EEVblog Electronics Community Forum

Electronics => Projects, Designs, and Technical Stuff => Topic started by: mikeselectricstuff on April 06, 2015, 05:33:09 AM

Title: LED Filament clock display
Post by: mikeselectricstuff on April 06, 2015, 05:33:09 AM
Messing with filament strips from LED lamps.
Clock design page : http://electricstuff.co.uk/ledfilclock.html (http://electricstuff.co.uk/ledfilclock.html)
LED filaments (http://www.youtube.com/watch?v=H_XiunR-cAQ#ws)
Title: Re: LED Filament clock display
Post by: Lightages on April 06, 2015, 06:39:09 AM
Those LED filaments look like they have lots of potential uses and also fun to play with. Thanks for sharing.
Title: Re: LED Filament clock display
Post by: PeterFW on April 06, 2015, 07:37:26 AM
Messing with filament strips from LED lamps.

Very interesting, thanks for posting and making a video about them!
Title: Re: LED Filament clock display
Post by: davidf2281 on April 06, 2015, 07:52:52 AM
Love the look of the clock.

Just had a quick gander on ebay but no immediate hits for filament LEDs under 99p -- any recommendation for sellers?

David.
Title: Re: LED Filament clock display
Post by: mikeselectricstuff on April 06, 2015, 07:59:10 AM
4 for £1.65 :
http://www.ebay.co.uk/itm/1W-COB-LED-Filament-Light-Source-for-LED-Bulb-Globe-Light-Pure-Warm-White-DIY-/301379430602?pt=LH_DefaultDomain_3&var=&hash=item462b9d28ca (http://www.ebay.co.uk/itm/1W-COB-LED-Filament-Light-Source-for-LED-Bulb-Globe-Light-Pure-Warm-White-DIY-/301379430602?pt=LH_DefaultDomain_3&var=&hash=item462b9d28ca)

10 for £3.59 :
http://www.ebay.co.uk/itm/4-10x-1W-DC50-60V-17mA-COB-LED-Filament-Light-Bulb-Globe-Candle-Light-Source-DIY-/301456288920?pt=LH_DefaultDomain_3&var=&hash=item463031ec98 (http://www.ebay.co.uk/itm/4-10x-1W-DC50-60V-17mA-COB-LED-Filament-Light-Bulb-Globe-Candle-Light-Source-DIY-/301456288920?pt=LH_DefaultDomain_3&var=&hash=item463031ec98)

100 for $43 :
http://www.aliexpress.com/item/50-60v-37-1-6mm-1w-COB-led-light-source-for-bulb-filament/32265251230.html (http://www.aliexpress.com/item/50-60v-37-1-6mm-1w-COB-led-light-source-for-bulb-filament/32265251230.html)
Title: Re: LED Filament clock display
Post by: davidf2281 on April 06, 2015, 08:01:09 AM

Lovely, thanks Mike.

David.
Title: Re: LED Filament clock display
Post by: davidf2281 on April 06, 2015, 08:08:26 AM

Also, keep up the excellent work with the YouTube vids -- they're v entertaining.

David.
Title: Re: LED Filament clock display
Post by: jimon on April 06, 2015, 08:33:43 AM
Ordered some of them, will see what is possible with them :)
Title: Re: LED Filament clock display
Post by: IanB on April 06, 2015, 09:37:56 AM
Strangely, I've been looking for those "filament" style lamps here in the USA and I have not seen any yet. Also, it seems (from watching Clive) that Poundland sells many more interesting items than I have seen in the likes of Dollar Tree here. I guess when it comes to bargains the greater value of the pound leads to more options than a lowly dollar permits.

As far as driving those filaments, would a VFD driver like the MAX6291 be of any use? It seems the voltages are similar.
Title: Re: LED Filament clock display
Post by: mikeselectricstuff on April 06, 2015, 09:48:22 AM
Strangely, I've been looking for those "filament" style lamps here in the USA and I have not seen any yet. Also, it seems (from watching Clive) that Poundland sells many more interesting items than I have seen in the likes of Dollar Tree here. I guess when it comes to bargains the greater value of the pound leads to more options than a lowly dollar permits.

As far as driving those filaments, would a VFD driver like the MAX6291 be of any use? It seems the voltages are similar.
probably - not looked at it, but it's Maxim, so expensive, if you can even get it.
TPIC6B595 is cheap - you can also use the slightly cheaper TICP6C595, which ISTR has a slightly lower clamp voltage, but still useable.
Hell if you get the voltage spot-on you could probably even drive them direct from a PIC output, tough that would  be a rather bad idea.
Title: Re: LED Filament clock display
Post by: DanielS on April 06, 2015, 09:59:18 AM
Strangely, I've been looking for those "filament" style lamps here in the USA and I have not seen any yet.

As far as driving those filaments, would a VFD driver like the MAX6291 be of any use? It seems the voltages are similar.
They might not be available in NA (at least not through official channels) due to not meeting safety requirements: no fuse, no resistor to limit current if the diode bridge shorts out, possibly nasty stuff if the current-limiting capacitor shorts out too and likely more.

As for the driving, as Mike said, you could probably drive those using regular 5V TTL chips by tweaking the supply voltage since most of the voltage gets dropped across the LED string: simply set your supply voltage to NxVF + 3V, then set RS to 2.7V/IF allowing 0.3V VCEsat on TTL outputs, this way the outputs should not get pulled much beyond VCC.
Title: Re: LED Filament clock display
Post by: mikeselectricstuff on April 06, 2015, 10:05:11 AM
Strangely, I've been looking for those "filament" style lamps here in the USA and I have not seen any yet.

As far as driving those filaments, would a VFD driver like the MAX6291 be of any use? It seems the voltages are similar.
They might not be available in NA (at least not through official channels) due to not meeting safety requirements: no fuse, no resistor to limit current if the diode bridge shorts out, possibly nasty stuff if the current-limiting capacitor shorts out too and likely more.
But no reason they couldn't be made safe with minimal extra cost -just a fuseable resistor and maybe a bleed resistor across the cap.
I don't recall having seen them in shops in the UK yet, though quite a few on Amazon
Title: Re: LED Filament clock display
Post by: IanB on April 06, 2015, 10:16:28 AM
I don't recall having seen them in shops in the UK yet, though quite a few on Amazon
That was the reason for my Poundland comment. I thought Clive had found some in there, though I may be thinking of something else.
Title: Re: LED Filament clock display
Post by: Hero999 on April 06, 2015, 08:05:30 PM
I thought these LED filaments require a relatively high voltage >50V so powering directly from a MCU pin is out of the question.

I don't believe Clive found an LED filament lamp in Poundland. I think there's some confusion over a video he did about a cheap LED lamp from Poundland, with a light sensor on it, which is very inefficient because it used a big resistor and consumed just as much power, whether it's on or off.
Title: Re: LED Filament clock display
Post by: Psi on April 06, 2015, 08:23:56 PM
Could use this dcdc module to generate 72V from a more reasonable voltage
They're $14 each from digikey and come in 5V input and 12V input versions.
$14 isn't that cheap but it saves you needing to make a HV supply yourself for a one-off project.

Output is -72V @ 42mA  (42mA if only loading one output, it also has -24V and -48V  output).
You could drive 42 of these led strings at 1mA each with one module
They're isolated so you can easy use them around the other way for +72V instead of -72V

http://www.digikey.com/product-detail/en/NMT1272SC/811-1584-5-ND/1927145 (http://www.digikey.com/product-detail/en/NMT1272SC/811-1584-5-ND/1927145)
http://power.murata.com/data/power/ncl/kdc_nmt.pdf (http://power.murata.com/data/power/ncl/kdc_nmt.pdf)
Title: Re: LED Filament clock display
Post by: mikeselectricstuff on April 06, 2015, 10:21:38 PM
Could use this dcdc module to generate 72V from a more reasonable voltage
They're $14 each from digikey and come in 5V input and 12V input versions.
$14 isn't that cheap but it saves you needing to make a HV supply yourself for a one-off project.

Output is -72V @ 42mA  (42mA if only loading one output, it also has -24V and -48V  output).
You could drive 42 of these led strings at 1mA each with one module
They're isolated so you can easy use them around the other way for +72V instead of -72V

http://www.digikey.com/product-detail/en/NMT1272SC/811-1584-5-ND/1927145 (http://www.digikey.com/product-detail/en/NMT1272SC/811-1584-5-ND/1927145)
http://power.murata.com/data/power/ncl/kdc_nmt.pdf (http://power.murata.com/data/power/ncl/kdc_nmt.pdf)
72V is a bit marginal,but you could stack it on top of the 5V rail  for 77V which would be ideal

Some cheap ones here - $6.88 - presumably surplus
http://www.componentsdirect.com/murata-manufacturing-co-nmt1272sc.html?utm_source=supplyFrame&utm_medium=buyNow (http://www.componentsdirect.com/murata-manufacturing-co-nmt1272sc.html?utm_source=supplyFrame&utm_medium=buyNow)
Title: Re: LED Filament clock display
Post by: ConKbot on April 06, 2015, 10:45:28 PM
I thought these LED filaments require a relatively high voltage >50V so powering directly from a MCU pin is out of the question.

Diodes always drop a forward voltage drop, it just drops as diode current drops. It certainly wouldnt be robust, or even good design practice, but if you're only feeding them a few mA, and if the Vf @ 1mA of the string was 70V, and the vF at 0-100nA (whatever the leakage of the uC pin is) is 66V, with a 71V supply, and 1V across your dropper resistor when its on, then the MCU should only see 5V on its input pin.  But you're relying on a whole lot to go right there. Could it work with a bit of luck in the voltages of the diode string working out and tweaking your supply just so? Maybe. And dont count on much surviving if a few of the dies on the LED string short out  ;D
Title: Re: LED Filament clock display
Post by: Hero999 on April 06, 2015, 10:52:06 PM
I thought these LED filaments require a relatively high voltage >50V so powering directly from a MCU pin is out of the question.

Diodes always drop a forward voltage drop, it just drops as diode current drops. It certainly wouldnt be robust, or even good design practice, but if you're only feeding them a few mA, and if the Vf @ 1mA of the string was 70V, and the vF at 0-100nA (whatever the leakage of the uC pin is) is 66V, with a 71V supply, and 1V across your dropper resistor when its on, then the MCU should only see 5V on its input pin.  But you're relying on a whole lot to go right there. Could it work with a bit of luck in the voltages of the diode string working out and tweaking your supply just so? Maybe. And dont count on much surviving if a few of the dies on the LED string short out  ;D
Oh I see.

I agree it's not good design practise and LEDs have a negative temperate coefficient which makes it worse.
Title: Re: LED Filament clock display
Post by: amyk on April 06, 2015, 11:10:14 PM
I feel like the Chinese are going to start selling a version of this that runs everything off a capacitive dropper... and uses an ASIC that integrates the HV constant-current drivers and clock logic.
Title: Re: LED Filament clock display
Post by: Hero999 on April 07, 2015, 02:48:47 AM
How about using a mains transformer and a capacitive voltage multiplier to power the LED filaments?
Title: Re: LED Filament clock display
Post by: DanielS on April 08, 2015, 10:29:31 AM
I agree it's not good design practise and LEDs have a negative temperate coefficient which makes it worse.
Depending on how many mA you want to put through the string and how many volts that resistor might need to drop, it should be possible to replace the current-limiting resistors with a JFET as a single component (roughly) constant current element.
Title: Re: LED Filament clock display
Post by: TerminalJack505 on April 08, 2015, 10:58:02 AM
If you're just needing to boost a DC voltage and don't need a lot of current (< 30mA) then you could probably use the scavenger (http://www.electro-tech-online.com/articles/the-scavenger-a-joule-thief-inspired-boost-regulator.594/) boost converter to drive the filaments.  Just swap-out the 5V Zener for a ~75V Zener.

I've only played around with it in the simulator but it should work.
Title: Re: LED Filament clock display
Post by: Hero999 on April 09, 2015, 12:27:43 AM
I agree it's not good design practise and LEDs have a negative temperate coefficient which makes it worse.
Or you could add an emitter resistor to the driver transistor.
Depending on how many mA you want to put through the string and how many volts that resistor might need to drop, it should be possible to replace the current-limiting resistors with a JFET as a single component (roughly) constant current element.
(http://www.eevblog.com/forum/projects/nixie-tube-clock-without-the-nixie-tubes!/?action=dlattach;attach=142825)
Title: Re: LED Filament clock display
Post by: PedroDaGr8 on April 09, 2015, 02:06:29 AM
Strangely, I've been looking for those "filament" style lamps here in the USA and I have not seen any yet.

As far as driving those filaments, would a VFD driver like the MAX6291 be of any use? It seems the voltages are similar.
They might not be available in NA (at least not through official channels) due to not meeting safety requirements: no fuse, no resistor to limit current if the diode bridge shorts out, possibly nasty stuff if the current-limiting capacitor shorts out too and likely more.

As for the driving, as Mike said, you could probably drive those using regular 5V TTL chips by tweaking the supply voltage since most of the voltage gets dropped across the LED string: simply set your supply voltage to NxVF + 3V, then set RS to 2.7V/IF allowing 0.3V VCEsat on TTL outputs, this way the outputs should not get pulled much beyond VCC.

I've seen these used all over the place here in San Diego, starting mid last year. I have even seen them for sale in a few stores but usually at around $5-10/bulb.
Title: Re: LED Filament clock display
Post by: cimmo on April 09, 2015, 07:31:59 AM
This looks like it'll do the job:
http://www.ebay.com.au/itm/321664475968 (http://www.ebay.com.au/itm/321664475968)
"100W DC-DC Boost Converter 10-32V to 60-97V Step Up Power supply module" - ($5.36USD incl ship)

And Aliexpress has "10pcs/lot 1W COB LED Filament Lamp Bulb Globe Candle Light Source Warm White Pure White DIY" for $4.69 USD (incl ship)
http://www.aliexpress.com/item/10pcs-lot-1W-COB-LED-Filament-Lamp-Bulb-Globe-Candle-Light-Source-Warm-White-Pure-White/32274032926.html (http://www.aliexpress.com/item/10pcs-lot-1W-COB-LED-Filament-Lamp-Bulb-Globe-Candle-Light-Source-Warm-White-Pure-White/32274032926.html)

Title: Re: LED Filament clock display
Post by: NANDBlog on April 09, 2015, 08:15:00 AM
So far your video is the best of the filament LED teardowns. I have half dozen of these bulbs at home, I call them blinky. There is a 50 or 100 Hz visible stobing effect coming out of them, it disturbs me a lot. I ordered E14 fitting, maybe the cap inside is too small, I havent looked into one yet. Anyway, I'll try to feed it with rectified mains (with a bigger cap) and see how it goes :-BROKE
Title: Re: LED Filament clock display
Post by: DanielS on April 09, 2015, 11:05:30 AM
I ordered E14 fitting, maybe the cap inside is too small, I havent looked into one yet. Anyway, I'll try to feed it with rectified mains (with a bigger cap) and see how it goes :-BROKE
If they are the same type as Mike's, the capacitor is a series current-limiting element on the AC input side instead of a decoupling cap on the rectifier bridge's output.
Title: Re: LED Filament clock display
Post by: IanB on April 09, 2015, 11:43:46 AM
So far your video is the best of the filament LED teardowns. I have half dozen of these bulbs at home, I call them blinky. There is a 50 or 100 Hz visible stobing effect coming out of them, it disturbs me a lot. I ordered E14 fitting, maybe the cap inside is too small, I havent looked into one yet. Anyway, I'll try to feed it with rectified mains (with a bigger cap) and see how it goes :-BROKE

I think it shouldn't be any worse than the flicker from fluorescent tubes. Are you particularly sensitive to lamp flicker, or do you find these lamps especially bad?
Title: Re: LED Filament clock display
Post by: ajb on April 09, 2015, 01:22:57 PM
So far your video is the best of the filament LED teardowns. I have half dozen of these bulbs at home, I call them blinky. There is a 50 or 100 Hz visible stobing effect coming out of them, it disturbs me a lot. I ordered E14 fitting, maybe the cap inside is too small, I havent looked into one yet. Anyway, I'll try to feed it with rectified mains (with a bigger cap) and see how it goes :-BROKE

I think it shouldn't be any worse than the flicker from fluorescent tubes. Are you particularly sensitive to lamp flicker, or do you find these lamps especially bad?

It depends a lot on the effective duty cycle.  If you've got a few of those LED filaments in series so that the Vf isn't far from the AC peak voltage, duty cycle is going to be very low with a very steep rise and fall, making flicker much more obvious.  I'm not sure what the output waveform of a fluorescent tube would like like, but I suspect it'll have longer rise and fall with a reasonably long on-time, which is why most people aren't much bothered by them.
Title: Re: LED Filament clock display
Post by: NANDBlog on April 10, 2015, 02:42:53 AM
So far your video is the best of the filament LED teardowns. I have half dozen of these bulbs at home, I call them blinky. There is a 50 or 100 Hz visible stobing effect coming out of them, it disturbs me a lot. I ordered E14 fitting, maybe the cap inside is too small, I havent looked into one yet. Anyway, I'll try to feed it with rectified mains (with a bigger cap) and see how it goes :-BROKE

I think it shouldn't be any worse than the flicker from fluorescent tubes. Are you particularly sensitive to lamp flicker, or do you find these lamps especially bad?
I've found there seems to be two types of these laps, one has some white plastic near the base. I assume it is there to cover the more electronics built in.
If I move my hand fast, I see 6/7 distinct hands before me. It generally drives me crazy, I cannot stand the light. I might be a bit sensible for this though, I found even a multiplexed LED screen (with MAX7219 500Hz) on lower brightness levels annoying.
Title: Re: LED Filament clock display
Post by: IanB on April 10, 2015, 03:29:48 AM
I've found there seems to be two types of these laps, one has some white plastic near the base. I assume it is there to cover the more electronics built in.
If I move my hand fast, I see 6/7 distinct hands before me. It generally drives me crazy, I cannot stand the light. I might be a bit sensible for this though, I found even a multiplexed LED screen (with MAX7219 500Hz) on lower brightness levels annoying.

It is possible to build these lamps with a smoothing capacitor after the rectifier and that would eliminate the flicker. But it would have to be an (expensive) high voltage electrolytic and it would be an additional possible point of failure, especially if it were a crapxon branded capacitor with a low temperature rating. Possibly such a capacitor could be added as an aftermarket modification?
Title: Re: LED Filament clock display
Post by: mikeselectricstuff on April 10, 2015, 04:10:39 AM
Quick comparison of light output waveform fluorescent (inductive ballast) vs. LED filament
 
Title: Re: LED Filament clock display
Post by: NANDBlog on April 10, 2015, 04:43:16 AM
FYI, I've found a datasheet
http://www.runlite.cn/userfiles/js9jaw90nxmvc1372646905.pdf (http://www.runlite.cn/userfiles/js9jaw90nxmvc1372646905.pdf)
On this guys blog: http://electronupdate.blogspot.ca (http://electronupdate.blogspot.ca)
A base of an E14 fitting is like 14mm x 15 mm (from memory) so you really dont have much space for electronics. What I expect to be happening, is rectifying mains, smoothing capacitor, and constant current linear drive. The two filaments are in series, so it would turn on at 140V.
I dont want to mess with the bulbs and fittings, what I have in mind is to convert my lamp to output DC or at least somewhat smoothed DC. I hope the rectifier inside are properly rated for this. If Mike is right and they have a capacitor on the mains directly, then I'm in trouble, and I'm actually not sure what to do with thees other than break them for the filaLED (made up word, seems apropriate).

Quick comparison of light output waveform fluorescent (inductive ballast) vs. LED filament
Those scope screens are worth a thousand word.
Title: Re: LED Filament clock display
Post by: IanB on April 10, 2015, 05:36:17 AM
Clive has torn down a bunch of these lamps on his bigclivedotcom YouTube channel. Mostly he finds a series dropper capacitor on one arm of the mains input followed by a bridge rectifier and then the LED string directly on the DC side of the rectifier. So the LED strings will only light up when the mains voltage cycle exceeds the forward voltage of the LEDs. For a large part of the cycle the LEDs will be off, which is consistent with the scope trace Mike showed.

With the series dropper capacitor arrangement the lamps will not work if you feed them DC, since the dropper capacitor will become a blocking capacitor. To make an improvement the lamp would have to be disassembled and the internal electronics modified.
Title: Re: LED Filament clock display
Post by: NANDBlog on April 10, 2015, 06:33:04 AM
Clive has torn down a bunch of these lamps on his bigclivedotcom YouTube channel. Mostly he finds a series dropper capacitor on one arm of the mains input followed by a bridge rectifier and then the LED string directly on the DC side of the rectifier. So the LED strings will only light up when the mains voltage cycle exceeds the forward voltage of the LEDs. For a large part of the cycle the LEDs will be off, which is consistent with the scope trace Mike showed.

With the series dropper capacitor arrangement the lamps will not work if you feed them DC, since the dropper capacitor will become a blocking capacitor. To make an improvement the lamp would have to be disassembled and the internal electronics modified.
Yes, that is what I was afraid of. Technically, if you would make an external say 150V power supply and a PWM from it at say 1Khz, the series capacitor would be a short circuit, and the parallel would be properly sized  but I think it is not really worth my time.
Title: Re: LED Filament clock display
Post by: DanielS on April 10, 2015, 07:20:05 AM
Quick comparison of light output waveform fluorescent (inductive ballast) vs. LED filament
Those scope screens are worth a thousand word.
The result was easily foreseeable: between the ballast's inductive nature keeping the arc alive as voltage drops and the phosphor storing UV lights before it can re-emit it as visible, there is not much down-time on FTs with a good ballast and tubes.
Title: Re: LED Filament clock display
Post by: ajb on April 10, 2015, 08:07:20 AM
Quick comparison of light output waveform fluorescent (inductive ballast) vs. LED filament
:-+

The fluorescent waveform is a hell of a lot smoother than I would have expected!  What sort of lamp was that?

The shape of the LED waveform is interesting, particularly how slow the decay is.  Is that due to the junctions heating up over the course of the on period, lowering the Vf?  It would be interesting to see the voltage and current waveforms along side.
Title: Re: LED Filament clock display
Post by: DanielS on April 10, 2015, 08:43:14 AM
The shape of the LED waveform is interesting, particularly how slow the decay is.  Is that due to the junctions heating up over the course of the on period, lowering the Vf?
The thermal impedance and specific heat would not have enough bandwidth for LEDs to warm up and cool down that fast.

The reason for the stepped waveform on the LED is most likely simply the LED turning off after the AC peak (rapidly diminishing current through the series cap until IF=0) and the step on the way down must simply be the phosphor layer's stored energy radiating out.
Title: Re: LED Filament clock display
Post by: mikeselectricstuff on April 10, 2015, 08:48:31 AM
I wouldn't read too much info into the waveformshape - I just used a random photodiode in PV mode with an arbitary load resistor, so linearity, rise & fall times  are probably  highly questionable
Title: Re: LED Filament clock display
Post by: DanielS on April 10, 2015, 10:13:41 AM
I wouldn't read too much info into the waveformshape - I just used a random photodiode in PV mode with an arbitary load resistor, so linearity, rise & fall times  are probably  highly questionable
Regardless of exact scale and transfer function, the general shape definitely looks like there is current for about 4ms per half-cycle, then phosphor glow for about 3ms (LED turns off, no more blue, hence the sharp-ish drop to 2/3) and nothing of significance for the remaining 3ms.