DIY LED driver for reef lighting
- Thread starter der_wille_zur_macht
- Start date
der_wille_zur_macht
Team RC
Baghyyy,
There are two spreadsheets already - I posted a link to one of them on the last page. It is a modified version of a spreadsheet Terahz gave me, so you could also ask him if you can have his.
At the surface you can just change Rsense to get whatever current you want, but IMHO it is important to do an analysis from scratch for your exact conditions to get optimize other important components (inductor, Coff, Cin, Cout).
There are two spreadsheets already - I posted a link to one of them on the last page. It is a modified version of a spreadsheet Terahz gave me, so you could also ask him if you can have his.
At the surface you can just change Rsense to get whatever current you want, but IMHO it is important to do an analysis from scratch for your exact conditions to get optimize other important components (inductor, Coff, Cin, Cout).
I have one spreadsheet.
you can give me the other?
I can please explain how this spreadsheet? thanks ...
I have 48v power supply
4 LED 12v 900mA
What should I change the sheet to see how it works?
thanks...
terahz
1x10^12 Hz
Good news. Problem for analog dimming is solved. All that it needed is 2 resistors to be changed couple of orders of magnetite to reduce the impedance to IADJ. After I got it to work, I noticed that the range is not at 100% (more like 95%) but that's not the end of the world. I'll test a few different values and update the BOM when I find a good combination, but for now this is still pretty good. BOM is now updated with the working values.
EDIT: also, Mouser now carries the LM3409HV so all parts can be ordered from there, however Future is still almost 1/2 the price, so if you're ordering more than 2 sets, it will be cheaper to order the ICs from future.
EDIT: also, Mouser now carries the LM3409HV so all parts can be ordered from there, however Future is still almost 1/2 the price, so if you're ordering more than 2 sets, it will be cheaper to order the ICs from future.
ghellin
New member
I am running the 8 cat 4101 driver board using 2 24v 6.5A PSUs from web-tronics and the lowest voltage I can get it 27v and up to 31 volts by adjusting the pot. When I bought the PSUs it said 24V +/-10% I cant even get close to 24v any ideas. I know a new power supply is the easiest option but its not a cheap one. Any help would be great!
der_wille_zur_macht
Team RC
New PSUs? If those don't meet the promised spec I would return them and get my money back.
Or, switch to LM3409's.
Or, switch to LM3409's.
Krakonosh
New member
I looked at Web-tronics web site and it's interesting that the specs don't have the actual output voltage specified, it's only in the name of the product.
I work in telecom industry and it looks like this PS was meant to be used in 24V battery systems with 27V float voltage, just like when we refer to 48V systems that actually run at 54V.
I guess this is not really the answer you were looking for, but if you need to drop few volts you'll have to put few beefy diodes in line or buy better PS.
This PS is not exactly hi end device.
Look at Meanwell SDR-240-24.
I'm planning to use these in 48V version in my upcoming build.
Cheers.
Check it out- Just soldered it up this morning. It's my version of the LM3409 based Led driver. It's set up for PWM dimming & 28 volt/2000 mA operation.This particular design also includes a 5 volt reg for providing power to a Nueventics "SynJet" led cooling module. I designed this driver to power the big BridgeLux BXRA-C4500- 50 watt emitters, but I suppose it will also work with the Cree XML's. I haven't powered it up yet, but I'll keep ya'll posted as to what happens during the first "smoke test".
der_wille_zur_macht
Team RC
This is the power supply I am using, in 24v format. PFC, relatively efficient, powerful. I snagged several of them on eBay last summer. They pop up from time to time on there for about $50 each which I think is a steal given their features.
mpja always has a wide selection of power supplies:
http://www.mpja.com/Single-Output-Supplies/products/3/
The potrans 24v units from them have been used by many, many people.
mpja always has a wide selection of power supplies:
http://www.mpja.com/Single-Output-Supplies/products/3/
The potrans 24v units from them have been used by many, many people.
der_wille_zur_macht
Team RC
Looks like an interesting design. No Cout for the LEDs?
What sort of power draw does the SynJet have? It looks like you're using a 7805 to drop 28v down to 5v? Unless it's only pulling a few mA that thing is gonna get HOT!
I flat out forgot to include it in the design. I'll take some measurements using this board and will add the output cap on the next board run, if needed. The 5 volt reg is rated up to 35 volts and has a fairly low drop out voltage. I've used them before with some in some other circuits with similar voltage/current demands and they always ran barely warm to the touch. The "Synjet" only pulls 150mA so that's negligible. It also reacts to the same PWM control signal that will be usd to control the LM3409. So in effect the "Synjets" will speed up/ slown down, following the intensity of the leds. I'll report my findings later - either it will be a success or a lot of smoke. We'll see tonight!
kcress
New member
If you're willing to screw with the supply you can probably get it down where you want it. That pot is usually in series with a resistor. The resistor is used to get the center of the POT to the nominal voltage. In this case they've probably set the "nominal" just as Krakonosh suggested - high for battery charging of a specific product. That could be why they're 'surplus'. I'd figure out what resistor that is then measure the output voltage, then without changing anything I'd add a high value resistor in parallel to that 'setting resistor'. Pick one like 47k (47,000) ohms or 100k. You don't want the output to change dramatically as that could damage something. You just want to see a small shift so you know which direction you need to change the original resistor's value.
If adding this high parallel value lowers the voltage you've got an easy fix. Just tinker with the parallel value until the voltage has dropped into your zone. If adding the test shunt increased the supplies output then you'd need to raise the value of the existing resistor. That means adding some more by replacing the existing one or putting some resistance in series. If you need to increase it tell us the existing resistor value and the pot value and we can probably figure out the correct value you need.
der_wille_zur_macht
Team RC
What matters here is the device's power rating, not it's max voltage. The minimum drop is not important, because in this application we can calculate that it's dropping 28v - 5v = 23v. At 150mA that's 23 * .15 = 3.45w of power it's dissipating. That sounds like a lot to me, but as you said, time will tell.
ghellin
New member
I do happen to have some 4.7k resistors laying around, would those work I don't have any 47Ks but I can buy some. Also the resistor on the psu is through hole what would
be the easiest way to run two resistors parallel?
Thanks again everyone I am trying to keep costs down wifey is not happy LOL.
kcress
New member
Folks I have to comment on what I'm seeing with a lot of these layouts.
When you go to surface mount the first thought is, "Oh wow! Look how tiny I can make my design." While this is true with non dissipating components like capacitors and resistors not dumping a bunch of power, or logic chips that are always room temperature, it's NOT the case with power elements! In the case of power elements the actual board area needed can be MORE than the same part in thru-hole packages. This is because all the freebie cooling aspects of thru-hole parts like leads surrounded by air and a tab sticking up that can radiate and convect are now missing with the surface mount parts.
Any surface mount part that is possibly dumping a bunch of power needs to be on a large contiguous piece of board so the board can be the heatsink. A part like the 7805 on the board above should have an area larger than a postage stamp to cool itself with. I would have a space at least as large as the package body surrounding that part on the left right and top. I would also mirror that same area on the back of the board. To complete the proper layout you then put in about 8 LARGE vias surrounding the part and close to the body. That is the only way to get the maximum capability out of those parts. It will also pay dividends on life expectancy.
In the above design that rule would apply to the upper left two parts also. Try to keep the surrounding areas the same size as the packages.
I know the urge is to see how many boards you can fit into a Seeeed studio price category or how miniature you can make your design but with power designs that is the ultimate penny wise pound foolish move.
der_wille_zur_macht
Team RC
I suppose this would be a good time to post my version of the design:
I had intended this to have better power dissipation characteristics by having a large area on both sides of the board that the three main power handling components were connected to (lower right) as well as trying to consider the instructions in the circuit layout section of the datasheet.
kcress, any comments?
I had intended this to have better power dissipation characteristics by having a large area on both sides of the board that the three main power handling components were connected to (lower right) as well as trying to consider the instructions in the circuit layout section of the datasheet.
kcress, any comments?
Kress,
Thanks for your input with regard to the proper way of designing power circuits utilizing SMD components. This is my first attempt at building a driver using the LM3409, so I'm bound to make some engineering mistakes along the way. This driver will either work as I envisioned or go up in a disappointing cloud of smoke. No worries- I'll simply keep redesigning it until I get it right. IMHO It's all about learning, and @ $10 per board run, I'd rather waste a few bucks doing this, than blowing it on some "off the shelf product" that does not quite meet my needs. As far as the 7805 goes- I did include a bunch of thermal vias under the package, so its got the whole bottom plane of the board for cooling.
Thanks for your input with regard to the proper way of designing power circuits utilizing SMD components. This is my first attempt at building a driver using the LM3409, so I'm bound to make some engineering mistakes along the way. This driver will either work as I envisioned or go up in a disappointing cloud of smoke. No worries- I'll simply keep redesigning it until I get it right. IMHO It's all about learning, and @ $10 per board run, I'd rather waste a few bucks doing this, than blowing it on some "off the shelf product" that does not quite meet my needs. As far as the 7805 goes- I did include a bunch of thermal vias under the package, so its got the whole bottom plane of the board for cooling.
kcress
New member
DWZM; That looks better than most and is certainly a little better than marginal. I would still mess with your design a bit more because the heat doesn't actually travel well thru the cladding. Because of the geometry the heat will only effectively move about about 1/2~3/4" from the hot part unless the copper is thicker than usual. (hence Alumiclad boards). What this means is that having a large area like the bottom right is not actually as effective as you might think. That's because the remote points are doing little for heat dissipation.
On your board there I'd try to ease the transistor more into that space. OR I'd change the board to be a little wider than it is now to get room on the left of that transistor. You can then reduce the bottom dimension. The result would be a better thermally performing board for no more area.
Of course it always depends on the dissipation. If that board has been run fully loaded for long enough to reach thermal equilibrium and the part was not too hot it's fine as is.
O2: Good for you. Nice philosophy. I often consider my trials as school I'm not paying a specific tuition to.
In your board's case you may find you need a fan moving air smartly across the board to make up for a lack of copper. It depends on the actual dissipation that part will normally have. In a pinch you can go to a hardware store or a gutter company and get a small piece of brass shim stock or copper gutter and make a little heatsink. Cut a little piece the size of a postage stamp and solder it to the tab sticking up off the board. Make your own heatsink as it were.
On your board there I'd try to ease the transistor more into that space. OR I'd change the board to be a little wider than it is now to get room on the left of that transistor. You can then reduce the bottom dimension. The result would be a better thermally performing board for no more area.
Of course it always depends on the dissipation. If that board has been run fully loaded for long enough to reach thermal equilibrium and the part was not too hot it's fine as is.
O2: Good for you. Nice philosophy. I often consider my trials as school I'm not paying a specific tuition to.
In your board's case you may find you need a fan moving air smartly across the board to make up for a lack of copper. It depends on the actual dissipation that part will normally have. In a pinch you can go to a hardware store or a gutter company and get a small piece of brass shim stock or copper gutter and make a little heatsink. Cut a little piece the size of a postage stamp and solder it to the tab sticking up off the board. Make your own heatsink as it were.
kcress
New member
You want a resistor that is about 10~15x the one on the board now. That's why I can tell you outright, I don't know what it is.
You can parallel it it with alligator jumpers or tack the leads to it with solder.
Permanently paralleling it can be just tacking on the parallel resistor to the existing leads.
