Looking for a specific DC CC driver for LEDs

[karimwassef]

The last quote sums up the risk:



The switchers are not synchronized (there is no way to sync multiple LDDs), and load sharing will not be split very evenly, but it is possible to run in this configuration. Its also likely an EMI mess but thats a given for most of the long cable run LED builds here :spin2:

Also the output resistor is just for a model and need not be included. Depending on the LDD construction the diode may also be optional (haven't really looked). I'd suggest a 2A+ Schottky diode for each LDD, rated for at least 50V.

Yes. I was planning 5A Schottky rated for 100V without a resistor. As far as sync, the dimming option is PWM driven. If I use the same PWM signal for all three, it should drive to the same driving clock (this is a guess). As far as ripple and EMI, It's a question of input and output capacitance. That could impact loop stability, I know, but a little capacitance should clean up any induced ripple?

I guess I could use 4 ($20 per LED), and that would give me up to 3.5A with margin.

 

[karimwassef]

your bulbs..but if is anything like (and it probably is) paralleling power supplies.. the risks seem to outweigh the rewards.. Besides as shown, "proper" drivers can be built.. for fairly cheap..and relatively simple.

I agree. The $20 estimate x 24 = $480 for the meanwell LDD-1000H x 4 with diodes. That's the baseline.

What do you think the total BOM cost would be to assemble 24 drives at 100W each on a single board?

The input would be a bulk 36V DC source rated at 2.7KW max pushing 75A. The output is 24 DC pairs of wires carrying 3.5A each at 34V DC. For control: 24 individual 0-10V (or PWM) input channels.

 

[theatrus]

Yes. I was planning 5A Schottky rated for 100V without a resistor. As far as sync, the dimming option is PWM driven. If I use the same PWM signal for all three, it should drive to the same driving clock (this is a guess). As far as ripple and EMI, It's a question of input and output capacitance. That could impact loop stability, I know, but a little capacitance should clean up any induced ripple?



I guess I could use 4 ($20 per LED), and that would give me up to 3.5A with margin.

The PWM signal will be synced, but the switchers will not be synced. They have an internal oscillator and don't have any relation to the PWM signal (except to be likely latched to the next cycle on turn on/off).

Proper high frequency decoupling will help prevent out of control ripple (not a standard electrolytic). Hand waving a start point, maybe a 2-10uF X7R ceramic.

Will it work? Yeah probably. Do some tests with a dummy load first and look for peaks with a scope under various scenarios. The upside of a current control scheme is if all four units go weird and don't limit the current per cycle correctly, your average max current in theory should be 4A.

Is it cheaper? Unsure - the LDD units are a bargain but there is a lot of duplication of parts.

 

[oreo57]

your asking for a custom board.. but there is no reason not to breadboard.. or "old tech" assemble them
Assuming theatris is correct, some "off spec" building of an Allego based driver would be on the line of a few $'s..

AFAICT there is no part over $1... and very few parts..

As to the $6 chips and asst. hardware.. well lots of parts but nothing too expensive..
I don't build these things for work or as sport so don't have my finger on pricing.. but as I said, I'd be surprised if the breadboard assembled drivers would be over $20 each..

 

[theatrus]

Don't assemble a switch mode, especially a few amps worth, on veroboard. Low impedance paths in the switch nodes and proper thermals are required.

Thankfully boards are cheap these days (dirtypcb, even OSHPark)

 

[theatrus]

I put a basic LM3409, 4A output design into TI's Webbench, which gave roughly this BOM.

The cost for all 24 channels (with no spares, no overage, no PCB, and about $5 total (not per board) missing in popcorn parts like tiny caps and setting resistors) runs to $168 (subbing in Samsung MLCC caps as a cost savings, though one is X5R, oops):

Also worth looking at Coilcraft for alternate inductors.

 

[theatrus]

A pack of 30 7x7cm (should be enough space to mount connectors, though at this stage you might consider multi-up boards with a PWM driver like a PCA9685 on board) DirtyPCBs boards, with DHL shipping, would set you back $90 total (once Chinese new year is over)

 

[theatrus]

And also need 2x the input and output caps - adjust cost accordingly. (+$15)

All in all, thats about $11-12 a channel not counting time, errors, connectors, wires, headaches, waiting, ...

 

[karimwassef]

Ok, so the simple LDD parallel option with diodes is $480.
The custom board is ~$240 after a lot of time and effort.

So the difference is $240 savings for about 8 weeks of work? Hmmm - I need modules I have a job already.

That is a little selfish though because once it's designed and proven in, all other DIY reef keepers would benefit from the work done.

I guess I'd say that if you can makem, I'd buyem for $16 a piece

With the growth of multichip LEDs, you might have a grass roots business here!

 

[theatrus]

Spent about 30 minutes mocking up a base layout for this:

Footprint is about 2x1.5in, sans any external connectors.

 

[O2Surplus]

I put a basic LM3409, 4A output design into TI's Webbench, which gave roughly this BOM.

The cost for all 24 channels (with no spares, no overage, no PCB, and about $5 total (not per board) missing in popcorn parts like tiny caps and setting resistors) runs to $168 (subbing in Samsung MLCC caps as a cost savings, though one is X5R, oops):

Also worth looking at Coilcraft for alternate inductors.

Please correct me if I'm wrong, but that schottky diode needs to be rated for far more than 1 amp. Also look for a MOSFET with the lowest RDS that you can find. 4 amps is serious current and that MOSFET will run really hot if it's not spec'd correctly. I' be built quite a few LED drivers using the LM3409. Saw a couple of them literally re-melt the solder at the MOSFET. LoL. The driver continued to function just fine, as long as the PCB was level. If not level...the MOSFET would literally slide or fall off the board.

 

[theatrus]

The figures felt a little low to me as well - I'm re-running the numbers in the datasheet to find the flyback current.

As for the switching FET, current one is 14mOhm of RDSon, which is *less* than the current sense shunt .

More importantly is the gate charge of the FET and gate drive capability of the LM3409 (usually lower RDS on means higher gate charge). This can severely limit the speed of the switch and can lead to massive power dissipation.

 

[theatrus]

Recirculating diode recalculated and over-specced to a V8P10: http://www.vishay.com/docs/89005/v8p10.pdf

 

[theatrus]

Its pretty much ready for a test coupon board - after I sleep on it and decide if I find more bugs or not

 

[oreo57]

Way Off topic but related:
I need a suggestion on a driver for high voltage, low current LED strings..

i.e 90V DC 120mA.

Just a suitable driver chip would be good enough.

NOTE: starting w/ 110V line current.

 

[theatrus]

Way Off topic but related:
I need a suggestion on a driver for high voltage, low current LED strings..

i.e 90V DC 120mA.

Just a suitable driver chip would be good enough.

NOTE: starting w/ 110V line current.

Do you already have a power supply sourced? 90V is in the "preferred" territory of at least an isolated flyback converter. The high ratios for a direct buck/boost converter here are rough on efficiency.

I've run an LTC1871 up to about 170V (at only 15-20mA output though) from a 12V input, if you have a 24-30V input and are targeting 90V you could probably get 120mA to work from that combo. Current control however will be pretty poor with that setup (its not a true constant current control scheme, just a current limit per cycle)

Perhaps this will be to your liking:
LT3796: http://www.linear.com/product/LT3796

Voltage disclaimer: do not taunt high voltages. 90V is more than enough to cause lots of problems, and will require special board layout and part selection considerations.

 

[oreo57]

Well technically like these (direct line drive w/ bridge ect):
but a "wee bit" higher output voltage..and lower mA's..
http://www.alibaba.com/product-deta...34_75a99f03239b8771aa63c84b820903c6&pid=85386

note: some of these "tube drivers" do list current as "settable"..
Most top out at 80V out though.

BTW: This isn't my project.. I'd never bother w/ such a high high V(f)
My biggest concern is V(f) approaching line voltage.. regardless of type (AC/DC)
Technically they could be run directly off line voltage w/ a Bridge rect. and "cleaning" circuit..and set more to constant voltage than current..Inelegant though.

Will look at the chip .....HEY...

Sorry o/p will free this up in a bit.

 

[theatrus]

Based on the layout that looks like a high voltage output flyback converter, just with a much higher voltage than is typical. Would work for this case, though I'm very hesitant to actually plug anything like that into mains

You can also use a 1:1 isolation transformer + rectifier + caps for an unregulated 141V DC. Don't suggest it.

 

[kapelan]

Some questions:
1. 30 drivers each of them 3A and 36V would be 3000W
This power is enough to burn out any tank. If it is developed to multiple tanks it would make sense to split it by 3- 4 drivers.
2. If all drivers on the same PCB the current would be for each driver 3A. Total for 30 drivers it will come to 100A. It cannot be done from any PCB. It is different technology. Either it should be separated or find some cool PCB developer that can do 2-5mm sick roads.

 

[theatrus]

Some questions:
1. 30 drivers each of them 3A and 36V would be 3000W
This power is enough to burn out any tank. If it is developed to multiple tanks it would make sense to split it by 3- 4 drivers.
2. If all drivers on the same PCB the current would be for each driver 3A. Total for 30 drivers it will come to 100A. It cannot be done from any PCB. It is different technology. Either it should be separated or find some cool PCB developer that can do 2-5mm sick roads.

I don't think a single PCB solution is in the works here (for the obvious problems you mentioned), but yeah I'm not sure what I'd do with 3kW of lighting either. It looks to be significantly bigger than my 90G for sure