DIY LED driver for reef lighting

[TheFishMan65]

I talked with an EE. Yes my idea will work, but THE ABOVE SCHEMATIC IS WRONG. The 1K potentiometer should have the wiper to the adjust pin (through jumper if needed) and the the currently unconnected pin should be connected to the resistors the wiper is currently connected to.

I will update the schematic if anyone is interested, but I have seen no comments?

 

[der_wille_zur_macht]

TheFishMan, can you email me your eagle files? I'm curious to look at the schematic on a larger scale. I must need new glasses, because I'm having trouble looking at it in your screenshot.

 

[TheFishMan65]

Updated Schematic

Updated Schematic

I got a minute to update the schematic. Hopefully it is easier to read.

What parts are people really interested in?
LED Banks
I allowed separate addressing of each bank. This would allow one bank only for moon lighting. Or you could brighten your tank east to west. Is it important? Or do we only need 2 channels blue and white?
I think the board should probably support 4 banks. DZWM I think you did 3 - was that just because of the real estate issue?
Dimming Modes
As I mentioned earlier this supports almost all dimming modes I have hear about. Which ones do people really want?
Price
Prices are from digikey.
The big price is the 1366 at $4.38.
The MCP4725 (D2A) is $1.19.
The MCP23008 (I/O expander) is $1.22.
I think the LM7805 (voltage regulator) is $0.43. Couple different versions?
I would leave off the headers and solder a wire where needed.

DZWM I hope you don't feel like I am kidnapping your thread. If you want I can start a new one.

 

[stugray]

I see a few oddities in that schem:

1 - The A0 pin on the MCP4725 goes to a node that seems to be floating.
The GP0-7 on the expander pins appear to have a jumper that can connect to this node or is that just a header connection to plug in an IO cable?

2 - The Hardware address input on the MCP go through nice jumpers, but then your net on the other side ties them all to 3 parallel pull-ups instead of three independent ones.

3 - you might want some kind of input isolating resistor on that PWM input between the jumper & the 2N706.


I didnt look 100% through your input logic for the various different inputs yet....

Otherwise it is an interesting alternative with I2C built in

Stu

 

[der_wille_zur_macht]

No worries about hijacking. The more ideas the better!

I'm not sure I'm following all the stuff you have on there for the different dimming modes, so I'm gonna take a look at the Eagle files you sent before commenting any more. I understand you want analog, PWM, and I2C all selectable as dimming options, but what's the port expander (MCP2008) on there for, with all it's outputs tied together to the address pin on the DAC chip?

One thought though, if you're going to have that 7805 on there to create +5v, you probably want some decoupling caps on it.

 

[TheFishMan65]

OK, I threw it together kind of quick

OK, I threw it together kind of quick

Stugray,
1) The port extender is used to select the D2A, The D2A only has on address line. Let's say you want four. They are all at address one plus the base adress of the chip. Use the port extender to connect to 4 different D2As. Then select one or more that you want to write from the port extender. Then write the D2A with IIC bus. The reason everything appears to be floating is that is a 2x8 header, put a jumper across a set of pins to select the address. The port extender could actually support 8 D2As, but I think 8 banks of LED would be too many. I am thinkg 4 as I menioned above.

2) Your right this is wrong. The jumpers are there to allow you to change the address if you have multiple port extenders. The pull ups are on the wrong side of the jumper. Each address should default to pulled up so the address is base + 7. The jumpers can then be used to pull address lines down and get all 8 combinations.

3) You might be right and it may not hurt. However I took that idea from page 21 of the data sheet. It depends on the exact circuit driving it (they show 3). That was sort of a quick demo of how it might be added.

DSWM
The whole idea is a resistor network. If I have a 3K resistor tied to 5 volts then the 1k pot to ground. If the pot is turned to one end the wiper (tied to adjust pin) is at ground and no output. IF we turn it to the other end we have 5 volts dropped across 4k ohms. So now the wiper is t 1.25 volts and we get 100% currents. All the other modes use different resistors for different voltages such that the upper end of the pot is at 1.25 volts.

In order to have microprocessor IIC functionality I replaced the 5V with the output of the D2A. This means the processors has full control of the current, within the range set by the pot. So the processor could be written to go 100%, but the resistor could actually limit to 50%. Advantage are we can test without burning out the LEDs or blinding ourselves. It is also possible to DIM the light if you get a new coral without completely taking it apart or changing the code, or using an interface. Those less electrically inclined may like this option.

The port expander is strictly used for allowing multiple banks of LEDs to be addressed indepently,

It probably needs more than a capacitor. I was trying to get the idea out there to see if people thought it was worth pursing.

 

[OTCHU]

Just did take a very fast look at your schematic .... you don't think that the 24V will not hurt the 7805 ?? Should be able to take it but like I know this regulator, he will need a pretty good size heat sink, no ?

 

[TheFishMan65]

Not sure either, but I don't think so it is not supplying a lot of current. I haven't added up the current draw yet so I am not sure what it will be dropping.
[EDIT]
If I am reading the data sheets right
4725 draws max of 24mA
23008 1mA
4 resistor drop 1.6mA * 4
so total of 31.6 mA say 100 mA
Power = .1 * .1 * 19 = .19 watts out of a worst cast 19 watts (19 volts at 1 amp)

 

[Drake1]

der_wille_zur_macht just curious how the larger board is going, did you get a test board made yet? what did you decide on the 50v cap
jeff

 

[newhobby1]

EDIT:
Linking to other forums is not permitted, and posting to advertise a product you are selling will result in either a restricted account or complete loss of posting privileges.

 

[kimimaro_kiddo]

Hi der_wille_zur_macht,

is your dual driver homebrew tested and running? how was it? any updates?

 

[der_wille_zur_macht]

der_wille_zur_macht just curious how the larger board is going, did you get a test board made yet? what did you decide on the 50v cap
jeff

Hi der_wille_zur_macht,

is your dual driver homebrew tested and running? how was it? any updates?

Other than the prototype Simon cut me, and the 7 or 8 one-off NCP3066's I've built, I don't have either up and running yet. The board house I was going to use (seeedstudio) is on a 3-week-long Chinese New Year vacation, so that's stalled things. Their prices are just *so* good I'm going to wait vs. go somewhere else.

I'll post an update as soon as anything happens but sadly it looks like it'll be a few weeks. I might try a one-off of the ZXLD1366 just to pass the time (as if I didn't have enough else going on :lol: )

 

[SpacedCowboy]

Other than the prototype Simon cut me, and the 7 or 8 one-off NCP3066's I've built, I don't have either up and running yet.

I've not got anything running either - I'm in the middle of building my 3D printer at the moment.

Given that I'll be able to fabricate my own plastic parts by drawing them on the computer, I have an idea for how to attach the LEDs without any soldering onto the LEDs themselves It'll be just like lego

Simon.

 

[va7zt]

Hi All,

I have 12 of the 15 drivers I need all done now. So far they have all worked great. I had one do some sizzling, snap and a puff of smoke yesterday but it ran fine after that for over an hour so I think it may have been a little loose piece of solder. I am still waiting on the LED's from the far east with no idea when they may come in (ordered in Nov but still out of stock!). Once I get the other 3 drivers built, I will mount all the blues on the heatsink and start wiring then make a decision on the outstanding order of LED's.

Any suggestions on soldering these little guys? So far I have just soldered 8 together without breaking them off the mould they come on. They have not been easy to get wire to stick to the pads because the heatsinks work so well. What are others using for a soldering iron? I'm using a 25W but I am thinking it won't be enough after they have been mounted to the heatsink.

One interesting thing though, so far after building each driver, I have hooked them up for at least an hour. I'm running them at 700ma. Even at that draw, they barely get warm. Makes me wonder if there really is a point to having a 55lb heatsink to attach them all to. From what I have seen these heatsinks are massive overkill. At least at the draw on mine.

Don

 

[der_wille_zur_macht]

After some head scratching and number crunching, I'm probably going to abaondon the ZXLD1366-based driver. Even with penny pinching and running half a dozen scenarios (different voltages, different components, no dimming, etc.) it always costs more per LED than the NCP3066 I started this thread about, so I'm going to stick with that - running 8 LEDs per IC, on 24v, in the dual-driver format. FWIW, I was including components, PCBs, connectors, and power supplies in the calculations. The NCP3066 design I've already posted is around $1.10 per LED. The BEST I could get the ZXLD1366 was more like $1.50 per LED. 40 cents per LED might not seem like a lot, but I'm gonna have ~200 LEDs on my tank and I don't see that the ZXLD1366 offers enough advantages to justify spending $80 more on it.

If anyone wants the EAGLE files for my ZXLD1366 designs lemme know and I'll put them on the google code site.

As soon as the board house is back from vacation I'll order up a batch of the 3066 dual drivers, so stay tuned.

 

[SpacedCowboy]

If anyone wants the EAGLE files for my ZXLD1366 designs lemme know and I'll put them on the google code site.

If you wouldn't mind putting them up, I might still be interested - purely for space reasons

Cheers,
Simon

 

[TheFishMan65]

SpacedCowboy,
I am going to continue on a design using the ZXLD1366. If you (or anyone else) are interested in coconspiring to do a design I welcome the input.

 

[der_wille_zur_macht]

I uploaded my "most economical" version of the ZXLD1366 driver, which was three of them on a board with common dimming. Also uploaded a BOM with digikey part numbers and prices. I'm still interested in following/supporting the effort if others want to use it, it's just that I won't be using them on my big tank.

Also uploaded a new version of the Eagle library, since I've added some odds and ends (including the ZXLD1366 itself).

Got a bunch more PMs about the PCBs for the NCP chip dual driver - the board house is on vacation for another week or two. As soon as I have an order in I'll make a note and go back through my PMs contacting everyone who has expressed interest, so hang tight until then.

URL for the downloads:

http://code.google.com/p/hpled/downloads/list

 

[SpacedCowboy]

I uploaded my "most economical" version of the ZXLD1366 driver, which was three of them on a board with common dimming. Also uploaded a BOM with digikey part numbers and prices. I'm still interested in following/supporting the effort if others want to use it, it's just that I won't be using them on my big tank.

Ok, I had a look at that

The resistor choice puzzles me slightly - you're going for 0.18R, which (as far as I can see) puts 1.1A through the LEDs in the failure mode of PWM = On constantly, even without any ripple effects or resistor variances, that's over the limit.

When I was looking at it, I was going to use 0.27R (=0.75A through the LEDs). There's a 0.2R resistor (RHM.20SDKR-ND for 1A nominal current) which is $0.203 in 10's, or a 0.27R resistor (RHM.27SDKR-ND) for the same. Either are cheaper than the 0.18R resistor in the BOM.

You can also get the ZXLD1366 cheaper at Mouser - it's $3.35, not $4.38.

That brings your total down to $18.29 from $21.84. Given that you're using a 10uF cap, I'm assuming 12 LEDs/circuit, so the cost/LED drops by about 10 cents. Still perhaps not enough to compare to the NCP3066 version, but it's compact size appeals to me

Looking at a total cost breakdown, it's still more expensive:

- components for three drivers = $18.29
- 3.3A, 48V PSU (285-1816-ND) = $35.48
- 2"x2" PCB, silkscreen, soldermask = $1.92
- Total of $1.54/ LED

... although (for my purposes) I'd be more tempted to use a 10A 48V PSU from MPJA (16756 PS). Slightly cheaper for the 10x 12-LED circuits I'll need, and it's enclosed and has its own fan etc.

Simon

 

[der_wille_zur_macht]

Simon,

Value of .18R was calculated from the design spreadsheet tool available on the product page at the manufacturer. What value are you using for ADJ pin voltage? If ADJ is floating (full on signal from PWM) it'll be 1.25v, and .18R gives an average LED current of ~525mA. At the max of 2.5v, your calcs would be correct.

Stop trying to win me back with your cost analysis. :lol: I need to just make up my mind and go one way or the other here. To top it all off, I just spent my lunch break reading about home-made reflow ovens. :lol: As if I need another way to complicate this!