[WIP] 15mm PAM2803 w/ ATtiny13A rough layout / possibility

Flashlight Modding and DIY Parts
77

Ah, thanks HQ. I feel really off balance after my time off from the forum. :frowning: You did specifically talk about that in post #57, I just didn’t absorb it when I looked at your post. I’m not 100% sure what’s up with the weird GND path for the existing MCU on the FT driver. IIRC we may have seen something similar in ImA4Wheelr’s HX-1175b1 thread? I dunno and I’m not going to try and hunt it down right now. I do strongly suspect that I’ve seen similar wiring on another Chinese driver. For me it’s good enough that you noticed the same thing, tried it with the ATtiny, and ended up changing it to a direct GND.

On a side note HQ, do you really work with your Eagle set to a white background or do you just do that for screenshots? I just tried it and ouch it didn’t feel good - especially with both the bottom and top stuff enabled at the same time, where I spend a fair amount of time. Maybe I’d get used to it. :wink: :slight_smile: For anyone who doesn’t use Eagle: the default is black like in the screenshots I sometimes post.

I don’t yet understand how the lack of the pulldown/bleed resistors would burn a PAM2803. Hopefully when the parts to build this thing show up I’ll be really motivated and setup the scope. (I laid out a lot of cash for that thing, we should get some use out of it…) With the scope on 1xAA we can observe the behavior without the pulldown/bleeder. Then install them (stack the bleeder on the output cap and airwire the pulldown) and check everything with the scope again. With any luck this will demonstrate any spikes they may help minimize without burning the PAM2803.

78

I don’t have much to contribute here, other than to say I also use a white background and it definitely was the default when I installed Eagle.

79

In the board editor? Crazy! Since my memory is on the foggy side I had to do a sanity check - sure enough, in the past black was the default. A quick Google Images search for eagle+board+editor corroborates this. I wonder why they changed it?

80

I should point out that all I posted are more or less singular events. It was one PAM2803 that released smoke, one successful rebuilt FastTech driver, one built HQb15v1 that did not work as expected (3 modes, mode change works, but flickering in all modes). I can’t explain what happened or went wrong. It’s not that I was setting up test series here…

.
The PAM2803 that died was on a Nanjg110. I had only 2 more at hand at the time and did not want to lose them. Then the FT board arrived and I went with them - including the reistors.

.
The resistors (that I dubbed R2 and R3) are, as @wight pointed out, only present on the boards with MCU. They were not needed when I tested the BAT60A on a single mode Nanjg110 (post#35). But I don’t know what these resistors are there for on the FastTech driver.
R2 (“18D” = 150kOhm) is between LED+ and GND.
R3 (“103” = 10kOhm) is between FetSource/PAMpin3/Rsense and FetGate/PWMsignal.

.
The HQb15v1 I did model after the successful FT conversion. Can’t say whether I misdesigned it, a part is defect or I did a bad (rework-station-) solder job. Next weekend might show, hopefully.

.
I started using Eagle in summer ’14, installed it on 2 PCs and have never seen anything but a white background. Needless to say, I’m simply used to it.

.
I’m aiming for 1AA and 2AA operation, 2AA for several Mini Maglites which can so nicely be converted to a triple using a modded P60 pill. Very family friendly for those I don’t want to burden with lithium-ion cells.

81

In the past editions it was black and I always switched to white so I could see where the cursor was. Images seem more striking on a black background so I switch to that for screen shots.

82

Heh, maybe the “more striking” bit is a concentration aid for me. :wink: I use the default cursor which is “inverted” from whatever it’s hovering over (black = white, red = blue, green = pink, etc). Since I was curios I checked it out on both backgrounds just now. I actually find the cursor easier to see on the black background. Did you check out the “Large” cursor while you were in there?

83

No, just the background color. I’ll have a look next time I’m in there.

84

Liftoff, we have a liftoff…

15mm boost driver (v1) working

.

Partlist and used Firmware below

Rebuilt on a 2nd board, running fine.
2x 2AA and 3x 1AA (Eneloops) drained so far, about 200 mode changes, all good.
It’s a HQb15 de-luxe with Coilcraft inductor and high value capacitors.

Board is the same layout as in post#72 above.

.

Oshpark link for HQb15 v1 EDIT: Obsolete as HQB15 v2 is finished, see post#96


.

Driver height with this inductor is 3.7mm (0.2mm less than with the FT inductor). Now take the new Oshpark 0.8mm boards and you’re down to 2.9mm. That’s a whopping 3.0mm less than a Nanjg110.

Enough space around the Attiny13a for the clip after driver assembly.

.

.

Partlist
(source in brackets; see post#25 for more info on the parts)

Switch: PAM2803, SOT-23-6 (FastTech driver)
L1: 2.2yH, Coilcraft XFL4020 (electr. supply)
D1: BAT60A (electr. supply)
C1: 10yF, 0805, X5R (electr. supply)
C2: 20yF, 0805, X5R (electr. supply)
C3: 4.7yF, 0603, X5R (electr. supply)
R1: 0.120Ohm, 0805 (FastTech driver)
R2: 150kOhm, 0605 (FastTech driver)
R3: 33kOhm, 0605 (FastTech driver)
MCU: ATTiny13A-SSU, Package 8S1 (electr. supply)
OTC: not yet
FET: (FastTech driver)

C2 and C3 are by design in parallel (C2 for PAM2803, C3 for ATtiny13A). In this built I used both caps and with high value.

.

Firmware

An adapted Dr. Jones MiniDrv firmware for starters.
Frequency changed to 18kHz, on-time-memory, 9 modes

.

.

Output

Eneloops fresh from charger (no resting time)
With 2AA holding ~780mA for quite some time (R1 (=Rs) = 120mOhm)
With 1AA starting >550mA, continuously going down, but surprisingly slow in comparison to Nanjg 110

.

loose threads…

no OTC yet
adapting R1 to change output
output at PWM1 is 16mA with fresh cells (1AA and 2AA)
trying different components (inductor, FET)
17mm version

15mm V2, will be a fine tuning of the layout after having built them twice
V1 has an overall size of 15.4mm, the GND copper layer by Oshpark is only 14.8mm. V2 will be increased to 15.6mm.

(EDIT needed after Drupal overhaul)

85

Wow! Wow! Can’t wait to see the progress!

86

Looks great. Will you be testing efficiency?

87

Nice!

88

Maybe add one or two more gnd vias at the fet for the next version.

89

@HarleyQuin

That is some really outstanding work! I’m excited to watch as you develop the design.

90

Good work! Looks like your motivation has been boosted a little. :wink: Did you use fresh components or are these just moved over from the badly behaving board in post #57? Clearly it is a different inductor of course. Have you developed any idea of what went wrong with the last build?

Now I’m pretty excited while waiting for parts to show up. My layout definitely needs some work to accommodate the two ‘extra resistors’ discussed recently… unless it just works and then I don’t care. :wink:

Efficiency is kind of a moot point if you can’t find a driver with these capabilities elsewhere. The PWM modes + PAM2803 are the same way DQG handles the situation IIRC. It’s also difficult to “test efficiency” when you’re doing PWM like this, except on 100% PWM (always on / high mode).

That said, I’m sure we’ll look into that at some point. (at least for 100% PWM)

91

Once we get a handle on this stuff the next questions are:

  • whether we can use PAM2803 on a larger PCB (17mm etc) to drive a P-Channel MOSFET. (so the internal MOSFET or FET or whatever in the PAM2803 would act as a gate driver I suppose). If we can get that to work we may be able to squeeze a little higher current out of a larger board but still work with 1xNiMH or 2xNiMH
  • whether we can remove or tamper with “VOUT” to get >5v output. Of course this is automatically a problem for the ATTINY and so might not be a great idea… but it’s an interesting question and if it came down to brass tacks a few diodes of the appropriate type could drop the ATtiny’s supply back from say 6.5v down to a safe 5.5v. That might be a workable way to get ~6v out of 2xAA or 1xLi-ion. Realistically I don’t think we’ll be doing much along those lines. Boosting from 2xAA to 6v isn’t going to be efficient at all and will be harder to spec components for I think, besides the fact that we’d be running stuff out of spec. Boosting from one li-ion to 6v can already be handled with a different boost IC.

Really I’m just interested in the second bullet point from an academic standpoint. The first though… that could get fun!

92

Hi Alex, just looking to get a baseline since hopefully there will be others to compare it with at some point.

93

Thanks for the encouraging comments.

True.
Knowing there is test-personnel standing by :wink:
It’s only that wife and kids are now wondering why I’m so absent minded again, not paying attention, mumbling strange words like sense resistor and redundant components.

.

I can’t provide data as HKJ does in his tests. Best I can do is I-in vs I-out. Even this has already proven to give different results depending on the eneloop used (hot from i4 charger or resting from Opus-charger). My testing procedure isn’t remotely accurate enough to invest time here.

.

The working second HQb15 v1 has no parts of the first one. I laid the 1st one aside after already checking continuity and having resoldered where I could with the iron - to no effect.
The 2nd one was a complete fresh start. New board, new FT driver, new components. Testing continuity on the complete board, testing the FT driver before desoldering (hadnt done that with 1st one) and taking unused additional parts.
My best guess for the 1st one: Bad solder connection / not enough solder paste somewhere. Perhaps a faulty component as a distant second.

For now I’ll leave R2 and R3 in the design, but I’m couriously expecting your test results.

.

Could you kindly explain the reason behind this. Unless I’m mistaken (and I could very well be) in this boost circuit the FET does not draw current directly from GND. It’s coming via Pin3 and Pin2 of the PAM2803. I might thicken this path, though.

V2 inbound, thatfor I’ll skip V1 link.

94

@HQ - sorry if this seems obvious or I’m missing something but if the negative connection to the battery tube is on the spring side then current from L- goes through the pwm controlled fet from drain to source and then through the board. Other components are grounded as well but the vast majority of the ground side current follows this path and just as the positive side needs short, generous traces/vias along the Vin - Led+ path, so does this. There are other gnd vias sure but straight through from the fet is the shortest path. Not a big deal at low current but still something I would normally suggest.

95

This is where I have more interest.

But it’s great to see the drivers built by this community growing!

96

@RBD
Thx for explaining your thoughts. The through the board part was indeed obvious to me. It was the at the FET part, which I translated to: bottom right of the driver, that made me wonder. I think the current goes from L- to FET to PAMpin3-pin2 to the left center of the board, so vias at the FET sounded odd.
Anyway, I added vias. Doubled them, actually. Even made them symmetric…