ASTROLUX MF01 Mini - common issues thread

So does the duty cycle of a single 7135 chip have any effect on the heat it puts out?

I was thinking the FET didn’t generate much heat, but we’ve seen some cheap ones burn up with high amperage running 100%. So would the heat it puts out be dependent on duty cycle?

In the case of the 7135’s the duty cycle will increase the heat simply because you are also increasing the current / power.

With the FET’s, a good one should have basically no heat at normal power levels (yes, it technically does produce some heat as all FET’s have some resistance but just not enough to really matter much for anyone asking a question like this). On hot rods where we are driving 10-20A, then the FET can produce some heat if it is high resistance and even die. The only cases that come to mind of the FET dying were at currents of ~20A.

Although this is more a case of simply pushing the FET past it’s design specs then it getting too hot directly.

The duty on the FET will increase the heat since it also increases the total power / current but technically it will actually get a bit more efficient when you reach turbo 100% duty as you no longer have the switching losses. Once again though this is a case of if you have to ask, then you most likely don’t need to worry about this as it will only matter in fringe cases.

Quick question, what exactly is a linear-Fet driver? Is that a thing?

I don’t know it exactly as well but I found this:

High efficiency LED driver. All modes (except moon) are implemented without PWM , linearly, using a current source based on a MOSFET transistor, which in practice means 1.5-2x longer work time at low brightness levels compared to classic PWM drivers keying the AMC7135 current sources or transistor.

The controller has a turbo mode (current 8A), about 3200lm for 4x XP-L, implemented by a MOSFET transistor, with hardware stabilization of the upper current value.

Parameters:
• linear5A current source implemented on MOSFET RDSon 0.95 mΩ transistor controlled from ATtiny85V with feedback
• modes (brightness levels) implemented without PWM, with maximum LED efficiency
• diameter 17mm - flashlight standard
• supply voltage range 2.8-4.5V
• power supply: 1x 18650 Li-ion cell, preferably without protection
• maximum LED current 5A, in turbo mode 8A (limited by the capacity and battery charge level)
• full current stabilization at Vf LED + 0.1V supply voltage, with the latest brand cells to be used 80-85% Li-ion capacity
• no PWM , flicker, noiseless operation
• total efficiency 2.8-4.5V power range is over 90%

and

Thanks for that. It’s all much clearer to me now. :beer:

As posted above it is the same style that LED4Power uses and the Texas Commander was designed around. Basically it uses the FET to do what the 7135’s do.

The issue is that the FET is not really designed for this so it is much harder to deal with the heat it produces since it does not have a dedicated thermal pad like the 7135.

Thats why I suggested putting the FET on the mcpcb to give it a better heat path, which ended up working out quite well from what I have seen from LED4Power.

circling back around, i’ve found on my full copper version that i can get it to stay at turbo until i can barely hold the light by setting the room temp at 14 and the throttle at 60

so it’ll definitely do it at least in ramped mode

So you say that in copper version there isn’t the step-down problem in ramping mode. What emiter type is you unit. I think i saw someone else saying that the copper head or all copper doesn’t do that but i am not sure, can anyone confirm if those versions are free of the bug present in the coloured ones.

I never quite fully understood the bug to begin with, but my full copper MF01S mini operates approximately identically to my MF01S; it ramps down once hot, but at an imperceptible speed, and seems appropriate. Only after a minute and being very hot will you notice that it must have throttled; a double click back to turbo will confirm it.

It will get so hot you can’t hold it. It does appear to get brighter once it’s cooled off a bit. This on a black and red Shockli 5500 26650

I did reset the resting temperature (found it way too high - like 33deg C at room temp) and set the throttle to 60

I don’t think there ever was a bug but of course this is merely conjecture as I don’t have any other mini other than this one full copper.

Last friday I’ve modded my MF01 mini, as mentioned in post #5, the result is very satisfying, no throtteling down any more in any ramp level (also top ramp) till the host (body) of the light reaches the setted max. temperatur. On max ramp 130 (20 clicks from turbo) it takes about 3 Minutes till it reach the max temp (60°C) and begin slowly to step down.
A level around 1200lm can be kept constantly (@amb t20°C), this makes the light very useble now.

Background of the mod idea:
After reading a comment form Lexel on TLF regarding the heating of the AMCs when not in turbo mode and the bad thermal contact of the driver pcb to the host, it was clear to me why there is such a throttling down after few seconds from max ramp.
So, if the light is operating not in full FET mode (turbo), the AMCs becomes hot very quickly, this heats up the whole driver pcb where the mcu with the temp sensor is located on the backside. Therefore the quick step down in max ramp, when the host is not even at 45°C, the driver pcb with the mcu is already at 60°C.

I replaced this plastic cover:

by one out of aluminum that looks like this:

Here is a link to a drawing, in case someone is intrested: https://imgur.com/jHjj1Xz

and contacted the green marked surfaces with thermal paste:

The Design was very accurately defined, so after fixing this heat sinking cover plate, the gap to the AMCs is 0,1 mm, there is no mechanically contact to the parts of the driver board except the big outer GND-Surface.

Afterwards it looked like this, additionally I used a thin plastic washer for isolating the aluminum cover to the battery plus pole and used a magnet:

After this mod, the AMCs and the GND surface of the driver pcb is very well contacted to transfer the heat to the flashlight body. The Battery tube is now also directly connected to this heat sinking cover plate if it’s screwed together with the head, this was not the case with the plastic cover.

This mod would be even better if the heats sinking cover plate where out of copper and CNC-machined precisely like this. Only the costs are relatively high for a single production.

Here is a link to a drawing, in case someone is intrested: https://imgur.com/dyBRHRq

I hope this mod info is interesting for some and that I could explain it understandably.

Okay, now I see. There is a plastic cover that prevents the battery tube end from pressing down on the driver (like I was thinking).

So does the battery tube screw down into those two screws?

What is the function of this plastic cover?

Not quite, it remains a small distance of about 0.5 mm. The screws are sunk in the plate a bit, and the upper surface of the plastic cap has still a gap of 0,3 mm to the the battery tube end. In my heatsink plate design I closed this gap, so there is a proper direct contact from tube end to the plate on the whole circumference of the tube.

It protects the driver pcb from direct contact with the battery, fingers, dust or other objects. It holds the screws for fixing the pcb. That’s it.

Nice mod, that should be made standard in the next run of lights honestly.

I bet you could sell some of those to other BLF members as well.

My earlier post.

So there is nothing metal that comes into contact with that copper trace the red arrows are pointing to?

I guess there is only a certain percent of vias to carry the heat to the other side of the driver where it touches the metal. Now everything makes sense.

I think a quick fix could be to just add a metal ring, aluminum, copper, etc…, just to connect the edge of the driver to the battery tube. Give the heat some mass to soak into so the MCU doesn’t get too hot, too quick. Touching the top of the 7135 chips may not be necessary.

Yes, there will be definitely a positive effect. Was also one of my first thoughts. I tested it with one aluminum ring without the two screws and without thermal paste. The ring was 3.7 mm high, so that there was proper contact to the battery tube. The results where better but I was not statisfied yet, step down on max ramp was around 1 minute. Ich thought there is not enough mass for the heat and designed the described plate.
Sure, a copper ring (Outer-Ø 29,8mm, Inner-Ø23,5mm, height 3,7 - 4,0 mm) will show better results with some thermal paste. But I would recommend to add two bores for the two screws to fix the driver and the ring.
If there is interest, I could provide a design proposal with a drawing for such a ring, for somebody who want to try it.

Unfortunately, I have no possibility of producing such parts, I’m depending on others who have such toolings.
I can provide the 3D-CAD data and drawings, maybe somebody is able to manufacture a small batch for a reasonable price.

I’m trying to think of how the average Joe could fix their lights performance quick and cheap.

Maybe ditch the plastic piece and get some better screws that can transfer heat. Copper screws? That may not do much.

Maybe make a flat C shaped copper bar from some copper sheeting and thermal epoxy it on top of the 7135 chips?

Maybe cut a section of iron pipe that is 4mm thick and drop into place where the plastic and screws go. Then the battery tube will tighten against it?

None of these ideas are all that great, but they can be done without a cnc or milling machine. That’s what I’m trying to avoid.

Ah, the only good fix might be a machined piece, but I doubt they would be cheap to produce.

It would be nice if we could just turn off the thermal stepdown and use our hand to adjust the brightness if it gets too hot.

I think I can add few toughs to this thread.

  1. AMC chips are not as rigid as everyone used to think. They start dimm or turn off after overheat. Some of them a better than others, not sure if any brand ones are still available. Those that can be found from ali are the worth.
    (Most of you know - Im not a fan of 10-sec output tests, and one of the reasons - they never show real usage issues. You wont recognize that some of AMCs have turned off with 10-sec hotrod. Use power supply and 1-hour test to see this).
  2. They can transfer heat to the central pad and even 1V difference between cell voltage and Vf will cause 0.35W heat that does not seem too much for this way. Yes then they heat up drivers outer ring.
  3. Next thermal way is too hard for analyzing. Copper coated through holes on the outer edge are possibly to small to transfer heat to another pcb side. I would guess chips №5 and №6 are heating mcu from the backside.
    Once again, see #1 - AMC chip can lower output by themself, without program step-down.
    We are not able to set different temperature range for components mounted on one pcb inside enclosed aluminium case. With different mode sequence you will get different parts and components overheated (i.e. one part could heat up other after one situation and vice versa in different situation).
    To make this thermal exchange more stable (dont forget that most components have properties that change with temperature raise), we need to limit temperature delta. Inside one pcb or whole host. Using thicker copper layer, alu core based pcb, thermal pads or grease, extra parts to transfer heat to the host. Using any sort of thermal materials between pcb shelf and driver components may looks strange - in first suppose led will heat up driver and not vice versa. But in real life, you wont be able to cover any component with “thermal shield” and leave it colder than whole host - this is not possible (unless you are making 10-sec test). And temperatures mentioned in this thread can rarely cause any issues - usually there is one or several true overheated (over 85 C) components on driver board.
    Now few questions for MF01 mini users:
  4. Have you tried using thermal pads between MCU and shelf?
  5. Have you tried more straight solutions? Simple ring that faces driver outer ring with one side and tube with other side? (Not sure this is possible, probably they should be electrically isolated).
1 Thank

Kiriba assumption is correct, problem are badly placed (no 5 and 6,but possible only no 5) AMCs directly below MCU on the other PCB side. And if there are any GND vias under MCU which connect top and bottom GND area, it's a recipe for - measuring AMC chip temperature instead of driver board temperature. 0.35W of heat dissipated in single AMC is more than enough to cause false MCU OTP triggering if that AMC is in thermal contact with MCU. While AMCs are soldered to GND ring, that heat path is not as nearly as good as it intuitively looks - 35um thick copper trace has very limited "reach" after which it doesn't conduct heat better than FR4 core itself. So it's a basically PCB design flaw, but even with AMCs far away from MCU, they are still on same PCB, so problem would be not completely solved, but definitely reduced.

Removing AMC no5 and/or 6 would show if all this is correct (I don't own this light,but I suppose staciking them onto other AMCs is not possible due to height).

All drivers (linear,switching) generate heat, I noticed similar issue back in 2014 with now ancient LD-1 driver, mid. mode would trigger OTP while light is still cold, but high mode was fine.

That's why decided not to use internal MCU temp. sensor and use external NTC sensor on LED MCPCB, which is the simplest and best way for accurate light temperature measuring - all my MCPCBs have place for NTC and in case of 1S lights it needs only one additional 30AWG wire, I expected BLF drivers would switch to this logical step, but for some reason till this date no BLF drivers have NTC support.