Good work. How high can you take the drive current with the larger inductor? Also, do you know the value of the stock inductor or any of the ones you tried?
I don't know the value of the stock inductor, but it seems to saturate somewhere around 6A...at around 6A-7A depending on the driver the output ripple increases by around 5-10 times regardless of anything else I do to the driver...which tells me that the inductor core is getting saturated. The actual inductance the stock inductor seems to be enough but the core can't handle the current. I had a few inductors laying around that were all 15A+ but only in the 10uH range and they weren't enough. The one that is working well is off of a 600W boost converter so I don't know its exact value, but I have several 20uH+ inductors on the way so I can get the best value/size ratio tuned in. Stay tuned.
The new oscilloscope actually ended up costing me more LEDs because I would have otherwise given up long before I did! The scope gave me hope of being able to resolve the issue. After a while, looking at the scope image I could usually tell on the second-to highest level whether the LED was going to blow on the highest level before I ever clicked it up but I usually clicked it up out of sheer hope and desperation anyways! 
I’m building a light and plan to use this driver (7A) stock with an MT-G2.
Do you forsee any problems?
I haven't tried it with an MT-G2 but others here have been pretty successful with it. Are you planning on 3S input? I could give it a spin tonight...although I will shed a much larger tear over an MT-G2 loss than I will over an XM-L/XM-L2.
Very detailed testing and at some sacrifice, appreciate the hard work and time spent.
I’ve also blown about 8 emitters with this driver, have a de-domed XP-L running stably at 7.2A now. I used even larger sense resistors and used a pair of R070. Vishay-Dale from Mouser if I remember correctly.
May or may not be looking to use my second driver. Am still pretty blown out from all the xperimenting on this one. 
Try some Luminus SBT-70’s on for size 
Yes. 3S input.
I didn’t even consider this driver might be a problem with an MT-G2 until I started reading these “over my head” discussions.
This is strange to me. Are you still feeding the PWM in through the chip enable pin rather than off time selection pin? (3/EN vs 5/TOFF)
After glancing at the QX9920 datasheet again, I guess the enable pin is being used either because:
a. using the TOFF pin is annoying because it requires extra components
b. they are abusing an undocumented soft-start function?
c. It’s possible I am misunderstanding the TOFF pin’s purpose. I don’t really understand the application the datasheet shows.
Wight, I'll admit that I don't understand what's going on with it but it seems to work. I am feeding through the EN. I didn't even know that we had a datasheet on this thing!
It probably got buried in the thread somewhere (#104), but I’m pretty confident we are looking at the QX9920 controller. http://www.yuketek.com/pdf/QX9920.pdf
Actually while getting ready to link you to it I re-read the introduction in the datasheet (through an online translation) and I see that they clearly specify using the EN pin for PWM. Whoops! I’m glad I already suspected myself, softens the blow ;-). Anyway I really want to know about that TOFF pin in this case. Maybe that pin is useless, or maybe tweaking that pin could soften the spikes. Currently I really fail to understand it’s purpose.
Will you please post screenshots from your scope showing the output from this driver? That would be really useful to me. Especially a detail shot of the spike. Thanks in advance.
Ouchyfoot wrote:I'm building a light and plan to use this driver (7A) stock with an MT-G2. Do you forsee any problems?
I would go with what ever RMM advises, but I can say that I've been driving one at about 11.5 amps sporadically with this driver using 3S King Kongs with no issues. I say sporadically because I don't use High mode much. For all I know, I may be slowly beating the emitter to death, but it hasn't shown any signs of distress yet.
The voltage spike should be a poof situation, either it poofs or it doesn’t. So I think you should be good to go.
I don't have any screenshots but when I'm testing the new inductors I will take some. All I know is that 0.9V+ output ripple is A TON! The issue we run into with buck drivers is that there really isn't very much space inside of a flashlight to fit big components. The only way that I know of to downsize components is to switch at a higher frequency which allows for smaller capacitors and inductors, otherwise if we want high current we've got to have large components.
Sounds good. You’re correct about switching at a higher freq. This is also undesirable due to increased switching losses.
I still suspect that high current is doable with a small inductor. IIRC in my LD-29 thread I said I measured the freq the buck circuit was operating at as around 400-500kHz (I posted a more specific number over there). That driver puts out 3A in stock form. It gets >80% efficiency and the output is absolutely smooth. HKJ’s review shows almost no ripple, I saw the same thing in my own testing. If we can generate a smooth 3-4A output at that freq with small 7.4 x 6.6 x 3.0mm SMD inductor it seems that it should be possible (maybe not with the QX9920) to generate two or three times as much current with similar ripple using an inductor which is two or three times as large.
RMM,
Since you have an oscilloscope, you may be interested in this idea. When I first worked on this driver, I mistakenly replaced the large diode with a 70N02 FET. I was driving a MT-G2 around 11.5 amps and the driver behaved nicely and did not generate heat on that area of the board. The heat generated only in the area around the voltage sense resister bank. Wight noticed the error I had made. When I replaced the FET, the driver became considerable hotter due to the large diode generating heat too.
I have no idea what was happening in the driver with the FET for a diode. Maybe it just blocked current flow through the inductor. I've been meaning to look at the traces to consider what was happening. The big tab pad for the diode connects to the toroidal inductor which is connected to the positive LED pad. The Ground and Gate pads share a pad that connects to the big tab pad of the FET pad. I have know idea how an N-channel would behave with positive power applied to a FET's Ground and Gate pins and Negative to the large tab. That is something I should check with a DMM.
Who Who knows what it was doing to the emitter as that the MT-G2 is one tough hombre. That is something you have the equipment to measure should it sound like something worth exploring.
That’s something I might try with my Y3. RMM said he put two 25v 22uF capacitors on the output to absorb the spike when you turn it on.
And then he said that it didn’t work. Keep reading.
Couldn’t hurt though… It seems it was a minor success anyway. Maybe a bigger capacity cap. Or mess with a resistor to let it fill slower. Like 1-2 secs till it reaches voltage. I used to do a lot of that with relays, to add a delay to relays. Also had 1 farad cap storage (huge bank of caps back in the 90s) to keep the draw from the alternator consistent. This stuff is all coming back to me. But the big cap remember really helped smooth things out in the voltage department and kept the voltage stable from the regulator. It might work if it was just a bigger cap- I’m not sure if they make ceramic caps that big though, or if there’s enough room. I might try it though, like with a regular japanese cap to test and see if the spike is there on the lower end before and after.
I mean it can’t hurt to put it in there. The cap would make it kind of cool too cause the light would fade out