Nuwai Q3 Zetex circuit

[koala]

Hi all,

I just got my Q3 today. Rip it apart and I got SX1K Lux III. I am surprised with the designed of the Zetex circuit. Looks like the designer omitted the input cap, output cap and output diode. Look at it again, its' a Zetex circuit in an unusual configuration. The inductor henry is very small, probably less than 10uH. The sense resistor mark R020, so it's 0.020ohm ?

The red positive and blue negative are cell input. The LuxIII negative output(OUT -) is the lower junction between the 310 and transistor. The circuit board is double sided, nothing much on the other side. Output tracks are very thick suitable for transporting more electrons :naughty:

I proceed to test the efficiency of the circuit with my power supply the data is a below...

Current In at 2.9v : 700ma
Current Out(at Luxeon) 2.5v : 470ma

Current In at 2.6v : 650ma
Current Out : 390ma

Current In at 2.4v : 620ma
Current Out : 340ma

The voltage measurements are closed circuit voltage. A xxxK bin LuxIII is probably driven at around 1.5watts when using a 3v cell. I tried adding a 1000uF output cap, no workie the circuit refuse to regulate. I will try modding the sense resistor later. Is there a way to improve the efficiency of this circuit?

EDIT: MrAl teaches how to use a 0.1ohm resistor to determine the current flowing through a circuit by reading the voltage across the resistor. With this method I obtained different results.

Current In at 2.9v : 810ma
Current Out(at Luxeon) 2.5v : 610ma

Current In at 2.6v : 740ma
Current Out : 490ma

Current In at 2.4v : 690ma
Current Out : 420ma

I find measuring current with 0.1ohm resistor in series produce a very good consistency. I can duplicate the exact results if my power supply is setup properly. I also tried a 0.01ohm 1% resistor suprisingly the results are the same except I am limited to 1 dot digit when reading mV range.

I used a Dale 0.1ohm and 0.01ohm with 1% tolerance metalchip smd resistor for the test above.

Vince.

 

[MrAl]

Hi there,

The most likely cause of low efficiency with the Zetex
circuit is wrong inductor. You can try other inductors
but you'll have to be careful that the output current
doesnt rise too high, which requires a sense resistor
change.
You might also note that i myself never recommended this
circuit because the output ripple current usually exceeds
the peak current spec on the Luxeon. Since the peak is
exceeded there's some risk associated with using this
circuit. People have used this kind of circuit anyway,
however, so there is also a chance the peaks arent
hurting the Luxeon, but there's no way to be sure without
much testing.

If you'd like to add an output cap you first have to
add the series Schottky diode. You cant just add an
output cap or the circuit wont work at all and may
destroy the transistor.

To add the diode, you'd break the connection to the
Lux and insert the diode with cathode connected to
Lux anode. Cap would go right across the Lux.

Let me know if this isnt clear.

BTW, just so you know, the diode and output cap
addition usually LOWERS the efficiency but it
protects the Luxeon by keeping the peak current
down below the rating of the Luxeon.


Take care,
Al

 

[koala]

Lots of space in there to fit a a good inductor. Off to look for one now. Will test it with existing sense resistor.

Oh btw, according to the data above, I would guess the efficiency of the circuit to be around 73%-75%.

[ QUOTE ]
You cant just add an output cap or the circuit wont work at all and may destroy the transistor.

[/ QUOTE ]
/ubbthreads/images/graemlins/hahaha.gif /ubbthreads/images/graemlins/grin.gif that's what I did, no workie without a diode. Lucky I didn't blow the transistor.

Thank you sir - your explaination is clear.

 

[MrAl]

Hi again Vince,

I calculated something around 60 percent with the numbers
you supplied, which is quite low.

Oh, you tried the diode already -- that's great too.

If you can find the old Zetex thread, we got around 85%
with a hand wound coil and i had tried several store
bought models that didnt work as well (80% or less).
The cores we tried are also in that thread somewhere too,
so you can find a good one. Only a few turns were required
(like 6 if i remember right) so it's not hard to wind.

Good luck with it, and please update us when you try
other inductors and/or cores...

Take care,
Al

 

[wojtek_pl]

[ QUOTE ]
MrAl said:
[...]
If you can find the old Zetex thread, we got around 85%
with a hand wound coil and i had tried several store
bought models that didnt work as well (80% or less).
The cores we tried are also in that thread somewhere too,
so you can find a good one. Only a few turns were required
(like 6 if i remember right) so it's not hard to wind.
[...]

[/ QUOTE ]
Can't find that thread... Do you remember which one was that ?

 

[koala]

Zetex ZXSC300 single-cell LED driver

Best I could improve with my collection of inductors is ~76% with current draw of 660ma at 2.9v and output of 470ma.

 

[wojtek_pl]

Thanks koala !

 

[rdshores]

Did you try decreasing the sense resistor? Looks like it's a R020(20 milli ohm)You could use the 32 ga wire trick that the CPF used CPF ZETEX when they fine tuned their circuit.

 

[mike101]

How do you measure the I(out)? Are you using the "LS Simulator and Ripple Detect Network" provided in here http://www.towner.freeservers.com/zetex.html ?

 

[koala]

To measure I(out), just connect a dmm in series with the load. Set your dmm to Ampheres mode.

 

[MrAl]

Hi there,

Thanks for the link info Vince!

We had used a 0.1 ohm resistor to measure current because
we didnt want to deal with the variations introduced
with different meters and to avoid long leads bothering
the circuit.
Measuring the voltage across the 0.1 ohm resistor the
current is then the voltage times ten.
For example, if you read 0.1 volts then the current is
0.1 times 10 which comes out to 1 amp.

Take care,
Al

 

[mike101]

What is the LED impedence while working? Would the extra 0.1 ohm resistance increase the total impedence significantly, so the circuit has less current flowing through the LED?

 

[mike101]

koala, per MrAL's post I think the way you measure the I(out) through the LED is increasing too much impedence to the circuit load. Sometimes I found 3~4 ohms existing on the contact of two prodes of my cheap DMM. It affects a lot the accuracy and makes you get a lower current value when you are measuring a low impedence load. Can you do the measuring again with MrAl's method?

 

[koala]

I don't have a 0.1ohm, I do have 1ohm, 2.2ohm will that work?

 

[mike101]

I think that the 1 ohm in series is better than the DMM in series with the LED. Or you can connect ten 1 ohm resistors in parallel to obtain 0.1 ohm resistance.

 

[pbarrette]

Hi Koala,

MrAl mentioned that the efficiency goes down with the addition of the diode+cap which is perfectly true. What he failed to mention, however, is that since the LED is being fed stable current from the cap instead of as a pulse train, the LED brightness goes up significantly.

Also, as MrAl (I think it was you Al) pointed out to me a long time ago.. It is very difficult to measure the true current output of the circuit (minus diode+cap) with a DMM. This is because the circuit is generating pulses of current instead of feeding a continuous current. What you are actually reading is a rough average of the current being pushed through the lux, dependent upon the frequency of the pulses and the sampling time of your meter.

What is actually happening with this circuit is that the LED is being overdriven with high current pulses at a rapid rate. The acutal current pulses may measure well over 1A at 3V, but since they are pulses, with zero-current spaces in between the pulses, the average current-per-time-unit (say.. current measured over 1 second) is lower.

When you put the diode+cap into the circuit it improves the circuit a lot. The circuit efficiency goes down, yes. But since you would then be supplying the LED with a continuous current, the LED will appear brighter at the same input voltage/current. This is because, now, instead of pushing a train of 1A pulses, you can push a steady, say, 700mA. You also get the added benefit of getting more accurate current readings from your DMM since the sample rate of your meter isn't affecting the reading.

In fact, a quick look at the datasheet tells the story wonderfully. The datasheet reference circuit minus diode+cap is listed as the "Maximum battery life" circuit, while the circuit with diode+cap added is listed as the "Maximum brightness" version.

In all of my ZXSC300 circuits, I add the diode and cap. I may lose a bit of battery life, but the brightness is massively enhanced and you aren't pushing those overspec current pulses through the lux.

So.. I guess what I was really trying to say is this. Don't worry so much about increasing the efficiency of this circuit if what you are really looking for is a brighter light. Add the diode and capacitor and fiddle with the sense resistor instead. Overall efficiency will go down, but the light output will go up. And, the light color will likely lose a good deal of that bluish or greenish tint that I'm guessing you have. That particular tint associated with being overdriven.

Be careful when modding this circuit though. If you apply power, and the LED doesn't light.. Cut the power immediately. That little FMMT617 transistor can get fried pretty quickly. If the LED doesn't light, check your connections. If the problem isnt there, try increasing the Rsense value.

If you blow a transistor, or the IC trying this out, let me know. I may have some spares I can drop in the mail for you.

Also, I have been getting pretty good results with 100uH inductors. I was using "on hand" 22uH inductors originally, but they were causing some heat issues due to reaching saturation. Those problems are gone with the higher inductance and inductors rated for the current generated.

Oh.. And if you go with the sense wire for a replacement Rsense, make sure you twist the wire loop formed to prevent power being inducted from the inductor to the Rsense wire loop.

Oh yeah.. And if you are using a tantalum cap, you can reduce output ripple a lot by putting a smaller (2-4uF), low ESR ceramic cap in parallel with the tantalum cap.

One more thing.. For your diode, use a shottkey rated for the current with as low a Vf as possible. I found a 1.5A rated shottkey with a Vf of ~0.23V on digikey that was cheaper than the zetex part. The lower the Vf the better as it helps your efficiency.

Ok.. I guess I've ranted enough on this one. I've got a lot of experience with this circuit, so I just wanted to share my $0.02-US hoping that it can help you get a better light.

So, hope this helps,
pb

 

[MrAl]

Hello again,

Thanks for the refresh on the info about the Zetex circuit
pb...i still like to think about that circuit from time to
time.

I just wanted to mention to koala...
that a 1 ohm resistor isnt
very good for measuring output current because it drops
a lot of voltage. A trick, however, might be to try the
1 ohm resistor and then increase the input voltage by
the amount that the 1 ohm will drop in the output circuit
(about 350mv for 350ma output). For example, say your
input is exactly 3.000 volts with no 1 ohm resistor in
the output to measure current. Then, turn off, insert
1 ohm resistor, then turn back on and increase input
voltage to whatever the 1 ohm drops (normally 350ma)
so this means increase input V to 3.350 volts.
Take a voltage reading across the 1 ohm resistor and
that should be approximately the correct current that
would be measured at 3.000 volts input and no 1 ohm
resistor connected.

If anyone else wants to try this they can take readings
with the 0.1 ohm resistor (and no increase in Vin) and
with a 1 ohm resistor (and increase Vin by 350mv or so).
We can then compare results using both resistors to
measure current in the output.

Take care,
Al

 

[koala]

Ok.. Ok.. I will go get a 0.1ohm resistor tommorow. Don't worry about the 1.0ohm it's too messy for me.

Does anyone know how much tolerance precision(%)is a 5watt resistor with a marking of 5W 0R1 J? Is J marking 5% ?

If it is 5% I could be getting 0.095ohm to 0.105ohm? Sorry my local store only sells low ohm values in big watt package. /ubbthreads/images/graemlins/icon23.gif

-vince.

 

[rdshores]

Yes the J is 5% tolerance.

 

[Krit]

Hi pbarrette , very good data, Thanks.