der_wille_zur_macht
Team RC
Sounds good! Which of the ICs discussed in this thread are you planning on using?
DIY LED driver for reef lighting
- Thread starter der_wille_zur_macht
- Start date
CAT4101's. Low cost and low part count won me over.
der_wille_zur_macht
Team RC
As promised:
Awww, it's like seeing an old friend. :lol:
TheFishMan65
New member
It must be Christmas. Like everyone else I got a present in the mail. My boards are in. I will try and post pictures tonight.
DWZM was correct that those smaller resistors are hard, but by the 8th one I think I had it down (and I haven't soldered in years). Fine point on the soldering iron and a good pair of tweezers.
DWZM was correct that those smaller resistors are hard, but by the 8th one I think I had it down (and I haven't soldered in years). Fine point on the soldering iron and a good pair of tweezers.
der_wille_zur_macht
Team RC
I do small SMT stuff with "blue tac" - the sticky putty stuff you use to hang posters, etc. I have a tool that's basically like a dentist's pick and I put a TINY little dot of the tac on the tip of the tool. It's just sticky enough that you can touch it to a small SMT component and it'll pick it up. Then you can hold the component in place on the board. For the actual soldering process, what I typically do is tin one of the pads with a blob of solder. Then, hold the component up against the blob, and touch the tip of the iron to the blob. When the blob melts, you slide the component into place and remove the iron. Then you can pull the probe/tweezers away and solder the other pad "freehand" since the part is held down by the first pad.
For big SMT parts, i.e. the IC itself, you can put a tiny blob of the blue-tac on the PCB itself right next to where the part is supposed to go, then mash the part against it and it'll hold in place. Then both hands are free to hold the solder and iron. I usually pull the tac away after one pad is done, because, again, once that first pad is done you're home free; the part will be held down on it's own. Getting the part aligned and tacked down with that first pad is half the battle, after that even tiny little parts are easy.
I tried to do some photos to show some of these methods since people had asked a few pages back, but my good camera is on vacation this week with the wife and kids, I'm stuck home with my cell camera which isn't high-quality enough to photograph stuff this small. Really, I started liking SMT once I got over the hurdle of figuring out how it's done. It's quicker and easier than through-hole, you can just whip across a board without fidgeting or turning it over, etc.
In general I like the blue-tac method better than tweezers because you don't have to regulate pressure - I find tweezers hard to use because if you pinch too hard or not hard enough you can drop the part. With my blue-tac dentist's pick/probe tool, I just touch the part and it's picked up. IME it's easier, at least, but I have shaky hands and bad eyes.
For big SMT parts, i.e. the IC itself, you can put a tiny blob of the blue-tac on the PCB itself right next to where the part is supposed to go, then mash the part against it and it'll hold in place. Then both hands are free to hold the solder and iron. I usually pull the tac away after one pad is done, because, again, once that first pad is done you're home free; the part will be held down on it's own. Getting the part aligned and tacked down with that first pad is half the battle, after that even tiny little parts are easy.
I tried to do some photos to show some of these methods since people had asked a few pages back, but my good camera is on vacation this week with the wife and kids, I'm stuck home with my cell camera which isn't high-quality enough to photograph stuff this small. Really, I started liking SMT once I got over the hurdle of figuring out how it's done. It's quicker and easier than through-hole, you can just whip across a board without fidgeting or turning it over, etc.
In general I like the blue-tac method better than tweezers because you don't have to regulate pressure - I find tweezers hard to use because if you pinch too hard or not hard enough you can drop the part. With my blue-tac dentist's pick/probe tool, I just touch the part and it's picked up. IME it's easier, at least, but I have shaky hands and bad eyes.
der_wille_zur_macht
Team RC
The single instance of the driver has been running flawlessly, nonstop for 48 hours. Tuned for the smallest drop across the chip, the IC is barely warm. I was done with work an hour ago, had a quick dinner, then I set up two more 6-LED test beds consisting of three blue Rebels and three cool white Rebels each.
Soldered the other two driver circuits on the board, and here it is:
Worth noting that the blue Rebels drop more voltage than the whites do, so the two ICs with blues in their strip have a higher Vf demand, so I had to turn the power supply up a bit. This resulted in the driver running only whites (with a lower Vf) getting warmer. It's still comfortable to touch for several seconds so I'm not worried, but it makes me wonder about how carefully we'd have to match the power supplies to the LED strings in practice.
Soldered the other two driver circuits on the board, and here it is:
Worth noting that the blue Rebels drop more voltage than the whites do, so the two ICs with blues in their strip have a higher Vf demand, so I had to turn the power supply up a bit. This resulted in the driver running only whites (with a lower Vf) getting warmer. It's still comfortable to touch for several seconds so I'm not worried, but it makes me wonder about how carefully we'd have to match the power supplies to the LED strings in practice.
stephenrwright
New member
I have been thinking about the differences in the Vf I have seen on the LEDs I have played with and though I should test each LED and mark its voltage drop and build string with the total drop as close to all the other strings as possible, so each driver can have the smallest drop possible.
stugray
Premium Member
So DWZM,
Do you think you need to "spin" the board for the 4101 version or is it good to go?
I think I would like to build some as well.
I'm just wondering what changes you would make after building & running one for a while.
Also, I started playing with Eagle, and I am curious if we could add pads for the BAE0910 as posted on the:
http://www.reefcentral.com/forums/showthread.php?t=484912&page=15
Thread.
Pic:
http://owfs.org/index.php?page=BAE910
If we put pads for just the chip it could be made to provide 3 separate PWM signals for the three strings and still have room for one of the AtoD inputs and we could measure the temperature of the driver board. The beauty of this chip is you load the PWM code to it over 1-wire and it just runs....
I even know a way to have three separate temp sensors glued to the driver chips and one AtoD would read the average of the three sensors ;-)
Hint: look-up AD590
Stu
Do you think you need to "spin" the board for the 4101 version or is it good to go?
I think I would like to build some as well.
I'm just wondering what changes you would make after building & running one for a while.
Also, I started playing with Eagle, and I am curious if we could add pads for the BAE0910 as posted on the:
http://www.reefcentral.com/forums/showthread.php?t=484912&page=15
Thread.
Pic:
http://owfs.org/index.php?page=BAE910
If we put pads for just the chip it could be made to provide 3 separate PWM signals for the three strings and still have room for one of the AtoD inputs and we could measure the temperature of the driver board. The beauty of this chip is you load the PWM code to it over 1-wire and it just runs....
I even know a way to have three separate temp sensors glued to the driver chips and one AtoD would read the average of the three sensors ;-)
Hint: look-up AD590
Stu
der_wille_zur_macht
Team RC
Stu,
Spinning as we speak. I'm correcting the caps that weren't grounded (though it ended up being absolutely no problem to fix when assembling the boards) and tweaking the package I used in EAGLE for the IC itself since it's not QUITE the right shape/size for the actual physical piece I got. And probably add a few more vias. None of this is critical of course, but I'll have it done shortly.
Adding that chip to the driver board would require a bit of rework since there isn't a convenient spot for it right now. Feel free to do so if you want, but I'm gonna run without it.
Actually, since Seeedstudio sent me so many extra boards, I don't even need to order any more myself so I'm set at this point. I ordered 5 under their old "open source" plan and they sent me 12! I've heard this happens from time to time since they panelize many customers' designs onto a single blank; so if they have to re-do the blank because of a failure on another person's board they just send everyone else the extras. So I lucked out I guess.
At any rate I'll post the EAGLE project on google when it's tweaked.
Spinning as we speak. I'm correcting the caps that weren't grounded (though it ended up being absolutely no problem to fix when assembling the boards) and tweaking the package I used in EAGLE for the IC itself since it's not QUITE the right shape/size for the actual physical piece I got. And probably add a few more vias. None of this is critical of course, but I'll have it done shortly.
Adding that chip to the driver board would require a bit of rework since there isn't a convenient spot for it right now. Feel free to do so if you want, but I'm gonna run without it.
Actually, since Seeedstudio sent me so many extra boards, I don't even need to order any more myself so I'm set at this point. I ordered 5 under their old "open source" plan and they sent me 12! I've heard this happens from time to time since they panelize many customers' designs onto a single blank; so if they have to re-do the blank because of a failure on another person's board they just send everyone else the extras. So I lucked out I guess.At any rate I'll post the EAGLE project on google when it's tweaked.
der_wille_zur_macht
Team RC
That wouldn't be a terrible idea, though in practice I'm guessing you'll need some wiggle room beyond what you measure to be the smallest possible drop on your test bench. I turned these LEDs on about 2 hours ago and the Vfs of the strings (plus the voltage put out by the DC power supply) has varied a tenth of a volt or so. I'm guessing this happened as things heated up/broke in/whatever.
It's fascinating to me that I have essentially two strings that are spec'd the same (three CW and three blues each) and the difference in Vf across the strings at the same current is .3v!
TheFishMan65
New member
IRR the resistance goes down as the LEDs heat up.
I got mine up and running with a power resistor. I am measuring input voltage, output voltage, and highest chip temperature.
V In V Out Temp current ma
5.04 4.540 088 0.89
6.13 4.620 122 .9
6.62 4.660 149 .9
7.05 4.650 169 .9
The temperature was with a IR temp guage. The 4101 does not have the tab soldered yet. Just incase I have to remove the chip I have not soldered the tab yet. Been running 15 minutes at 7.05 Vin with no temperature change. Ambient is 67. Measurement in fahrenheit.
Current was measured from the power supply on board meter (not calibrated and old).
It looks like we need a little extra voltage to get the full current out. Anyone else notice this?
I got mine up and running with a power resistor. I am measuring input voltage, output voltage, and highest chip temperature.
V In V Out Temp current ma
5.04 4.540 088 0.89
6.13 4.620 122 .9
6.62 4.660 149 .9
7.05 4.650 169 .9
The temperature was with a IR temp guage. The 4101 does not have the tab soldered yet. Just incase I have to remove the chip I have not soldered the tab yet. Been running 15 minutes at 7.05 Vin with no temperature change. Ambient is 67. Measurement in fahrenheit.
Current was measured from the power supply on board meter (not calibrated and old).
It looks like we need a little extra voltage to get the full current out. Anyone else notice this?
der_wille_zur_macht
Team RC
I'm seeing that it'll do full current with ~.6v drop across the driver circuit. That's pretty close to inline with the datasheet spec of .5v.
FYI to Stu and others interested, the EAGLE project for my triple board is up here:
http://code.google.com/p/hpled/downloads/list
Released under Creative Commons BY-NC-SA 3.0 (in other words, you're free to use, distribute, and modify it for non-commercial purposes, as long as you give attribution).
FYI to Stu and others interested, the EAGLE project for my triple board is up here:
http://code.google.com/p/hpled/downloads/list
Released under Creative Commons BY-NC-SA 3.0 (in other words, you're free to use, distribute, and modify it for non-commercial purposes, as long as you give attribution).
TheFishMan65
New member
I amgoing to borrow another meter tomorrow and measure the current as voltage goes up. So maybe I will get results more inline with yours. As I said the meter is not kind of an unknown.
Still running about 169 (after 1/2 hour) the board is 145 so I don't have bad transfer to the circuit board.
Still running about 169 (after 1/2 hour) the board is 145 so I don't have bad transfer to the circuit board.
TheFishMan65
New member
dwzm,
How did you mesure the LED current verses the input voltage? As I go from about 5.5 to around 10 volts in the current goes from about.916 ma to .936 ma.
How did you mesure the LED current verses the input voltage? As I go from about 5.5 to around 10 volts in the current goes from about.916 ma to .936 ma.
der_wille_zur_macht
Team RC
Set the DC power supply to a certain value, testing (under load) with the multimeter. Then, shut it all down, swapped the multimeter over to the output side, turned it on, and measured current.
I only went up to about 3v drop because I don't think I'd ever run it with more than that. Below .5 - .6v drop, the current changes drastically (as expected). Above that it changes veeeery slowly (a few mA per volt) which IMHO constitutes "stable operation."
I would expect the result you're seeing - 20mA rise for a 4.5v change in input current. If you look at the datasheet, there are a few graphs which confirm this - as voltage drop and temperature increase, the LED current goes up a hair. So it strikes me that despite the "change" in current, you ARE operating in a stable, expected mode at that point and the chip is regulating current as you'd desire it to; vs. below the minimum drop where regulation isn't happening at all and current varies much more sharply.
I only went up to about 3v drop because I don't think I'd ever run it with more than that. Below .5 - .6v drop, the current changes drastically (as expected). Above that it changes veeeery slowly (a few mA per volt) which IMHO constitutes "stable operation."
I would expect the result you're seeing - 20mA rise for a 4.5v change in input current. If you look at the datasheet, there are a few graphs which confirm this - as voltage drop and temperature increase, the LED current goes up a hair. So it strikes me that despite the "change" in current, you ARE operating in a stable, expected mode at that point and the chip is regulating current as you'd desire it to; vs. below the minimum drop where regulation isn't happening at all and current varies much more sharply.
der_wille_zur_macht
Team RC
Convert to gerber.
I was on the verge of packaging the gerbers up in that zip too, but since people might take it different places, and different shops will change their requirements from time to time, I decided not to.
Basically, it's a two step process:
1) Get the DRC (design rule check) file from your vendor of choice, or at least get their design requirements - things like minimum trace width, etc. and run a DRC check. This board is very "big" in terms of traces and spacing so there's really no chance it'll fail, so you can skip this.
2) Get the gerber requirements from the vendor, or even better, get an EAGLE CAM file from them. The CAM file will have settings for EAGLE to generate gerbers in the format that vendor wants.
For Seeed, their EAGLE requirements are here on the prototyping product page:
http://www.seeedstudio.com/depot/fusion-pcb-2-layer-5cm5cm-max-10pcs-p-513.html?cPath=64_12
I see the DRC link there but I don't see a CAM file linked there which is weird because they used to have one. I just put the one I used up on the google site so you can get it there.
So, you need to get that CAM file, put it in your "cam" folder (something like /program files\eagle\cam\) then open the EAGLE project, hit the CAM editor (looks like two blue filmstrips), then hit file>open and open the CAM file, then hit process job. It'll write a bunch of files to the project directory, so you can then go there and zip them up.
I was on the verge of packaging the gerbers up in that zip too, but since people might take it different places, and different shops will change their requirements from time to time, I decided not to.
Basically, it's a two step process:
1) Get the DRC (design rule check) file from your vendor of choice, or at least get their design requirements - things like minimum trace width, etc. and run a DRC check. This board is very "big" in terms of traces and spacing so there's really no chance it'll fail, so you can skip this.
2) Get the gerber requirements from the vendor, or even better, get an EAGLE CAM file from them. The CAM file will have settings for EAGLE to generate gerbers in the format that vendor wants.
For Seeed, their EAGLE requirements are here on the prototyping product page:
http://www.seeedstudio.com/depot/fusion-pcb-2-layer-5cm5cm-max-10pcs-p-513.html?cPath=64_12
I see the DRC link there but I don't see a CAM file linked there which is weird because they used to have one. I just put the one I used up on the google site so you can get it there.
So, you need to get that CAM file, put it in your "cam" folder (something like /program files\eagle\cam\) then open the EAGLE project, hit the CAM editor (looks like two blue filmstrips), then hit file>open and open the CAM file, then hit process job. It'll write a bunch of files to the project directory, so you can then go there and zip them up.
TheFishMan65
New member
Yes, I noticed that just above .5 volts over the output voltage the current jumped to about 900 ma. I was just qualifying the smaller current changes (I thought somewhere you said no current change - you have bad eyes I have bad memory). I hadn't thought to read the data sheet. So that is good we are seeing the same thing.
I am not sure how much we have to worry about the different forward voltages of the LEDs. I went back through the thread and found three cases where LEDs forward voltage was compared. The differences were .13, .12, and .22 volts from smallest drop to largest drop in a string. So if we take 7 (XP-Gs) as the most we can run with a difference of .25 volts that would be a total of 1.75 volts. Probably a pretty slim chance of all the lowest and highest being on their own string - but worst case analysis. Add to that a .6 overhead we have 2.35. That is pretty close to my last case of 7.05 volts in and 4.65 voltage out running at 170 degrees. That is 77 degrees Celsius with a thermal shut down of 150. I talked with an EE and general discussion if the thermal shutdown is 150 safe operating is probably 100-120 Celsius.
Now just to qualify this. 77 Celsius is hot - get a tender spot and you will get burned. Also a lot of the heat is being dissipated to the boards so the other components near by better be able to handle the heat also. I am also making the assumption heat generated by the chip is only related to current (pretty close to max in my testing) and the voltage drop. From the data sheet I don't see a drop from 7 to 4.5 being any different then 24 to 21.5 in terms of the heat generated, but I might be wrong.
I also think my board could go 2.5-3 volts more on the input before I got to the 100-120 range. Of course I was only running one driver and it was sitting on a table. There is the option of using a thermal pad that is not conductive and heat sinking the whole board.
I am not sure how much we have to worry about the different forward voltages of the LEDs. I went back through the thread and found three cases where LEDs forward voltage was compared. The differences were .13, .12, and .22 volts from smallest drop to largest drop in a string. So if we take 7 (XP-Gs) as the most we can run with a difference of .25 volts that would be a total of 1.75 volts. Probably a pretty slim chance of all the lowest and highest being on their own string - but worst case analysis. Add to that a .6 overhead we have 2.35. That is pretty close to my last case of 7.05 volts in and 4.65 voltage out running at 170 degrees. That is 77 degrees Celsius with a thermal shut down of 150. I talked with an EE and general discussion if the thermal shutdown is 150 safe operating is probably 100-120 Celsius.
Now just to qualify this. 77 Celsius is hot - get a tender spot and you will get burned. Also a lot of the heat is being dissipated to the boards so the other components near by better be able to handle the heat also. I am also making the assumption heat generated by the chip is only related to current (pretty close to max in my testing) and the voltage drop. From the data sheet I don't see a drop from 7 to 4.5 being any different then 24 to 21.5 in terms of the heat generated, but I might be wrong.
I also think my board could go 2.5-3 volts more on the input before I got to the 100-120 range. Of course I was only running one driver and it was sitting on a table. There is the option of using a thermal pad that is not conductive and heat sinking the whole board.
der_wille_zur_macht
Team RC
I guess to be 100% specific what I had meant was "acceptably small current change, indicating normal operation."
To be specific about the differences in drops, I'm worried just as much from an efficiency perspective as a heat perspective. I realize it's only a few watts difference but I guess I'm just picky like that. And, I'm not as much worried about getting different Vfs on different strings of identical LEDs (though as indicated above the differences were surprising to me) but rather the difference between, say, a string of XP-Gs and XR-Es run on the same DC power supply. THAT will be a significant difference, both in terms of several watts of efficiency and potentially enough to overheat the IC, especially once it's crammed in a project box with 30 other drivers, or whatever the end situation will be.
So, if people are doing large builds and are concerned about efficiency, they should probably "group" similar strings of LEDs on unique DC power supplies. And, at the least, everyone who's making these should tune the DC power supply down to the point where it's near the minimum voltage that'll still power the drivers in their normal operating mode. This will take 30 seconds and I see no reason not to do it.
Again though, I'm not losing sleep over these concerns, just commenting so people are aware.
