Now I have usually many things on my plate but I hate it when things just fail to work.
Sometimes this can be just unlucky and i'm just another MTBF statistic that falls outside the histogram,
For those of you who understand such statements you know where I am,
Anyway the case in point is the LED spotlight that is pictured above. Originally I bought 3 of the items and to be fair they have lasted quite well for over a year now and the two remaining are still going strong. One of them unfortunately gave up the ghost and was faced with getting another.
Price wise they are about a tenner each in the UK and for what you get you cant complain. They emit a nice 10W soft light and are ideal for a porchlight or as I have done scattered them around the garden as backlights.
They are subtle and pleasing.
Most people would accept that and move on …..I mean why did it fail...do we care.....?
It clearly says LED not replaceable..might just as well say no user serviceable parts inside...well its a red rag to a bull for me...anyway even if they aren't we can always have a peek ...is it a good design?
Right out with the screwdriver ...time to take a deeper look...…...
Now at this juncture its time to put my preaching head on, and just like the other items I have commented on involving the mains and voltages above 60V take extreme care.
I am not responsible for anyone getting hurt with this stuff and unless you know what you are doing then don't. Its as simple as that.
If you must and you are curious enough then always disconnect the item from the source of the supply and be wary of any un discharged voltages that may persist, they still hurt.
If you must connect any meters or suchlike to measure absolute voltages then turn off the mains, connect the test meter and switch back on. Work always with one hand and better still with a suitable rccd.
Earth is earth on the casing but on the pcb diagram is not if it is isolated by a diode. If using a scope then float the test piece via an isolating transformer or else no more scope. Don't float the scope or be tempted to lift its earth pin, its bad practice and can be forgotten if left.
You have been warned....death can be fatal!
So on with the show.
Step 1: Unscrew the Offender
Now the back of the fitting has a screw type cable cover which underneath reveals 3 terminals as per the photo.
Marked accordingly LNE I removed the cable and connected another mains cable that I know was definitely live.
Connected up.... not a sausage...typical.
Turning the fitting over I note it is secured with 4 x 3 hole phillips looking security screws which aren't.
I think you can buy a tool for this but needless to say I didn't have one. Possibly this is part of CE spec. They are countersunk within the back diecast enclosure just to make you cuss louder!
Half an hour later and with some choice words the front glass was removed to reveal a reflector with a centre diffuser over the top of some leds by the looks of it...funny that.
Reflector was held by a couple of phillips and the pcb was held in an insulated cover.
Connections to the pcb are via a spliced and clamped 240V joining junction. You can see this on the right of the picture. Note also the earth that comes in and is bolted to the die cast enclosure.
I have highlighted the two areas on the diecast enclosure that hold the goodies.
The area with the blue border holds the led's which are secured with 2 screws and a layer of heatsink paste, I will describe these more fully later. The area in red is where the pcb sits.
So why this not work!
Step 2: The Nitty Gritty
So the mains enters as line and neutral[Red/Blue] and connects to the pcb one end.
There are also two wires which leave the PCB coloured red and blue which connect to the led block.
Lets do some quick checks to see if this is a quick fix such as a blown fuse etc.
With the meter on ohms check the fuse[red block] which is on the track directly after the live feed red. Dead short which is good news for once.
As the fuse has not blown it means whatever takes power has not gone short, or the device which is switching the power has not gone short or any other device across the HV rail which maybe good, however if there are any semi's involved it may be bad.
What's next ...bridge rectifier full which feeds into a magnetic filter block consisting of two caps and two magnetically coupled coils. That's two coil formers in a common ferrite enclosure. This works in two ways, keeps the common noise from the outside at bay and will keep any switching noise internal, again probably a UL or CE requirement. The caps are sitting across the mains and are rated accordingly at 400V which you can see in the photo above. Anyway this wont stop it working so what happens next. Looks like it runs into some resistors and a chip of some sort...no doubt a buck regulator as the mains is considerably high to power led's directly.
Lets check up to the fat mains cap sitting across all of this. Its rated at 400V @105 deg c so will have the rail sitting across it which these days is 220V AC RMS or 220 x 1.414[root2]=310V DC ISH.
A short digression here. Mains AC is nasty enough as it rolls through at 50Hz but at least its gracious enough to go through a zero point every 10ms, 300V of DC does not it just thumps very hard...how do I know?...don't ask.. so unless you are going through a cardiac arrest at the very moment of contact its very unwelcome. Come to think of it I expect you are likely to have one if you do....play safe at all times.
Now we could at this point tack a light bulb across the cap and see if it lights up ok but will play it safe and check with a meter the diodes in the bridge rec and the connections through the filter. All OK so now what...…..
Now in the past I have known these mains dc electrolytics to die, either due to excessive heat which creates a higher resistance and then more heat. Give away is the can which expands but this looks OK...see photo.
Right I may have overlooked the obvious here ...what if the leds are bust...of course they glow incredibly bright and create shed loads of heat...lets have a look at this block
Step 3: The LED Block
So here we have the LED block. It consists of 9 LED's which I guess are rated at 1W each although i'm not totally sure. This is odd as the backplate states 10W but I suppose they are referring to the little marketing loophole which refers to power consumed rather than the equivalent LED power. Anyway 9W it is with a percentage turned into photons....more efficient than incandescent so lets move on.
Note the plus and minus connections to the board.
Now I know that high power led's have rather different forward voltages than the bog standard led varieties and looking at the google spec for 1W devices it looks like it needs at least 4V to get them to emit light.
So I have a string of leds in series which I need to test using a power supply. 9x4 =36v...my power supply only does 30v with the wind behind it so need to split them to test.
Take a look at the construction of the plate that the top pcb is glued to. I have included some side pictures.
Its aluminium with the pcb glued directly on top to get rid of the heat but without having a direct connection to the substrate of the led I doubt its too efficient. We can get a heat gun on it later to see how hot it gets.
From photo bottom row first 5 in a line. Current limit the power supply to 100mA and they blaze away when we get to 16/20V
Top row 4 No go...ah hah...got a broken led.
Incidentally while checking these diodes I made up a quick bodge of a 9v PP3 battery and placed it across the diodes to check them individually. Make sure you have the polarity correct of course.
Just thought of another project...switchable led checker......stop it...
Lets do a one on one led until we find the culprit....hmm looks as if it has cooked if we look at the photo.
Now it so happens I have a spare led that's rated at 0.5W and wonder if that will work in the current setup.
I'm not going to do anything fancy here in removing the old one with a heat gun so just remove the old one and glue in the new one. The new one is rated at 100mA with 150mA max so if the others are running at 150 mA we may be in with a chance. What have we to lose...apart from another LED. Hang on though what is setting the parameters for the current in the leds and just how the hell are we getting from 310V DC to an approximate 45V of DC across the led string....for a switch mode supply it must be a buck regulator step down , but also as its cheap and cheerful completely non isolated.....be afraid...be very afraid!.
Lets dig deeper.
Step 4: The Driver
From earlier findings we ran up against some resistors after the rectified DC which appeared to be associated with some kind of driver chip. At this juncture I usually try to find out the name of the chip or in some cases its obvious by the components used around it. For this one its easy as its clearly marked MT7812 which is marketed by Maxictech or it was until it was superseded by others. its interesting that this board is marked version 2.3 from 2015. The datasheet is very comprehensive and gives you some application info. lets see if I can relate this to what we have here.
From the datasheet we have the full wave rectifier linked directly to the reservoir capacitor C1. In our case this is first fed through the filter network before hitting C1 . The inductor filter legs I measured as approx. 1.82mH for each leg in parallel with two caps of 220n and 150n respectively.
The resistors RST1 and RST2 are 200K each and C2 is a chip cap of about 1.5u. on this board the centre junction of RST2 and C2 has a Zener to ground rated at 14V. RST1 and 2 set the Zener current for this and would be selected to hold up the 14V to pin 3 of the IC even at VMIN which is estimated at 290v. R1 and 2 are the over voltage protection resistors at 330K and 12K. Well checking all of these seems to be fine and could check the Zener by feeding with a low voltage to check its zennering so to speak although given the dead LED I doubt if its an issue. I might check later.
Well so much for the inputs what about the o/ps and feedback?
Lets go back to our board and see if anything differs from the schematic.
First of all the really interesting bit is the pin 8 resistor Rcs. Reading the notes and looking at the internal construction it appears that the resistor here sets the current in the LED path independent of any voltage considerations and looking at our board there appears to be two resistors sitting on pin 8. One is 20 and the other is 1.3 ohm. In parallel this looks like 1.22 ohms as the 1.3 ohm resistor dominates. plugging this into the equation for peak inductor current gives 327mA.This will give a led current of 163mA which is slightly above what is rated for the 0.5W led so we could increase the resistance to slow down the current. Perhaps aim for 120mA to be on the safe side. If I had found a 1 watt led on Ebay perhaps this may have been a better option? Anyway its 10p so lets splash out.
Step 5: Equations Equations
Here is my new board with led lashed in. not very elegant but what do you expect:-) Note its a lot bigger which may help to rid some heat but doubt if my contact to the underside is particularly good. Lets fire it up with the DC power supply and see if it dies. So I connected 5 in circuit and at 34V managed to get 118mA through them. Looking at the spec for a 1W led ...this is not a COB device but a straight slab I have highlighted the salient bits on VF and current in YELLOW . VF like all leds wanders around depending on what time of day it is ...actually not really ...more like how hot it is and how much current you are trying to shift through it. The 1W version likes 140mA and will accept a peak of 260...wow that's near a 100% markup...I don't think I would try my luck at that as the MTBF would probably plummet. On the other hand my poor little substitute has 100mA as running and 150mA max and chucks out 45lm. Forward voltage not specced. The other spec has lm missing for some reason.
I know.... I have a cunning plan lets measure the original led and try and push 150mA through it .
I'll advance the volts until we get no change in brightness...apparent ….and measure the current.
Now this was interesting...@6.6v forward voltage we were shifting 150mA and was fairly bright, to get an appreciable difference you have to push it to over the 200mA level and then destruction is starting to loom. With the naked eye 120mA seemed no brighter than 150 so it would seem they are probably better off running at the 125mA level. That's not what the resistors are set to so perhaps we can mod this a little to get everything to last a little longer without compromising the light level too much. Lets plum this into excel to see what it makes of it.
We can use the calculation formula as in the data sheet.
Excel seems to think that given a running current of 123mA and a peak of 246 we should run with a resistor value of 1.625 ohms. So what is this made up of...well plumming into the online calculator coz i'm lazy
gives 30 ohm in parallel with 1.8=1.7ohm..that will do. Gives us a running current of about 120 in the inductor...
Is that it are we done?....well not quite we have some checking to do...we have opened a can of worms!!
Step 6: Confession .......Its a Dogs Life
Now in one of my other instructables I may have mentioned that we have taken in a Cocker Spaniel that was destined for the dogs home. Now she is incredibly affectionate and tears around like a whippet but she is also incredibly naughty. Dont be fooled by the looks....she will eat anything, if its on the floor and loose its fair game.
Pushed through the letterbox even better.
Now what has this to do with fixing this light and what about the confession?
Hang on i'm getting there.
Now during the testing and finding out the original duff led was to blame I carried out some testing and inadvertently increased the voltage across one of the other leds and blew it away...yes dead deceased...no longer more. They dont like it up em you see, The smoke escaped and its shuffled off. so I thought I would just plumb in another from the 50 I had purchased from ebay some time ago. You have guessed it, my wife informs me that the dog has eaten the whole string, well,chewed them resulting in a binning. The wife omitted to tell me as she thought I would not miss them...typical...so back to Ebay and i'll order some 1 watt ones.
This won't detract from the mission though so while we wait for them to arrive along with the new resistors lets have a look at the magnetic lump that feeds this string of leds.
Step 7: In-duc-tance Not Tape.
Remember referring to the schematic that the led's are fed from the HV rail through an inductor to the Drain of a fet inside the driver chip. Being an inductor and applying the full monty across it will generate a triangular current through the inductance. Do remember also that this is not an isolated circuit here and all we have to play with is the full on DC voltage from the HV rail. So current ramps up to IPk which in our modified version will be close to 250mA giving us 125mA average. It can't exceed this as the chip will sense and turn off the fet...so what governs this rate of rise? Now if we have a fast ramp the frequency will increase and the way to slow down the ramp is to add some inductance...hang on this must mean that the frequency is inversely proportional to the inductance. Take a look at the equation in the data sheet, frequency is definitely inversely locked to inductance and IPk but its also directly locked to the voltage across the LED string and the incoming voltage....so if the incoming voltage goes down then the frequency goes down...does this matter?
The answer is not a great deal but there are limits to do with the conduction times of the chip and surrounding components such as the recovery diode and the avoidance of getting into discontinuance mode. In an ideal world we would like a triangular waveform that is nearly continuous across the ramp up and down.
So lets look at some likely numbers.
Frequency that the chip can handle is 30 to 80Khz so that sets our boundaries.It also sets the size of our incoming filter although the roll point should not be affected too much.
Vin Min can be 10% lower than our 310V so lets put it at 285V.
What about our LED string...we had 9 leds that I measured as 6.6V forward drop, incidentally this fits within our spec of these 1 W leds at 5.8 to 7V...so lets use 6.6V.What about the L? hmmm where do we start...I know lets start with a value for the inductance from 100uH and sweep it forward to see what type of figures we get for the frequency, after all we have enough consts for this...in programming parlance.
Step 8: What's the Frequency Kenneth?
Some of you in the UK will remember that track which shows your age....damn it shows mine as well...moving on.
So this thing will work in the middle at 55Khz so is this the best policy?
Well for those in the know this equates to a period of about 18microseconds or the same amount of time my bank account is in the black each month...no I joke ...its pico seconds:-)
So what are the constraints.
From the data sheet Toff Min must be greater than 1.5us
Toff Max but not be greater than 400us
Max on should not be greater than 55us.Lets run the numbers.
Looks like at these parameters there are only some values that sit within the limits.
At the end of the day you need to squeeze the magnetics into a box and the driver for this is the higher the frequency the lower the inductance and there fore the smaller the coil...hooray....that's in Henries as well!
So what we got...2.4mH ----->5.8...what if we gave the chip a breather and sit it at 55Khz...that's 3.4mH.
At this point I am going to defer to the old ferroxcube program to spit out some numbers...lets look at the EFD series as they are small and low profile
Lets start at 2.4 mH which would have the chip running at its top level of around 78K. It means the inductor will be small though. however if vin goes up so does the frequency which may violate some of the constraints.
So bung in some numbers such as EFD series/inductance value/current and hit go!
boom we have a suggestion as EFD15 core material 3F3 with 125 turns. Its also 15mm across which is the same as currently mounted
This is a gapped ETD with wire size 0.224 and an RDC of 2 ohms....now this is interesting as it implies that the existing inductor on the board [I could not measure its inductance as it would mean prising it off the board] but I could measure its resistance as around 5 ohms. This would imply that it had a lot more turns.
Ok lets try the mid point at 3.4mH. Nope puts the core size up one value.
Ok lets try just under at 2.9mH...bingo EFD15 153 turns at about 3 ohms 0.2mm wire size. Frequency 64854 Hz.
So now lets up the ante and sweep the input volts what happens to our 2.9mH value. lets try at the nominal 310V.
Well as predicted the frequency has increased to compensate but only to 66231 Hz Ton and Off are within limits.
Ok so final check is to overload it at 341V. frequency 67K still within limits.
what happens if we increase the Vfv of the leds to 7V?
Again as expected our frequency climbs to 70Khz but still within limits with 11usecs of off time.
Step 9: And All the Kings Horses and All the Kings Men.......
Probably might struggle putting this back together , but lets have a go.
So what have we learnt apart from its always quicker to buy a new one.
Well it looks like the root cause of the failure was definitely one of the leds in the series string. If one dies they all die as the circuit goes open. You need to replace with the right LED as the design is critical in respect of series current and voltage especially when driving directly from the mains. In the case of the ones here they are definitely the 6V to 7V 1 watt variety at 150mA. Shopping around it looks as if they were originally backlight leds for tele sets. They are not expensive but if you shell out £5 then its half the cost of the flood however I have bought 50 which should enable me to repair these ones for some time or maybe I can look at another design perhaps.
I will include here the closing schematic of the board as traced by me with component values plumbed in where possible. At this moment I don't have the value of inductance for the on board inductor but I might try and get it off the board. Its soldered on both sides which is a pain and it may get destroyed in the process. we have proved that we have some leeway with the design as its stands and is self regulating as long as the inductance behaves Once the led's arrive I will repair and plug it in and grab some images with the scope so look back in a few weeks and I will update it. also I have attached the excel spreadsheet with some numbers for you to play with. Allegedly Maxitech have their own design program for this chip but I could not find it. Would be good to see if my numbers match theirs!
Well I hope you enjoyed this waste of time but you may have learnt something perhaps. I know I have.
as always ping me a message if you thought it was interesting or maybe even useless.
Well I bit the bullet and decided to remove the inductor from the board. yep and you guessed it the worst happened. The bottom winding on the coil snapped meaning I had to painfully remove all the wire from the bobbin and rewind it. Now without a winding machine this is tough and I counted 300T or just shy of .The wire i measured as 0.17mm .To fit in the winding window with worse than 50% winding ratio is also a killer, however I managed it and measured the inductance on this coil which was 2.49mH@5.8 ohms. This gets more exciting if we plumb these figures into the spreadsheet. With an Ipk of 250mA we are in trouble with insufficient henries causing a swing towards 80Khz cut off... I don't like it. If we put back at a higher peak say the 327mA as per the resistors all is fine and dandy with the frequency dropping back to the 60Khz marker. This is a tight design with not much leeway for dropping the current through the leds unless we add in some more inductance say aim for the 3mH marker.This needs a bigger core and more size and space.
Anyway for now we do not have to worry about such things because as long as the inductor holds up....might not I rewound it,,,,then we just need to replace the dead leds with some new ones of the correct forward voltage and current and we're done. Oh of course you need to fit them the right way round being diodes of course.
OMG Another Can
For those that have followed some of this diatribe without falling asleep you may have noticed something if you were paying attention...probably not I know I wasn't.
Well when we measured the existing inductor it was about 2.49mH according to my inductance meter. Now this is a cheapie so I have another which measured 2.84...hmmm anyway the crux of the matter is that it had close to 300 turns on it...and its in an e-core package of about 14mm x 14mm x 10. I also noticed that this core is ungapped and the two e-cores are held together with some insulting tape[sic]. Yes ....and....well 300T of 0.17 mm wire would seriously saturate a core of such miniscule proportions especially ungapped. Try this using the ferroxcube program or just sit down and input the numbers for flux density. Obviously gapping takes time and effort and this thing is cheap so what's going on. Why are there so many turns when the ferrocube proggie says use around a 100 and gap the core. Interestingly if you don't gap the core the number of turns go down but the wire size increases along with the core size...more cost and space.
Clearly its a government plot to......no no that's the other forum.....no I would guess that this core has a distributed gap by virtue of its material. Although its slightly ferrite its definitely lacking in the permeability arena. This trade off for size is substituted for core loss as we need more turns to create the required inductance due to the increased reluctance., this results in the small wire size to fit the window and the resistance increases accordingly. This copper core loss is I^2R= 6R x 163mA=150mW....ah ha...know where some of my 1 W went to now...heating up the enclosure in an attempt to trash the led string and get me to buy another.....well it worked I suppose.
How about I just bolt a 10W COB diode to the back of this flood and power directly from the mains...now theres a thought ...watch this space.
Beanhauler November 2018