there here ( what MR.X using ) LED

[knna]

The driver is a (24V max) constant current device (500mA I believe). Actually most CC drivers have a CV first stage.

I agree that connecting a high current low voltage driver to several string in parallel is not the most efficient way, but most LED lamps do so. At least those certified, because shock test are way easier to pass that way. When you work with high voltage (anything over 60V), you need a double isolated PS, way thicker dielectrics (worse thermal conductance), earth grounding...nothing of that is required with low voltage.

But connecting several strings in parallel requires an additional way to balancing them, in order to avoid initial difference in brightness and specially, problems if a LED fail. I was talking with EVO's designer about that. There are good solutions to that.

They had diminished the problem of initial difference in brightness by grouping strings by Vf. Likely the copper path is not a problem for a circuit running at 500mA, differential drop in voltage is very low for any standard copper thickness used in PCB. Except if a trace is damaged, of course.I didnt ask for the copper thickness used on the PCB, but standard 1oz often works fine, 2oz or higher is for sure enough to keep voltage drop difference between string completely negligible.

About the sensibility of human eye to difference in brightness, it is not the same looking at reflected light than looking directly to the light sources, and it makes a huge difference if you compare the two light sources one along the other than if not. When you compare them side to side, eye/brain find easily difference in brightness and chromaticity.

Color tone is easier to understand in an example, although eye/brain adaptation mechanism is very similar. Put a lamp of a given CCT,for example a 3000K warm white, turn of and try a 4000K white lamp. It is difficult to notice the difference after a brief period of adaptation. But put turn on them at the time side to side, and you immediately notice the difference in tone.

Anyway, I agree no providing a means to balance current between strings in parallel is a flaw. However the lamp may work perfectly still so. It would be required to know the actual difference in brightness, voltage and current on Habeeb's lamp to have an answer on this, if he can/want to solve it himself or prefer to ask for a change.

Im not involved on any means on the development of this lamp, just I know the people who did it, they are good friends but nothing else. I never would have released a lamp with a circuit like that. But they told me they had tested it and that difference are small and reliability good, with first units working for near 2 years with minimal light loss. I believe them, so probably it is a minor problem, but any one having one should check it for themselves and see if it is so. I dont know if Habeeb's lamp is in good condition or not, or if his lamp is representative or not. Actually, I never had an EVO lamp in my hands.

I just try to help Habeebs on this. If he give me Vf measurements, which dont need anything else than a multimeter and a couple minutes, I could say more accurately and suggest the best to do.

 

[Shafto]

knna, check out the datasheet for the driver, I posted it.

It's a 12V 5A driver, not 24V 500mA. A circuit board trace even at 2oz. would have to be quite large for 5A.

While I agree that it's possible to put together an array with series parallel strings, it's never as good, and there will always be imbalance in current and brightness among the LEDs. This will vary among each array produced. That's why Habeeb is seeing issues while others will not. It's an inconsistency problem that will cause issues with performance and longevity. It will be hard to compare these with other light sources, or even to each other, because no one will be alike another.

The problem can worsen as the LEDs age, or even correct itself in some situations, but that's just the thing, inconsistency, you'll never know.

I think for now it's the best LED product on the market, but there's still a lot of room for improvement.

About the human eye, as with anything, not just light, it's easier to compare two things side by side to see the differences than remember one thing to compare to another. I will phrase it better this time. -If you can see the difference with your eye, there's a large current imbalance.

 

[knna]

The version used is the 24V, 2500mA (sorry, I missed the 2 on past post).

I fully agree with the inconsistency of such design, more when it is easy to solve with minimal components count and cost. However if the groups are well matched in Vf and the lamp not run too hot (which seems to be the case), it may work decently anyway. Main problem I see on that design is if one LED fail, but that is the case when ask for the guarantee is a must. Total LED failure mostly happen at the beginning of the lamp life but it is very rare for LEDs working well into specs past the first 1000h (usually any single LED failure happens within the first 100h).

The key point on doing this design able to work decently is the relative low current used as compared to their max rating and the large area for dissipating heat. That limits the current unbalance and ensure it keeps all LEDs running into specs and link the current unbalance directly with the differences in Vf of each string, so if they are small, the lack of accurate current control not becomes a severe problem. It could be so if the initial difference in Vf is noticeable.

So all depends of how large is the difference in Vf. For an string about 23V, a difference of 1V for the whole string at a given current (500mA in this case) is large, about 4%, and can lead to a difference in current of 30% and over. On the other hand, differences of 0.5V mean a relative difference of 2% and that won't result on a large difference in current (say, 20% as max). Usually LEDs from same batch are well matched on Vf and varies little from one to the another so it is to expect not excessive differences,specially if strings are grouped by Vf, so differences in LEDs in a string gets averaged. For sure not optimal, but may work decently.

Ive build several strings from the same batch of LEDs and usually their difference in Vf is well below 0.5V for a 24V string. Often, about 0.1-0.2V. But you dont have any guarantee of that so always it is possible than a string have a larger difference in Vf and if connecting it in parallel, then the difference in current (and brightness) is noticeable.

What I did was suggest EVO's designer how to solve that current unbalance and he told me they are going to fix it on next version. But until that is done, people using them must check well the difference they see in brightness when getting the lamp. But likely, with the lens it is not easy to see because the angle affects how bright you see each group: as you move yourself and see each group from a different position, you see different brightness from each. I trust more on measurements, either electrical or optical.

Looking forward for more info from Habeeb. Although I noticed after writing it that measuring voltages wont give any clue, as all the groups will run at same Vf, but different currents. And for measuring current you need to open the circuit, not always possible with a PCB.

 

[AlexTrebek]

I live in Canada.
Evo 50 with shipping retails for 320$ CAN.
Evo 70 with shipping retails for 370$ CAN.
Evo 90 with shipping retails for 420$ CAN.
For anyone wondering about prices.

 

[whazzup]

Sorry for the off-topic but something completely different: I always see people worrying about inrushes, but never when they use LED. Make not mistake about it, also switching power supplies can have a huge inrush current. 60A@230V according to the specs in the mentioned little 60W Switching Power Supply.

 

[knna]

Sorry for the off-topic but something completely different: I always see people worrying about inrushes, but never when they use LED. Make not mistake about it, also switching power supplies can have a huge inrush current. 60A@230V according to the specs in the mentioned little 60W Switching Power Supply.

Yep, but this is only a problem for the PS electronics, not for the line, as such inrush currents are below ms. The time it takes input cap to be charged, so only the PS/driver notice it. And usually the inrush current is about 1A.

However, what may be a concern is the low power factor of some drivers, specially the cheap ones. Meanwell has for example a range of drivers with active power factor correction, but those which havent it often has low power factors. Not a problem usually at home, but if you are using heavy loaded lines, use a driver with active PFC.

 

[Shafto]

The version used is the 24V, 2500mA (sorry, I missed the 2 on past post).

In the first post you can see the driver included. It's the 12V version, ELN-60-12D. Maybe now they are using the 24V version.

 

[knna]

In the first post you can see the driver included. It's the 12V version, ELN-60-12D. Maybe now they are using the 24V version.

You are right , its the 12V 5A version. I just knew what designer told me, which makes more sense because each group of 9 LEDs has a Vf a little over 23V. Probably they have changed that.

Im struggling trying to guess how did they designed it to work from a 12V source. No way of splitting 6 reds and 3 whites on 2 equal strings. It would mean 3 paralleled strings of 1 white and 2 reds each on each group. So short strings way more likely will suffer from current unbalances and different brightness.

While I reasoned it could work decently from a CC 24V source still not having a system for balancing current, I dont think the same with strings below 8V.

 

[Shafto]

knna, I think they maybe used 2 strings for each group of 9, but I'm not certain either.

The only way I can think of is 5*2.25V = 11.25V for one string of red, and then, 3*3.4V + 2.25 = 12.45, but it's not a very even split at all.

The 24V driver would make more sense indeed.

Maybe they are using only 8V strings, but the driver cannot do this, it can only go down to 10.8V in CC mode. Even the 9V version of the driver can only do 8.7V in CC.

 

[knna]

The 12V driver has an operating range of 6-12V, so it wont have any problem driving 9* 8V strings and an average current of 550mA each (down to 420mA by adjusting SVR2 to the minimum). The 10.8V is the minimum adjustment for SRV1 potentiometer, its done to reduce losses in the driver when it is going to work with voltages below the max.

 

[Shafto]

Of course.. I was about to ask if you had found a more detailed datasheet, for some reason I didn't notice the operating voltage line at first.

Strings of 3 it is then.

Larger strings are usually balanced better because they often average out, but there's also the ability to go much further out of spec if you happen to stack a bunch of high or low Vf LEDs in the same large string. Checking Vf of surface mount LEDs before soldering them into place is also a hard thing to do, to know how to arrange them for even Vf distribution.

 

[alkalien]

We do strings of up to 54V and did test a few of them before connecting. The difference in Vf combined has allways been arround +-0.1V at most. I wouldn't mix bins though.

Edit: I did even try to meassure the Vfs while running the strings parallel until I realised how stupid that was. That was of course after I partied because the voltages seemed super equal...

 

[habeeb]

well, you guys have lost, me.. but glad your describing it and talking about it... knowledge is good..

well, here are some shots I took of the power difference, keep in mind difference of light reflecting off the meter housing / where the light meter was placed... I tried to find the highest number moving the meter around.. this is obviously not scientific, but shows you what I'm talking about, and how I feel it's un-acceptable to be this varied..

once again I spotted this by my eye when putting your hand under the arrays' you can see the power difference..

first ( next to power coming in )


middle ( keep in mind it has two lights on either side shinning on the meter )


last ( farthest from power source )



lucky for me, I have two units, so I just "flip" one and then have high power on either side with the middle hopefully coming in at a middle ground.. but i think this is not quality we have come to expect from someone who has taken the time to produce what we think to be a good example of LED tech..

also the numbers on the meter are from the last setting on the unit..




keep the info rolling guys..

 

[GP73LPC]

wow, that is quite interesting...

so one end of the bar puts out much more lumens....

kinda got me re-thinking the EVO's now

 

[knna]

Unacceptable difference, Habeeb

With low voltage strings, small absolute differences in Vf does a large difference in percentage. At about 7.7V each string, just 0.1V difference is about than 1.3%. Small differences in percentage leads to large differences in current. Over 3% is difficult to manage and the problem aggravates as the lamp gets hotter.

I've noticed the brightest string is the closer to the input wire, so there is the possibility the problem is the final solder points between the 3 pieces are not well done. It is very plausible because those solder points are done when the lamp is already mounted on the heatsink, so it is difficult to get the area hot enough. You can check easily if it contributes to the problem. Just take a voltmeter (or a multimeter on V on DC position) and put the pointer at each side of the two solder points. A cold solder point might add electrical resistance. If so, you will see some value on the display (it should show 0.0V if the solder point is good) and depending of the value, it may explain the problem and would be very easy to repair (just rework the solder point).

If not, I would ask for a change for a unit working fine. If I understood well the designer, now they use 24V strings and it should minimize the problem. Anyway, I will call him tomorrow or on Monday and suggest him to add a current balancing system ASAP. It is sad a good unit fail from a problem which can be solved cheap and easily.

 

[habeeb]

^ knna, thanks for the info..

umm with the voltmeter.. wondering if I did it right? I set to DC v. then when I touch the to ends together it should read zero correct? I touched the solder points on the first cluster right? tried two different times, with a on/off before the second try,to see if I did it right, and it read both times - 7.08 . does that sound right?

 

[knna]

Yeah, it sounds right. But I'm interested on the 4 solder points between the cluster. The lamp is made of 3 pieces, the central piece has 4 visible solder points, 2 at each side. Put the voltmeter pointers at each side of each solder point, I need to know the value of the 4 solder points.

The + or - sign just depend on the order you put the pointers.

I need the value in volts, so I need to know the scale of the voltmeter. It can be volts, or volts x10, x100, +1000, meaning the value of 7.08 may be volts (if you put the pointers on the ends of LED cluster, although it sounds a little low, but possible), or 0.7V (if you put the pointers on one solder point between pieces, there is the problem, it is a very high resistance), 0.07V (possible, but I dont believe the voltmeter shows up to 1/10000 of volt sensibility).

If you have doubts about the scale, put a pic of the vmeter.

Happy Christmax!

 

[knna]

I talked the other day again with EVO's designer, he told me currently they use 24V strings and that it avoid the problem of current unbalancing. It happened on the first batch which uses 12V drivers, the unit of Habeeb, especially when some final solder points were cold.

So check them as I explained on last post so we can see if you can fix it easily. Otherwise, call them and ask for a replacement.

 

[GP73LPC]

thanks for that update knna.

that's good to know the problem has been corrected...

 

[Shafto]

Just to be clear, the problem hasn't been corrected, it's been scaled back.

The problem is still there, it's just maybe not so bad that you can tell with your eyes any more.

The correct way to do it was outlined earlier. A single series string of LEDs for each driver.

I'm not trying to be a jerk, I just want consumers to know what's up so they can start to demand a quality product from vendors. Doing it wrong to save $10-15 in driver components on each unit would never fly with my company.