I posted some brief details of my LED project a few months ago (http://www.ukaps.org/forum/threads/does-new-lighting-have-to-be-expensive.36226/#post-389067) and have finally found some time for a more detailed write up.
I needed some lighting for an old tank that I've put back in service. The tank and hood have been through the mills over the years; I worked out that they were bought in September 1990 - so not far off 25 years old🙂. It had been stored in my dad's garage for the last 10 years or more, so I was a little nervous when I first filled it... but so far, so good. The tank is 3feet long, 18inches deep and 12inches front to back.
The plan was for some decent, high quality (i.e. efficient) LED lighting that was cool running (without fans), dimmable, light weight (I didn't want a kilo of heat sink hanging on the underside of the hood opening) and, most importantly, not too expensive. I wasn't too bothered about the overall brightness levels - was just aiming for good enough!
I also decided early on that this would be part of a larger diy tank controller project, so the timer/dimming would be controlled by a raspberry pi running some software that I wrote myself (I'll cover the controller in a separate thread).
Anyone who has ever looked at a DIY LED system will tell you that there are 1001 ways to build them, so rather than bore everyone with the various options I considered over a number of weeks, I'll just cut to the chase.
I opted for 10 Cree XM-L2 cool white LEDs, pre-mounted on aluminium stars. £35 from eBay (http://www.ebay.co.uk/itm/10W-5W-3W...t=LH_DefaultDomain_3&var=&hash=item8032929890).
No lenses needed as these will only be a couple of inches about the surface of the water.
(one led already removed for testing)
Running at 1amp, these in theory (i.e. per the specs) consume 3watts each while generating between 412 and 468lumens of light output (depending on the temperature). So a 10 LED array should produce over 4000 lumens for 30watts - or somewhere between 130 and 150 lumens per watt, which is pretty efficient.
I couldn't find any specifics but estimate that these LEDs are somewhere in the region of 50% efficient - so would expect them to be producing roughly 1.5watts of heat energy each, which is a factor to consider when it comes to heatsinks...
I explored many options for heatsinks, even considering plain aluminium bar, but finally opted for purpose built Ohmite individual LED heatsinks (http://uk.farnell.com/ohmite/sa-led-113e/heatsink-led-12-7mm-black/dp/2097677?CMP=i-bf9f-00001000).
These actually worked out to be quite economical (£27 for 10 delivered) and are ready made for mounting the LEDs and mounting to a supporting structure. They have a rated thermal resistance of 5.33°C/W - so in theory the roughly 1.5 watts of heat being pumped into them at 100% output should generate a temperature rise of roughly 8°C above ambient - I was expecting at least double that though, as the hood will be enclosed with little air flow, and had planned for anything up to 20°C rise - which would keep the heatsinks at or below 50°C even on the warmest of summer days.
The heatsinks' LED mounting holes are designed for M3 threads (or whatever the US equivalent is), so you will either need to use self-tapping screws (not easy to find on the 'highstreet'), bolt through with M2 bolts (not recommended) or tap for M3 bolts. I had a mixture of stainless steel and nylon M3 bolts in my bits box, as well as some nylon M3 washers and, most importantly, some M3 taps - so an hour or so of tedium later, they were ready for mounting.
Be careful if tapping the threads yourself; as the hole is 'open' on one side, the tap wants to take the path of least resistance and break out, leading to a wonky thread. Take your time and apply even but firm pressure - I messed a couple up (but there are 'spare' holes). Also make sure you change taps if one goes blunt and clean up the hole and thread when finished (i.e. remove burs that might make a mess of mounting the led).
Actually attaching the LEDs is then pretty straightforward. I had plenty of CPU thermal paste lying around but it is readily available. Be warned that most thermal pastes are toxic to both fish and invertebrates - so don't go nuts with it; you only need the tiniest blob per led which should just squeeze out the side of the LED star when it is screwed down (there are some horror shows on youtube that demonstrate how not to do it 🙁).
Don't forget to use nylon washers if you are using metal screws or nuts. Shorting the power to the heatsink may not be fatal to your build - but it may take some time to figure out the problem.
I wired up one LED and heatsink for testing. As reported elsewhere, take care soldering the wires to the LEDs. I soldered them after attaching the LEDs to the heatsinks but I did 'pre-tin' the contacts to make it easier and I was using a fairly powerful soldering iron. If you try to solder the contacts with a low powered, fine tipped soldering iron after attaching the heatsink, you may struggle to melt the solder as the heat is drawn away - which is exactly what the heatsink is meant to do!
To drive the LEDs I opted for Meanwell LDD drivers which are widely used for good reason; they are compact, efficient, relatively cheap and are dimmable. To keep costs down, I used a second hand 20volt/90watt laptop power supply (£2 from a local ebayer). As the power supply is only 20v, it won't drive all 10 LEDs via one driver - each LED needs about 3volts plus a couple of volts 'dropped' via the driver, so I split the LEDs into 2 groups of 5. Each group will draw about 16-18 volts (I've never actually checked), so there is a little headroom at 20v. Power consumption wasn't an issue - including the controller, the finished project consumes around 36watts; well below the 90watt rating for the power supply.
I mounted the drivers on a Coralux 4 driver board. The board and 2 LDD-1000H drivers were £20 delivered from California (in 3 days!). I opted for the 4 driver version to allow for future expansion.
Note that this board needs a 3-5v input to force the driver on. The drivers are designed to be on if there is no input to the controller pin but this can cause flicker when used in conjunction with digital controllers (e.g. when they are booting, the LEDs might be on at full power), so on this board the input is 'pulled' low and must be actively pulled high. Other driver boards are available which do not take this approach or allow it to be configurable via a switch.
Once complete, the driver will be controlled via a PWM (Pulse Width Modulation) signal from the raspberrypi, but for testing I just connected it to the 5v output of the UBEC that will power the raspberrypi.
My first impression was - wow! As ever, photos just don't tell the whole story. These things are bright!
To mount the 10 heatsinks to the hood, I used a length of aluminium bar (about £7 from a local supplier for 2m). This allowed me to build the whole thing as a removable unit, rather than attaching each heatsink to the hood. The bar was marked out and drilled and I cut some electrical bolts to the right length, with 4 longer bolts that bolt right through the heatsinks, ali bar and on to the hood. This turned out to be the most time consuming part of the project as I couldn't get screws that were just the right length - so ended up cutting them individually to length with an angle grinder...
The wiring itself was fairly straightforward and has been covered in detail elsewhere - but do make sure you check your connections as you go, checking for continuity and grounding issues with a multimeter.
The final switch on was always going to be nervy...
I had a spare digital temperature sensor which I attached as close to the centre of one of the heatsinks as I could manage. To date this reads at about 18-20degrees above ambient - which is reasonable, though the centre of the heatsink is probably running a little hotter.
Finally, as an afterthought, I created a splashguard from a spare piece of 2mm acrylic in the shed. Cutting acrylic takes a little patience with standard hand tools - otherwise you will start to crack it. I bent it over a piece of 2x2 using a heat gun; while this worked, I won't be trying it again😉 - trying to keep a 3foot long bend line sufficiently hot/warm to bend is nigh on impossible. The end result was usable - but only because it's out of sight🙂 There's a reason they make specialist tools to bend acrylic sheet.
The splashguard was definitely a good idea. I hadn't planned on using condensation trays (I can't stand them) and quite a lot of condensation forms on the splashguard overnight - which could wreak havoc on exposed electrical connections so close to the surface.
All told, excluding the controller side, I've spent in the region of £100; though I had various bits and pieces to hand. It's been up and running for around 4 weeks now and works without issue. There are a few things I would have done differently (mostly small details with the mounting and wiring) but generally I'm very happy with it - I'm not sure if I could have bought an equivalent unit for that hood, and even if there was something, I'm quite sure it would have cost at least twice as much.
and as I mentioned in the previous thread, virtually all of the components can be reused when (not if!) I decide to change things around.
regards
Mark
I needed some lighting for an old tank that I've put back in service. The tank and hood have been through the mills over the years; I worked out that they were bought in September 1990 - so not far off 25 years old🙂. It had been stored in my dad's garage for the last 10 years or more, so I was a little nervous when I first filled it... but so far, so good. The tank is 3feet long, 18inches deep and 12inches front to back.
The plan was for some decent, high quality (i.e. efficient) LED lighting that was cool running (without fans), dimmable, light weight (I didn't want a kilo of heat sink hanging on the underside of the hood opening) and, most importantly, not too expensive. I wasn't too bothered about the overall brightness levels - was just aiming for good enough!
I also decided early on that this would be part of a larger diy tank controller project, so the timer/dimming would be controlled by a raspberry pi running some software that I wrote myself (I'll cover the controller in a separate thread).
Anyone who has ever looked at a DIY LED system will tell you that there are 1001 ways to build them, so rather than bore everyone with the various options I considered over a number of weeks, I'll just cut to the chase.
I opted for 10 Cree XM-L2 cool white LEDs, pre-mounted on aluminium stars. £35 from eBay (http://www.ebay.co.uk/itm/10W-5W-3W...t=LH_DefaultDomain_3&var=&hash=item8032929890).
No lenses needed as these will only be a couple of inches about the surface of the water.
(one led already removed for testing)
Running at 1amp, these in theory (i.e. per the specs) consume 3watts each while generating between 412 and 468lumens of light output (depending on the temperature). So a 10 LED array should produce over 4000 lumens for 30watts - or somewhere between 130 and 150 lumens per watt, which is pretty efficient.
I couldn't find any specifics but estimate that these LEDs are somewhere in the region of 50% efficient - so would expect them to be producing roughly 1.5watts of heat energy each, which is a factor to consider when it comes to heatsinks...
I explored many options for heatsinks, even considering plain aluminium bar, but finally opted for purpose built Ohmite individual LED heatsinks (http://uk.farnell.com/ohmite/sa-led-113e/heatsink-led-12-7mm-black/dp/2097677?CMP=i-bf9f-00001000).
These actually worked out to be quite economical (£27 for 10 delivered) and are ready made for mounting the LEDs and mounting to a supporting structure. They have a rated thermal resistance of 5.33°C/W - so in theory the roughly 1.5 watts of heat being pumped into them at 100% output should generate a temperature rise of roughly 8°C above ambient - I was expecting at least double that though, as the hood will be enclosed with little air flow, and had planned for anything up to 20°C rise - which would keep the heatsinks at or below 50°C even on the warmest of summer days.
The heatsinks' LED mounting holes are designed for M3 threads (or whatever the US equivalent is), so you will either need to use self-tapping screws (not easy to find on the 'highstreet'), bolt through with M2 bolts (not recommended) or tap for M3 bolts. I had a mixture of stainless steel and nylon M3 bolts in my bits box, as well as some nylon M3 washers and, most importantly, some M3 taps - so an hour or so of tedium later, they were ready for mounting.
Be careful if tapping the threads yourself; as the hole is 'open' on one side, the tap wants to take the path of least resistance and break out, leading to a wonky thread. Take your time and apply even but firm pressure - I messed a couple up (but there are 'spare' holes). Also make sure you change taps if one goes blunt and clean up the hole and thread when finished (i.e. remove burs that might make a mess of mounting the led).
Actually attaching the LEDs is then pretty straightforward. I had plenty of CPU thermal paste lying around but it is readily available. Be warned that most thermal pastes are toxic to both fish and invertebrates - so don't go nuts with it; you only need the tiniest blob per led which should just squeeze out the side of the LED star when it is screwed down (there are some horror shows on youtube that demonstrate how not to do it 🙁).
Don't forget to use nylon washers if you are using metal screws or nuts. Shorting the power to the heatsink may not be fatal to your build - but it may take some time to figure out the problem.
I wired up one LED and heatsink for testing. As reported elsewhere, take care soldering the wires to the LEDs. I soldered them after attaching the LEDs to the heatsinks but I did 'pre-tin' the contacts to make it easier and I was using a fairly powerful soldering iron. If you try to solder the contacts with a low powered, fine tipped soldering iron after attaching the heatsink, you may struggle to melt the solder as the heat is drawn away - which is exactly what the heatsink is meant to do!
To drive the LEDs I opted for Meanwell LDD drivers which are widely used for good reason; they are compact, efficient, relatively cheap and are dimmable. To keep costs down, I used a second hand 20volt/90watt laptop power supply (£2 from a local ebayer). As the power supply is only 20v, it won't drive all 10 LEDs via one driver - each LED needs about 3volts plus a couple of volts 'dropped' via the driver, so I split the LEDs into 2 groups of 5. Each group will draw about 16-18 volts (I've never actually checked), so there is a little headroom at 20v. Power consumption wasn't an issue - including the controller, the finished project consumes around 36watts; well below the 90watt rating for the power supply.
I mounted the drivers on a Coralux 4 driver board. The board and 2 LDD-1000H drivers were £20 delivered from California (in 3 days!). I opted for the 4 driver version to allow for future expansion.
Note that this board needs a 3-5v input to force the driver on. The drivers are designed to be on if there is no input to the controller pin but this can cause flicker when used in conjunction with digital controllers (e.g. when they are booting, the LEDs might be on at full power), so on this board the input is 'pulled' low and must be actively pulled high. Other driver boards are available which do not take this approach or allow it to be configurable via a switch.
Once complete, the driver will be controlled via a PWM (Pulse Width Modulation) signal from the raspberrypi, but for testing I just connected it to the 5v output of the UBEC that will power the raspberrypi.
My first impression was - wow! As ever, photos just don't tell the whole story. These things are bright!
To mount the 10 heatsinks to the hood, I used a length of aluminium bar (about £7 from a local supplier for 2m). This allowed me to build the whole thing as a removable unit, rather than attaching each heatsink to the hood. The bar was marked out and drilled and I cut some electrical bolts to the right length, with 4 longer bolts that bolt right through the heatsinks, ali bar and on to the hood. This turned out to be the most time consuming part of the project as I couldn't get screws that were just the right length - so ended up cutting them individually to length with an angle grinder...
The wiring itself was fairly straightforward and has been covered in detail elsewhere - but do make sure you check your connections as you go, checking for continuity and grounding issues with a multimeter.
The final switch on was always going to be nervy...
I had a spare digital temperature sensor which I attached as close to the centre of one of the heatsinks as I could manage. To date this reads at about 18-20degrees above ambient - which is reasonable, though the centre of the heatsink is probably running a little hotter.
Finally, as an afterthought, I created a splashguard from a spare piece of 2mm acrylic in the shed. Cutting acrylic takes a little patience with standard hand tools - otherwise you will start to crack it. I bent it over a piece of 2x2 using a heat gun; while this worked, I won't be trying it again😉 - trying to keep a 3foot long bend line sufficiently hot/warm to bend is nigh on impossible. The end result was usable - but only because it's out of sight🙂 There's a reason they make specialist tools to bend acrylic sheet.
The splashguard was definitely a good idea. I hadn't planned on using condensation trays (I can't stand them) and quite a lot of condensation forms on the splashguard overnight - which could wreak havoc on exposed electrical connections so close to the surface.
All told, excluding the controller side, I've spent in the region of £100; though I had various bits and pieces to hand. It's been up and running for around 4 weeks now and works without issue. There are a few things I would have done differently (mostly small details with the mounting and wiring) but generally I'm very happy with it - I'm not sure if I could have bought an equivalent unit for that hood, and even if there was something, I'm quite sure it would have cost at least twice as much.
and as I mentioned in the previous thread, virtually all of the components can be reused when (not if!) I decide to change things around.
regards
Mark
