I started this thread as a kind of discussion but no one answered…very sad
Now I have spend some time in learning all these electronic layout stuff and ordered smd parts and the PCB.
——————
I just assembled my first advanced prototype of a simple direct drive driver. Man these smd parts are fiddly, I guess I have inhaled some resistors and FETs while hand soldering the board. The FETs are 2.5x3mm big…
I used the nanjg style, tiny13A controller drives two sot23 FETs.
One problem which stole me hours:
It made whatever it wanted but not what it should |(
was solved by adding a bigger capacitor between GND and VCC
I just ordered some 10uF 0805 capacitors to replace the bulky one.
Another problem was the soldering mask, I made a mistake with the restricted area so the outer GNDring is covered, no big problems as it can be easily grinded away with a dremel in less than half a minute.
——-
The FETs I used are rated for 4A and have low onresistance below 80mOhm max. On a 18650 I get over 5A and on a normal Chinese 14500 I get 3A max. I have no imrs yet.
Looks very promising.
By the way I just killed one XML on a heatsink with a power supply and this driver while letting it run several time on 5.3A… Maybe I should have used a copper sinkpad and AS instead of a thick layer Fujik….stupid former me
hi Werner, it happened a few times to me lately that I started a thread and it just gets burried in no time, no matter how interesting it may be (well, at least I think it is interesting, but it may be just boring...), there's so many people starting threads every day that half a day not following the forum can make me miss very nice stuff.
I am not into electronics but I will be following your attempt at making a nice 15mm driver, there's so many nice hosts that only take 15mm, and so few good 15mm drivers around, and only 1 or two that are dedicated li-ion drivers.
Yes isn’t it amazing how heavy this small babies are :bigsmile:
I use irlm2502 n Channel mosfets Rate for 4.2Amax and 3A if they are hot.
So capable of 6A with ease, I Used two to reduce the on Résistance and because this was my first board so I planned in some flexibility. I got them cheap from china, the good thing on these FETs is they all have the same footprint so you can use whatever you can source locally with the values u need.
I did no destroying test of a single FET, yet but I guess I will do. What I can confirm that one FET can handle 3A for 10 minutes. No further long time tests made yet. I also didn’t try to stack them but I guess it could be done too.
I twice tried to build a direct drive driver by hand but it ended in a disaster, because everything has to be so small. From now on I just order a PCB from china, seed studio PCB fusion charges 18$ for 10 boards and 15 arrived. Takes over a month but saves me so much hassle with soldering. Layouting and reading manuals in a warm room on my laptop is way better than frustrating soldering championships in the cellar.
Soldering a soic 8 chip by hand to wires is thousand times harder than soldering it on a perfect footprint with solder mask.
I get 4.3A with a ncrA and over 5A with a lgD1…but I am not sure if I have duty cycle 100% because my soic8 clip is on strike and in my failure finding process I played a bit with the code…
And I use 5cm long jumper wires for a connection to the DMM and Battery, so it may be even a bit more with fresh cells and perfect conditions. In real life flashlight application the switch will steal some voltage too so I would be fine with everything over 2A through the LEdD rom one of my Keeppower14500.
I hope the smd caps get here fast, but it will take at least two days(one day posting next day delivery).
If it works with them I will put this driver in the golden two dollar zoomie to give it the power it’s deservess.
Sometimes i feel like replying to a post but stop myself because i figure would the OP want a reply along the lines of this is interesting but i have nothing to add
I can’t agree more with this, i don’t check every new thread anymore, too many to follow, and i am amazed that some threads that i hope will provoke lots of thoughts end up almost empty
Nice work Werner and l apologise for not commenting as I do read these posts. Its just that I dont have any idea whats going on and would not have any technical input to whats going on so I just read and shutup.
You might want to try OSHPark for your PCB fabrication as well. They only offer one thickness (1.6mm) but their boards are nicer quality, and you can see how they will turn out before you order. They're also very cheap
So the IRLM2502 responds well to PWM? Very nice.
One thing I don't under stand is why you need such large capacitors between VIN and GND. All that capacitor should be doing is protecting the ATTINY13 from variations in the power supply. All my own circuits use this same setup and have no issues with a 0.1uF capacitor. Why do you think you need 10uF? Genuine question as I can't provide any answers lol.
I am wondering too, maybe some of the more educated electronic gurus has a hypotheses…
What I haven’t tried yet is to power the small cap circuit from a battery, maybe this issue only occurs because my experimental power supply can’t provide enough juice in a short time.
Or they send me the wrong value smd capacitor, you can’t tell which value they have without the whole reel. That’s why I ordered a used reel of the 10uF from eBay
——-
Yes the oshpark UI is much nicer you can directly upload the eagle file and see what it will be. But if I remember correctly they were a bit more expensive from the money/pieces rating.
On the other hand it was not so much trouble to export the gerber files, if I had watched on them with open eyes I would have seen that everything is coated with soldermask…
I have chosen a very thin board, and the quality is superb. I soldered and desoldered on a board the whole day and not even one pad got torn. I have destroyed nanjg pcbs with much less abuse.
If you can, add a diode between battery + and the uC/decoupling cap. This should make it run reliably without the big cap.
Also, be aware that the big cap has to handle the ripple current of a square wave at potentially several amps - unless it’s a low ESR cap designed for the application (robbed from a SMPS or something) it might pop. Buy some MLCC ceramics - you can get 47 and 100uF sized caps in 1206/1210 packages, X5R dielectric, rated for 6.3V. These can take tons of ripple current and they’re small.
Just one stupid question to see if I got this right: the only loss in the driver comes from the resistance of the FETs, so it should be fairly efficient, and should not get too warm even under full load, right?
I definitely want some of these when you’re ready!
@gmarsh
I planned to use a diode even have some laying here, but 15mm are very tight so I thaught the polarity protection isn’t necessary because I am smart enough to use a battery in the correct direction.
Do you think that there might be some negative ripples from wire inductance or what do you have in mind?
Yep the only loss is in the switching phase, which is so fast that it is very little, and the power drop over the FET. And of course the standard warming of the and path resistance. Didn’t got noticeable hot until now…
It should be very near direct drive, but with modes. At least this is my plan
The diode isn’t for reverse battery protection, it’s to keep VCC stable.
When your FETs switch on, the battery voltage drops to the LED forward voltage, and VCC suddenly changes with a very fast dv/dt. Doing this to a CMOS uC can cause all sorts of weirdness, especially to analog bits in the uC like the oscillator, POR, brownout, ADC, etc. With the diode in place, the uC runs off the decoupling cap during dropout.
You can get diodes in small packages, I use SOD-323 as my general small diode, SOD723 is smaller but still hand solderable. Either will be smaller than your electrolytic cap
Also, you’ll have conductive loss in your FETs in addition to switching loss.
Say you’ve got 5A in your LED, 2.5A per FET. (2.5^2) * 0.045 = 280mW in each FET. And this is at 25C - Rds(on) goes up with temperature (see figure 4 in the datasheet) so you could be pushing ~400mW of dissipation in each FET. Almost all of the heat in these FETs comes out of the drain pin, and you don’t have a large thermal area to get rid of it.
Are we talking about that normalized chart?
70°C is below 1.25….
So 280mW*1.25=350mW
I also used no R4 so that the switching is fast and switching losses are small, not sure if this is wrong…?
Anyway compared to wire and switch resistance the FET onresistance is very small.
I have sod323 shottky diodes laying here for use in bigger drivers. The space is really small even for these. I had them in the layout but removed it.
Do you think that without diode the capacitor is discharged through the led?
This stuff is hard to imagine for me, with all those time relevant things I can’t tell which state is on which time…if the FET lets some current flowing for a small time or if the FET is closed immediately so no discharge through the led.
It’s tiny, and available on ebay, cheap, that’s good. Current though, pretty small.
Although buying parts from ebay/china means buying worse than what the specs of the original say.
IRLML2502, even smaller than 7135… and not compatible pin wise, not cool.
Making your own boards?
What do you measure on? 5A on? An LED? Short circuit? Resistor? What battery as power source?
For how long? 10s? 10min?
Yeah the capacitors are basically useless if you don’t have them away from the LED.
Nanjg105c has a diode as well so the capacitor filtering power for the uC does not get discharged by the LED.
Why a pull down resistor? Not enough to have it on the uC only? Nanjg105c does not use a pull down for the 7135s. I’m wondering if I need it or not now haha, hopefully not but it’s easy to add anyway.
I don’t see a space for R, C, D, for a mod so you can detect temporary and long term off state and to avoid having to idiotically use double clicks to switch a mode when a flashlight is on for a while. I guess you plan to use the timeout after start that will remember the last mode but will operate as classic ATtiny13 that needs double clicks because it has no idea if it was off for 0.1s or 10min.
Nice work, keep it going, yes not many people here fiddle with drivers unfortunately.
Advice for next time, just copy the Nanjg105c schema 0:)
So you’re saying it’s direct drive in Hi but also has lower modes? If used in a smaller light running an IMR10440, it’d make possibly 2.5-3A? Now that DOES sound interesting! I see what you mean Matt, I do see what you mean.
I wonder how reliable these little jems are, that you’re using?
I fit in alongside MRsDNF, I’m extremely interested but don’t have the knowledge to be of any help. When I saw your thread and the “direct drive” I didn’t realize you were making it along the lines of the original East 092 in the HD2010, with the lower modes but direct in high. Having the modes certainly put’s it in a different “light”.
I like to Order smd parts from eBay, not so reliable source but the most seller ship in envelopes which is cheap. If I order from an electronic dealer the shipping is around 7-8$ and often I have to order for 10-15$ minimum…not to mention the big ones like RS or mouser and so on they charge even more…so with me getting smarter everyday I have to order everyday ya something new…
So I just ordered some parts from eBay some even from china and pcbs from china.
I just thought if this may end in a disaster I didn’t loose much money…
Of course I can build bigger circuits but there are plenty out there in 17mm, the nanjgs are awesome for the 3$ they cost you can’t beat them. For bigger diameter you could also just use this board and wire a bigger FET in to220housing on top but I guess for the most people that wouldn’t even be necessary as 4-5A is enough, but there is no adequate power for those 14500 lights. For example the zoomie has <1A tailcap current…
I used “nanjg layout”, so firmwares are compatible.
Of course direct drive for me is kind of lame for a real world application.
I ordered an efest10440 and one 14500 so I can measure that in some days. I also will mount a new test led to a heatsink for longer time tests. All tests are made with a XML on a 25mm star from stlv6…
