DIY LED driver for reef lighting

[perikaruppan]

Kress:
"Slow down a bit. Your description is making no sense to me. Nothing "gives out resistance".
This is when i measure the resistance across the resistor connected to pin 4 of the CAT

Will reduce the length of the Cord and work on that.

The LEDS string that i tested is working. When Just one string is connected there is no problem, the current is set to 700mA even when the Vin is increased. The same string i moved through the different channels and it worked in every channel the same correct way. But the problem rose, when i started to add more strings to the PCB and the current started to DROP and i had to increase the VIN to get to 700mA and that in turn started to put more heat through the CAT.


Fishman:
I have a common input on the CAT board, the bottom LEFT on the Board to which i have connected the SMPS power supply cord. The measurement was done at the SMPS which was higher, then it dropped when i measured at the input of the CAT Board -- this is where it was 20.2V.

Pin 1 is 0
Pin 2 is +5V
Pin 3 is 0
Pin 4 - will measure and tell
Pin 5 will measure and tell (should measure from this pin to ground right?)

I did check the TAB and it is well soldered to the ground. I can feel the heat through the holes on the back side of the PCB.

 

[TheFishMan65]

What gauge wire from the power supply to the board? I think it is too small.

Pin 1 at 0 shouldn't produce any light
Pin 5 yes measure with ground reference

If you look at you board above there is a black rectangle around where the CAT mounts. There is not ground plane there. The heats gets to the gray area, but it has no where to dissipate to. If the black was ground the heat could move outward. Does this make sense?

 

[perikaruppan]

I cut the wire from 6m to 2' and now the voltage from the SMPS is 21.1, and at the input of the CAT is 20.8 so there is still a 0.3V drop there...

And i attached just 1 string of LED and the current is at 0.89 A (for a 680ohm resistor = 800ma), thats the lowest 21.1V i could go on the SMPS...

tried the same string of LED on all the other CATS and same result.. less heat.. but more Current for the resistor!!!...

The GREY box with those holes are the ones that were done for heat dissipation on teh PCB back side... the GREY box is linked to the ground plane, but small lines on the top of the GREY...

But didnt understand about the "The heats gets to the gray area, but it has no where to dissipate to. If the black was ground the heat could move outward. Does this make sense? "

 

[TheFishMan65]

What gauge wire from the power supply to the board? I think it is too small. That is a lot of voltage drop for such a short wire.

Sounds like it is not regulating are they all doing this? Double check your resistors. Can you swamp one with a lower resistor to to see if it regulates?

The heat goes from the CAT to the grounding pad, from the pad through the solder to the gray ground plane and through a little bitty teeny tiny trace to the rest of the ground plan. In the future you want lots of these trace or I just made it solid. Get it now?

 

[SpacedCowboy]

The GREY box with those holes are the ones that were done for heat dissipation on teh PCB back side... the GREY box is linked to the ground plane, but small lines on the top of the GREY...

But didnt understand about the "The heats gets to the gray area, but it has no where to dissipate to. If the black was ground the heat could move outward. Does this make sense? "

What Fishman is saying is that the heat is:

1) Radiated away from the top of the CAT4101 into the air
2) Conducted to the topside of the PCB (coloured gray) area
3) and that's it.

You say there's a tiny connection to the rest of the ground-plane at the top of the gray area (and I can just-about see it), but that's not what you want.

Think of the heat conduction as water flowing through a pipe, and imagine you have a really wide pipe where the water starts to flow, and a really narrow point later on in the pipe. If you start filling up the pipe at the wide end, you'll start to get a backlog of water because only a small amount can get out of the narrow end. This is what you have here...

The CAT4101 is generating a huge amount of heat which we're trying to dissipate. Most of that heat is being transmitted through to the PCB. The thermal capacity of the grey area on the underside will quickly saturate, and you have a tiny tiny spot at the top where the heat can be conducted away to other areas of the PCB.

The problem is that you'll reach a steady state in the grey area very quickly - if only a small amount of heat can be transmitted *away* through the bottleneck at the top, then only a small amount of heat can be additionally *added* from the CAT4101. The physics isn't quite like that, but it's a good way to think about it. You've put a roadblock in the way by only letting heat migrate through that little channel at the top.

This isn't *entirely* the case of course, FR4 will conduct some heat away, but copper is far and away better at it, so if you're having problems I would re-do the board and get rid of that isolation boundary around the underneath of the CAT4101. It's all GND anyway, so there's no problem...

On a slightly different note, when you say "single-sided", does the board have plain copper on the other side and plated-through vias ? Because if it doesn't, those vias won't be doing much - in fact you're probably losing some heat capacity.

My experiments with the CAT4101 seem to indicate you need about 1.5" by 1.5" of bare 1oz copper surface to have it run just-too-warm-to-touch-comfortably (which is well within its operating parameters). If you haven't got copper on the other side of that board, with plated-through vias for heat conduction, I could totally understand why the chip is misbehaving - there's little-to-no heat conduction on the top surface, after all.

Simon.

 

[perikaruppan]

each of the two wires + and -, is about 1mm thick.. not sure of the gauge.. its all copper multistrand... and the electrical people said it will easily take upto 240v/6A...

i will try replacing with a smaller resistor and see what happens.

Now i get the thing on the ground plane. in mine, from the grey ground area, through those tiny holes onto the other side of the board.... am i right?

 

[perikaruppan]

spacecowboy - now i really understand the system of heat dissipation...

there is no copper on the other side of the board.. the board is a single sided one with copper and green mask.

Will just take some images of the board and then you can help me out as to what needs to be done....

 

[SpacedCowboy]

each of the two wires + and -, is about 1mm thick.. not sure of the gauge.. its all copper multistrand... and the electrical people said it will easily take upto 240v/6A...

This is where I go to find out about wire and current limits. 1mm diameter is ~0.039" - is that including any insulation, or without. If it's with insulation, you need to measure it without, but assuming you measured just the copper wire, the chart says it's about 18-gauge wire, and you can comfortably transport 2.3A.

There's quite a bit of wiggle-room in those figures (they're very conservative) and 4 x 0.7A is 2.8A. I think you'd be ok with that, even if it's not ideal. In a perfect world, I'd bump the wire down a gauge or two, but 0.5A will just warm your wire a little bit more.

*If* however, your 1mm is with insulation (and your statement that the resistance of the wire is 2 ohms over 6m makes me suspicious), then those wires are way too thin. Doing the maths:

2Ω over 6m
=> 0.33 Ω/m
=> wire gauge of 30
=> maximum power transmission of 0.142A

Even taking into account that the estimates are conservative, that's way way too low. *If* your 1mm is with insulation, that is.

Simon.

 

[TheFishMan65]

Sounds like it is low anyway I am assuming SpacedCowboy that that is 4 CAT at 700ma = 2.8 amps, BUT the board has eight so 8 * .7 = 5.6 amps.

 

[perikaruppan]

The wires are not insulated wires. i am actually planning to drive about 8 CATs, 4 @ 1A and 4 @ 700mA... and when used the calc at the bottom of the link... it for 18AWG/24VDC/18Feet wire for 7A load, there is a 1.6V drop in voltage... so that indicates the drop in voltage and change in lesser drop of voltage when i reduced it to 3 feet.

another problem...

from the power supply to the LED there is a distance of about 12 feet... so, and vertically it is 5 foot atleast... so these are my options:
1. have the SMPS and CAT at the ground level closer to each other use a 18AWG itself, and then take the outputs from the CAT through a 6 foot cable to the LEDs.... those wires are the ones used with printer cables, 16in1 cable... will that work? will be taking two cables per CAT...

2. Have the SMPS at the ground level, and CAT on the LED System with short wires from the CAT to the LEDs and long wire from SMPS to CAT and to minimize the drop of Voltage use a 12AWG wiring? but will be using the same multicable wires from the CAT to the LEDs.

which is better?any suggestions..

i guess i am going off topic with my problem though!!!

 

[SpacedCowboy]

Sounds like it is low anyway I am assuming SpacedCowboy that that is 4 CAT at 700ma = 2.8 amps, BUT the board has eight so 8 * .7 = 5.6 amps.

[editing previous post]

Well, if you're driving at that power I still think you'd be better off with a lower-gauge wire, but if the voltage-drop is ok, I guess it's your call

Simon

 

[perikaruppan]

so if there is a voltage drop in the wires, it is not a problem? apart from lesser efficiency?

Here are the images of the board... single sided...

with the CAT

Without the CATs

Back side with the Holes

The 16 Strand wire that i am planning to use from the CAT to the LEDs.

 

[SpacedCowboy]

Some comments then:

1) If the voltage drop is acceptable to you, then no, I don't *think* there's a problem, you're just paying money to heat wire. It is, however, not something I would do. I'd get larger wires.

I sleep easy at night because I try to stick within guidelines that I know are conservative. Sorry, sounded a bit preachy there, not intended - just trying to say that I don't know much about pushing more than the recommended amount of current through a wire, because I don't do it.

2) Regarding the wires-to-the-LEDs. The best way I know to identify the current-carrying capacity of unknown wires is to basically do what I did above. Measure the resistance over a known length, convert that resistance to ohms/kilometer, look up the closest gauge wire that matches in the table for that link I gave you, and see what current is recommended as a maximum. That'll tell you if you can send 1A over those cables. I suspect you'll be ok (1A isn't too much), but it's better to check.

as for the board, there's a few things I would change:


3) Get rid of the isolation border around the ground pad on the CAT4101 part in your CAD program. All it's doing for you is restricting the heat flow. Not good.

4) I'd make the second-side copper-plated, without it those vias are doing nothing for you. The whole point of the vias is to have a copper path (which is why they need to be plated-through vias) from the top to the bottom of the board. Heat can then travel to the bottom of the board through the (many, to reduce that bottleneck effect) vias. If you don't have copper on the bottom, I don't see the point of the vias, personally.

5) Assuming you can't/won't change to a double-sided board, I personally would spread those CAT4101's out more. The lowest I went was a 1.5" by 1.5" board for a single CAT4101. That board was double-sided and had vias, but the copper on-top was mainly covered, so it effectively only had ~2.25 (1.5 x 1.5) square inches of radiative copper heatsink. It was uncomfortably warm to the touch after running at 0.7A for a week or so. It's hard to get a scale from the photos, but I don't think (even if you do (3) above) you have that much copper per CAT4101.

Simon

 

[kcress]

perikaruppan; Thanks for the info a ways back. SpacedCowboy and TheFIsh are giving you very good advice.


I will add/modify their suggests with these:

1) Ideally you would mount the CAT boards IN THE FIXTURES not remotely. That would allow you to use short wires from the CATs to the LEDs. You should alter your system design to accomplish this or you will have lots of problems.

2) Then. The wire from the supply to the cat boards should be 1.2 to 1.3mm (the copper part).

3) Yes you screwed up the layout. You should not have "heat isolation", (the thin little connectors), on the pads under the CAT chip. Both sides of the board should have large ground areas around the CATs that have no isolation what-so-ever.

4) You must use double sided board for both this reason and to get the maximum ground to each chip.

5) Using your board with the CATs-on-the-bottom orientation is the worst possible. The heat trying to convect UP off the CATs can't since it's under the board instead of on top of the board.

6) Ideally your supply would be within a foot of the boards. If you don't have it that close you have to do additional things.

7) "Additional things" means you have to add bulk capacitance onto your board so small instantaneous current spikes caused by the CATs can be supplied. Otherwise for them it's like they're rowing a boat with oars that have a floppy rubber section in the middle.


Fear not though. We can salvage what you have.

8) To successfully salvage your board do the following.

9) You need to add bulk capacitance. I'd suggest a 220uF 50 or 63V Aluminum Electrolytic Capacitor. They're polarized so make sure you put the (-) side to ground.

Just scratch off the solder mask, (if there is any), and solder the leads to the copper.

10) Using the same process solder wires everywhere I have them in blue. This is because your grounds are sadly under capacity.

11) For the cooling of the CATs you can completely avoid new boards by doing this. It will actually be better than the boards would be anyway. I've drawn it on one of the CATs.

a) Get some brass shim stock. It should be very easy to find in your area. Try to get some that's like stiff paper. Cut it about the size I show. You could actually go a little larger too. You could use super thin copper sheet too if that's easier to find. Brass shim is used to adjust bearing clearances. Hardware and automotive places may have it.

b) Once you have it cut, tin it with solder, just in the middle of where the pad will be. Just the center! Just a little blob.

c) Then set it as shown with the solder blob down and heat it from the top with a powerful soldering iron. It will be soldered down to the pad now.

d) Now drop the CAT on top and solder it as normal. Do the pins first. Then solder the big pad down to the brass. You will need the same powerful soldering iron because the brass is an effective heatsink.

e) Now, where I show the dotted lines, bend the brass up in the air around the sides of the CAT. The end result will be two nice tall heat sink fins that will dump a tremendous amount of heat off the CATs.

I know this will be a little work but it should fix all your issues with what you have already.

 

[perikaruppan]

thanks a ton for all your help...

spacedcowboy:
1. voltage drop is not acceptable, will go with thicker wires..
2. Will measure with the wire and determine the capacity.. thanks again for this info.
3. i am thinking of doing a Double Sided, with just the ground on the Other side... as doing double sided boards here are difficult.. but if it is just going to be copper on the other side, i dont think it will be a problem... and can have the vias connected that way..

Kress:
1. Planning to mount them on the fixture only...
2. Will go for more safety as spacedcowboy said, will do a 1.5 to 2mm thick wire from SMPS to CAT Board.
3. Well only after getting advice from experts like you all, we know that we have screwed up!!! and thanks for that eye-opener.
6. Supply will be at the ground level, and LED fixture at about 6' height from ground taking into the stand, tank ht into consideration... so the wire from SMPS to LED will be about 6-7 foot long... for which i will be using thicker wires as indicated by you..

7. Will add the 220uF as indicated by you...
8. If i do a double sided board, with just copper on the other side, with these via's connected to them, including the ground via's from front to the back, making the other side ENTIRELY ground... instead of wiring them as indicated by you... will the cooling be alright?
9. If that will not be alright ,or will be OK... can i just add a normal heatsink to the CAT's TOP as suggested in some posts?

 

[kcress]

Sorry perikaruppan; I missed your response, until just now. I wondered why you'd never gotten back and came looking.

I know of no commercial heatsinks that you can add to that package style. So my suggestion is all I can offer. With the exception of, adding a fan would let you keep what you have as is. You'd definitely want to make sure the air was forced across the boards. Something like a box with the board in it and the fan on one end blowing thru the box. Or, your new proposed board design.

If you do a new board:
1) Cover everything mentioned.

2) Add that big cap.

3) Make that one + power supply trace double sided.

 

[perikaruppan]

thanks a lot kress.. currently to run my 48 LED panel... i just got some 1" heatsinks and stuck it using Artic Silver onto the top of the CAT to keep it normal..
and as u suggested made a box... fixed a 3" fan on one side and slotted holes on the other side... and the Heat sink is aligned in such a way that the air flows through the heatsink...

since my CATs are pretty close to each other, the heat sinks touch each other, and been running the panel for about an hour and the heat sinks are warm but not hot!!!.. so i think that is holding it...

I will re-do the PCB totally with all that you have mentioned... i was planning to make the + and GND of the Supply double sided to serve better...

But i still find a problem...
When i increase the voltage there i increase in the Current when measured across the XPGs....but not the XPEs...

Just a small comment on that
1. use 2 seperate boards to drive the XPGs and XPES, the resistors on the XPE Boards are all 768 Ohms, whereas in the XPG it is 3 561 Ohms and 1 No 768 Ohms, there is a mix of XPG and XPE's in this board due to the ratio order of XPG:XPES (2:3)...

Will that play a factor in not-limiting the current?

Regards and again thanks a lot for your advice...

 

[perikaruppan]

Finally checked it again and again.. for the XPGS the Current is at .99A (Should be 0.98A for the resistor) and for the XPEs it is 0.71A (should be 0.7A)... and any increase in voltage upto 26 i hit.. the Amps went upto 1.03 and 0.72A... so i guess the CAT is regulating it properly now after all your advice!!! and cooling of the system...

from this, without the cooling the CAT was performing randomly... and with the cooling it is working alright i suppose!!!....

i hope i am right...

thanks again a lot for your advice kress,spacecowboy and fishman..

 

[kcress]

Sounds like you've got it now.

Those small changes with voltage are pretty typical. There's even a spec for that. "Supply regulation".

Thanks for the results.

 

[perikaruppan]

thank you again...
but when i tried doing two cats in Parallel, that is, pin 5 (Output) of two cats were connected to gether to get double the current, but that didnt happen... can you tell me why?

I am doing tooo many experimentations....

The 700mA board reaches saturation at a very less voltage about 22V, but the 1A board takes upto 23.5V to reach 1A and remain constant.. and since both the boards are operated by the same SMPS i am supplying 23.5V to the 700mA board also so more is getting burnt in that board...

so thought of doing all with 600mA boards, and for the XPGs connect 2 in parallel to get 1.2A and XPE blues will run at 600mA... so lesser voltage, lesser loss and better efficency...
and when i started to do that, the current didnt double!!... can you tell me why?