Yeah I saw the forum and I should definitely expose my problem on it.
Well I tested at any current really, as you know the minimum current of my drivers is very low, virtually zero, so I don’t think the load is the problem. I should still test with the buck driver removed and see what happens.
I just assumed it would work as well as the MP3432, which works excellently at low voltages, as long as its Vin pin is >2.7V.
Hey @thefreeman, any chance you will order a small production batch of drivers that are compatible with FW3X/FW3A host?
Well there are several points that makes it more complicated :
- it’s double sided so the setup fee is significantly higher, increasing the price per driver in a small batch. Because it would be produced by a fab I could use 0402 passives to save space but that wouldn’t be enough to make it one sided.
- for the FW3X, since it’s a boost driver, it would be configured to 9Vout (3 LEDs in series), and the MCPCB would have to be modified to 3S like I’ve done several times on different MCPCBs, but are people willing to do that ? That may decrease the number of people interested in the driver.
- physically it’s the same driver in the FW1A, potentionally increasing the demand, but then it would be different Vout/Iout variants, which require a couple of different resistor values. For example 6Vout for B35AM and XHP50, 12Vout for XHP35, technically there is the B35AR (12V) but the availability is not as good, and the XHP50 can be wired in 12V with adequate MCPCB, the problem is that they are often non DTP, or if they are DTP the middle connection goes through the middle pad, shorted to ground, meaning shorted to batt- , it’s a problem that people tend to not be aware of and end up with a non functioning modded light.
The fab can’t do several variants in one batch, so there is a new setup fee for each variant.
What I could do is order one configuration and modify for the other ones myself, it’s just more work, still much less than fully assembling the drivers myself.
- because the driver cavity is only ~2.5mm, the inductor choices are only XAL6020 or XGL6020, both of which can’t be sourced with JLC, which I considered as 1st option since they have the lowest setup fees. I would have to reflow the inductor myself.
So that’s why I would prefer to go with a single sided driver for a first production batch, a buck driver would have only one configuration, which would be simpler.
I did not expect that there will be so many hurdles to implement a suitable driver for the FW3X/FW3A type host.
I must apologise to you for taking up your time to explain the technicalities of the drivers to a person like me with zero knowledge on electronics.
Interested to know what existing flashlights can fit the one sided driver?
Don’t worry, I don’t mind at all.
If you’re asking about BST21, which is the single sided boost driver with 21mm component clearance, it can fit in probably many lights, for example the WT1M/WT3M (25.5mm), D4S (31.2mm),DM11, KR1, PL09, some Convoys 21700 with large drivers M21D (26mm IIRC), MT01/MT04 and Astrolux equivalant, probably a number of sorfin and Astrolux 21700 lights.
But as you can see they have different driver diameters, which would mean different production batches, or maybe one larger diameter could be produced ( or 2) then people would file them down to the suitable diameter for their light, but are people willing to do that ?
Assembling and testing the MP3432+MP2145 NiMH-li-ion boost+buck driver :
The MP3432 was used specifically (instead of MP3431) because it has a passthrough mode, meaning it stops boosting when the input voltage is higher that the output voltage and behave like a piece of wire, with some resistance (high side FET+ boost inductor = 10mΩ+8.2mΩ), this output voltage being the boosted voltage, not the LED voltage, presently the boosted voltage is set at 3.37V, then this serves as the input voltage for the constant current buck driver.
If you look at Vout at 3Vin and 4Aout, it’s lower than at 4Vin, that’s because the 3.37V boosted voltage is not high enough to allow 4Aout at ~3.28Vout, which is the LED Vf + drop across the longish wires and current shunt, used for accurate current measurement. It could only do ~1.7A.
So I used a lower resistance shunt there, lowering the output votage needed for 4A. A 519A at 4A has a Vf around 3.2V, so it should be OK, if a higher Vf LED is used then the boosted voltage should be set a bit higher to allow full output at any (li-ion) Vin.
I was a bit worried about the passthrough mode because it only works in PSM, meaning the switching frequency decreases in the audible range at very low output and could generate noise, like my first driver with the TPS61288 did, but I hear nothing so that’s great. Plus now the power consumption will be lower at very low output vs USM (ultrasonic mode, freq>30kHz).
Considering I used 4020 inductors (4x4x2mm) the efficiency is really good. Though the fact that it peaks at relatively high output current means that I might have used a too small inductance (680nH, 8.2mΩ) for the boost converter. It would be good to test with XGL4025 (4x4x2.5mm), like 900nH (7.4mΩ), or 1uH (8.5mΩ), it could maybe break the 90% barrier with NimH voltages.
How much I would have such driver so well done even for my old Nitecore Tip or others square lights
Another great-looking design. Thank you for sharing, especially the test data. Really great performance even on an NiMH cell - 6W with a fresh cell and close to 5W even as it nears empty.
I had to read back through the thread a couple pages to remind myself of the previous FWAA mods. If I understand right:
The latest FWAA driver is the cascaded boost-buck design. This one works on both AA and 14500, where as the previous version had an issue with the boost chip unexpectedly not working at low voltages.
The first FWAA mod thefreeman did used a boost only driver, so it was matched to a triple MCPCB that was modified to feed the three LED’s in series (~9V) instead of parallel (~3V).
Both designs are very similar in efficiency, with seemingly a slight advantage in certain conditions in efficiency for the boost-only driver, and more obvious advantage in maximum power, but of course, the cascaded boost-buck offers the most flexibility (suitable for both the FWAA and FW1AA, for example).
Thanks.
That’s right, the first FWAA NiMH/li-ion driver was boost only, set to 9V for the 3 LEDs in series : thefreeman’s HDR Anduril 2 high efficiency drivers - update : FWxA boost driver - #319 by thefreeman
The second one was boost+buck based on the TPS61022+MP2145, the TPS61022 is very small and require no extra passives, still with decent internal FETs (12/18mΩ), but unfortunately it doesn’t startup with NiMH : thefreeman’s HDR Anduril 2 high efficiency drivers - update : FWxA boost driver - #352 by thefreeman
And the latest here boost+buck with MP3432+MP2145, the MP3432 being larger and requiring more passives I had to use 0402 ones, but it has lower resistance FETs (6/10mΩ), so it’s more efficient/powerful in the end.
Once I receive XGL4025 inductors I’ll make new measurements, maybe push it to 5 or 6A on li-ion and a bit more on NiMH as well.
Another way to increase the efficiency a bit would be to use the TPS62867 for the buck converter, which is a bit better than the MP2145, but aside for availability problems, the layouting coupled with the MP3432 is kind of difficult, while the IN/OUT/GND pins of the MP3432 and MP2145 conveniently match.
Pretty impressive work!
So hoping for an improved Fwaa driver with boost!!
So I’ve been working on making it production ready :
- now single sided for lower assembly setup fee, the FWxA having 20mm clearance I used BST21 as base, a 6024 inductor instead of 7030 and a few 0402 passives to shrink it to 20mm
- for the different output variants (6V5A, 9V3.33A, 12V2.5A) I left the resistors to be changed as 0603 for easier swapping.
- I used a much less costly inductor that still has very good characteristics, only very slightly higher DCR (8-9.6mΩ instead of 6.8-8mΩ for XGL6020)
And efficiency measurements :
Hi, excellent job!
Do you think it could be possible to make it fit this flashlight?
It has got a magnetic switch that dimmer the intensity.
Well aside from size constraints, the magnetic switch requires additional hardware, and the firmware would have to be adapted for it (unless it simulate a simple on/off momentary switch)
Any chance you will have a functioning Fwaa driver (3v LED) available? I fried my v2.2 driver so have a dead Fwaa ti at the moment…
Well I have the original ones, both SOIC and QFN versions. I don’t have any of mine for selling at the moment.
Ok thanks. Please let me know if you make any to sell so I can fix my light. 
The efficiency looks great for the size!
Installed in the FW1A with an XHP50.3 4000K sliced and RGB aux LEDs :
3 Thanks
Very nicely done!
Those tiny wires drive me nuts.