der_wille_zur_macht
Team RC
Here's one of the builds I used these drivers on:
http://reefcentral.com/forums/showthread.php?t=1761547
http://reefcentral.com/forums/showthread.php?t=1761547
DIY LED driver for reef lighting
- Thread starter der_wille_zur_macht
- Start date
DWZM
After reading the Soundwave's monster thread, I was anxious to see someone to design a simple, practical custom driver that could be integrated with a controller. I think there were a few claims on the moster thread, but I don't remember any follow ups... that is, until now. So, looking forward see how it goes. Hopefully will use it in my next set-up (I am also 3-4 months away from my big dream tank... for some time already
Since you've asked for comments on the design, I do have something that bothers me.... First of all, I am not EE, but did design and build a few buck/boost converters using Linear chips a few years back. They were parts of power management for an autonomous robot and had very strict efficiency requirements and made me spent quite a bit of time simulating the designs in the Linear's (free) simulator.... Anyway, as far as I understand, the most problematic part of the switching regulator (in your set up) is a start-up. This is the time when it may have extreme spikes of current and voltages, which waste a lot of power, and shorten lifespan of inductor and capacitors. That brings the points I want to make:
1. Trying to build extremely efficient regulator is possible, but may strain other elements (capacitors/inductor) to the point that their life will be very limited (in my case, I ended-up with estimate of something like 2 years at designed performance). Basicaly, you need to check how your design stresses the capcitors, and check it against perfomance/aging graphs of the later.
2. The dimming mechanism chosen will be shutting down, and re-starting the regulator with PWM frequency, with all the issues above (i.e., performance and lifespan). An immediate consequence is that the efficiency should definitely be lower than what you've designed it for, although it is hard to say by how much. A simple way to test it is to measure the temerature of the IC, capacitors and inductor working with Enabled pin always on, against driven with PWM, to see how much power is dissipated.
The later point is probably why, as kcress said:
"Dimming is usually accomplished by full on/off PWM of the regulated current."
Anyway, thanks for sharing. You've done a lot of the later here, and a great inspiration for any DIYer!
After reading the Soundwave's monster thread, I was anxious to see someone to design a simple, practical custom driver that could be integrated with a controller. I think there were a few claims on the moster thread, but I don't remember any follow ups... that is, until now. So, looking forward see how it goes. Hopefully will use it in my next set-up (I am also 3-4 months away from my big dream tank... for some time already
Since you've asked for comments on the design, I do have something that bothers me.... First of all, I am not EE, but did design and build a few buck/boost converters using Linear chips a few years back. They were parts of power management for an autonomous robot and had very strict efficiency requirements and made me spent quite a bit of time simulating the designs in the Linear's (free) simulator.... Anyway, as far as I understand, the most problematic part of the switching regulator (in your set up) is a start-up. This is the time when it may have extreme spikes of current and voltages, which waste a lot of power, and shorten lifespan of inductor and capacitors. That brings the points I want to make:
1. Trying to build extremely efficient regulator is possible, but may strain other elements (capacitors/inductor) to the point that their life will be very limited (in my case, I ended-up with estimate of something like 2 years at designed performance). Basicaly, you need to check how your design stresses the capcitors, and check it against perfomance/aging graphs of the later.
2. The dimming mechanism chosen will be shutting down, and re-starting the regulator with PWM frequency, with all the issues above (i.e., performance and lifespan). An immediate consequence is that the efficiency should definitely be lower than what you've designed it for, although it is hard to say by how much. A simple way to test it is to measure the temerature of the IC, capacitors and inductor working with Enabled pin always on, against driven with PWM, to see how much power is dissipated.
The later point is probably why, as kcress said:
"Dimming is usually accomplished by full on/off PWM of the regulated current."
Anyway, thanks for sharing. You've done a lot of the later here, and a great inspiration for any DIYer!
der_wille_zur_macht
Team RC
aryth,
I'm interested to learn more about what you're discussing regarding designing for long lifetime and efficiency.
To clear up one potential misunderstanding, when you say Linear with a capital L, I'm assuming you mean Linear the chip manufacturer, not linear regulators? Which Linear parts did you use? Can you go one more level deep in explaining the trends you saw? i.e. did switching to bigger or smaller components change efficiency and lifetime in a certain manner? I know some components (i.e. capacitors) are spec'd with a given lifetime at a given temp (i.e. 10,000 hrs @105c). These drivers aren't running that hot - I can touch all components on the board and detect no significant difference between the components and room temperature, regardless of dimming or running wide open. What does that do to lifetime projections?
As always I appreciate the comments - this thread was intended to be a learning process for all (including me), not a presentation of a 100% optimal finished solution.
I'm interested to learn more about what you're discussing regarding designing for long lifetime and efficiency.
To clear up one potential misunderstanding, when you say Linear with a capital L, I'm assuming you mean Linear the chip manufacturer, not linear regulators? Which Linear parts did you use? Can you go one more level deep in explaining the trends you saw? i.e. did switching to bigger or smaller components change efficiency and lifetime in a certain manner? I know some components (i.e. capacitors) are spec'd with a given lifetime at a given temp (i.e. 10,000 hrs @105c). These drivers aren't running that hot - I can touch all components on the board and detect no significant difference between the components and room temperature, regardless of dimming or running wide open. What does that do to lifetime projections?
As always I appreciate the comments - this thread was intended to be a learning process for all (including me), not a presentation of a 100% optimal finished solution.
stugray
Premium Member
aryth,
"but may strain other elements (capacitors/inductor) to the point that their life will be very limited"
Yes that is a concern. So...
der_wille_zur_macht,
You didnt choose any tantalum capacitors did you?
They typically look like a tic-tac with legs and are polarity sensitive ( they will have a + on one leg )
If you stress those with too much inrush current they can EXPLODE & catch fire.
Stu
"but may strain other elements (capacitors/inductor) to the point that their life will be very limited"
Yes that is a concern. So...
der_wille_zur_macht,
You didnt choose any tantalum capacitors did you?
They typically look like a tic-tac with legs and are polarity sensitive ( they will have a + on one leg )
If you stress those with too much inrush current they can EXPLODE & catch fire.
Stu
der_wille_zur_macht
Team RC
Nope. No tantalum capacitors. Ceramic for the small stuff an aluminum for the large stuff.
I know it's possible to do better with specs on some of the components I chose (2k hour life caps vs. 10k). But I still don't know what sort of effect this will have in the real world, given I'm operating under much "nicer" conditions than the lifetime test.
I suppose if you kept the peak current low enough (through choice of average current and inductor size) you could forgo the big cap on the output side. One less thing to fail.
I know it's possible to do better with specs on some of the components I chose (2k hour life caps vs. 10k). But I still don't know what sort of effect this will have in the real world, given I'm operating under much "nicer" conditions than the lifetime test.
I suppose if you kept the peak current low enough (through choice of average current and inductor size) you could forgo the big cap on the output side. One less thing to fail.
DWZM,
Yes, I meant The Linear Technology - company. I was building DC/DC converters and charger, so it probably does not matter what parts I've used. Nevertheless, I remember they had dedicated Led drivers (chips), though likely all SMD packages. I would look anyway, as they definitely are one of the best analog designers/chip makers in the world, and have very good documentation, and simulator.
With this DIY Led driver, I totally agree that we don't need the absolutely optimal design. Actually it was my point that maybe while designing (with the OnSemi spreadsheet), you should not try to chase the best performance, but a more gentle switching regime, to ensure longer life (and better chance of it working in the first place, which is an achievement in itself with switching regulators... in my experience ). Judging by the fact that it works, and there is no any obvious overheating, you've likely done it well already. In particular, you may do without the second tantalum capacitor, the Stu was referring to, in this application, although I would check the stability of the current on the output using oscilloscope, or simulating this or similar circuit with only one, electrolytic output capacitor. Btw, you really should try the Linear's simulator. It is very simple, useful, and a lot of staff built in, including many designs of switching converters (voltage and current) based on their chips. (hmm, disclaimer: I don't work for Linear, or anywhere near the electronic industry )
...but, the question of lifespan is still important. I.e., I would not want to build a led rig, with idea it to be maintenance free for 10 years, only to learn that I have to replace all the drivers every 12 months or so.
The only factor, I remember/could think of now, which is important here, given your already working design, is the input and output capacitors. I don't remember now why, but I think the output capacitor was the bottleneck, and you really want to get the best one that fits the design. Again, don't remember now all the details, and don't have time to get into this right now, really sorry (common, just check timestamps of my posts). Off the top of my head, you need to ensure (1) that capacitor is fine with the operating mode of the switcher (frequency, and RMS of voltage and/or current), (2) its internal resistance is as small as possible (at the operating mode/frequency! - it usually will be different than the one stated in the capacitor's spec). Both of the above will give you the real lifespan of the capacitor. I remember it was all a bit involved, but some manufacturers provided a free design software that gave you the answers, including the lifespan....
Actually just googled one of them: Kemet. They have some program here:
http://www.kemet.com/kemet/web/homepage/kechome.nsf/weben/kemsoft
If it does not give all the answers, or not that friendly try another company (I've used at least two programs from different companies and one of them was very good, one of them was Kemet's - but don't remember which one
)
In my experience, it was very important to select the capacitors right, as their characteristics, and lifespan change a lot depending on mode of operation and other parameters (and to my great surprise to ambient themperature - although it should not be an issue in this particular case)
Ah, exactly what I was talking about, I would use the 10K capacitor. The lifespan gets reduced exponentially, when the capcitor starts to fail, due to normal aging, or due to dificult working mode.
I know, I write too much, too late, and remember little... hope it helps anyway, or wait another 3-4 months, and I will start MY build
Yes, I meant The Linear Technology - company. I was building DC/DC converters and charger, so it probably does not matter what parts I've used. Nevertheless, I remember they had dedicated Led drivers (chips), though likely all SMD packages. I would look anyway, as they definitely are one of the best analog designers/chip makers in the world, and have very good documentation, and simulator.
With this DIY Led driver, I totally agree that we don't need the absolutely optimal design. Actually it was my point that maybe while designing (with the OnSemi spreadsheet), you should not try to chase the best performance, but a more gentle switching regime, to ensure longer life (and better chance of it working in the first place, which is an achievement in itself with switching regulators... in my experience
). Judging by the fact that it works, and there is no any obvious overheating, you've likely done it well already. In particular, you may do without the second tantalum capacitor, the Stu was referring to, in this application, although I would check the stability of the current on the output using oscilloscope, or simulating this or similar circuit with only one, electrolytic output capacitor. Btw, you really should try the Linear's simulator. It is very simple, useful, and a lot of staff built in, including many designs of switching converters (voltage and current) based on their chips. (hmm, disclaimer: I don't work for Linear, or anywhere near the electronic industry )...but, the question of lifespan is still important. I.e., I would not want to build a led rig, with idea it to be maintenance free for 10 years, only to learn that I have to replace all the drivers every 12 months or so.
The only factor, I remember/could think of now, which is important here, given your already working design, is the input and output capacitors. I don't remember now why, but I think the output capacitor was the bottleneck, and you really want to get the best one that fits the design. Again, don't remember now all the details, and don't have time to get into this right now, really sorry (common, just check timestamps of my posts
). Off the top of my head, you need to ensure (1) that capacitor is fine with the operating mode of the switcher (frequency, and RMS of voltage and/or current), (2) its internal resistance is as small as possible (at the operating mode/frequency! - it usually will be different than the one stated in the capacitor's spec). Both of the above will give you the real lifespan of the capacitor. I remember it was all a bit involved, but some manufacturers provided a free design software that gave you the answers, including the lifespan.... Actually just googled one of them: Kemet. They have some program here:
http://www.kemet.com/kemet/web/homepage/kechome.nsf/weben/kemsoft
If it does not give all the answers, or not that friendly try another company (I've used at least two programs from different companies and one of them was very good, one of them was Kemet's - but don't remember which one
)In my experience, it was very important to select the capacitors right, as their characteristics, and lifespan change a lot depending on mode of operation and other parameters (and to my great surprise to ambient themperature - although it should not be an issue in this particular case)
Ah, exactly what I was talking about, I would use the 10K capacitor. The lifespan gets reduced exponentially, when the capcitor starts to fail, due to normal aging, or due to dificult working mode.
I know, I write too much, too late, and remember little... hope it helps anyway, or wait another 3-4 months, and I will start MY build
der_wille_zur_macht
Team RC
I appreciate the comments. Over the last few days I've been digging through digikey searches again, looking at specs for some of the parts, at it looks like at least from a lifetime perspective there are improvements to be made. For instance, the big output cap has a 2k lifespan at 85c. Here's one, same manufacturer, but 5k hours at 105c:
http://search.digikey.com/scripts/DkSearch/dksus.dll?Detail&name=P13459-ND
And it's actually cheaper. . .
And you mention tantalum caps - for clarification, there aren't any - just ceramics for the little guys, and electrolytics for the big guys.
I may just go cut the big electrolytic cap off the output side of one of my prototypes and see how it behaves. I'll be running at low drive current so no chance of going above the peak without it in my case.
http://search.digikey.com/scripts/DkSearch/dksus.dll?Detail&name=P13459-ND
And it's actually cheaper. . .
And you mention tantalum caps - for clarification, there aren't any - just ceramics for the little guys, and electrolytics for the big guys.
I may just go cut the big electrolytic cap off the output side of one of my prototypes and see how it behaves. I'll be running at low drive current so no chance of going above the peak without it in my case.
terahz
1x10^12 Hz
I've had good experience with Nichicon HE series caps. I've build a few audio amps and power supplies for them and used these caps in both. Just checked the 100µF one is rated for 7000 Hrs @ 105°C
http://search.digikey.com/scripts/DkSearch/dksus.dll?Detail&name=493-1605-ND
EDIT: just noticed they are twice more expensive...
http://search.digikey.com/scripts/DkSearch/dksus.dll?Detail&name=493-1605-ND
EDIT: just noticed they are twice more expensive...
der_wille_zur_macht
Team RC
kcress
New member
What you guys may be missing on these cap Life Numbers is the operating temperature. You may have noticed that your TVs, ovens, and car stereos last longer than say 2000hrs, (1/4 of a year!!!)
This is because that Life Number is only for the cap running at that label temperature. Think about it.. Do you think your caps are running at a temperature above boiling water? (105C) Is your flesh instantly seared when you accidentally touch one?
If you dig around in the data sheets you should find the temperature/life curves. They embrace the chemical dictum of "double the reaction rate for every 10 degree rise". Luckily this pays us dividends in reverse. If you run a 105C rated cap at room temperature its life will be something like 200,000 hrs. (20 years 24/7)
So keep your caps at a reasonable temperature and they will greatly exceed the life of the LEDs. And if you keep them cool, 1000hr/85C caps will even serve you well.
This is because that Life Number is only for the cap running at that label temperature. Think about it.. Do you think your caps are running at a temperature above boiling water? (105C) Is your flesh instantly seared when you accidentally touch one?
If you dig around in the data sheets you should find the temperature/life curves. They embrace the chemical dictum of "double the reaction rate for every 10 degree rise". Luckily this pays us dividends in reverse. If you run a 105C rated cap at room temperature its life will be something like 200,000 hrs. (20 years 24/7)
So keep your caps at a reasonable temperature and they will greatly exceed the life of the LEDs. And if you keep them cool, 1000hr/85C caps will even serve you well.
I did some research today. Aparently the lifetime of the capacitor is a well known issue in power electronics. I found a couple of articles:
http://powerelectronics.com/passive...apacitors/power_internal_construction_boosts/
this one basically says what I did, but gives actual details of what, and why it happens.
This one is even better. It is right on our topic:
http://www.lrc.rpi.edu/programs/solidstate/pdf/Han-SPIE2009-7422.pdf
It discusses the lifespan of led drivers, and failures due to aging of the capacitors. I skimmed over, and it seems that commercial drivers are good for 10K-15K hours. We should try to get the DYI drivers close to this, otherwise, it may be cheaper to buy one.
Merry Christmas!
:celeb2:
http://powerelectronics.com/passive...apacitors/power_internal_construction_boosts/
this one basically says what I did, but gives actual details of what, and why it happens.
This one is even better. It is right on our topic:
http://www.lrc.rpi.edu/programs/solidstate/pdf/Han-SPIE2009-7422.pdf
It discusses the lifespan of led drivers, and failures due to aging of the capacitors. I skimmed over, and it seems that commercial drivers are good for 10K-15K hours. We should try to get the DYI drivers close to this, otherwise, it may be cheaper to buy one.
Merry Christmas!
:celeb2:
Kcress,
I understood that this is a bit different when high current/voltage ripples run through a cap, i.e, in power switching applications. These spikes create high temperature, but for very short times within the capacitor and cause failures in small areas (the "hot spots"), while overall, the temperature stays reasonable.
Anyway, I don't think we can design the perfect driver, but the rules of thumb I would follow:
1. use the best in/out capacitors I can get, and
2. do not mount the drivers permanently, and use connectors, to allow quick replacement (even with commercial drivers)
kcress
New member
I've never heard of this and not seen any issues in my designs or for that matter anyone else's with respect to noticeably short lived electrolytics.
Occasionally someone screws up and a bad load of them hit the market. Around the turn of the century this occurred with one cap that was used in millions of pc mother board power supplies. For a while there, 90% of all mother board failures were considered to be from that problem.
SpacedCowboy
New member
Have to say I'm with kcress on this one. I've been building electronics devices for the last two decades or so, and I've not run into any issues with capacitors yet.
Some major films have been filmed using my electronics modules, (and still are). If they can survive harsh environments ranging from deserts through to the antartic (often both!), I reckon they can survive the relatively pleasant environment above an aquarium, with only slightly higher humidity than usual and an ambient temperature.
I can see the argument in the paper linked to, but in real life, I've never experienced it. The driver is very cheap, in any event (roughly $3 per 6 leds) and in my case they're all replaceable modules, but of all the things that could go wrong, I don't think capacitor destruction is high on the list..
Simon.
der_wille_zur_macht
Team RC
I've been "offline" thanks to the holidays. Once things settle down I'll put something out.
BTW I appreciate the cap discussion, so thanks everyone. I might just build with the longer life parts anyways, since they're not always more expensive (Panasonic's 7k hr caps in this size are actually CHEAPER than the 2k hr caps I used, by a few cents.)
BTW I appreciate the cap discussion, so thanks everyone. I might just build with the longer life parts anyways, since they're not always more expensive (Panasonic's 7k hr caps in this size are actually CHEAPER than the 2k hr caps I used, by a few cents.)
der_wille_zur_macht
Team RC
Yes, but I haven't looked up appropriate numbers. I'd just look for similar component values for the caps that were longer life. I'll do a "revised" part list when I put the schematic up.
