EEVblog Electronics Community Forum
EEVblog => EEVblog Specific => Topic started by: EEVblog on October 09, 2013, 11:01:47 am
-
What's inside a modern 18W LED Fluoro T8 replacement tube from German company Mueller Licht?
LED tube design Safety: http://youtu.be/saYtnaBp4QA?t=16m30s (http://youtu.be/saYtnaBp4QA?t=16m30s)
Datasheets:
http://www.nxp.com/documents/data_sheet/SSL2109T.pdf (http://www.nxp.com/documents/data_sheet/SSL2109T.pdf)
http://www.philipslumileds.com/uploads/409/DS205-pdf (http://www.philipslumileds.com/uploads/409/DS205-pdf)
http://www.mueller-licht.de/ (http://www.mueller-licht.de/)
(too short to be this LED)
http://www.lextar.com/promote/2013_GILE/PDF/Lextar_SMD_series-13.5.pdf (http://www.lextar.com/promote/2013_GILE/PDF/Lextar_SMD_series-13.5.pdf)
EEVblog #533 - LED Fluoro Tube Teardown (https://www.youtube.com/watch?v=PFcC315vuQ8#)
-
Just wondering what the 10ohm input resistor is for, its in the nxp? datasheet schematic. Also why did they put the fuse after the MOV.
The data sheet for the light said >0.9 power factor, that seems a bit of a stretch considering the lack of PFC correction.
edit: apparently it is for current inrush. By my calcs it will use 0.8 of a watt, which does effect the efficiency a bit.
-
Those tiny LEDs are pretty amazing. I've used similar ones made by Nichia ( NSSL157AT / NSSW157AT ) and was afraid to overheat them, but it didn't happen. I made the effort to add copper-pour heatsinks on both sides of the PCB, which turned out to be overkill. The single-sided solution is very sufficient in terms of cooling.
-
What was that you said about being surprised there wasn't a resistor between each row of parallel LED's? Is there a problem putting LEDs in a parallel/series configuration like this without resistors?
So if I wanted to do have a string of LEDs and the supply voltage didn't allow them all to be in series, I thought I could do exactly what those fluro tubes did and have some in series and then put those in parallel. Didn't know there should be a resistor between each set of them.
Sorry for the newbie question but I'm dying to know.
-
Yes and no ;-)
Current regulation for each string is safer, but more expensive. If for some reason one LED blows in a paralleled string and fails shorted, more current will pass through that string. This will probably cause more LEDs to fail.
This appnote discusses calculation for current-sharing resistors for paralleled LED strings.
http://www.ti.com/lit/an/slva325/slva325.pdf (http://www.ti.com/lit/an/slva325/slva325.pdf)
-
LEDs (especially when not from the same batch or even different types) have a slightly different voltage drop. If you connect them in parallel the LED with the lowest voltage drop draws a higher current. Due to the higher current, the LED gets warmer than the other LED which reduces the forward voltage and further increses the current. To avoid this, a resistor is added.
These resistors are often omitted because they waste power, add additional costs and they are not always neccessary: The LEDs mounted on such boards come from the same reel, so they probably have a similiar forward voltage (you can get LEDs sorted by their forward voltage). White (and blue) LEDs also have a higher internal resistance than green and red ones.
-
Dave, just in case if you want to know how it is pronounced right :) at 0:10: Müller-Licht GmbH, Bremen (https://www.youtube.com/watch?v=iV271znFIL4#ws)
But I guess the company is not as big as for example Osram, I never heard of it.
-
dave, they probably underpower those leds to keep heat-dissipation under control
even if they could get all the heat from the led into de aluminium-profile it would still be inside a plastik tube inside a case mounted under the ceiling
no chance of getting rid of the heat without active cooling
-
LEDs (especially when not from the same batch or even different types) have a slightly different voltage drop. If you connect them in parallel the LED with the lowest voltage drop draws a higher current. Due to the higher current, the LED gets warmer than the other LED which reduces the forward voltage and further increses the current. To avoid this, a resistor is added.
These resistors are often omitted because they waste power, add additional costs and they are not always neccessary: The LEDs mounted on such boards come from the same reel, so they probably have a similiar forward voltage (you can get LEDs sorted by their forward voltage). White (and blue) LEDs also have a higher internal resistance than green and red ones.
Thank you for that explanation.
So, since some of the LEDs are in series, they must all take the same current and that's understandable and okay. But, if you hook them up in parallel, one set may have an internal resistance higher or lower than the others and will then consume more or less current than the other strips. I get that now. So how will adding a resistor help? Wouldn't you need to customize that resistor to be an exact and customized value for each Fluro tube that you create? You would measure the internal resistance of each strip and then put custom resistors in series to each strip to make them all the same.
If, in your manufacturing process, you just put the same value resistor on all four strips, some strips may still consume more current because of the total resistance after considering the LEDs plus the resistor will still be different.
-
Again, I suggest to read the appnote. It is not difficult to understand.
Or in other words: RTFM.
-
LEDs (especially when not from the same batch or even different types) have a slightly different voltage drop. If you connect them in parallel the LED with the lowest voltage drop draws a higher current. Due to the higher current, the LED gets warmer than the other LED which reduces the forward voltage and further increses the current. To avoid this, a resistor is added.
These resistors are often omitted because they waste power, add additional costs and they are not always neccessary: The LEDs mounted on such boards come from the same reel, so they probably have a similiar forward voltage (you can get LEDs sorted by their forward voltage). White (and blue) LEDs also have a higher internal resistance than green and red ones.
true and not true.
LEd's are binned according to color temperature during test. the are also very tihgtl controlled in terms of Vf.
If you go to the local store for 20 led's you have no idea where they come from. you may have 5 from this and 9 from that.. depending on wheat the guy has in his little drawer.
Users lik Mueller buy led's in bulk. these are tightly controlled and their specs ARE known. you can bet money that mueller went to their le dmanufacturer and have filtering done on tolerances... so in these designs you can get away without the balancing resistors because you have supply chain control. and any resistor saved means more profit for them.
as a hobbyist : can't do that.. put resistors.
-
LEDs may be underpowered because of three reasons:
1. As smashIt said, keeping thermal under control.
2. Current sharing. If they would be supplied with 240mA (60mA*4) then they need perfect current sharing not to go over 60mA. If current sharing is not ideal then they must limit the current in the string that gets the max current. Assumption then is that one string gets 60mA and other strings get less than 60mA. Then, in total, you need to supply less than 240mA. I guess if they know the forward voltage distribution in a batch they can estimate what is the probability of getting a decent current sharing. Maybe its a good idea for a new massive number of measurements video :)
3. Lower current = more lm/W.
-
I commented on the previous video, that I wonder what difference the diffuser make, I think from the ceiling is far enough to get the light spread and not bright spots on your desk.
The plastic tubing is mostly to user protection since it's not isolated.
-
I wonder what difference the diffuser make, I think from the ceiling is far enough to get the light spread and not bright spots on your desk.
The diffuser is there to lower the maximum luminous emittance (lumens emitted per area). Lowering this value makes the light source easier on the eyes when you look at it directly. You don't have a number of very bright spots but a big, reasonably bright, surface. This is why OLED panels may be a win in the future. They don't require specific optics and diffusers to do just that.
-
I thought they would be using something like IMS PCB's (http://en.wikipedia.org/wiki/Insulated_metal_substrate#Insulated_metal_substrate (http://en.wikipedia.org/wiki/Insulated_metal_substrate#Insulated_metal_substrate)) for applications like this but thats perhaps more expensive than making a aluminium 'cooling clip' like they did now.
-
Hi Dave,
while I appreciate your fondness for German engineering, I think that the only german part of this LED tube is the umlaut in the name.
Müller Light (which I never heard of before) seems to be mostly a reseller and distributor. I could not find any indication on their website that they do any electrical engineering at all. Or as they write themselves: "our quality standards are 'Made in Germany'". Just the standards, nothing else.
In addition to the use of some cloned chip, the use of this ugly serif font for the silkscreen (which the Chinese seem to be so fond of that they use it everywhere) indicates to me that this LED tube has not seen any german engineering and is just another "One-Hung-Low" product, including the LEDs which I doubt to be real Luxeons.
But then again ALDI is not the place where you would expect to find high quality stuff -- just reasonable quality at competitive prices.
Cheers,
Thomas
-
Probably true.
I've had very bad experience with Müller Licht incandescent and halogen bulbs. Didn't last more than 3 months and many of them were already broken when delivered. Cheapest crap you can think of.
-
My friend said his work had to gut all the old light fixtures when they changed over to led tubes recently. I couldnt understand why they didnt put smarts in the tube (like we just saw). What would electricians be doing if not simply swapping in new tubes?
-
while I appreciate your fondness for German engineering, I think that the only german part of this LED tube is the umlaut in the name.
Dave,
I too think you went pretty soft on the design. The meter-long mains-level trace millimeters from the LEDs (called out in a text overlay on the video) and the use of a cheap brand of filter capacitors are two things you would have been much harder on had it been a cheap Chinese product. I would only go so far as to say the design passes the Australian safety requirements ... otherwise it's only marginally good.
-
The meter-long mains-level trace millimeters from the LEDs
I'm not sure if this is a problem. It's filtered and rectified mains. Will mains transients also occur on it? And the LEDs are not isolated from mains anyway.
-
(..)so in these designs you can get away without the balancing resistors because you have supply chain control. and any resistor saved means more profit for them.(..)
Hi. Here is how I see that problem from control theory perspective:
The terminal temperatures variation of these LEDs cannot be controlled with the use of tighter/more stringent tolerances. It is because of an internal property of the differential equation that describes the process. No matter how hard you try, the resolution of (linear) approximation of those (nonlinear in general) equations has an unstable element (with positive exponent).
A basic concept (http://en.wikipedia.org/wiki/Linear_differential_equations)
That is - the tolerances control how much positive the exponent is but can never make exponent negative (and thus stable) so that the terminal temperatures are more and more even with time. Same applies to BJT transistors connected in parallel - if you do not provide some governor then eventually they die one by one.
Of course it is not that without some explicit control the difference of temperatures rises infinitely because some other phenomena also take part in that equation (like thermal coupling, limited temperature rise of the LED itself, copper tracks that have some resistance etc).
The fact is that the more variation in between LEDs terminal temperatures, the less efficient/competitive the design really is. Resistors are one of the most popular solutions because the attractive tradeoff in between losses and temperatures. Alternative designs could connect LEDs in series or drive two strings with 50% PWM etc but the resistors are typically "be$t".
I suppose that one of the reasons for the poor temperature/current stabilization could be that the EU directive about energy efficiency requires that it is 18W/1700lm light source only for 10 minutes of testing and not during the marketed life of 30000h.
-
The data sheet for the light said >0.9 power factor, that seems a bit of a stretch considering the lack of PFC correction.
The existing ballast coil would act as a passive PFC. Combine that with what would normally be an undersized filter cap and the power factor can be pretty good.
-
Cheapest crap you can think of.
Buying it through Aldi makes it pretty reasonable however.
Meets Australian standards and you can take it back for a full refund after 2 months for any reason makes up for not top notch quality.
And as Dave said, they will probably do the job quite well for quite some time.
-
Dave,
What about the RF interference and/or electrical interference it creates with comparison to old-style neon lights?
Would it be possible to investigate this?
Would it be thus better to light your lab with such lamps instead of the classic daylight neon tubes?
@see: http://www.signindustry.com/neon/articles/2004-02-16-RadioFrequency-RC-.php3 (http://www.signindustry.com/neon/articles/2004-02-16-RadioFrequency-RC-.php3) & http://www.audio-technica.com/cms/site/d0c406516fd8b42b/ (http://www.audio-technica.com/cms/site/d0c406516fd8b42b/) for description and examples of the problem they used to cause
-
I would only go so far as to say the design passes the Australian safety requirements ... otherwise it's only marginally good.
It's only marginally priced.
-
Fluorescent lamps are infamous for EMI issues since they deal with high voltages at high frequencies. LEDs are DC so the EMI they produce is comparable to that of a common switching power supply. In applications where the EMI must be kept as low as possible, LEDs can be driven with a linear power supply and still get very good efficiency.
-
With all the chatter over the quality Dave, now that these have been up for few weeks, are you still getting the same light output from them, From experience with some 12V strings from china :-[ I placed under the counters in the kitchen, I am noticing a distinct reduction in brightness after a month or two, the voltage supplied is stable and the same as day one, but it defiantly seems dimmer. I do not have a meter to test the output but based on the fact that when they went in, we did not have to use the older tube lights or the central incandescent at all and now am finding it is almost every time I'm in the kitchen to do anything... there going down fast.
If yours are truly top notch German engineering rather than the cheaper china made brand, the proof would be in the longer term light output stability ?
Up for a re-measure Dave, and quieting down the crowds, although every couple of months a re-test would be nice also, maybe just posted in this thread ?
-
Thanks for this fine teardown. I run the lights for a year both in continuous operation and intermittent operation. They work well and I like the neutral and flicker-free light.
So I like to see what is in.
as a hobbyist : can't do that.. put resistors.
In my opinion you can omit the series resistors in the 4 LED strings for statistical reasons.
In each string we have 24 LEDs in series. So the statistic helps to reduce or the tolerances of the stings. This follows the "Central Limit Theorem".
The more components are connected in series in a string, the lower the deviation in the total resistance. So, the single strings gets more equal to each other by increasing the number of components in a string.
To see this, you can play with the following Java-Applet:
http://www.chem.uoa.gr/applets/appletcentrallimit/appl_centrallimit2.html (http://www.chem.uoa.gr/applets/appletcentrallimit/appl_centrallimit2.html)
Probably you have to add the url into the sitelist of the JAVA control center first.
Select a distribution using the radio buttons, e.g. normal distribution or Gaussian and increase the sample size up to 24. So you can see how the relative tolerance decreases.
It is interesting to see that this works not only for Gaussian distribution but also with "odd" distributions.
-
I wonder if they will be/are marketed with the same name in the UK. The Müller dairy have been selling Müller Light yoghurt (http://www.mullerdairy.co.uk/our-products/mullerlight) here for many years.
-
In my opinion you can omit the series resistors in the 4 LED strings for statistical reasons.
In each string we have 24 LEDs in series. So the statistic helps to reduce or the tolerances of the stings. This follows the "Central Limit Theorem".
The more components are connected in series in a string, the lower the deviation in the total resistance. So, the single strings gets more equal to each other by increasing the number of components in a string.
To see this, you can play with the following Java-Applet:
http://www.chem.uoa.gr/applets/appletcentrallimit/appl_centrallimit2.html (http://www.chem.uoa.gr/applets/appletcentrallimit/appl_centrallimit2.html)
Probably you have to add the url into the sitelist of the JAVA control center first.
Select a distribution using the radio buttons, e.g. normal distribution or Gaussian and increase the sample size up to 24. So you can see how the relative tolerance decreases.
It is interesting to see that this works not only for Gaussian distribution but also with "odd" distributions.
Sorry but this is a newbie mistake.
You can't assume a gaussian distribution because chances are the LEDs resistors or whatever are from the same batch, or come from the same manufacturer, made with the same machines etc.
This much I know but I am sure that someone else who knows more about this can explain more fully.
-
Sorry but this is a newbie mistake.
You can't assume a gaussian distribution because chances are the LEDs resistors or whatever are from the same batch, or come from the same manufacturer, made with the same machines etc.
This much I know but I am sure that someone else who knows more about this can explain more fully.
Looks like the gaussian distribution describes resistor variations very accurately, even from the same machine and the same batch:
https://www.youtube.com/watch?v=1WAhTdWErrU (https://www.youtube.com/watch?v=1WAhTdWErrU)
-
Or not.
http://www.eevblog.com/2011/11/14/eevblog-216-gaussian-resistor-redux/ (http://www.eevblog.com/2011/11/14/eevblog-216-gaussian-resistor-redux/)
Let alone sorted/binned LEDs.
-
Sorry but this is a newbie mistake.
Really is it? May be yes. The question is by whose side? ;-)
You can't assume a gaussian distribution
I totally agree. But this is not necessary. The Central Limit Theorem does not need a Gaussian distribution as an input. You can take nearly every distribution as an input, the Gaussian distribution will be the OUPUT. Please take a few minutes playing withe the applet in http://www.chem.uoa.gr/applets/appletcentrallimit/appl_centrallimit2.html (http://www.chem.uoa.gr/applets/appletcentrallimit/appl_centrallimit2.html) . It ist amazing to see that almost "all gets Gaussian" if the number of members in the chain is high enough. And 24 seems to be high enough. The convergence is faster if you have already Gaussian input data but it also works with oher input distribution.
-
But this is not necessary. The Central Limit Theorem does not need a Gaussian distribution as an input.
Ok I'll concede this, It seems I didn't read your earlier post properly.
You are only looking for similarity between the strings, you are not looking to reduce the overall tolerance.
I was using discussion in this thread as a basis for my statement.
https://www.eevblog.com/forum/projects/tolerance-of-resistors-in-series-and-parallel/ (https://www.eevblog.com/forum/projects/tolerance-of-resistors-in-series-and-parallel/)
but I will admit that statistics is not my strong point.
At uni I remember sitting in stats and thinking fondly about jumping out of the window. Whereas I enjoyed almost all the other subjects.
As far as the LED flouro design goes, it does seem to pose a number of risks. Especially in the grouping of the strings that will mean they have slightly different thermal operating points. Also I doubt a pick and place machine is a truly random selector.
But on the other hand, the tubes seem to work, they are probably running comfortably within spec. So they are moot points.
Any knowledge of a failure rate?
-
I currently have a number of the 2 ft tubes operating for short periods a day for the last 6 or so months and have not noticed any light output reduction, perceptible at least!, or outright failures. So will just keep going until there is a failure, either light output or completely dead, and report such here.
Other than personal experience I suspect it will be unlikely any one will get hold of any statistical info relating to these 'cheapo' lights!.
-
Bought four of these last autumn but only installed one at the time. That was a single fixture with a conventional ballast. There was a slight hum which I suspected came from said inductor/ballast, but it was quite ok.
Now I installed two in a fixture that had a dual electronic ballast. I rewired the fixture so both LED tubes are fed directly from mains. All four produced very audible hum, but in some it was more of the 100 Hz type, in others, it was higher order harmonics. Interestingly, the one that was quietest in the conventional setup was loudest when connected directly to mains.
Looking at the video, it seems the 1 mH inductor from the data sheet is there (actually, I can see two black inductors) but the 10 R inrush limiter has gone missing. I suppose they omitted it because your average conventional ballast will have at least that resistance. So I wonder if putting that resistor back in will get rid of the noise.
If all currents were sinosoidal, at 18 W, the lamp would be drawing 78 mA, hence the resistor would dissipate 780 mW or 4.3% of rated power. It is probably even worse because the capacitor gets charged only at peak voltage, so this could easily be 1 to 1.5 W.
-
So I wonder if putting that resistor back in will get rid of the noise.
It is an easy try, so it is worth while to check this.
But I do not expect too much. Noise is created in some cases by capacitors but mostly by inductors.
I assume that the LED driver has some kind of closed loop. So introducing an additional resistor you will decrease input voltage and the controller will increase input current.