This is not real due technical abilities. For 3-5pcs you can buy alu-based blank and make them by your own.
OSH Park are making their work great. But they have three type of boards, one type of mask, one type of copper finish and thats all. Prices per sq inch are not cheap ($3.15 for three led switch boards, you can order full assembled lightening switches from ali little cents more). I can suppose that OSH Park have different process with small size blanks and this way they can keep cliche for all steps.
More usual way, is using big size blanks. There is almost no way to find them in Cu. With this way all steps and all machines are made just for this size, there is no way to bring smaller size piece and ask to make your pcb, anyway it will be made from full size blank. Thats why minimum reasonable order can not be less than 0.5 or 1.0 sq. m.
One more problem with metall core pcbs, is milling. Square boards can be cutted other way, but round ones require milling with complete different tools and rates from those that can be used with glass fiber boards.
P.S. Most alu-based blanks are one-side, this can be problem. From the other side, soldering one-side driver on aluminium board can be fast and repietable in medium quantities.
kiriba ru that's the point though. Most people just want a few (three or even less), and yes $3 a board is probably outrageously expensive per area, but no compared to your option of doing building them yourself, a few bucks when all you need is a few small boards anyway, isn't a big deal. There's no better option really. You want to spend hours making a $3 part? Well ok, I guess that's somewhat what folks do around here, but well, there might be some limits. Group buys could work, but really I don't think that fills the needs. If you could get RMM in, that would help, but doesn't solve international shipping.
OSH Park uses full size blanks. They put multiple orders (from multiple customers) on each blank, until it is full. That way, they can offer small sample sizes. They do have a couple of different choices available, not just one. And the pricing starts at $5.00USD per square inch for three copies (so really $1.67/in² ) shipped worldwide. It isn’t the cheapest way to get boards made, but more convenient for small sample size than most other shops.
Board is just one small part.
You can share design and smb will be able to order one pcb.
Then he needs to order mcu, diodes, smd resistors (sold in 100 or 1000 packs) and etc.
Then he needs to order mcu clip, cable, adapter for programming.
All this work looks silly, hundreds of blfers are making same drivers by one when smb could make 100.
And yes, if we will look forward to buck and boost drivers, we will see that proper components are made by 2-3 companies in the world, they dont have wide distribution and there is almost no way to get equil component in different countries (I cant make meteor driver copy because parts that can be ordered from Moscow are different from those that are selling in Kyiv).
100's or 1000's? There's exactly one of the drivers in my light, and before I made it, nobody knew if it would work. Sure, then group buys come, designing lights with thorfire. But then people want the dirver from one light in a different light. You can't get custom and mass produced too and there's just not that much market for boards at all to sell more than a couple of designs retail, fully built. Given the price of some expensive lights, I wouldn't say it actually to me thaaat long to construct this either, R&D notwithstanding.
For you own light, you dont need high quality mask or finish. You can use “laser iron” technology (print layout with laser printer and copy this image to the copper layer with old heavy iron).
I suppose you dont understand what are you talking about. With normal technology, second blank adds 20-30% to the price of the first one. This means that 70% is cost of clishes for all steps.
OSH are based on small orders of shared projects. Most buyers are not able to make their own design and just order existing projects. If you will fill big blank with big amount of different designs, you will need to make new cliches several times a day.
I think he knows what they say they do, and he also know the end result, that is good for us, and that nothing else really competes with for its purpose. I'm not sure what point you're trying to make. Sure we could iron our own. Ok. If you want to make up 100,000 of your favorite boards and start setlling them cheap from a proper online store front, I'm sure people will buy them (not sure 100,000 will). We'd love you.
A 2nd order filter would be an overkill here. The opamp is heavily over-compensated to ensure stability in any use. A side-effect is that almost no ripple makes it past the opamp.
Also too much math in those links . I just kept the RC cut-off frequency two orders of magnitude below the PWM frequency, ~50 Hz IIRC.
That's not a second order filter in that post. It's a first order. It's the same design you have here and the same one I put on the Texas Buck. The only difference being R6, which doesn't seem very exciting. yeah, for that buck, we needed to have the ripple tightly controlled. It turns out a bit more complicated than just RC, becuase you've got a different R for charge and for discharge, and the ripple fraction ends up depending on the actual output level, but yeah, basically of course if you're in the percent ballpark on that estimate, you're not going to end up with 30% ripple of course (except with very low PWM levels).
If this is true, it might explain why OSH Park makes a minimum of three copies of each board design. It makes sense to me. Three blanks, one cliché. But, you’re right. I don’t know what I’m talking about. I only know what they claim on their website. And I don’t even care how they do it. Only that it is convenient for me. That’s why I asked if you have a link to a fab shop that does it exactly the same way as OSH Park, but with metal PCB instead of FR4. That’s really all I want. Everything else is unnecessary argument from you, because I don’t care how any of it actually works.
Reminding myself of that post and being only one letter more mathematical than you DEL, what it shows is that the right RC to use is actually
R3*C*G where G is the divider ratio R4/(R3+R4)
This slightly better estimate gives you exactly the right answer as PWM goes near zero, (and only better at higher values actually, so I had that a little backwards). So just one letter more math really. If you're trying to stay at 1% level though with the simpler estimate, but the divider is a 5x divider, your low PWM's end up closer to 5% ripple, so yeah, that can matter (maybe not for this op-amp though).
The issue here is a matter of cost, complexity and practicality.
Naturally having the LED and FET separate is better for heat spreading but how much better? Based on my experiences I would say it will have a fairly minor effect after the first 1-2 minutes in most hosts. The 30 second numbers will most likely suffer some due to the mcpcb heating up faster but once it is hot the results will not be much different.
A heat sink on the FET would help for the first few seconds like you said but it will still desolder itself after that if you push it too hard. It needs a heat path to get rid of the heat.
The issue here is we do not have any kind of thermal safety in the FET so it will keep heating up until it melts if you let it. I de-soldered mine several times during testing due to thing when the light otherwise appeared to work fine. one of those times the wire actually bridged to ground and had I been using a battery instead of my current controlled power supply it would have caused a direct short for the battery and could have exploded. Needless to say, that just won’t work.
So the only option is a better heat path. The only option for this that would be retrofittable into almost any light is putting the FET on the mcpcb. Anything else would not be universal enough and a royal pain to use.
Also to whoever asked about resistance, there would not be anymore resistance then the normal wire that goes to the LED. You are just putting the FET at the opposite end of it.
I have looked into this and it was my first thought as well. The issue is not even about getting them made, I found several places that can do it (although they are pricey compared to normal PCB’s). The issue is that they are all limited to 1 sided PCB’s, which makes sense. We simply can’t fit any driver design that would need an aluminum pcb on a single side.
Plus we would have the issue of getting the V+ through the PCB and mounting a spring.
If you know of cheap 2 sided aluminum pcb’s, please do share.
Yes, these were the first thing we considered for this driver actually. It is basically a glorified opamp in reality. It does basically the same thing that the opamp we are using does but with much less user control and a larger package.
The other major issue with them is that they have a “ramp up” “feature” that doesn’t allow them to be PWM, thus making them basically useless unless you only want a single mode.
The real nail in the coffin was that to run more then the rated current you need an external FET, which runs into the same issues we have here. No thermal pad and thus no way to heat sink it.
That is unless someone knows of an FET with an electrically neutral thermal pad?
The reason that 7135’s work as good as they do is because they have a thermal pad to get rid of the heat. They also have built in thermal management to keep things from overheating. This makes all the difference.
If you looked at the datasheet I think you missed the part where you can use an external N-channel MOSFET with it to increase the current capacity. It works quite well.
. I just kept the RC cut-off frequency two orders of magnitude below the PWM frequency, ~50 Hz IIRC.