PDA

View Full Version : CF v4 + Phoenix default config... wacky??!



Novastar
06-03-2008, 09:16 PM
I'm opening this thread because it appears Obi-Dar and Eastern have run into some issues with the defaults for:

* CF v4
* Using Phoenix Font

For some weird reason:

* Poweron light stuff works
* Poweroff light stuff works
* Idle light--fades out and doesn't return!!
* Clash light stuff works
* Lockup light stuff works

Ummm... I've got to be honest here folks: I don't get it. :)

BUT... I'll have Erv and myself look into it!!

eastern57
06-03-2008, 09:39 PM
I'll put this here too (I sent this to obi-dar and Nova)




Here's my thoughts and observations:

1) The phoenix pyre is special because it seems to eat up a LOT more power - I've noticed this consistantly, when I have a relatively low battery, this sound font triggers the Low Bat flashing LED, but not the other fonts.

2) In addition, with a full battery, phoenix pyre will flicker/pulse normal, but after a few minutes will resort back to "pulses all the way down and only lights up on impacts or other triggers".

3) My understanding of the "flickerp" setting was, if set to 5 (as in pp font) the blade will pulse between 95% and 100% of brightness - but it seems that at 5, it varies between 5% and 100%. This is also the same with the other sound fonts.


So in summary, I believe that, yes, the V4 still works as intended, but the specific combination of the Phoenix pyre sound font, low battery, and the [perhaps] misunderstanding of the flickerp setting, the overall power of the LED is just not enough to keep it alive - unless forced to by means of clash, blaster, etc.



eastern57 :)

erv
06-03-2008, 11:01 PM
this font does not eat more power, I'd even say less power since it's not constantly on.
It looks like a bug in my algorithm, like the destination value of the current control to get it back to 100% is getting wrong after a while.
About flickerp : It's being bad explained in 3.0 manual I think, now I guess that with the 4.0 it's better.
The value is the "dig". The bigger, the deeper the "dig" is.
With 80, you dig between 100% and 20%, you have a 80 span. We finally stayed on that cause it easier to remember, big number, deep pulse.
Now, to trouble shoot this one. I've notices that in some case, if you put 100%, combine with too much flicker you can get negative somewhere, I haven't got the time to really debug this one.
So, a combination of 95 for the pulse (100-5%) + heavy flicker can lead to this bug, but... I can't see why it would happend all the time.

low battery thing : when pumping back and forth the brightness, the battery cannot follow in a relevant way, and the voltage has ripples instead of being quite stable (dim down, less current, battery "breathes", voltage is climbing up etc). I'm also thinking that using the pulse mode with a version prior to 3.0 might lead to strange things since the low battery indicator isn't working the same way. I haven't thought about that. So, about the post above, if you have a 2.63, it's sort of normal that you have this low bat thing going on. Sorry for not having seen that before.


Trouble shooting & debug : set flickerd to 0 and see how it goes. If it's the same, change flickerp and turn it to 50 instead of 95.
Make sure the batteries are fresh, and let us know, it's an interesting bug.
The weirdest thing is that I've personnaly tried this font on li-ion batteries and it worked fine. Nova too.
so...
1 billion $ question : what battery setup are you using in those sabers ???

Novastar
06-04-2008, 12:57 AM
I wouldn't be sorry for any of this Erv... I'm constantly *AMAZED* at how few bugs are in CF!!!!

Great job!

And hey--thanks to Obi-Dar, Eastern and myself and yourself... we CAUGHT THAT LITTLE CREEPY CRAWLER!!!!

Now, we just take time to figure on how to smash that little guy, heheh

erv
06-04-2008, 01:40 AM
*evil voice*
HAMMER TIME !!!!!

Mad Hatter
06-04-2008, 07:19 AM
I'm having strange results with the default pulse as well. (Not just phoenix pyre.)

I haven't seen the blade turn dark and stay that way, but the pulse does drop all the way off on several fonts before coming back up. It's also different with different LEDs.

Episode II font looks great with the 10W red. (2 dies in series @ 1A on the main, and the other 2 in series driven by a 1 amp buckpuck controlled by the shimmer/clash outputs and 2 power extenders.) The red has a nice steady pulse that doesn't go too deep.

However, if I switch to blue or green 10W (wired the same way), the pulse drops all the way off on the low end. Other fonts with pulse drop off all the way with every color LED.

It isn't a matter of battery power. (7.4v 2600mAh li-ion.) The novastar font, which doesn't pulse, works great on all colors. And the battery has no problem with the extra 1A shimmering/clash effect (which looks spectacular btw!).

I was able to tweak the blackstar font to have a low depth pulse with the red, but even that drops off all the way with blue and green. I'll play with it a bit when I get a chance, and see if I can get a functional pulse on the green.

erv
06-04-2008, 08:04 AM
pulse mode relies on the LEd you're using, ie the forward voltage. PWM output from the luxeon driver will change from a led color to another, leading to different timing and "digging" results for the pulse mode. Those have to be tune depending on the color of the LED.
If it does not get deep enough, increase flickerp. This is going to slow down the pulse too. Increase a bit flickerg (grain) to balance that side effect.

and...
welcome to the world of Multidimensionnal parametrization !
:mrgreen:

if I was a really bad bad guys, I would say that it's SO complicated in express purpose to force the user to edit his config files... Not true (but nice try)

Erv'

Mad Hatter
06-04-2008, 08:50 AM
Thanks Erv.

The problem that I'm seeing is that even a very low flickerp will dig all the way down to nothing.

I know that I'm dealing with higher than normal forward voltages. Have you had a chance to play with the 10W blue or green with two dies in series on the main output?

Using the red 10W, EpII font (32 flickerp) looks great. With the same LED, blackstar (10 flickerp) digs all the way down to nothing.

Hmmmm.... the flickerd is much higher on blackstar... I think I just don't understand the parameters.... :)

eastern57
06-04-2008, 09:00 AM
1 billion $ question : what battery setup are you using in those sabers ???

4.8V 2600mah nimh (4AA) - now where's my money?

seriously, thanks for setting me straight. Apologies if I put out bad info - I'll stick to observations only - diagnosis we'll leave to the experts. I'll keep changing the setting - to see if I can sense any disturbances.

But that whole bit on multidimemsional parametrization...

whoa...

Eandori
06-04-2008, 09:12 AM
Episode II font looks great with the 10W red. (2 dies in series @ 1A on the main, and the other 2 in series driven by a 1 amp buckpuck controlled by the shimmer/clash outputs and 2 power extenders.) The red has a nice steady pulse that doesn't go too deep.

However, if I switch to blue or green 10W (wired the same way), the pulse drops all the way off on the low end. Other fonts with pulse drop off all the way with every color LED.
Mad Hatter... what LED's are you using? 10 watts?

Red Luxeon III is about 3.4v at 1.5A for 5.1 watts of total power.
Green\Blue Luxeon V are about 7.0v at 1.0A for 7 watts each of total power.

What is the 10 watt LED's you are using? What is their forward voltage/current?

Mad Hatter
06-04-2008, 09:28 AM
I'm using the new LedEngin 10W LEDs (http://www.mouser.com/Search/Refine.aspx?FS=TRUE&Ne=688453+688632&N=2093270+4294758820+4294722512+4294722529+4294722 442+4294628698+4294628695+1323038&Ns=P_Power+Rating%7c1%7c%7cP_SField). There is a thread about them here (http://forums.thecustomsabershop.com/showthread.php?t=4637).

The LEDs have four independent dies under a single dome, and each die has its own pads for custom wiring. I'm using 2 in series overdriven at 1A on the main CF output. The other 2 are wired in series powered by a 1A buckpuck. The buckpuck is controlled by the REF and CTL wires connected to a power extender, which is gated to the shimmer output through a potentiometer (to tune the voltage input, so the buckbuck will shimmer like the main blade). Then that power extender is hooked to a second power extender that is triggered from the clash flash - effectively giving a high-power 1amp shimmer effect on clash. (Big thanks to Erv for helping me figure all that out! :))

Forward voltage on the red is around 5v with 2 in series. Blue and green are around 7.6 or so by the specs. Putting the meter on the outputs when the CF is flickering shows less than 7v, which I'm assuming is an average.

Malaki Skywalker
06-04-2008, 10:36 AM
DUDE! A 10 watt and a CF!? :shock: Please post a video or pics of this if you can! :p

Mad Hatter
06-04-2008, 04:26 PM
It's not finished yet, but I'll start a thread soon to show the progress so far. I've been spending most of my time on custom circuits, and haven't really started serious work on the hilt.

Mad Hatter
06-09-2008, 01:27 PM
Okay, I've played around with the settings some more (with the 10W green), and there is definitely something up.

If I set flkrs, d, p and g all to 1, I get a very slow pulse that dips all the way down to nothing and back.

I raised flkrs and d up to "normal" values and it adds the flicker like you'd expect. I also varied the flkrg setting, and it behaves like it should.

However, the flkrp setting behaves vary strangely. Its almost like I have only 2 settings - 100% and around 60%. A setting of 1 gives 100% fluctuation. A setting of 35 gives 100% fluctuation. Setting to 50 or 80 gives about 60% fluctuation. Setting to 90 or 95 goes back to 100% fluctuation.

Now, I'm sure that there is a little variation in those results. (I'm probably getting two ranges of about 50-60% and 90-100%.) But it definitely isn't behaving as it should.

I'll have to switch back to red and see what I get. I've definitely seen a good pulse with the red on the default EpII config. Is it possible that the algorithm can't handle the high (~7.6v) forward voltage?


Edit: 36 seems to be the lowest proper setting with the green LED. The pulse is a little excessive at 36, but maybe not so much as to make me disable the pulse altogether. Setting flkrp to 35 or less causes the pulse to ramp all the way down.

Novastar
06-09-2008, 03:10 PM
THANKS for doing this kind of testing Mad!!! This is the kind of thing that will help me in the future (hopefully) regarding tweaking config.txt values for the NSCFCD... and just to help people in general.

Ever since I convinced Erv to try to "mix" both the pulse AND the flicker settings (as an option)... it's been QUITE tricky for us to figure out exactly how the behavior would go. Why? Well... where do we begin:

* Erv has JUST implemented this "combining" of the flicker & pulse (new concept)
* Every LED behaves differently due to design, current & voltage specs
* Some LEDs "react" more when you increase the current from X to Y... some less
* Some LEDs will "not light up" up until a certain voltage that is close to their ideal forward voltage... and when they do, they don't "ramp" from being super dim to bright... they can SOMEtimes just sort of (how do I say this)... start at 20% brightness (instead of 1%)...
* Erv's board is polling for the voltage the LED wants, so... I imagine that differing fwd voltages like 3.4... 3.9... 2.95... 6.85 and whatever... will have a drastic effect on the ramping, and therefore: the pulse + flicker
* Even I have problems getting the config.txt values "just the way I want them" for my fonts, given MY settings/LED/batteries, etc.

The moral of this story is... there MAY actually be *NO* way for poor ol' N-Star here to set some "ideal" parameters for the default v4 flicker & pulse settings--without them being WRONG for person X or Y. :(

This sucks quite a bit in my opinion, as... I really wanted to have everyone see how intriguing the light driving has become (and see the MANY different ways it can be setup for EACH sound font)... but it might be impossible since in this case: it is NOT "one size fits all"!! :)

Finally: if you are having SERIOUS v4 issues when both pulse AND flicker are on... disable one or the other by setting it to "0". Test and play, then see about changing them around again if you like.

Thanks again, Mad... :)

Mad Hatter
06-09-2008, 04:08 PM
Yes, disabling the pulse works fine. The blade looks good with just the flicker. And it isn't too bad with the pulse at 36.

Unfortunately, there's no middle ground. :( I tested every number from 1-35, and they're all the same - full drop off. (Well, it looked like 4 was a bit less, but it was still more than 36.) I'm assuming it's an issue with the algorithm, and not the LED, since these numbers should be less pulse, not more.

I didn't test all the higher numbers. I know that 50, and 80 both seemed to work (though 80 wasn't nearly 80% drop off). But 36 seems to be subtlest pulse I can get with this LED.

I haven't thrown the red LED back in, but I checked the default config that came with my CF, and the EPII font has flkrp of 32. I remember that looked good on the red. The default Blackstar had 20 (or so) and that dropped too deep. I tried lowering it, but couldn't get a good number. In hindsight, the red must have a drop off point somewhere between 20 and 32. (I was trying to decrease flkrp so it woudn't pulse so deep, but in this case it looks like I need to raise it.) I'll hunt for the lowest usable number on the 10W (~5v) red later.

Edit: I tried disabling the flicker to see if that fixed the pulse, but that doesn't work. Setting flkrs to 0 disables the pulse as well. Setting flkrd to 0 while s, p and g are higher still tanks at a flkrp of 35.

Novastar
06-09-2008, 08:23 PM
Hmmm... that's a bad sign right there... regarding if 36 is one of the only settings to produce a "nice" result.

The Sound CD (for v4 configs) uses the pulse numbers the way they were INTENDED... which may not necessarily reflect the ACTUAL result... we'll have to see.

Oh well... at least it's not something that will completely prevent the usage of CF for all versions... or even with v4 itself.

I don't even have a v4 anyhow, so... hard for me to test, heheh!

erv
06-09-2008, 11:21 PM
I'm going to rework a V4 proto that was unfinished and I'm going to try those values to see how it goes.
I have some 10W ledengin on my side, so I'll be able to test any configuration
Erv'

erv
06-11-2008, 12:48 AM
ok, in a nutshell (like sometimes I'm not writing a novel) it's a bug.
Total mea culpa, I got stuck with a variable type lenght in my code when calculating the target of the pulse ramp. Ramp is supposed to pulse between 100% and (100% - pulse effect strength). Value above 36 or so might not work and lead to big "digs" instead of subtle ramp.
So basically, if you are willing to get a real pulse, use values above 40 or 50.

I'm sorry about that, I didn't test with mini values, I was trying the pulse effect with values like 60 or 80 and didn't noticed.
It's a minor bug so I can pay for getting all the V3 and V4 back for free, but if you want to get your board upgraded and you pay the shipping, I'll happily do it like I've always done.
But for now, I'm moving next week and I might add a few things in 4.1 (next batch, when the missing part is available again but I don't know when), so please wait before getting this fixed, you'll get the benefits of a new version.
Erv'

Mad Hatter
06-11-2008, 05:52 AM
Thanks Erv. I'll ship it back to you to fix, but I'll wait for the upgrade.

As for new versions, any chance of getting a clash flash signal during blade lockup?

erv
06-11-2008, 06:21 AM
added feature to the wish list. That makes sense too. I mean, until now, only a VERY few people are using the clash flash so, I'm glad to receive any extension of this feature.
Will be in 4.1, it's in my todo list now !
Erv'

Obi-Dar Ke-Gnomie
06-14-2008, 10:59 PM
Sorry folks, I've been meaning to post in this thread, but I've been really busy lately.

I think that the problem I experienced was caused partly by the configuration settings mentioned above, and partly by the fact that my batteries were low. Just for the record, I'm using 4 NiMH AA batteries.

If you remember, I shot one video where the LED acted wonky, then in my second video, (which was shot immediately after,) my low battery light came on. When I have freshly charged batteries, everything works just as it should.

I'm not going to bother ripping out the board and sending it back. I'll play around with the settings, and keep the batteries charged, and it should be fine. Maybe I'll bump up the low battery indicator setting to give me a little more warning that I'm running low.

Thanks for the help!

Novastar
06-16-2008, 03:52 PM
Don't worry, Obi-Dar... you guys helped Erv and I to discover a bug. And sure, it's a relatively SMALL bug, but... the entire point of newer versions are to improve / build upon things.

And again... I'm ALWAYS very surprised at how bug-free Erv's stuff has been. Especially once things have gotten ****REEEEALLLY**** complicated, what with the new light-driving options and so forth.

Anyhow... I've been helping Erv test the changed parameter, and it appears to work exactly as it initially would have if the parameter was simply (unsigned) or something simple like that.

In essence... the bug was almost as small as "forgetting a semi-colon" way back in the day of compiling C++ programs with the standard UNIX command-line stuff way back in the day. ;) And FORTUNATELY, the bug is not so catastrophic as was forgetting a bloody semi-colon back in the day...!

Novastar
08-13-2008, 08:07 PM
Anyone else have a problem with this lately? Just wondering...

erv
08-14-2008, 02:14 AM
I don't think a lot of people are messing so much with parameters (me included, when working on the boards, design or algorithm, I'm rarely trying fine tuning the values, I try 0, low value, a "quite high value" and that's it).
that's why bugs sometimes stay burried until an advanced user starts digging it !