Blue Shark Advisory.

[dat2zip]

I'm going to make a speculative guess. It appears something has changed and the only thing that I can think of is Linear Technology has made a revisional change to the die either die shrinking to improve yield or some other die change. It's possible they shrunk the switch element transistor and retained the same specifications. Where before the switch element had a lot of safety margin it may now be hard to achieve full specifications. This has happened to me before with the Badboy converter IC. There were the good early years where the idle current was high, but, the internal switch was beefy and could handle 2A without issues. They made a die change to reduce "idle" current to improve lower power based designs which increased efficiency numbers. The internal switch as no longer beefy and exhibited this same similar symptom with thermal overload on loads that it used to handle without issue.

I got the Red Shark raw fab boards today and assembled one and mounted it to a test heatsink. I was not able to get past 3.2A on the input side without using a strong fan blowing on the fins of the heatsink (on high). When I ran the fan on high near the heatsink the heatsink temperature was not more than a few degrees higher than ambient.

There is one specifications that is not given in the datasheet for this particular footprint style and that is the die to power pad thermal transfer ratio. I'm going to guess it's in the 10C/W range as the new Red Shark has 16 vias in the power pad and they are silver filled. This was a very expensive proto board, but, there is nothing better than silver filled vias for a standard PCB process unless you want to go ape bonkers on cost. Then there are other alternatives that I'm sure would exponentially raise the raw board cost.

As it stands now the local fab shop quoted me $1200.00 and $1800.00 for 20 boards to have the vias on the power pad silver plated. I did not use the local fab proto shop even though they are one of the best in the industry and can do anything and then some.

I was able to get the cost down. I won't mention who, where or how, but, I'm going to have to trust they are silver filled. I can't see them and I'm not going to perform some metalurgy test to see if they are silver filled or not.

I have tried solder fill, silver paste fill (70% silver) and now 100% silver filled vias. There is not much more that can be done with the current design to improve the situation.

The thermal transfer from the bottom-side of the board to copper C to heatsink is around 2C/W or so if the copper C is mounted correctly.

Even turning down the output to 800mA I was still able to put it in a situation where it thermally shutdown.

This one situation is:
2 li-ion batteries and 6 LEDs in series. At 800mA the input reached 3A around 6.6V. This is still in the range of the li-ion cell being able to deliver heavy current demand. This will shut down thermally under this condtion.

For a 6 LED in series configuration you need to use at least 3 li-ion batteries and keep the dead cell voltage above 7V or so so the Blue Shark is not taxed to thermal overload (that is with 800mA LED drive current).

Either that or use two Blue Sharks each set for 500mA. This is a safe condition for a 2 li-ion cell driving 6 LEDs each converter driving 500mA each to the string.

I am going to take the blue shark off hold status. I'm sorry if this change affects your build and please contact us so we can work on any resolution that you currently have or will have due to this change.

Remember the silent BE prefix on all converter boards still applies. All converter boards are on the "Bleeding Edge" of technology and thus some risk is given.

The absolute safe max input current is now around 2.5A. This is significantly different than before, but, as of now I don't believe it is something we have control over.

I am going to start work on new design that will not have these issues. I will post any updates as they become available.

 

[dat2zip]

If LTC did change the die there is one option and that would be to scour the world and see if I can find unused excess supply of pre-changed date code ICs that would have the prior characteristic. This is speculative and just a guess. I have no proof except the fact the performance and max power the Blue Shark can do has changed in ways I can not seem to explain.

Wayne

 

[Justin Case]

The absolute safe max input current is now around 2.5A.

For my MZXR-7 with the Blue Shark dialed down to 810mA drive, my SF M6 still shut down after about 13 min of run time. Tail current draw for fresh cells was less than 2A (about 1.8A IIRC). Granted, I didn't measure what the draw was when the cells were heavily drained, but even if they were down to say 3.3V each, I don't think the draw would reach 2.5A. And I do recall that after 13 min of run time, my 3SxAW17670 cells were not drained down to that OCV.

With 6SxAW17500 (each cell charged to 4.10V on my iCharger 208B+), I get a tail draw of about 0.85A and at least 23 min of continuous run time without thermal shut down (I suspended the test at that point). I put the MZXR-7 head on my bench supply to see what value of Vin corresponds to Iin = 0.85A, but I forget now what that voltage was.

 

[dat2zip]

For my MZXR-7 with the Blue Shark dialed down to 810mA drive, my SF M6 still shut down after about 13 min of run time. Tail current draw for fresh cells was less than 2A (about 1.8A IIRC). Granted, I didn't measure what the draw was when the cells were heavily drained, but even if they were down to say 3.3V each, I don't think the draw would reach 2.5A. And I do recall that after 13 min of run time, my 3SxAW17670 cells were not drained down to that OCV.

With 6SxAW17500 (each cell charged to 4.10V on my iCharger 208B+), I get a tail draw of about 0.85A and at least 23 min of continuous run time without thermal shut down (I suspended the test at that point). I put the MZXR-7 head on my bench supply to see what value of Vin corresponds to Iin = 0.85A, but I forget now what that voltage was.

The problem is there is no exact statement or rule that actually is in fact the absolute truth.

If the heatsink the Blue Shark is mounted to gets hot then your margins are worse than if the heatsink stayed near ambient temperature.

Since the heatsink temperature directly affects the actual usage it will be based on the heatsink temperature more than anything else.

If the Blue Shark heatsink gets over 40C then you will be lucky if you can maintain 2A on the input.

Remember, the tests I did was with a fan blowing on the heatsink to keep the heatsink near room temperature.

If I turn off the fan and or use a smaller heatsink the Blue Shark with go into thermal overload much sooner and at a much lower input current threshold.

So, it's actually near impossible if one person says I have xxx batteries and yyy LEDs and they work great... The next person with the same batteries and the same LEDs could have a situation that never works.

It all depends on where you mount the Blue Shark and if the Blue Shark can get adequate thermal transfer to a cool medium.

A general rule of thumb is never mount a high power driver to the same heatsink as the LEDs are mounted to. The LEDs are for sure going to be the largest heat generator and the heatsink the LEDs are mounted to will always get hot. Mounting the Blue Shark to the LED heatsink is potentially a recipe for disaster.

Let me clarify.
The absolute safe max input current is now around 2.5A. assuming the heatsink is maintained at near ambient temperature. If the heatsink the Blue Shark is mounted to will get hot either from the Blue Shark or some other source of heat you will need to derate the 2.5A accordingly.

Every build configuration is different. Users need to do worst case testing to ensure they meet these guidelines. Even if another user has identical batteries and LED configuration does not mean that your build will work as similar or have the same success as the other person. It goes the other way as well. Just because another person is not successful does not mean that your build will fail as well. If you take proper precautions and proper thermal management you may be successful where others have failed.

 

[Justin Case]

I agree totally. Just providing another data point with my build.

IMO, the major weakness of my build is the thermal coupling of the LED's heat sink with the driver's heat sink (the universal contact piece). Seven XR-Es throw off a lot of heat that needs to be managed.when that heat also gets into the driver's sink, things get hot very fast.

Here is the evolution of run times I got before thermal shut down:

1. Stock config: Blue Shark trim pot set for max drive current, MZXR-7 asembled according to manufacturer's instructions. Got about 2-3 min run time before shut down.

2. Implemented various top side copper fins and heat pipes to pull additional heat away from the switcher ICN inductor, and diode. 5-6 min run time.

3. #2 plus turned down the trim pot to about 800mA drive. About 12 min run time.

4.#2 and #3 plus attempted to thermally insulate the LED's hea sink from the universal contact piece using Kapton tape. About 13 min.

5. #2-#4 plus greatly improved Vbatt to Vload ratio by running with 6S x 17500 vs 3S x 17670. At least 23 min (could be more -- I stopped the test, the light didn't shut down on me to stop the test).

 

[dat2zip]

I assembled another Red Shark for reference with an older date code and tested the thermal performance of the two Red Shark boards.

Both boards basically behave the same. Both will go into thermal runaway if pushed too hard. Looking at the inductor current with 8V in and ~20V Vf of 6 LEDs for the load I was able to get close to 3A input and Ipk on the inductor was around 4A which is the true 4A limit. Isw limit is 4A, not Iin.
Input to output ratio was 36%. Adding a small heatsink on the topside of the iC the input was able to reach 3.2A and required forced air on the main heatsink otherwise it would thermally runaway. (testing on Red Sharks). Blue Sharks may not achieve the same results since they are not silver filled vias. This information is FYI only.

I know remember that I'm confusing testing with that of the Shark Buck. I did extensive testing on the Shark Buck and came up with safe operating numbers for the Shark Buck. The Shark Buck has less of a thermal heating than the Blue Shark.

All I can think at this time is some people are using higher power LEDs or LEDs with possible higher Vf. Some may just be too high of input to output ratio and pushing the converter too hard.

A safe number to consider for the Blue Shark with Copper C is as stated above or stated in another way a load value of 15-20W load. If you can keep the fresh battery state to be greater than 2:1 ratio to the LEDs Vf then you should be a safe zone.


The fundamental internal switch is a bipolar device and bipolar transistors will thermally run away. Once they get hot they get less efficient and get hotter making them less efficient. This escalates into a runaway condition which is what is happening with the IC controller used on the Blue Shark.

I might get the raw board for a new design that will replace the Blue Shark. It will use a MOSFET as the switch element and be external to the controller IC. This allows the design to be scaled by using the appropriate external MOSFET.

My goals are:
Input Voltage range: 6V-25V
Output Voltage: >Vin to 28V
Output current: (default) 1A
Converter efficiency (worst case): >89%
Converter efficiency target: 90%-95%
Physical Dimensions: 0.75" round
Components: Single sided
Heat Sink requirements: TBD (none expected)
Low battery detect shutdown: Yes (resistor divider)
Variable Brightness control: PWM and Linear
Trim pot on board: NO
Variable output: Daughterboard

DaughterBoards
#1: Trim pot daughterboard. Provides adjustment for ~0% - ~100%

#2: Microprocessor daughterboard
Two Thermal sensors: Either sensor can be mounted remotely to monitor heat souces and to provide thermal foldback based on heat.
PWM control output
Analog control output
UI - Enhanced Blue Shark interface. Cycling the power will toggle the UI as in the standard UI.
TBD: Preliminary information
The UI will use PWM to control high levels of brightness. At lower levels of brightness will be a combination of PWM and analog control. For example low might be 50% PWM and some value of analog to retain some white balance.

To help offset some of the costs for components I am considering offering protos as they become available. They come as is. There may be one to a handful built each round. For some the protos will as they are provide the desired control they are looking for and will not have to wait for production. Assuming the proto comes up and works I will start a new thread. Code name is Hammer Head. I think I've used this name before, but, never in a production board. So, if I made a proto before I am recycling the name.

The microprocessor effort will when completed be put into the base board for the makita box. The base board will not have any converter on it. Rather, a multitude of different drivers can be attached in conjunction with the microprocessor to provide multiple brightness levels.

Wayne

 

[dat2zip]

Update: 15JUL10

Non related topic...
Visited the city senior center and signed up for facility usage. They have a weight training center and it's free if you are over 50. Sadly, I am over 50. Not sure if I should celebrate or be saddened by the fact that I fall into a senior citizen class or not.

Also, I finally got prescriptions for both my eyes and exzema surrounding the eyes. They are finally getting better and I no longer have ghosting or multiple image vision. I can now work a full day and not have really serious eye strain and blurred vision. (YAY!!!!!)

Regarding the Blue Shark. I got the new Hammerhead fab boards the other day and have one loaded and it's giving me nothing but grief. I have high frequency oscillations on the gate drive signal on the rising edge. This is causing some losses which I might not be able to fix without rolling the PCB and getting new boards. On the other hand I have made two revelations that might affect the Blue Shark.

The first revelation is the diode on the Blue Shark. I did not know what the turn off time is and it is not published in the data sheet. Can't compare this specification from diode to diode as no one seems to want to publish this spec anymore. Not sure when this happened. Used to be in the data sheets. I can't find one mfg data sheet with this spec anymore. As it turns out I was troubleshooting the Hammerhead board (HH board) and noticed a long time for the diode to turn off. This causes momentarily the output to be shorted to GND via the MOSFET switch element and through the diode which should be off.

It then dawned on me that on the Blue Shark this period might be causing the internal switch to get hot. At least this is wasted power and in any case getting rid of it is a good thing.

I took one of my Blue Sharks mounted to heat sink that I know thermally overloads and changed the diode. Now it runs to full power and comes out of regulation when ISW hits limit which is around 3.2A on the input.

At limit the values I recorded were:
6.34V @ 3.226A
19.69V @ 0.891A

Under this condition the efficiency was 85%. Very impressive at this input current level. Before Efficiency fell below 70% under similar conditions. It's still running on the bench as we speak.

I have spent three days now debugging the HH board. These are challenges I like, but, drive me crazy.

The second revelation is the inductor. it turns out the AC component in the inductor is causing more heat than the DC component portion of the inductor. Even though this inductor is rated for 4amps or more (DC) it doesn't do will with the inductor current waveform at the frequency the converter board is running at. I had a weird tingling a long time ago when I noticed the inductor would rust. This implied some iron and iron is not good for high frequency inductors. Usually good high frequency inductors use ferrite of some form. This performs much better vs iron.

I might be able to reduce the inductor heat some using a different type/value/size.

As it stands now I believe the diode change solves the thermal runaway condition of the internal switch. The bench power supply is stable and I do not see the power supply current slowing increasing (a sign of thermal runaway in power transistors). I will do more testing and post updates as they become available.

Wayne

 

[dat2zip]

Here's an update to the Blue Shark.

The current Blue Sharks have a 6.8UH inductor on it. Previous boards might have a 4.7UH or a 4.7uH marked inductor. All will work with the two diode configuration. Some are less efficient as others.

Still the difference is small compared to not being able to get past input current of 2.5A with the CMS04 diode.

Depending on which inductor you have your board should perform similar to the graph below.

NOTE: Measurements must be incorrect for low level currents as efficiency should get worse (follow the curve ending at ~200mA). No biggie as this is normal for most converters of this type.

The far right numbers were extracted with maximum input current of ~3.2A.

The data for max power in the three inductors are:
4.7uH
Power supply: 6.91V @ 3.182A
Output: 20.218V @ 0.963A
Efficiency: 87.06

4.7UH Inductor
Power supply: 6.897V @3.19A
Output: 20.185V @ 0.944A
Efficiency: 84.53%

6.8UH Inductor
Power supply: 6.992V @3.147A
Output: 20.092V @ 0.946A
Efficiency: 84.233%

If you need more data I can provide it if necessary.

If you are curious the case of the letter "u" or "U" are two different models from this manufacturer. Why they did this is not documented on the site. It also appears that I have ordered in the past the "u" version and got "U" parts which has totally confused my inventory and stock.

In addition I do not plan on changing out to the 4.7uH inductor as I would like due to the manual labor required to change the inductor and also for the fact the part is not in stock and lead time seems to be 45 weeks at the moment. I know.... 52 weeks is a year. 45 weeks is almost a year out. Sick

I think I have 6 4.7uH inductors in my possession and quite a few 4.7aUH and some 6.8UH.

Wayne

 

[Justin Case]

For those of us with apparently problematical Blue Sharks, can you tell us what replacement diode you used to improve the performance? I would have naively thought that the supposedly inherent fast reverse recovery time would have applied for any Schottky diode. 84% to 87% efficiency at a Vbatt:Vload ratio of about 7:20 is impressive.

Also, I've sent an email request to the Inductor MFG engineers asking them what changes, if any, there may be in their inductors for the "u" and "U" marked versions.

 

[dat2zip]

The dual diode can be a Toshiba CRS08 or a diodes DFLS130L-7 SOD123 footprint diode.

Both can be had at Mouser or Digikey.

Wayne

 

[dat2zip]

I've updated the graph to include the Hammerhead board that would have replaced the Blue Shark had the two diode fix not been found.

Even at maximum load the Hammer Head utilizing an external MOSFET is around 92% efficient. Both Blue Shark and Hammer Head board (HH board) are 0.75" diameter boards.

To eek out 95% efficiency would be to eliminate the diode and go with a synchronous converter.

The HH design is scalable since the MOSFET is external a larger MOSFET and larger diode along with appropriate input and output capacitor changes you could make a heftier design capable of driving more LEDs and have a higher output voltage rating.

Test conditions for the HH data:
Vin ~7.7V-7.9V @2.8 max
Vout: Same 6 LEDs mounted to Cree bar as used in the Blue Shark testing.

Wayne

 

[Justin Case]

Here is the word from the inductor vendor:

[The vendor] uses various marking schemes including upper and lower case letters and underscores to differentiate different manufacturing locations, processes, and materials. This way, we can tell a number of things about the product just by looking at the printing. These markings allow us to trace that specific inductor all the way back to manufacturing date, production line, and material lots.

The difference you have noticed does not indicate any difference in performance, efficiency, or reliability. You can be assured that despite the marking change you have noticed, the performance will be identical.

In regards to the reference in the Candle Power forum, there should be no difference in the efficiency of the "u" and "U" in their circuit.
Perhaps they really had a "u" and a "U" where the "_" is an important indicator for material type which has higher core loss at high frequency, but a lower DCR.

There is not enough information in the Candle Power forum to make a conclusion why there was a difference in efficiency. Again, if you have a "u" and a "U", the material is the same and the performance should be identical.