Grey kit electronic thread: discussion, mods, observations, misadventures

While fiddling with my grey kits which arrived today it occurred to me that this whole grey kit exercise is potentially very unpleasant for Wayne Y who designed the circuit. Look at it this way: if you were an engineer that designed a part for the Space Shuttle, even though you have long since been paid for the job and were satisfied with your work, it would be hard to not be personnally affected by the recent Shuttle disaster. Similarly, Wayne, who undoubtedly worked hard on this design and did a good job, may find it unpleasant to read about what the unwashed barbarians are doing to his circuit. He may never forgive Peter for what he has done. As a consideration to Wayne, I suggest that we start one thread to be the repository of our comments, observations, and misadventures with the electronic portion of the grey kits. By doing this, there will be a single thread that Wayne can choose to avoid reading if he so wishes. Let's use this thread for this purpose.

Hehe...I'll go first.

I have already let the magic smoke out of the voltage regulator chip on one of my grey kits. Ugh! Just missed seeing the bit of metal crossing several pins...D'oh!

Now I need suggestions for another circuit to go in it's place.

Maybe the "Grey Kit" opens up a "mini-market" for Wayne to do some limited "drop-ins" with colors or even standard LSs for versatility? I don't know how many folks would be interested, may not be worth the "trouble"?

OK, I'll go first.
I note that this board does not use an input filter capacitor. This is very unusual for a switching regulator. The reason this circuit works well without one is that used as intended, the battery is electrically very close to the circuit. The capacitance of the battery is thus able to serve the function of the missing input capacitor. If for your purposes you power the board from batteries more remotely located, or from a power supply as might be the case during various experimentation, the circuit may misbehave without an input capacitor. I found that adding 1uF at the input was adequate to permit stable regulation while operating from a bench power supply. I did this on the battery contact side of the board assembly. Using a Dremal type tool with an abrasive stone tool, I removed some of the solder resist from the positive and negative traces to permit installation of a 1206 case size capacitor.

On the front of the board there are two tiny components [0402 case size] between the IC and the inductor. This is an RC network to "compensate" the IC, i.e., insure stable operation. Mine where delivered with R=2K and C=30nF. Buried somewhere in a thread in the ARC forum is a post by Wayne reminding Peter that Wayne has recommended that the values be changed to R=0 and C=100nF. Looking at the datasheet for this IC, this change makes sense to me too even though the as built board worked OK under the conditions that I tested it. Implementing Wayne's recommmended change should make the circuit more stable through a wider range of variation of various variables. You don't even have to remove the existing components to do this. You can just bridge the two pads nearest the IC with a 100nF cap. I managed to get a 0603 case size in there. It was a bit ugly but it was what I had handy.

On the boards I received, the LEDs were mounted using a thin, double-sticky thermal gap tape. While this may be thermally less effective than something like Arctic Silver epoxy, it has the advantage that it makes changing the LED much easier if you wish to do so. One thing to watch out for is that the hole in the flex PCB for the emitter slug is only very slightly larger than the slug diameter. You must be careful to center the slug in this hole. If the slug is misaligned and one edge of the slug is resting on the flex PCB material it will compromise the thermal conductivity of the thermal joint.

Testing with a bench power supply demonstrated that the circuit regulates nicely down to a Vin of 1.6V and then rather abruptly goes into "moon mode" between 1.6V and 1.5V.

General comment: I like the overall design approach that was taken with the circuit and the integration of flex PCB with the heatsink. Wayne, Peter, and Company did some good work. I *can* see that getting the flex PCB aligned on the heatsink must be a real PITA. I guess the upside of this is that it helped create "grey kits" for our enjoyment.

Doug S

I concur about the input capacitor. One of my LS's that I modded to had a tendency to (I'm taking a swag here but I think I'm right) oscillate whenever I tried to measure current draw. I believe the leads from the ammeter was acting like an antenna and the circuit didn't like it. The compensation resistor should definitely be shorted out. The circuit's transient response is sluggish with it in there, but you only see this briefly at power up or when it transitions into moon mode.

The overall design approach was very good. There was a lot of thought put into that one and Peter and Wayne did a nice job. Yes the alignment is a PITA (I have some personal experience with that one) but Peter has some fixes with the Rev 2.

Originally posted by CM:
Doug S

I concur about the input capacitor. One of my LS's that I modded to had a tendency to (I'm taking a swag here but I think I'm right) oscillate whenever I tried to measure current draw. I believe the leads from the ammeter was acting like an antenna and the circuit didn't like it.

Click to expand...

<font size="2" face="Verdana, Arial">I would guess that the added inductance of the ammeter leads is the source of the problem.

I was thrilled to get a "grey kit" and I have to say I was extremely impressed with the design of the light. On the mechanical side Peter and the crew have made even the "unseen" parts with precision and a very rugged feel.

The aluminum slug with flex board, and the extra nicety of the alloy battery contact and plated case contact were a pleasant surprize. They show a lot of attention to detail.

In addition to Doug's comments (I agree ) here's my two cents. I like the boost chip selected, and fully intend on modifying the circuit for fun. Given the wide open board area, I might have fattened up the trace between the diode and the output cap. Not having the mechanical CAD drawings to check for interference, I may have switched the locations of the diode and output cap to have a less meandering path for the inductor current. It's easy to be the Sunday morning Qback .

Really great design job!

Originally posted by php_44:
In addition to Doug's comments (I agree

) here's my two cents. I like the boost chip selected, and fully intend on modifying the circuit for fun. Given the wide open board area, I might have fattened up the trace between the diode and the output cap. Not having the mechanical CAD drawings to check for interference, I may have switched the locations of the diode and output cap to have a less meandering path for the inductor current. It's easy to be the Sunday morning Qback

.

Really great design job!

Click to expand...

<font size="2" face="Verdana, Arial">You are right. It *is* easy to Sunday morning Qback; fun too. I agree with your suggested changes and apparently so does Wayne as he has already implemented them all in his Badboy design. This [the ARC design] would have benefited from wider traces on all of the power dissapating components to enhance heat transfer to the underlying heatsink.

BTW, that thin doublesided thermal tape used to mount the LED and flex PCB to the heatsink appears very similar to [and may be] Bergquist Bond-ply 100. If so, if they used the thinest available thickness which is .005" this would yield a slug to heatsink thermal resistance on the order of 14 C/W or a die to heatsink of 29 C/W. This appears to be another of the various reasons that Peter has backed away from the idea of a 5W in this host. Those considering hotrodding the board with the "as mounted" configuration of the LED should exercise some restraint.

FYI here is a link to Bergquist:
http://www.bergquistcompany.com/index_nonflash.cfm?CFID=875 700&CFTOKEN=26ee05b-acc055a5-6d7b-4c54-a83d-dc741c223f57

I have done some testing of the input/output efficiency of the Greykit circuit boards. You may find this data of interest. My boards were delivered with all components mounted except for the output filter capacitor. For this I installed a 10uf/10V X5R cap. I also added to the design a 1.5uf/16V X7R cap at the input to permit stable operation from a bench power supply. I changed the values in the RC comp network to R=0 and C=100nF. All measurements were made via added sense leads that were separate from any power carrying leads. I performed a fairly rigorous error analysis and also cross-calibrated the four instruments used for measurements. I am confident that the efficiency values computed are accurate to within +/-1%.
Here are results where the load is the 1W LED mounted on the PCB. The PCB was mounted in the flashlight housing minus the tailcap and the optic.

</font><blockquote><font size="1" face="Verdana, Arial">code:</font><hr /><pre style="font-size:x-small; font-family: monospace;">
Vin(V) Iled(mA) Eff(%)
3.80* 349 90.2
3.60 335 87.3
3.00 339 84.1
2.80 341 83.1
2.50 341 80.1
2.20 339 75.3
2.00 341 71.9
1.80 347 66.8
1.60 273 56.8
1.50 93 73.5 </pre><hr /></blockquote><font size="2" face="Verdana, Arial">*at Vin=3.80V the IC is not switching and the LED is being driven via DC through the series path of the inductor, diode, and sense resistor.
Note the very tight current regulation.
While not shown in the table above, the very worst efficiency occurs at the input voltage where the current is just dropping out of regulation.

I next simulated the case of driving a 5W luxeon by using the onboard 1W LED and adding a 10.1 ohm resistor in series with it. This results in a load voltage within the range of variation to be expected with an actual 5W luxeon. Note that the heat of the resistor was dissapated external to the flashlight with the result that the PCB was operating cooler than it would have if an actual 5W luxeon were mounted. I do not believe that this difference would significantly change the measured efficiency values. I found that the LED current regulation with respect to Vin was very poor. I increased the input cap to 4.7uF and this corrected the problem.
Here are the results:
</font><blockquote><font size="1" face="Verdana, Arial">code:</font><hr /><pre style="font-size:x-small; font-family: monospace;">Vin(V) Iled(mA) Eff(%)
7.45** 371 94.8
6.98 328 94.8
6.02 335 90.2
4.99 339 87.1
4.00 341 83.3
3.00 344 73.6
2.50 345 66.3
2.20 307 62.5
2.00 303 61.5
1.50 85 73.8</pre><hr /></blockquote><font size="2" face="Verdana, Arial">**at Vin=7.45V the IC is not switching and the LED is being driven via DC through the series path of the inductor, diode, and sense resistor.

In the case of the 1W data, I find the efficiencies to be impressive considering the difficulty of stepping up from such low voltages and the space restraints of the design. If you gave me unlimited board space I could easily design a circuit yielding >90% at 2.5Vin. Given the space restraints that he was working to, I would say that Wayne did a good job.
For those who wish to change the sense resistor value to change the LED current, you will see decreasing efficiency with increasing current and vice versa.

I have a little more data. I swapped the stock diode with a Toshiba CRS02 and rechecked the efficiency at 2.0 and 2.5V with the 1W LED as a load. The average efficiency improvement was only 0.6%. In the Greykit diode thread I had suggested several viable substitutes for the stock diode. The CRS02 was one of these. Apparently the efficiency gain is small but at least it is slightly cheaper and is in stock at Digikey.

Doug,

Thanks for the data. I think I have similar bench data when I was designing the converter.

That's good news on the diode. I've ordered some CRS06 which I think you mentioned. I'll have to take another look at the datasheets and see what's different between the CRS02 and the CRS06.

Any efficiency gains, even small are worth it.

Wayne

Doug,

Very good presentation of the data. I've been wanting to do this in a controlled environment like you did but didn't want to do it with a $70 light. These efficiency numbers are consistent with what's predicted in Linear Tech's data sheet. I agree with Wayne, any efficiency gains are worth it. If you add up all the small efficiency gains you can squeeze, they'll add up to something significant.

Originally posted by CM:
Doug,

I've been wanting to do this in a controlled environment like you did but didn't want to do it with a $70 light.

Click to expand...

<font size="2" face="Verdana, Arial">I hear that. One of my four Greykit boards met an untimely demise in the process of gathering the above data.

How to destroy your Greykit. This post is in response to some questions that I received off-forum concerning what mistakes to avoid.

Do not operate the PCB without all of the components installed. There are several threads that have pictures of the PCB including this one:
http://www.candlepowerforums.com/cgi-bin/ultimatebb.cgi?ubb=get_topic;f=31;t=000933
Checkout the photo and make sure everything is there.
Here are the various scenarios:
Operate without LED: Switching IC blows.
Operate without diode: Switching IC blows
Operate without sense resistor: Switching IC blows
Operate without output cap: LED will light but current regulation will not function properly which *might* cause LED damage.
Operate with diode installed backwards: diode blows followed immediately by switching IC if the diode fails open, IC *might* be ok if diode fails shorted.
Operate with input voltage polarity reversed: switching IC blows

I'm sure that people will find other ways to f*ckup but these are the ways that immediately come to mind.
Good luck.

Originally posted by Doug S:
Operate without output cap: LED will light but current regulation will not function properly which *might* cause LED damage.

Click to expand...

<font size="2" face="Verdana, Arial">So can I run it for the time being without the output cap? What's the chance that the LED will blow (or another component)? I tried it for short periods and I get a VERY bright purple white color that is huge! How long can I run this at a time?

I have installed the Toshiba CRS06 diode in both of my Grey Kits. I'm not currently capable of supplying the kind of data that Doug S.and dat2zip are as I'm relearning all the electronic knowledge I have forgotten over the last 15 years! I can say that for my stock 1W setups the diode performs perfectly. For those who can't wait and will not be modding their Grey Kits I would say that the recommendation of the CRS06 is solid.
As for the LT chip I blew by being careless, I have replaced it with a new LT chip. I used ChipQuik to remove the old chip and to clean up the contacts. Replacing SM components with out the proper equipment is a PITA to say the least. It took about ~20 minutes get the chip lined up and then soldered! It was all worth while when the LED fired up bright and clear. What was odd, when I tried the light using the old, damaged chip I was able to get even but dim light. I wonder if it was acting like a direct drive circuit?
On my second kit I had to reflow all of the solder joints (except for the LT chip) in order to get the circuit to work. I'm sure this is why the board never made it into a first run LS. Anyone with the same problem of a seemingly dead circuit board should try to carefully reflow the solder joints on all the components (not the LT chip!) before tossing the board.

Originally posted by logicnerd411:
</font><blockquote><font size="1" face="Verdana, Arial">quote:</font><hr /><font size="2" face="Verdana, Arial">Originally posted by Doug S:
Operate without output cap: LED will light but current regulation will not function properly which *might* cause LED damage.

Click to expand...

<font size="2" face="Verdana, Arial">So can I run it for the time being without the output cap? </font><hr /></blockquote><font size="2" face="Verdana, Arial">Well, you have already proved that. If it were me, I wouldn't push my luck

Good little board, Wayne. I was a bit surprised when I looked at it and even more surprised when I ran one. How come these run hotter than the old-old ARC-LS's? The brightness seems comparable. What am I missing?

Oh, I'm planning on designing a drop in replacement for the 'slug' assembly that will be buck/boost--so I can use a Li-ion battery. Might stick the battery management on there, too. But, I've got more room to work with as I have the whole battery compartment.