hipFlex

[georges80]

I've run some efficiency measurements with my son tabulating the readings (Vin, Iin, Vout, Iout). I then calculated Pout versus Pin and graphed the results.

I performed the tests with 1 x P7, 2 x P7 and 3 x P7 at 2.8A drive current. Obviously the P7's are in series, so a higher total Vf as more P7's are connected together.

So, here's graph, Vin is in volts. The production driver will be somewhat more efficient (maybe 1% or so) due to lower loss in the polarity protection circuitry (FET rather than Schottky diode).

1 x P7 (3.5V x 2.8A) = 9.8W output
2 x P7 (6.82V x 2.8A) = 19W output
3 x P7 (10.2V x 2.8) = 28.5W output

A 3 x P7 light with 4 cell li-ion battery pack would have the driver running around 93% efficient at 2.8A.

I'll have to run similar tests at lower current to see how the driver performs. Though, from a driver power dissipation point of view (keeping it cool), it's nice to see the efficiency increases as the load increases (i.e. it runs more efficiently with 3 P7's versus 2 P7's versus 1 P7).

The driver is more efficient when the input voltage is closer to the output voltage which is what one would expect with a switching converter so no surprises there.

Edit: Ok, ran similar tests with 1.7A output and the efficiency curves follow very closely within a couple of percent. So, it would appear that the main effect on efficiency over a reasonable range of current is Voltage In versus Voltage Out and Voltage Out. As can be seen from the above curves, efficiency is higher at higher output voltages (more P7's) and when the input voltage is closer to the output voltage.

cheers,
george.

 

[trout]

This is too much good news in a small time , fantastic .

When does the bad news start to arrive .

 

[georges80]

This is too much good news in a small time , fantastic .

When does the bad news start to arrive .

I built the first hipFlex prototype around mid July, so even though it seems the news is coming out quickly, there has been quite a bit of work that's gone on in the 'lab' to get to this point

Here's a pic of the first hipFlex proto (it had the uController and a few other components on the back side):

I've been looking into a higher powered buck converter for the past couple of years (on & off) and I've prototyped at least a 1/2 dozen different designs but have never been pleased with their performance or characteristics.

hipFlex is the culmination of all those prototypes and leverages the current UI-uni firmware to provide a consistent user interface across my various uController based drivers. By utilizing the same firmware source base I don't have to worry about bugs etc since the code is well proven. I have the same source file that generates firmware for the bFlex, nFlex, maxFlex, d2Flex and now hipFlex. All that changes is a few conditional compilation statements for the different current regulation scheme of the specific *Flex driver that is being programmed.

cheers,
george.

 

[trout]

Thank you George for all the hard work you have done

So non technical folk like me can have a product that works
straight out of the box , no messing with pots and resisters .

I cant wait for its release .

not just for bike lights .
I want to replace the halogen security lights on my house with some good
led replacements so the HipFlex & P7/MCE will be perfect for that too .


I suppose A remote control is out of the question

 

[dom]

Impressive stuff George
I'll be keeping an eye out for these when available as well.

I think you have the best drivers in the world

Cheers
Dom

 

[StevelKnievel]

This is exceptional George, I'm eagerly awaiting availability!

Steve

 

[TTGV]

Hello George,

What do you think about a 1.4 A current level, and may be a 2.1A one, to fit with the max curent needed with 2 or 3 lines mades of MC-E ?
I don't know the effect of 0.750A instead of 0.7A on MC-Es !? Very shortened life ? More blue light ?

Sorry for my cheese-made english
Very nice design at this point.
Best regards.
TTGV

 

[marcopolo]

George, do you by any chance have efficiency graphs for the bFlex similar to the graph you posted above?

thanks,

Marco.

 

[georges80]

George, do you by any chance have efficiency graphs for the bFlex similar to the graph you posted above?

thanks,

Marco.

No, way too many combinations to work through. And I have only take a few combo's for the hipFlex to give an idea of where things lay.

The bFlex/nFlex should be typically in the 85 - 95% efficiency range with a similar trend of higher efficiency with higher input voltage AND input to output voltage being closer.

Google LM2675 and you can look at the data sheet. They show efficiency curves for various input/output voltages at 1A load.

And, for the previous post about drive current. I plan to support a few (max) output current tables and will decide on those before releasing the production boards. +/- 50mA on drive current for a high power LED will make essentially no difference to it's lifetime/colour etc - HEATSINKING of the LED is MUCH more important...

cheers,
george.

 

[georges80]

Received production blank PCB's a few days ago and finally got some time to assemble one.

Hopefully I'll get some time tomorrow to fire it up and check how it compares to the prototype, though I can't imagine it working any differently.

I'll have to then finalize current tables and levels within each current table before making up some more.

Oh, just noticed I forgot to solder R2 in place - oops

cheers,
george.

 

[trout]

George :thumbsup:
As usual that is quality and I cant wait to see it up on the buy products page at Taskled.

I like the single sided approach , Makes mounting a doddle

:bow:

 

[georges80]

Fired up the production proto hipFlex and it is running well.

I am trying to decide on current tables to implement so I'm interested to hear opinions.

At this point the ones I'm thinking of are:

1000mA (shipping default and a safe starting point for the customer to configure from)
1400mA (for MC-E wired in 2P2S configuration)
2000mA (for folk that want to push their MC-E 2P2S)
2800mA (for P7 or MC-E wired 4P)

Any other thoughts? I don't want to support <1A, that is the domain of bFlex and nFlex and they are more cost effective for that current range and of course also physically smaller.

cheers,
george.

 

[bluecrow76]

Those numbers look right on the money! Is there any value in having a higher setting to overdrive a P7 or MC-E (3.0A or even 3.2A), similar to your 1.2A setting on the maxflex? Personally all I've played with so far is 2.8A on a D binned P7 using AMC7135 based drivers, so I don't know what happens above that... ????

 

[StevelKnievel]

Looks good to me! The 1A and 2A tables will also be good for running large arrays of single die LEDs.

 

[Teego]

+1 on the tables you listed. I'm looking into driving a triple osram ostar and this looks like just the thing to save me from myself and my own kludgy home made drivers.

I'll certainly take one, if not several.

 

[georges80]

Very little headroom, I just ran a test.

3 series connected P7's at 2.8A drive current.

Total Vf of the 3 P7's is 10.26V at 2.8A and was in regulation until Vin (measured right at the input of the driver) dropped to 10.89V.

The prototype uses a schottky diode for reverse polarity protection and I measure 0.268V across it at close to 2.8A input current.

The production boards will use a power FET instead (like bFlex) and it should only introduce around 0.1V drop at 2.8A input.

10.89 - 10.26 - (0.268 - 0.1) = 0.46V

So, it would appear the production hipFlex (for 3 P7's) will need about 0.5V of headroom at 2.8A output which imho is quite impressive at that current level!

cheers,
george.

Well, since I have the production PCB hipFlex running (with the FET polarity protection) I figured I should see how close to the mark I was in calculating headroom.

My test setup was 3 x P7. Measurements were made right on the LED+/LED- and VIN+/VIN- pads on hipFlex. I used a current clamp to measure the point where output current would drop (go out of regulation) and logged the voltage at the LEDs versus input voltage. From that I computed the difference or 'headroom' needed at the input to maintain regulation at the output at the specified output current.

At 1A there was 0.2V difference.
At 2A there was 0.31V difference.
At 2.8A there was 0.44V difference.

So, I was pretty close with my guesstimate (pats himself on the back)

I'm finalizing the current tables and the steps within the current tables and am close to having that part finished.

After that I just need to verify the input voltage monitoring hardware.

Getting close folk.

Oh, and as a teaser, I'll have pictures of a juvenile relative to hipFlex in a couple of days...

cheers,
george.

 

[hootsmon]

Well done George!

 

[georges80]

Oh, and as a teaser, I'll have pictures of a juvenile relative to hipFlex in a couple of days...

cheers,
george.

You can check out hipCC in the Homemade & Modified Lights subforum...

cheers,
george.

 

[rizky_p]

man i have been waiting for George to build this driver since my first P7 mag build.
I really Loves My Penta Cree Q5 running George's maxflex driver. I will convert my P7 using this driver as well. Cant wait.

 

[trout]

Fast work George
I just had a read on Taskled .com and you say to keep the switch wires short ie 4 inch .

Are there any issues with the length of the wires to the leds
Like if the HipFlex was in the battery pack and powering 2 light heads on the bars so maybe 18 inches long