moving?) but I didn’t get to sit down and actually try anything today.
Is there some way to measure cap’s? The ones I’m using are all known value and all come from new tapes so I can’t imagine they’re bad but if there was some way I could test them that could maybe help some.
The part I've been using for the off-time function is 0805, 1uF, 10v, 5%. Not had any issues with them. Even with a higher-Vin zener-modded driver there will never be more than the MCU's Vcc voltage available at the off-time cap, the MCU gets unreliable over 6 volts. So if the MCU is still working the 10v-rated OTC will never get more than 6 volts.
I think the one I used in the Digikey shared carts is the same as above, but 16v.
I have a cheap import DVOM with a 'capacitor' function, it seems to give readings close enough when checking known-value parts.
A big problem is the capacitor value changes with the applied voltage. Who knows what voltage the tester is testing at?
Capacitor issues aside, we know what the variable is that changed here and it isn't the capacitor properties/type of capacitor/heat/etc.--it is the capacitor location. I don't think that capacitor damage or installation is the issue here.
Regarding reflow temperature...I always use at least 350 and air speed of 5. I also solder 7135 chips with my iron set at 720, but I like to work fast. Everyone has their preferences. Total heat absorbed/temperature of the work piece is what matters.
In regards to your initial paragraph here… I don’t see how your “for instance” is actually supporting “the closer they are run to their working voltage spec, the more evil they become.” Can you elaborate? It sounds as if you may actually mean “the further away they are run from their voltage spec, the more evil they become,” but I just can’t tell. I’d like to hear more about this, but I also don’t see where it’s relevant to the problem at hand.
A quick search found this useful article:
http://www.maximintegrated.com/en/app-notes/index.mvp/id/5527
I try to use C0G, NP0, X7R or X5R dielectric (in that order). I generally avoid Y5V Z5U. And using the highest voltage rated part for a given case size and value will go a long way to minimizing capacitance loss with applied voltage.
I went back and re-read the OP. Cereal_killer mentioned to me previously about his offtime issues w/ the 22mm driver but I forgot about it (sorry!), and when I read the OP originally I somehow glossed over the fact that it was with one of my drivers. Sorry for totally ignoring this Cereal_killer, it wasn’t on purpose.
I’ve got a 22mm driver built up with the offtime cap in place (post #38). Mine works just fine setup for offtime with memory. This indicates that at least in the case of the 22mm driver it is not a layout issue. Cereal_killer, I don’t know if you added it later but I see that your picture in post #35 is of a 22mm board with the offtime cap position not populated.
Resurrected,
anyone have any idea what the problem is/was? I'm having similar issues.
Without re-reading the thread, I suspect that this may be related to the 22k resistor I was recommending. ToyKeeper did some tests which indicate that this resistor is a very bad idea, 19.1k is still the way to go.
Take a look at the posts I mention in my own post #642 here for more information. If this doesn’t seem like what you are experiencing, doesn’t apply, or I just need to re-read this thread let me know.