ZLT-CCR2 PCBs

[operator_smooth]

I have found a bug in the calculations I used on http://zhat.dk/cpf/zlt-ccr-2-graphs.html . The results were incorrect for discontinuous circuit operation. It should work correctly now.

 

[MrAl]

[ QUOTE ]
operator_smooth said:
The fmmt617 specs state that the turn-on time is typically 120ns and turn-off time is typically 160ns - that would only subtract 40ns from Toff?

[/ QUOTE ]

Hi again,

Could it be because im slightly overdriving the device
during turn-on...it's pretty much academic anyway as the
range for the Zetex 300 chip is 1.2 to 3.2us, so if there
was a problem with the Zetex spice model (from their site)
and the switch difference turned out to be shorter it
would just mean that im testing it at a shorter chip
off period spec (lower than 1.7us)...that's about it /ubbthreads/images/graemlins/smile.gif
As i mentioned before, it would be better to do a range
of values anyway or at least the two end points right?
The turn on time might be faster into inductive load too,
for the lower current value, while the turn off time
remains longer...but unless you're looking at exact
pulse times for some reason i dont think it matters that
much really. If i wanted to be more exact i would also
drive the device at some values closer to what the
Zetex chip puts out too.

BTW, on that site what did you use for the sense voltage
level? I'm finally getting around to looking at the
equations for this circuit again, in the middle of doing
a bunch of other stuff that came up. Still interesting to
me though, and i guess for you too. The Zetex circuit
is quite a little circuit even though there's not much
to it.
[ADDED LATER] Im getting different values than you
are so maybe it would be a good idea if you specified
how you have the circuit set up, and what values you
are using for toff and Vsense, etc. Im using 5.1 ohms
for the one resistor because people seem to be able to
get that value easier than 5.0 ohms.


Take care for now,
Al

 

[operator_smooth]

I am using somewhat less favourable values for Toff, schottky drop and efficiency.

I set Eff = 0.85 and schottky drop to 0.42V so the final efficiency becomes 0.85 * Vled / (Vled + Vschottky)

I use Toff = 2us

My Vsense calculation is an attempt to reproduce the typical curve in zxsc300 spec for Vsense(Vcc)

 

[MrAl]

Hi again,

Ok, if you want me to look over the equations you'll have
to be more specific...that is, i need to know what values
of vsense you're using for both 2v and 3v inputs.

And wait a minute......
You were saying that the tables i posted were not when the
battery runs down, and i had said that they would show
what happens if you set the output for a constant current
for three values (2,2.5, and 3v) of input of the original
non-regulated Zetex 300 circuit, but i think it does in
fact show what happens as the battery runs down in the
CCR-2 circuit also. This would happen because the resistive
bias circuit can maintain the output current to within about
1.8 percent (theoretically) of original setting if the
values of the upper and vsense resistors are chosen
correctly.
For example, lets assume that using a particular inductor
value L requires the following relationship between Ipk
and Vin in order to maintain 350ma output:

Vin=3 Ipk=0.51675
Vin=2.5 Ipk=0.64455
Vin=2 Ipk=0.8175

This means if the value of the sense resistor was CHANGED
for each level of Vin the output would hold a steady
350ma dc current.

Now given the same table:
Vin=3 Ipk=0.51675
Vin=2.5 Ipk=0.64455
Vin=2 Ipk=0.8175

if, using the bias circuit, we could find two resistors
(upper resistor and the sense resistor) of constant value
that causes the circuit to change it's Ipk exactly as
the table shows (the requirement for constant 350ma out)
that means we'll get the same results as if we had adjusted
the sense wire length each time.
Make sense?

It just so happens that i've been able to find values
that hold the current to within 1.8 percent of the set
point, which is pretty darn good i think.

As i stated in (an) other post(s), i thought it wasnt going
to help to try to adjust all three resistors, because two
of them work as a ratio together rather than control
anything individually (top and 5.1 ohm resistors).
This turns out to be proven once again, as adjusting the
5.1 ohm resistor and optimizing for the other two doesnt
reduce the percent error over the three input levels
(although now that i think about it i could try using
more levels (like 2.1, 2.2, etc., up to 3v) in the
optimization process.

Let me know your thoughts on these things...


Take care,
Al

 

[operator_smooth]

MrAl, what I was trying to say is that the peak current needed to maintain a constant output current is a non-linear function of input voltage. The resistor network in the CCR-2 circuit, however, causes the peak current to increase linearly when input voltage drops. This is why the peak current will not be what it should be to fully compensate for the decreasing input voltage if the voltage falls much outside the application range. Off course, this is not really a problem with a two-cell application, because once the input voltage drops below 1.6V it would be hard to suck any more energy at the required 1.2 to 1.3W for a 1W Lux out of the cells anyway.

I guess that depending on the actual live characteristics of the parts, the difference between the required peak current and the peak current in CCR-2 could be quite small for a limited range of input voltage. My experience with the CCR-2 using small inductors (6.8uH, 4.7uH) is that once calibrated, the output current change for input voltages ranging from 1.9V to 3.0V is not detectable on the 10A scale of a DMM, so the regulation within this range is pretty good, as you say. But when the voltage drops to e.g. 1.3V the output current drops by more than 20%. Because the peak current is not as high as it should be, it is possible to use an inductor with a lower sat. current rating without compromizing the CCR-2 efficiency.

I think that the reason why Zetex specifies inductors with that high current rating is to be sure that they are completely linear.

BTW, here is the function that I wrote to simulate the typical Vsense as function of input voltage:

<font class="small">Code:</font><hr /><pre>
// Theory: found using graph in ZXSC300 PDF file
function calcVsens(V)
{
if (--V < 1)
return 0.0175 + 0.0013*V
if (--V < 1)
return 0.0188 + 0.0012*V
if (--V < 1)
return 0.0200 + 0.0011*V
if (--V < 1)
return 0.0211 + 0.0009*V
if (--V < 1)
return 0.0220 + 0.0009*V
if (--V < 1)
return 0.0229 + 0.0008*V
if (--V < 1)
return 0.0237 + 0.0007*V

return 0.0244 + 0.005*V
}
</pre><hr />
This means that Vsense is assumed to be 18.8mV at 2.0V input and 20.0mV at 3.0V input.

 

[MrAl]

Hi again operator,

Oh i see what you are talking about now...i posted more
numbers than i should have probably /ubbthreads/images/graemlins/smile.gif

That's probably a better function for the Vsense variation
with input voltage than i used, but i see now that we are
thinking along the same lines as i used this:
Vs=0.019+0.002*(2.5-Vin) --18mv at 2v, 20mv at 3v input
where the range of input was limited to 2v to 3v only.

I didnt try to be extra careful about reading the spec
either, because i didnt think it would matter too much
with a small error because there were other things that
would come into question that would overshadow the results
anyway, so i went with the totally linear Vs response to Vin.

It's amazing how well this simple bias circuit works,
within a certain range, even without the diode...
isnt it? I got (theoretical) results to max about
2.3 percent of set value with either allowing the sense
voltage to vary as a linear function of Vin (as above)
or by making it a constant.
I've gotten very different resistor values however,
by allowing the toff time to vary from 1us to 3us.
The top resistor varied from about 1500 ohms to 1600 ohms,
depending on pulse width. In each case though, after the
resistor was set to the correct value, max error was only
about 2 percent (between presumed range of input V).
Now i guess i have to look at what happens when we select
a value that would work for 2us and see what happens if the
off time is really 1us or 3us and calculate a percent error.
That would really be interesting.
Also, what would happen if the assumed 19mv nominal sense
voltage was really 14mv or 24mv. Probably the simplest
adjustment procedure would still require adjusting the
sense voltage resistor, but maybe we can come up with
something that works best for a wide range of inductors
and possible chip specs.


Take care,
Al

 

[Leeoniya]

alright. found 2 inductors that are both small enough to fit my design and have low dcr and ok peak current capacity...

look at these...one is Sumida, other is TDK

digikey:
445-1204-1-ND...........SLF7045T-220MR90
308-1131-1-ND...........CDR105-220MC

i also found this link of some interest from TI which compares efficiencies of different manufacturer inductors
http://focus.ti.com/lit/ds/symlink/tps61010.pdf

Sumida and TDK are highest.
i'm not sure if this carries over to switching regulator applications like this Zetex one.

is 0.9A enough peak current or will i have problems? it has good low DCR of 0.061. the other has a peak of 1.7A but a DCR of 0.09. both are 22uH. which is preferable?

thanks
Leon

 

[MrAl]

Hi there Leon,

The best way to find out is to try both of them.
The 1.7 amp sounds best because the dcr is only a little
higher, and the dcr doesnt hurt the eff as much as the diode
and the transistor in this circuit anyway. I did simulations
using 0.1 ohms and got pretty good results.

The best way to start up is to get a 'standard' Zetex
circuit running, then add the bias resistors later.

Take care,
Al

 

[Leeoniya]

...and so construction begins with the arrival of my new digital camera (powershot s410) with its incredible macro capabilities.

the new inductors are on order. i will be punching round PCBs tomorrow morning and chopping the aluminum tubes.

need to source some arctic silver epoxy now....

full res /ubbthreads/images/graemlins/smile.gif

thats right kids...that's 0.625" dia slug of 10ga. sheet sterling silver, hehe.

...just had to show off my camera.
Leon.

 

[Leeoniya]

...best camera ever. both of these pics were taken under very bright, yellow/orange construction lights. a little fiddling with white balance makes everything look dead on. it was solid orange in thew first trials.

full res

 

[MrAl]

Hi there Leon,

Very nice pics! Cant wait to hear some results you get
using your new parts and with the new inductors. Looking
forward to hearing about that.

Take care,
Al

 

[operator_smooth]

Leon, is that slug sufficient to keep the emitter cool or does it rely on contact with a metal flashlight body?

 

[Leeoniya]

while i'm waiting for the 2 inductors which should be here tomorrow or tuesday, i bought a pipe cutter and cut myself an aluminum spacer from the 6061-T6 tube i ordered ID 0.495, OD 0.625.

full res

my original idea was to have the circuit board the same diameter as the slug and sandwiched between the silver slug and the aluminum spacer. but that will affect heat transfer from the slug to the spacer a great amount. instead i have opted to punch the round circuit board to the ID of the spacer and insert it into the end. this will make the grounding of the board a bit more difficult, but the heat transfer will improve greatly because there is no FR4 in between the spacer and the silver slug, and it will reduce the overall height of the pill by 1/32".

the double sided pcb (components only on 1 side) will be arctic silvered to the back of the slug and slug will be polished and arctic silvered to the polished spacer.

most of the space inside the spacer will be occupied by the inductor, and the entire remainig space will be filled with a thermally conductive potting compound.

full res

so to answer your question, no. the slug itself will not be sufficient. the heat of the LED will be dissipated to the aluminum spacer through the slug and the heat of the circuit will be absorbed by the thermal compund and also transfered to the aluminum spacer.

what you do with it from there is up to you. i'm plannig on mating it to the flashlight body, but i cant tell you before i try it whether or not it will actually be necessary, but then again, is all of this crazyness necessary anyways? to most, probably not.

i'll be at the shop tomorrow morning punching the round PCBs and drilling the grounding hole through the silver slug as well as the pcb. i'm hoping to have the board etched by tomorrow evening and ready to go for the inductor. i still have no arctic silver on hand. and none on order at the time, unless someone wants to donate. heh.

i was disappointed to discover that Lux3s have reversed leads. so you wont be able to adapt it without drilling the hole on the other side of both the pcb and the slug and relaying out the circuit. /ubbthreads/images/graemlins/frown.gif

Leon

 

[MrAl]

Hi Leon,

Wow nice pics there again! You know they say a pic paints
a thousand words, but with this stuff i think it's more
like 100,000 words /ubbthreads/images/graemlins/smile.gif
I appreciate you posting those so we can all see clearly
what parts you're using and your heat sink too.
Thanks very much!

Take care,
Al

 

[NewBie]

In the past, many cpf'ers would make their own converters, and various experts here on cpf would discuss the results, and folks would work on them together.

I thought I'd bump this thread, due to it's great content, and if there are any budding technicians/engineers that would like to fiddle around a bit, the Zetex 300 series is a great place to start.

There are alot more related threads, deep in the halls of cpf, with all sorts of construction information, and this one is simple enough, that you don't really need a board to make it. The parts are low cost, and you can even wind your own inductor if you want.

 

[NewBie]

deleted.