Started June 1, 2007...Last updated July 08,2007

Zetex 300 series refresh
First part of this is some boost mode testing and improvements, and the second part is a method of utilizing it in a buck mode, and testing the same, which you will find roughly half way down. There are a lot of pictures and graphs so be patient while they download, get a drink of water, or make a head call.


****BOOST MODE***


Well, I had a number of the old Zetex ZXSC300 and ZXSC310 chips sitting around in one of my numerous boxes of old parts, and was wondering what to do with them. Then it occurred to me that I had a number of dismal performing flashlights that had the same Zetex 300 series chip on board. In many years past, I've ran a number of tests with synchronous boost regulators, and had managed to get one chip to exceed 98% efficiency while using very nice passive components. I was curious, with the latest inductors, MOSFETs, schottky diodes that have hit the market lately, just how far could one push this old antiquated chip.

The ZXSC310 is an extremely low end converter chip, used in many very cheap Chinese flashlights, and runs under 0.36 dollars each in the US, and even cheaper in Asia. It doesn't take very long for even a slow person to realize why, since they are trying to maximize their profit margins. This helps to make it one of the most popular converter/switcher chips on the market for very low end flashlights.


So, I took one of my lights that used this chip and had the largest board, as well as the most internal room, in order to see just how far one could push it. You will notice the board has been heavily modified, has additional new traces on the PCB, additional parts, and at the moment is a mess, which I will clean up later, after I do additional testing and arrive at a more final result. In the background is a Fenix L2P I purchased, shown for size reference.




I decided to place the typical stock inductor on the board, so that one can get an idea of what the typical inductor size looks like, near the large inductor. In the background, is an ARC AA for size reference.




I did an efficiency plot of the stock circuit, to get a reference to go by:




Under a decent load, a typical CR123 cell (a CR2 cell would be worse), the battery voltage drops to 2.7 volts, and is pretty much spent by 2.0V. Using 2.5V as a reference, one quickly realizes the stock efficiencies were in the 44% range. I also used a newfangled oscilloscope to verify the measurements, due to ripple voltage present. One item of interest to folks is that I noticed that the LED gets considerably warmer when pulsed like this, than it would otherwise, even so, it put out less light. See my web page on PWM driving of LEDs for more information on this issue. Below is the stock circuit (not shown is the 0.1uF input capacitor).



Zetex claims very high efficiencies with this circuit in their reference design, but they were running their circuit at very low power levels (1/10th) in comparison to the tests here. The stock circuit which was used in this flashlight had a much smaller 10uH inductor and may have ran considerably into it's saturation rating, causing significant losses around the B-H loop. They were also running at considerably lower power which would make a huge difference. Output power in the stock configuration @ 2.5Vin was 1 Watt. See page 7 in datasheet linked below.
http://www.zetex.com/3.0/pdf/ZXSC300.pdf

The first thing I did, was to add a schottky, X7R 10V 10uF capacitors on the input and output, and switch over to a larger, in order to put the circuit in the more conventional boost setup. I like to use ceramic capacitors with higher than necessary voltage ratings, and prefer X7R over X5R, as there is less dielectric saturation and less dielectric absorption, and I like to use parts in the larger package, as it also lowers these problems. A 6.3V X5R ran at it's rated voltage in an 0805, can drop to as low as 2uF, from 10uF due to these issues. Here is the schematic:




Results are here, of note, with additional gain in efficiencies, the efficiency rose from 44% to 76%, while at the same time, the output power also went up to 2.15W (from only 1W):



Next, I increased the 0.1uF input capacitor to 10uF, and replaced the stock inductor with a Sumida CDRH74 10uH inductor. Here the efficiency @ 2.5V rose further to 81%, while at the same time the output power also rose to 2.4W (this is good, more power and higher efficiencies at the same time):



I then took the same circuit, and went thru a two dozen MOSFETs, and got an additional efficiency gain to 85%, with the same 2.4W of output power, so the input power requirement dropped. Most folks at first glance would choose the lowest on resistance MOSFET, but due to the ZETEX chip having a weak output drive, there is a trade-off between nanocoulombs required to drive the gate, the rise and fall times, and the on resistance of the MOSFET. A very low gate voltage MOSFET worked the best. See results below:




I then took the same circuit with all the changes so far, and moved to a larger powdered iron inductor (no buzz and higher ratings), and got an additional efficiency gain to 87%, plus got a gain to 2.45 Watts of output power. The inductor choice could be improved further, but it is becoming a smaller piece of the pie. A yet even larger Sumida CDRH127 ferrite 10uH inductor was also tried, but was found to be less efficient. You will note that this the powdered iron inductor is show in the first photograph up top, and one advantage of these is they are shielded and greatly reduce emissions, while having the powdered iron molded around the inductor windings dramatically reduces inductor buzz (however, ceramics capacitors have piezo properties and can produce noise when pulsed).





For the next test, the circuit was changed to the bootstrap configuration, where the Vcc and Stdn pins were moved from Vcc to Vout at the output capacitor positive junction as shown in Zetex Design Note 68. This gives the chip the advantage of being driven from a higher voltage (sourcing from the output), which allows it do do a better job of driving the MOSFET, especially from lower input voltages. You will notice there is no efficiency improvement @ 2.5V out, but not obvious in the first graph below is that the output power rose to 2.65 Watts, while holding the same efficiencies!





Here is the graph of Watts Output vs Voltage Input:



Here is the graph of Current Output in Amps (multiply by 1000 to get mA):



Following scope waveforms are the latest configuration @ 2.5Vin:





















As you can see here, there is 200mV of ripple on the output, 4.8% ripple, suggesting that additional capacitance could be useful on the output, beyond 10uF:



As you can see in the earlier graphs, the output of these Zetex chips is *FAR* from any resemblance of possibly even being considered to have something resembling a regulated output. One extremely obvious things a person could do is add bias from the input voltage on the sense resistor, in order to get something possibly resembling a regulated output.



I wanted to push the converter a little harder, so I changed some of the copper traces that were carrying heavy currents, into copper planes, and also added multiple vias thru the board. The copper planes also help to heatsink the MOSFET as well as the Schottky diode. Basically, I was switched from the stock LED, to a CREE XR-E, which has lower forward voltages, and the currents would be considerably higher. I also added additional ceramic capacitors on the input and output, especially on the output, to help with ripple current. Here is a photo of the change:



Here we have the efficiency, where you can see it is slightly lower, but still doing quite well, at the higher output current (due to lower CREE XR-E Vf):




Since these Zetex based converters are unregulated, you can see that the output Watts jumps, especially at higher input voltages, yet the circuit efficiency above still hung on well with the higher output wattage:




Below you can see the higher output currents, which is where the non-synchronous converters get less efficient. However, I'd selected out a few dozen Schottky diodes, looking at criteria like Vf drop, diode capacitance, and leakage current, then testing each one in the circuit, before arriving at the one I'm utilizing here. Note the considerably higher output current:




I noticed the CREE XR-E had just begun shifting towards blue near the end of the test above, but I didn't think about the current, until I processed the data. The poor CREE XR-E was being driven at [b]2,280 milliamps[/b], well beyond it's rating, where I typically see similar shifts on the Seoul P4 from 500mA to 1000mA.

I decided to get some more oscilloscope captures, with another scope. Here is the gate drive:




Here we have the MOSFET Drain during operation:




Drain waveform expanded to show rise time waveform:




Drain waveform expanded to show fall time waveform:




Sense resistor waveform:




Input ripple waveform:




Output ripple waveform:




Anyhow, the results aren't too bad, adding another dollar to the board cost, one could get rid of most of the change with a smaller inductor, yet still get a majority of the performance for under 50 cents. Oh, btw, that large inductor- yes, it fits into the original flashlight without mechanical modification of anything- fyi.


For those interested, a reknown CPF'er known as Mr. Al, has a mod he came up with for making the Zetex chip act more like a constant current regulator. His modification is shown below. Also, LED Dynamics patented a very similar modification to the standard circuit.
http://hometown.aol.com/xaxo/page3.html
LED Dynamics modification



****BUCK MODE***


When I first got this light a few years ago, I was happy with the output, but with the new much higher efficiency CREE LED flashlights I've bought, this light, just like the one above, had taken back seat to other better CREE LED based flashlights. Meanwhile, it sat there, collecting dust, and not being of much use. Then a few months ago, I started yanking the Luxeons out of various flashlights and swapping the LED out for better performing CREE devices.

A person can either use 2x output or more (depending on the original Luxeon bin), or they can adjust the converter to a lower output to match the original brightness, and usually gain considerable converter efficiencies, as well as much more efficient use of the battery power, since heavier loads waste power within the battery, and often get 3x the original runtime with the same brightness.

One additional item a person can do is to improve the efficiencies in the stock converter, often by making better choices for components (often designers are pressured into utilizing the cheapest possible devices to maximize profit margins), fixing board issues, and improving heatsinking to the circuit. Here I just made a few changes in choices of inductors, MOSFETS, schottky diodes, capacitors, as well as fixing board trace issues while gaining better heatsinking for the MOSFET and Schottky diode.

As such, the designer picked up one of Zetex's application notes and implemented their circuit for the ZXSC310 in buck mode. The Zetex chip when running in buck mode, is much more efficient, as the peak currents are much less, and components are not stressed nearly as hard.

Unfortunately, it is still unregulated, but one can also play the bias trick with the buck circuit, to get an output that gives a person a flatter runtime graph that would more resemble something that is actually regulated by design.

So, lets go over the testing I did by taking a look at the following graph:



Since a two cell buck mode flashlight puts the batteries under much less load, I'll be using 6V input voltage for my examples that follow. If you look above- the dark blue line, you will see the stock converter's efficiency graph shows about 85% efficiency.

The first thing I did was to put a better inductor in, and better capacitors. Unfortunately, the efficiency didn't pick up but by 1%, but in later graphs if you look, you will see that it gained the 1% efficiency as well as putting out 0.6 more Watts at the same time- not too bad.

The next thing I did was to install a very low on-resistance MOSFET in order to reduce the on-resistance losses. In this case, the ZXSC310 converter has a very weak drive circuit. In this case, with the MOSFET I used, the gate charge was high, which increases switching times and increases losses when compared to a MOSFET with low gate charge. So there is a trade-off which will be unique for each driver chip, operating frequency, as well as operating currents.

First thing you notice on the yellow line for this MOSFET is that there was an efficiency loss, which put us back to the original efficiencies, but gained us about 0.7W output over the original (see graph for this later on). Obviously, this MOSFET is not optimum for the operating conditions and driver present.

So, I chose a MOSFET on the other end, which had exceedingly low gate charge, with a high on resistance, but nothing all that high- somewhere around 75mOhms. This is shown in the light blue line on the graph. You will notice there is a nice 5% gain in efficiency over stock, as well as a 0.4W increase in output at the same time. Looks pretty good.

The third MOSFET (actually I ran several dozen MOSFETs through the tests total), I chose a device with almost as low of a gate charge as the previous one, but a device that had an on resistance between the other two. Looking at the brown line, you can see there is a gain of 6% over stock efficiency, as well as 0.45W more output.

I noticed the devices were getting warm, so I turned the traces into a plane for the MOSFET Drain, which lowers board resistances, as well as provides much better heatsinking for the MOSFET. What happens to a MOSFET, is that when it gets warm is that it's on resistance goes up, reducing it's efficiency. A similar thing happens to a Schottky diode, but in it's case, the device gets leaky, and just burns up power. The lowest Vf Schottky diode is not always your best choice.

So, I added better heatsinking and lower resistances to the board, by using copper planes instead of traces. On the red line, you can see the results, where the efficiency stays the same. But, if you look at the graph below, you will see that it held the same efficiency gain (6%), yet produces 0.9W more output. Very nice!

Finally, we get to the final part, where I tested a few dozen different Schottky diodes, the best of which is shown in green. Here you can see we gained 7.19% efficiency (91.7% total) over stock. We also picked up 0.9 Watts more output power at the same time!

Below are graphs of Watts Output vs. Input Voltage, and Current Output vs. Input Voltage:






Anyhow, for the techs and engineers out there, I have some oscilloscope "eye candy":
Here is the stock signal on the MOSFET gate, vs after the final modification (unfortunately they are not running at the same input voltage in photos). Notice how the ringing has gone away:





Here is the modified circuit gate drive risetime expanded in time:


Here is the input voltage, notice the ringing, the 10uf capacitor I added got rid of this issue (the stock circuit used a cheap 0.1uF capacitor). In the third photo, I expanded the vertical gain, and you can see the severe ringing/ripple has dropped from 3.2V to just a tad over 0.1V (100mV), which drops to 50mV @ 7.5V (switcher running at higher frequency):










Here is the stock Output ripple voltage, roughly 1.25Vp-p. The stock unit also had a cheap 0.1uF capacitor on the output, which I changed to a 10uF capacitor. This caused the output ripple to drop to ~0.11V (110mV) at low voltage, and when the switcher frequency picks up, it drops down to about 30mV @ 7.5V. One item I noticed, is that once I added the bigger capacitor on the output, the LED ran cooler- which is to be expected, this is a known problem with PWM'd LEDs that are pulsed at higher voltages but are off for a period. The LEDs also run less efficiently for light output, see my web page on this for further details. Likely the LED was also brighter, with less heat produced, but I didn't measure for this.








Here is the voltage on the Drain of the stock MOSFET, then the cleaned up signal on the modified circuit:








Here is the rise/fall time on the Drain of the modified circuit MOSFET:






Here is the voltage across the sense resistor, stock and modified:






I got a shot of the voltage across the inductor for anyone interested:




Here is the modified circuit I was testing on, before I cleaned things up and put it back in the flashlight:




And finally, beamshots of the flashlight with the modified circuit, plus one of the new XR-E emitters, compared to a unit that is still stock:








I almost forgot to include the Zetex reference for the step-down (buck) converter:

Zetex Buck Design Note


I hope everyone found the information useful. Over the past month had lots of fun doing the testing of various configurations (the majority of the information I have not posted-just the highlights).


Main site
Mirror:
Main site mirror