Not a thread - just finished some tests with the BLF17DD v1.0 using the de-domed XP-G2 S2 and got about 0.15A to 0.18A higher, depending on the cell (SAM 25R or LG HE2). Best #'s for each driver:
A17DD-S08: LG HE2 @4.23v: 3.78A @tail
BLF17DD: LG HE2 @4.22v: 3.96A @tail
Same host, same LED (didn't even remove the LED - it's screwed down), same 20 AWG wire lengths for LED wires, same 22 AWG wired driver spring setup (different spring but don't think that should matter), same batteries.
Conclusion:
Well, with just one sample of each, there can be variables I'm sure: quality of reflows, tight driver mount, battery charge differences, FET variations, etc.. Seems slight edge to the old BLF17DD which is kind of what I suspected, but since I only tried one A17DD-S08 up to this point, I can't say anything definite from this - too little data, not enough samples tested, don't 100% trust my methods, measurements, etc. It's not lab level for sure... I would trust djozz more than myself - absolutely .
Several other options are discussed in this thread. Did you look for those? Rufusbduck suggested one I really liked and I think there were a couple of others which sounded OK too.
Exactly. We would need very precise measurements and tightly controlled test conditions if we were actually out to prove that one is better than the other. But to prove that the info from the datasheet applies in the way we thought it would? No problem, you and Tom E have already done that, in spades!
All that said, anyone wants to play around with testing different FETs, here are 3 which all appear to have some superior characteristics:
PSMN0R9-30YLD - rufusbduck pointed this one out, it has very similar characteristics to PSMN3R0-30YLD but is better. Slightly lower Rds(on) and slightly lower Vgs. Also significantly higher “total power dissipation”. We expect the lower Vgs to ensure that the FET remains more “fully open” on single cell setups.
SiR800DP - comfychair pointed this one out to me in reference to this driver by PM weeks ago. I forgot about it until I went looking for FETs which I thought might work properly with the QX7136-based A17LDQX driver and it turned up again in that context. Much lower Rds(on) than our selections so far and even lower Vgs than RBD’s selection. Again, the low Vgs should help keep the FET fully turned on.
SiS414DN - Another one which turned up during my search for good FETs for the QX7136. Low maximum current handling, 20A “package limited”. I’m not really sure exactly what a package limit involves, since we know that PowerPAK® SO-8 can handle at least 50A (the SiR800DP is spec’ed for that) and the graph in the datasheet goes way beyond 20A. Even lower Rds(on) than SiR800DP, but only slightly. Similar Vgs characteristics to SiR800DP. Lower Qg than SiR800DP, but I doubt that that will help us any. Costs less than the other two FETs in this list.
Typically we look at Rds on as tied to gate voltage but after slogging through some of these data sheets I see that it also varies with drain current and temp just not as dramatically. Which makes sense since resistance in most things goes up with temp. To test more would you start with the third possibility in the hope of seeing gains at lower cost in spite of the current limit or with CC’s suggestion to find the possible upper limit?
It’s already been pointed out elsewhere (by comfychair probably, and maybe others) that we achieve diminishing returns once Rds(on) gets low enough. IMO it’s already “low enough”. The difference between 10 milliohms and a hypothetical 0 milliohms is probably under 0.2A for an XP-G2, assuming we are at/near/under 5A. That’s just not very exciting to me. (Consider that we are already <7 milliohms! How low do we expect to get?)
I wouldn’t have listed all 3 if I didn’t think that they were all just as worthwhile to test. That said, the maximum gain anyone can expect to see for an XP-G2 is probably around 0.1A, maybe 0.15A - and it doesn’t matter what FET package we are talking about, we’re that close to true DD here. The higher the current, the higher the gains, but there’s still very little to be had!
Sorry wight, I just know that a relatively few out there are looking for every last mA. I totally agree it’s not much and to most of us not worth the added expense but until my own head hits the wall something keeps me leaning forward. Although the low on state voltage of the one I suggested was what caught my eye at first, it’s the lower “elbow” in the curve that made me want to bring it to attention. We may not gain anything at all but at this relatively early point developing these drivers I’d still like to confirm what the different options offer.
Nothing to apologize for RBD. My point was that anyone looking to experiment should probably buy all 3. (Also that nobody should assume that I will do the experiment.)
RMM, vias help prevent the lead wires from ripping the pads off the PCB, if you solder the wires thru the via it is VERY hard to rip it out, especially if you over heat it while soldering and/or mod alot, makes the boards a bit more resiliant
Either way C_K and wight are one of the MASTERS of building and designing these lights…I learned alot but these guys excel at it!
You're right about that, the through hole vias are stronger, but haven't ripped an LED pad off yet so it really isn't an issue for me, but I did want to know what others thoughts were since we all have had different experiences building lights.
Generally when I’ve ripped pads off in the past (yeah, many times) it’s been because of misbehavior on my part. Ripping LED+/- pads off isn’t something I’ve done much of though. Generally I’ve damaged component pads due to lots of heat and pressure. Before I learned to add solder when removing components I frequently applied pressure to components while attempting to remove them. Between the extended heat dump from my efforts and the pressure it was common to damage a pad. I’ve also damaged SOIC pads after air-wiring directly to them and then manhandling the assembled PCB.
I don’t see it as a problem that needs to be solved, although I’ve been adding big vias when I had space and they didn’t seem to interfere badly with other things. Maybe I should re-evaluate? If it doesn’t have any downside I can see the advantage of a hole for wiring the LED directly to BAT+ or BAT- depending on the type of driver. These PCBs are small and if drilling is to be done it’s better/easier/nicer to let the fab do it IMO. People don’t have to use it, but for those who want to it’s already done… OTOH aesthetically I do not think that the large vias are beautiful and they do take up space that could be devoted to large SMD pads or other misc stuff.
I used to try to use 20 ga teflon insulated wire that is very stiff. Trying to stuff the driver in a tight space where the wiring has to be manipulated would pull the pad off the driver.
Now that I use 22ga Silicone insulated wire that never happens.
Oh, er, yeah. I’ve removed other pads by accident while un-soldering things. Probably for the exact same reasons. But those were generally pads I wasn’t planning on ever using again.
Through-holes are definitely sturdier, and preferred when possible. Attaching 3x22g wires (twisted and soldered together) to the LED+ and LED- contacts in my SRK wouldn’t have been very feasible without the through-holes on the BLF-SRK driver.
I asked Matt to put them on the Tiny10 for the sake of versatility and with so many doing spring mods having a single wire go from the top of the driver spring straight to led+ makes some sense(spring carries only mcu current) but yes, since I’ve switched to more flexible wire my pads are in less danger. 0:) Being able to solder the connection from a less crowded side can be nice too.