flip chip versus wire bond die efficiency
- Thread starter snarfer
- Start date
Help Support Candle Power Flashlight Forum
My understanding is that the flip chip is not a true top emitter when not encapsulated. That is a small percentage of the output from the die is emitted from the sapphire edges. The spec sheet also shows the radiation pattern with a dome encapsulant. The binning note mentions a highly reflective mounting surface so in a way they are recommending the best implementation for the flip chip and the most likely configuration where the end users measurements will correlate with the specs.
For a wire bond product like the Cree a manufacturer will flash test the array before encapsulation which will typically include a phosphor blend. This data will better correlate with Cree's specs.
For a wire bond product like the Cree a manufacturer will flash test the array before encapsulation which will typically include a phosphor blend. This data will better correlate with Cree's specs.
That sounds plausible. It says in one of the datasheets that about 10% of the output is radiated from the edges.
But they could easily measure the output including the edge radiation without encapsulating. So why use a different metric?
I ask this question because I've been testing flip chip dies without encapsulation and coming up with figures that are about 50% of what is written on the bin label.
It's a mystery where the lumens are going. I though it might be the soldering. Tried different types of solder, different substrates. No change. Tried encapsulating with methyl silicone and then phenyl silicone. Only 10% improvement. Tried a 99% reflective surface around the die. No change.
I seriously begin to suspect there must be some trick I don't know about. Why is it so much harder to get light out of a flip chip die than a wire bond die?
But they could easily measure the output including the edge radiation without encapsulating. So why use a different metric?
I ask this question because I've been testing flip chip dies without encapsulation and coming up with figures that are about 50% of what is written on the bin label.
It's a mystery where the lumens are going. I though it might be the soldering. Tried different types of solder, different substrates. No change. Tried encapsulating with methyl silicone and then phenyl silicone. Only 10% improvement. Tried a 99% reflective surface around the die. No change.
I seriously begin to suspect there must be some trick I don't know about. Why is it so much harder to get light out of a flip chip die than a wire bond die?
Wow, 50%. That's a lot of missing energy there. I assume you meant mW and not lumens though (even though that would explain the low numbers!). One of our early builds with the first revision flip chips had too much solder paste. Apparently the first rev had small solder balls on the pads which were not taken into account at the board house. The second revision chips eliminated the balls.
Do you have technical support at Philips?
Do you have technical support at Philips?
Yes, I've got support at both Philips and Cree. But they haven't been as helpful as you might expect.
And yes I'm talking about mW not lumens.
Interesting about the solder balls. We were using the first generation dies but we used SAC solder and lower temperatures so the AuSn solder balls didn't actually do anything. We thought that maybe the bond quality was the issue but we are putting the chips directly on the PCB so we cannot use eutectic bonding temperatures.
Finally we got some of the second Gen chips. They are specifically designed for SAC. I put one on a star using a hot plate and got exactly the same output as the previous. 300 mW from an emitter binned at 600 mW.
Have you ever measured the output without encapsulation? Could this be normal?
And yes I'm talking about mW not lumens.
Interesting about the solder balls. We were using the first generation dies but we used SAC solder and lower temperatures so the AuSn solder balls didn't actually do anything. We thought that maybe the bond quality was the issue but we are putting the chips directly on the PCB so we cannot use eutectic bonding temperatures.
Finally we got some of the second Gen chips. They are specifically designed for SAC. I put one on a star using a hot plate and got exactly the same output as the previous. 300 mW from an emitter binned at 600 mW.
Have you ever measured the output without encapsulation? Could this be normal?
DIWdiver
Flashlight Enthusiast
Who's talking about either mW or Lm? I don't see any of that.
This is really interesting and I feel like I'm missing something!
This is really interesting and I feel like I'm missing something!
DIW- the lm and mW reference is to the comment from snarfer about not knowing where the lumens went. I came off as being a little bit of a jerk although not meaning to with my back handed reference to the die being measured in mW. The multiplier to get lumens from mW at ~450nm would make a reading of 300lm pretty good. Not so good in mW for a 600mW die.
I don't have data for just the die at hand but I'll look into it tomorrow. We have used the flip chip in a four and 12 die array both first and second generation. Both performed a little lower in total output in our remote phosphor application than we expected but only by about 10-15%. Too much to be acceptable but not the 50% low you are seeing. Like I mentioned before the biggest issue we have had is with the solder paste layer being way too thick. It's not to the point of tombstoning but you can see the tip under a scope.
Are you testing these in an array or as singles? We have used Bridgelux die which have a five sided output like the flip chip. The wire bonds dictate a larger spacing but it is still possible to place them close enough that they would reabsorb the output from an adjacent die.
*just re-read how you tested the second Gen die and understand it to be a single. I know who to ask at Philips so I'll shoot her an email tomorrow as well.
I don't have data for just the die at hand but I'll look into it tomorrow. We have used the flip chip in a four and 12 die array both first and second generation. Both performed a little lower in total output in our remote phosphor application than we expected but only by about 10-15%. Too much to be acceptable but not the 50% low you are seeing. Like I mentioned before the biggest issue we have had is with the solder paste layer being way too thick. It's not to the point of tombstoning but you can see the tip under a scope.
Are you testing these in an array or as singles? We have used Bridgelux die which have a five sided output like the flip chip. The wire bonds dictate a larger spacing but it is still possible to place them close enough that they would reabsorb the output from an adjacent die.
*just re-read how you tested the second Gen die and understand it to be a single. I know who to ask at Philips so I'll shoot her an email tomorrow as well.
Last edited:
I guess you must be located in the US. We are located in Germany and the support is a bit different here.
Anyway, we are testing with single dies on stars, but our end product actually uses quite a few dies on one large substrate. We had a custom ultra-high CRI phosphor made and did a lot of other work, but then the efficiency was so low, we couldn't figure out why. Finally traced it back to the dies.
For our application it would be much better to use the flip chip parts because there aren't many suppliers who want to wire bond on a 46 cm wide substrate...
If you think your company might be interested in supplying us, please PM me your contact info.
Anyway, we are testing with single dies on stars, but our end product actually uses quite a few dies on one large substrate. We had a custom ultra-high CRI phosphor made and did a lot of other work, but then the efficiency was so low, we couldn't figure out why. Finally traced it back to the dies.
For our application it would be much better to use the flip chip parts because there aren't many suppliers who want to wire bond on a 46 cm wide substrate...
If you think your company might be interested in supplying us, please PM me your contact info.
I can't help but think this is either a light extraction issue or a measurement issue, but I am leaning towards a light extraction issue.
50% is a huge decrease, more than I would ever expect purely from extraction issues.
Are you using an integration sphere to measure output, or are you using a different means and then estimating back?
Semiman
50% is a huge decrease, more than I would ever expect purely from extraction issues.
Are you using an integration sphere to measure output, or are you using a different means and then estimating back?
Semiman
Well we did some rough estimation at first, but then we verified it by sending several samples to a calibrated testing laboratory...
Actually we had the same results from three different laboratories, if you include the two different phosphor suppliers we've been working with.
I suspect that it is some sort of assembly issue, but I never heard of the soldering process having such a huge effect on light output.
Actually we had the same results from three different laboratories, if you include the two different phosphor suppliers we've been working with.
I suspect that it is some sort of assembly issue, but I never heard of the soldering process having such a huge effect on light output.
