Haikelight MT03 II modding (driver)

Hey all,

Not long ago i received my MT03 II (as in mark 2) “Devourer”, which can do around 9000 Lumen out of the box.
It’s a neat looking thing and obviously very bright… :smiley:

Of course i was curious about the insides, and i also think it should be able to 12000 Lumen or more.
I already did the mandatory spring bypasses, but i need to do the driver too, and the heat management could use some attention too.
My questions mainly concern the driver.

Here’s a couple of pics with my comments:

This is what i found after disassembling the head.
Not enough cooling paste or maybe not enough pressure.
But that’s fairly easy to solve.
But what disappointed me is the thickness (or should i say thinness) of the shelf the MCPCB is on.
It’s only around 1.5mm thick, the MCPCB is also around 1.5 mm thick.
The MCPCB is held in place by 3 tiny screws in loose threads, so they can’t press the parts together properly
There’s also a big screw in the centre, sandwiching the MCPCB between shelf and reflector.
Not yet completely sure how, but i’m planning on adding a copper ring between MCPCB and reflector to incorporate the reflector in the thermal path.

Now the DRIVER.

Those coils are really tiny… :frowning:
Very few through holes for the minus to the opposite side.

There are basically 3 step down circuits, one for each XHP70.
They use 2x R100 (0.1 Ohm) in parallel, resulting in around 3000 Lumen per LED.
So it’s a 0.05 Ohm shunt / sense resistance for each circuit.
Adding an extra R100 to it would make that 0.0333… Ohm shunt.
And that’s the plan, basically.
But maybe i should settle for 0.04 Ohm ?

But those mini coils kind of bugged me, so:

I desoldered them and sourced some fat ones from a class-D audio amplifier PCB i had laying around.

I’m not experienced with modifying this kind of drivers, so i have a couple of questions:

Am i on the right track with this, to have it perform better, or would it work just fine with those tiny coils?
Where do i find R100 resistors in that size, or should i use larger ones? (that will get less hot)?
Anything else i may have forgotten to think about?

after seeing this light in real life action, it was already wow!

Thermal paste does not seem too little, just not distributed enough

I don’t know the answers to your questions, but I’m looking forward to your results and what the experts have to say.

What is it stock, 4 or 5A per XHP70 or so?

@ Miller:

Yeah, that was at your place in France recently. :slight_smile:
But what i eventually want to do is put XHP70.2 emitters in it.
They will have lower Vf so i can push some more current and still have full blast with partially discharged cells (because it’s a step down driver).

@ EasyB:

Not sure, but i guesstimate it’s 4 Amperes per emitter now.

Nice thread. Good info, pics and questions. I have very limited input to provide:

  • Seems the MCPCB or the shelf or both are not flat. That much thermal paste should be enough with excess squeezing out the sides.
  • Don't know if the bigger inductors will help, but seems like a good idea. One would assume they have thicker wire which should in theory flow more current and create less heat, but I'm just guessing.
  • Bigger sized resistors are fine. The can handle more current.
  • Regarding .04 verses .033, I like to reduce resistance to just before the lowest mode starts increasing current and before current maxes in any mode lower than turbo/highest mode.

Looking forward to hearing how your mods go.

O.k. , you covered most of the "bottle neck" of the driver.., the original coils were , for sure , at the upper edge of their capabilities..! I used some like these , trying to minimalize the height of some drivers and at 3 Amps I could feel the smell of burning , coming out of them !To be sure that what you want to use are better just make a comparative measurement of resistance , for sure the bigger diameter of wire will give you lower resistance .Temperatures above 80 decrees celsius will decrease , in time , the self inductance of the material , so in tests , keep an eye at the temperature of the coils !

Because this inductance are storing the energy during the time "off" of the mosfet and give it back in the circuit during "on" , I should suggest you to DIY by using some bigger thoroids ( as bigger as the space inside will give you) , and copper wire , 0,7-1.0 mm , as many turns you can fit . The core material should be " 52 " , is very good and stable in time (the cover color is a light green + blue on a side ).

About the resistance , at the actual value (0.050 ) the power dissipated right now is about 1W , so , the actual size should be enough . Stacking some more in order to lower the resistance (0.033 as you assumed...) will result in increasing the dissipated power of the resulting resistance.., so is better !

I'm not sure that I was too clear in what I said , but is late and I'm very, very tired...sorry...

Ok…I’m learning from this…does each set of R100 resistors control each led? Two R100s per led?

I compared the coils just now.

From here both seem to be pretty flat.
I’m planning on drilling out the tiny screw holes and put M3 screw and nuts in place, and the centre screw will also clamp the lot down when i add a copper ring.
And both shelf and MCPCB are only about 1.5 mm thick, so it will flex enough where needed to make good thermal contact.

I just ran 0.373 Amperes DC through both, and the big one has 7.9 mV Voltage over it, the tiny one has 35.6 mV Voltage over it.
R = U : I
Tiny one 0.0356 V : 0.373 A = 0.0954 Ohm
Fat one 0.0079 V : 0.373 A = 0.0212 Ohm
So with, say, 5 Amperes the Voltage drop would be U = I × R
Tiny one 5.00 A × 0.0954 Ohm = 0.477 Volts
Fat one 5.00 A × 0.0212 Ohm = 0.106 Volts
So then the heat produced in Watts (Power) for each coil would be P = U × I
Tiny one 0.477 V × 5.00 A = 2.39 Watts
Fat one 0.106 V × 5.00 A = 0.53 Watts
So yeah, in all it seems quite worthwhile to me to go for the fat coils. :slight_smile:

I’m inclined to go for larger ones too, but that won’t fit on the PCB, so i would have to put them upright or something…
But the same goes for the coils. The will have to stand on their sides and i’ll have to use some tiny strips of copper to connect them.

…i was assuming / hoping the other modes would just increase by the same percentage…

Me too ! :wink:

That’s right.
The 3 driver circuits seem to be similar to the Convoy L6 driver, which also uses a 10µH coil but it has 2× R080 (0.08 Ohm) in parallel (is 0.04 Ohm) as shunts / sense resistors, and that’s a 5 Ampere driver i.i.r.c.
I understand you can safely boost them with an additional R150 in parallel, making it around 0.033 Ohm.
(didn’t calculate that yet, fussy calculation…)

For calculate parallel resistance use this link : PARALLEL RESISTOR CALCULATOR

Easy and precise !

And do not forget to change the big capacitor (22 microFarrads/50 V) with a smd one , any value between 10-100 microF at 25-50V , more reliable and significant smaller. The value is not critical , is just a decoupeling capacitor on the power lines.

Anyway , congratulations for this piece of jewelry , I would like one ..,but is not sold in my country and I can not afford it .... I am curious about the final solutions for this mode , and I hope you'll keep us posted , right?

Way to go Jerommel,

I have two of the new perfect MT03’s also that Haike sent me as replacements for the bad ones from that GB and those first ones are what I’d love to get fixed myself, but my knowledge tells me that all I know about these drivers is that there is a board with some gadgets on them and they aren’t working proper so something is messed up and I’m actually fairly certain about that much anyhow,

I really would like to get them working safely then give them to someone as a present for someone who can’t normally afford a light like these.

Today actually I was looking around Texas Lumens new open for business online shop and thinking about what he might charge to get these first HaikeLites working properly,

So I just might send him a message tomorrow and ask about what it might cost to fix them up real good.
I sure wish I had a little knowledge about these electronic parts but a little bit is about all I’m probably ever going to know about these drivers but sure do admire your knowledge Jerommel about it all so please keep at it.

Thanks for sharing your project Jerommel… :slight_smile:

Thanks :THUMBS-UP: very helpful, bookmarked.
Now i see 0.08 + 0.08 + 0.15 in parallel makes 0.031, which is a little lower than 3 × 0.1 in parallel.

I can’t really afford it either… :FACEPALM: …but there was a coupon code floating around at the time, so i could buy it for $80 at Gearbest and i couldn’t resist. :LOVE:

Yes i will.

But how about the FETs they use?
It’s the “XORB 48” next to the diodes marked as Q1 to Q6 on the PCB.
They use 2 of them for each LED, but they’re very tiny.
From what i could find about them (thus far) is 5.8 Amperes max. continuous current, and i think they’re in parallel so that would mean 11.6 Amperes.
But the FX-30 driver (convoy L6) uses (as far as i can tell) 4 big power switching transistors, i assume they’re FETs too.
The FX-30 uses a big fat diode too, so maybe i should put extra diodes in parallel?

I like where this mod is going ! But this baby worth everything..! So , let the dogs out..

The new coils you have are clearly better , but you have to see how they gone handle the 5-6 Amps that you are aiming to, right? I'm still thinking that the ones hand made will do better , especially that they will give you the possibility to solder its in every position that suits the other changes you want to do ( I have in mind the driver with 3 coils from SRK...) but this is up to you .

The fly back diodes are only 5A so , some Schottky diodes , of 10A or so , will be fine .The problem is that the size will be greater and not fit in the space.., but you have the solution of stacking 3 more SS54 on the existing ones , and here comes an other problem , that you'll have to pair them , in order to have the same voltage drop across , at different currents .Is about the same thing that you've done with the coils , a graphic at 1A , 2A , 3A ...and so on , with the same dropping voltage across the diodes . That means that you'll have to take the ss54 out of the board and match them with other 3 diodes .This has to be done in order to spread the current evenly, between the 2 diodes in parallel.

About the FET , you're right , they are in parallel . The XORB 48 is equivalent with 3400L ,link here: mosfet 3400L 30V 5.8A N-Channel SOT-23-3 Suitable for Use as a Load Switch/in PWM, View mosfet 3400L semiconductor, OGFD Product Details from Shenzhen Goford Electronics Co., Ltd. on Alibaba.com

As you can see , Vdss=30V , Rds on=31mohm@4.5V , 45 mohm@2.5V , 28 mohm@10V , Ids=5.8A .If the data sheet is accurate , the power dissipation on these FETs should be around 0.3W (6 Amps X 0.045 ohm) , that makes them suitable for the driven currents , BUT, not only the Rds on , is the only one that contribute to the power loss , the shape of the pulse on gate is important , the time of transition between the state off-on is also very important. Here , at the transition are the biggest loss of power , dissipated as heat .The FETs are considered fast switched devices , but if the wave form that drives them are not square enough , there is the bottle neck...

There are a lot of very good transistors , with a extremely low Rds on resistance ( ex: TPCA 8028-H , with Rds on of only 2 miliohms !) but the problem is the size or the shape of its , that makes the replacement very difficult. I'm not saying that it is impossible , as far as I saw in your posts , you CAN do it , but , because of the limited space ,hard to do . Maybe , after soldering better FETs , will be necessary to pot the driver , in order to obtain the mechanical shock resistance ...

Now , I've notice that instead of making your work easier , I opened some new paths...

Anyway , the idea is to put the driver on the bench , and test it after each modification made , in order to see the improvements.., so , in order :

New coils , test , double diodes , test , new FETs , test , resistor mod , test.. Also , it is possible that the UI modes to change really hard , after modding...

I hope I didn't harm too much your choices , you are the only one that decide what to do. We can help.. if this could be named "help"... Cheers !

Thanks for the “help” :wink: Adrian.
Yes… There’s a lot to think about…

The potential UI problems are a little discouraging…
What to do when the UI doesn’t work good anymore?

Actually i don’t have enough knowledge for this… :TIRED:

Is it the LEDA1624 that makes the square wave?

With the 2S2P light, I'd prefer a FET based driver replacement. Custom modes, thermal stepdown, LVP - all should work. Maximize the power and flexibility, use a SIR404DP FET, good to go.

For me of course, I'd install Narsil of course with smooth ramping. I want to mod Narsil for supporting only one output channel - the FET, so don't have the 7135 problems with high amps, high voltage. Low and mid range modes aren't the most efficient, but they will work fine.

You are welcome , anytime !

The problem of loosing the UI should be the least of the problem , and easy to fix ... Lets go to the basic of the driver , o.k ?

I was looking to yours picture , more closely . LEDA 1624 seems to be a simplified version of a N-chanell FET Switching Regulator Controller , that is getting the PWM from the U2 circuit (with a flashed modes , factory set ). This one should have a duty-cycle according with the UI modes.

The duty-cycle refers to the total amount of time a pulse is "ON" , over the duration of a cycle. Longer time on , bigger the current on load. Logic , is it ?

This is a simplified loop : Duty-cycle generator , Switching Regulator Controller, FET ,Load (LED). Inside this loop you have feed backs that are comparing the "in" and "out" , in order to have a balanced system. These are not what we are interested in , so let ignore them , for the moment...

So , LEDA is driving the FETs and provides the feed backs from sense resistor , in order to "hold" the current within the limit imposed by the duty-cycle.

When you mod a driver by decreasing the sense resistor , the biggest increase of the current thru led should be in the high and turbo mode , where the feed backs capabilities will NOT be exceeded ,BUT in the lower modes should be very close to the original modes.

Increased currents in lower modes means that the inductors are going in the saturation zone , that should be avoided (massive heat , loss of magnetic permeability...). Because of this , I was pleading for strong inductors , hand made...

Do not discourage , You are good , I'd seen this , just be patient , make a plan of what you have to do , and do not expect miracles.., every hundred of lumens counts , is better than before , we are the ones that knows better...!!!!

See you around , we are here for you , I'm sure that more guys will step in , with better solutions than mine.., you'll see ...

That's why , when you mod a driver , this has to be done in small steps (that is not hard to do when you stack resistors in parallel !) . If you loose the UI , it means that you lost the feed backs , and you have to increase the sense resistor to the value where the feed backs are capable to hold the system stable ...

So , do not worry about lost UI , could be fixed , within decent limits .., you can not ask to a driver designed for 100W to deliver 200 W !!! Unless you change the power side , accordingly with what you want .The small signal side as able to go up to 600W , or more...

@ Tom E:

Do you mean a linear driver, constant current circuit with a FET?

I like the idea of a buck driver actually, and i’m not gonna use the light in high modes all the time, so in my view a pimped triple buck driver is the way to go.
And with the XHP70.2 in mind even more so.

I’m trying to learn more about these switched mode power supplies, so I find this very interesting. Could you explain this part more? I’m not clear on how the driver selects/controls the lower modes and why it would cause the inductor to go in saturation mode. I guess some drivers just PWM the highest mode. But I presume they could also just lower the FET duty cycle to control lower modes with constant current.

That's my man ! The buck driver is the way to go.., being a 2S2P configuration it means 8.4 volts . An XHP is 6Volt , so the efficiency of the buck driver will be more than 90% (as you know ,closer the voltage supply to led voltage , higher the efficiency !)

BTW , I've seen on the driver's board a NTC (bottom right ).It has something to do with the temperature , but with witch one ? Could be the general temperature of the light , including the leds , or the temperature of the board ? This light is timed step down or the output is self adjusting with the temperature ? Jerommel , what is the behavior of stock one ?

If it is a self adjusting, this is good , is gone work like a protection , not to worry about over heating !...

I didn’t notice any PWM on the lower modes (tested with rain drops), so it probably reduces the duty cycle like a proper buck driver should.