Oh, it was 95 at 20%.I remembered wrong.
thrunite T1 :
Uses the MP3429, 600MHz synchronous boost converter with integrated switches, ~12V output.
6030 0.68uH inductor
AON7423 RPP PFET, <5mΩ at –4.5V, <6.5mΩ At 2.5V
Rsense = 150mΩ, Vsense = 150mV
The efficiency isn’t as good as it usually is with this converter, it has quite thin traces, fairly high current sense resistor and too low inductance.
Input capacitance is quite low (one 0805 capacitor), I had to add a bulk capacitor for making the measurements as it didn’t work properly in turbo.
Are there any other major manufacturers or models for small boost converters other than MP or TI?
For high power converters (integrated FETs) I don’t think so, Analog devices’ highest switch limit boost converter is 7A (and it’s a dual phase, 6A for a single phase one), whereas TI has two 10A ones (TPS61088, TPS61178) and a 15A one (TPS61288) and MPS has several above 10A. But for lower power there are many manufacturers
What do you think would happen if using 2 boost converters in parallel or series? Is it viable to achieve higher output currents without hit to efficiency when using parallel setup for example?
For parallel I think you need some sort of load sharing mechanism otherwise one converter will deliver most of the current, for series I’m pretty sure the output needs to be isolated.
Using those converters the simplest thing is IMO to create a dual output driver and have 2 separate LED channels, which is possible with multiple LEDs but also with XHP50/70 since they feature 2 separate strings of 2 series LEDs, the MCPCB needs to be modified though.
Another way is to use a boost regulator (external FETs) like LTC3786 with low RdsON FETs for higher power, several regulators can be put together to run several phases for a very high power single output boost converter. But with this even just one phase takes more space. Loneoceans GaN driver for example uses a boost regulator and well, GaN FETs, probably uses that regulator actually.
Why would one converter deliver most current? In mechanics you could use two pumps to suck from one pool and both pumps do the same job and deliver the same amount. Is there some different principle in electronics that I don’t get? Sorry for off topic in your thread.
I’m honestly not sure, I think I’ve read that somewhere and that it’s was basically a bad idea but I haven’t looked much into it.
Sounds on topic enough to me 
I started working on a LTC3786 driver for SP36 a while ago but I haven’t started routing yet :
LTPowerCAD simulation :
This is capable of much more power than MP3431 based drivers.
Are you able to keep the charging function of SP36, or are you going to abandon it?
I don’t think I have sufficient space with the inductor I chose (1580 sized, 0.42mΩ), if I were to add USB-C charging I would probably need to use a 1010 inductor instead (like XAL1010 1uH 1mΩ).
I guess this is due to the output regulation.
With two pumps immersed at the same level and at the same power they both produce a similar pressure and the small difference there might be between the two doesn’t disbalance the current of each much. This is a open loop control.
With a DC-DC converter we want a constant output voltage (for example), if Vin and Iout is always the same then you could use a fixed duty cycle D = 1-(Vin/Vout), as with the pumps it’s an open loop control. But since Vin and Iout (causes voltage drop) will change then you won’t get a fixed Vout, so instead Vout is compared with a reference Voltage (Vfeedback), if *Vout<Vfb then D increases and vice versa, it’s a closed loop control.
The problem is that Vfb isn’t going to be identical between two converters, so one might regulate to 6V and the other one to 6.05V or something. So the 2nd one targets 6.05 V by increasing the duty cycle, the second one will keep decreasing its duty cycle because it wants 6V and 6.05V is always going to be too much and will end up doing nothing.
With the constant current regulation I think it will be the same.
That’s my understanding of it.
Edit : reversed that*
Might be. But if also feedback is parallel, then it should be same to both? Doesn’t it depend from led VF from which it’s stepped down with resistor. If both converters read feedback from that same source shouldn’t both boost converters still think they have same job to do? Might be worth trying if I ever find motivation to start planning my own driver.
“”TI Synchronization “”:https://www.ti.com/lit/an/slvaeg8/slvaeg8.pdf?ts=1653289335328&ref_url=https%253A%252F%252Fwww.google.com%252F
But im think if you want moore power better to use external switches controlled by IC directly .
This is unrelated, this is for synchronising the switching frequency.
Acebeam E70 CRI (FC40) :
Again it’s based on the MP3429, 12V output for this version, I believe the XHP70.2 version is 6V.
What is interesting is that they use a second higher value (0.25Ω, 2x 0.5Ω weirdly) sense resistor for the moonlight mode, like my drivers and Zebralight’s (usually they use 3). The low value high power sense resistor is 10mΩ + the ON resistance (3mΩ) of the NFET (AON7520) used to switch between the two.
The turbo output starts to decrease below 3.5V and reach ”high” output around 3.2V, this isn’t a limit of the boost IC but a firmware limit.
The reverse polarity protection PFET is an AON7423 (3333 package), this seems to be a common one, seen on Thrunite, Convoy and Acebeam drivers.
Inductor is 7.5x6x5mm, 1uH.
The efficiency is not very good in moonlight low and mid 1 because they decided to use FCCM (fixed frequency PWM) instead of PSM or USM (usually this one is used used) so relatively a lot of power is wasted at low output, bad choice.
Another strange choice is the tantalum capacitors for the input, I thought they were more expensive than MLCCs and two 1206 MLCCs would have been adequate here.
I forgot to take a picture but there is an aluminium post in the driver cavity above the boost IC with a thermal pad to improve thermal dissipation, that’s a very good decision especially with the fairly high output for this converter.
Nice! You still have more drivers to measure?
It would actually be interesting if you could do comparative measurements of classic BLF-type drivers with 1+n+fet topology and fet only topology.
I’ve seen pictures of theory, how having multiple different 7135 channels affects effiency, but never any measurements.
Freeman this is a really great contribution! Great thread and thank you for putting all the time into this and sharing with us.
I’ve got the SC64 LE left and then I’ll have been through all my flashlights with switching drivers
Problem is PWM affects LED efficiency so I would need compare lm/W instead of just the driver efficiency, it would be interesting but a bit more complicated.
Thanks.
Yeah,it is more complicated. Great work so far.