Oshpark Projects

I’ve been reading this application report from TI: http://www.ti.com/lit/an/slpa010/slpa010.pdf

It’s really informative on the subject. Of course the ringing occurs due to the rapid change in voltage or current on the power side of the FET, and then through some method or another ends up on the gate pin too. While their focus is on voltage margin for the FET parts (we aren’t close to the voltage limits of course) they do clearly illustrate that just a short trace length makes a big difference in ringing. They are showing it because we’d need to know that for using a scope to diagnose what’s happening at the FET.

I’m going to say that I think the SRK-DD board is probably doing the same thing as the resistor in the BLF17DD - softening up the turn-on/off. We take the same efficiency/losses hit and heat bump in that case.

We have no caps for the FET. The app notes are about tuning up synchronous buck converters, but it seems like some capacitance might help us here too. Of course capacitors are expensive and take up a lot of space.

I’ve read a good chunk of the document at this point. Granted that it does not directly apply to what we are doing, but it seems pretty cut and dry here: either get some capacitance very close to the FET or slow down the switching time.

EDIT: I don’t know which capacitance numbers mean what (input / output / reverse transfer) but the datasheets show significantly less capacitance for the AOD510, especially for Ciss and Crss. Maybe a simple 1uF input capacitor located very close to the FET would knock the problem out.

Talking about a 17mm DD family.....

As an end user of driver boards without experience on designing board lay-outs or the ability or will to program MCU's, but with a lot of enthousiasm for using the DD-boards....

for the small light mods I usually do, it would be really nice to have a 17mm (and 15mm) DD board with no components on the battery side (even not a few tiny resistors), and using (apart from the FET) only components scavengable from other boards like the 105C.

Here in Europe every Digikey or Mouser order comes with $25 shipping costs. I can ask Richard for a special delivery of components because he is a really nice guy but he normally does not ship outside the US (he did get me a batch of those Vishay FET's :-) ). The way I will mostly likely build up these drivers is scavenge a qlite, it has the modes I am happy with.

Not sure if it is possible to make a design like that....

I think it’s certainly likely if the lfpak56 gets a board designed around it. The only real difference is the added gate resistor. I like the singe sided idea myself with only spring and stars on the back. I think there’s room for the Zener as well but I’ll have to wait on either learning eagle or more interest in that FET. It does seem to be on ebay(UK).

My sentiments almost exactly. I am willing to sacrifice some beauty for functionality. We are the only ones that will look at the positive terminal and see the presence or absence of chips and have an opinion. I suppose it is analogous to a woman in pretty matching underclothes- perhaps no one will see them but she knows that they are there and it makes her feel happy.

I would like to take this opportunity to thank you guys again for all the work you are doing for these projects. I only understand a fraction of what you say but I enjoy the progress you have all made and shared with all of us. Thanks.

My latest driver project can be used that way. With just an FET you can ignore all the warnings about needing a custom firmware… just pay attention to the warnings about how hard to build it will be. As I say in the thread, it should not be quite as difficult as it looks - but it’s close. It does use all 0805 sized components, so you can use what you rob from the Nanjg 105c, but since I bypass the protection diode you’ll want to use a different value for R1. Pads are available for either DPAK or LFPAK56 MOSFETs. 17mm DD+7135 — linear regulated driver w/ FET turbo As you can see, there are nothing but 7135’s on the bottom. The entire bottom of that board could just be deleted and replaced with one big BAT+ pad.

For the extra 0805 components (R1, gate resistor, gate pulldown resistor) you can just get them from Chinese eBay vendors 100x for a buck or whatever.

Which LFPAK56 FET are you using?

This one.

Data sheet

What's the minimum-cost oscilloscope that'll give us a picture of what's going on with the gate signal? Are the sound card-based DIY things good enough? Are the Hantek 2-channel USB scopes worth the going price of $60-70? We don't need lab-grade precision, just something that shows the waveform, right?

In a like-for-like comparison between the AOD510 & 70N02 - identical SRK-DD drivers/firmware, same cells, same 3XP/XPG2 board, I measure 41.5mV drop across the AOD510, and only 32.6mV across the 70N02. So the specs in the datasheets that show the 510 having lower Rds(on) don't hold true in this lower-voltage application.

I don’t know, but the soundcard option is free. I think you want your sample rate to be at least a few times as high as the freq of the waveform you want to look at.

Is it possible/convenient for you to re-run that test on 5.0-5.5v input to the ATtiny13A? If doing that turns the tables we’ll know more about the situation.

In terms of the ringing problem, I thought to go and look at what the East-92 schematic does to help with that. It seems that a diode is used across the LED. Maybe this is similar to what I’ve seen referred to as a “freewheeling diode” elsewhere - the Schottky diode in a non-synchronous buck driver. When the sudden collapse hits (FET turns off) there is a path for the stray electricity to get away through… Anyway that’s the only reason I can see for putting a diode there. “Post #185”: East-92 4amp 17mm driver in the East-92 4A driver thread. Of course that wouldn’t come up in the TI document about synchronous buck converters since it would never apply!

I can short across the protection diode to raise the voltage on both drivers, would that be enough? Not sure I want to risk trying to rig it up with 5.5v powering the control side and 4v powering the LED side.

The SRK-DD board already has the provision for a big schottky at D2, between LED- and B+. It's not used on these two examples, as neither is using the inductor. I can add the diode to both, but since these both work properly I don't think that would tell us anything. Adding one to the 17DD would be... awkward.

If it’s easy to do it won’t tell us nothing… But it’s not ideal. The protection diode should only be dropping 0.2v.

In theory if you had a scope you could see the difference on that when you add the big schottky. Without that, the only way for us to observe the effect would be on the 17DD.

Hang on... can I just jumper something like 6v onto the gate pin from 2x CR123s, using the normal 18650s to the LEDs? With the MCU 'off'?

For the purpose of measuring Vdrop at 100% duty cycle? Good idea, but that will exceed the absolute maximum rating for “Voltage on any Pin except RESET with respect to Ground”: –0.5V to VCC+0.5V

So now that I think about it, here’s what you can achieve with jumpers if you are so inclined:

  1. jumper the protection diode to pickup 0.2v
  2. then jumper VCC to the PWM pin to pickup another 0.7v-1.0v on top of that

And that’s not nothin’. :slight_smile:

Thinking about it a little more, let’s go for a bonus round. Instead of doing both jumpers, just stick your DMM on the protection diode and confirm that it’s dropping <0.5v. If so you are good to go on jumpering directly from BAT+ over onto the PWM pin / gate pin.

I did see your project but did not realise that the board could be used like this :-) Sooo, if you replace the bottom by a big +pad and a fat -ring, the one-sided 17mmDD board would be almost there, but the reverse polarity protection would be lost because there is no space for the diode? But reverse polarity protection is a very useful feature (I do insert batteries the wrong way every now and then), I am not sure I would like the board without it.

(always complaining )

That board has only 7135’s on the bottom. D1 is on top between the mcu and the FET.

Ok, sorry, I switched on my brain and dived a bit into the board, now I see: the protection diode is there. So do I understand correctly that the place where the trace starts that the MCU uses for the voltage measurement for LVP is different from the 105C, namely on the other side of the diode, and this must be compensated either by the firmware or otherwise by using a different R1? Can this also be solved by cutting that trace and make a bridge from the trace to the other side of the diode?

Yes, you understand correctly.

That change in the trace layout was to allow the voltage divider to see the full battery voltage even after a Zener mod so that lvp could be made to work again with just a change in R1. It means that the 19.1k value for R1 isn’t quite right for single cell operation though.

I assume the BLF15DD V3.0 OSHPark listing here: https://oshpark.com/shared_projects/lQUEtZcJ, has the wrong value for the R4 resistor? Says 100 ohm, should be 130 ohm? Can Mattaus correct that?

So the "Revision 2" has no limiting resistor while the V3.0 does. This is a little confusing between V3.0 and Rev 2 - different nomenclature - I'm not even sure if the rev # pertains, because one version has an option while the other doesn't. I wish rev #'s could be put on the silk screening somewhere, because I have mutiple board rev's all over the place... Are there any recomendations for R3 values if we want to limit current? Is anyone doing that? There's no BLF15DD discussion thread, so no source of info on this board.