'Slow' response on flash button

[remco_k]

Yesterday I tested FreeStyler (together with my own build Velleman K8062/VM116 USB->DMX) for the first time on a real DMX universe, at home.
My 'universe' was'nt big, it just contains a 4 channel DMX dimmer. My goal: just to check if my Velleman interface actually works.
And yes! It does work, happy as a child I went on playing...

Some Sequences and Cue's later, I went on to the override buttons. There I noticed a pretty long delay between the actual keypress and the moment the light goes on.
Here's what I did:
Create a 1 step sequence, setting 1 light to on, saved it, adding it to the buttons and set the button to flash mode.
So when I press the button, the light goes on, when I release the button the light goes off again.
The problem is that there is too much delay between the actual keypress and the moment that the light goes on. I'd say its some times more than 100msec, sometimes less and sometimes it even completely miss a very short flash (the light stays out, while I'm pressing the key very short).

Somewhat the same problems I have when I make a sequence with 2 steps, with 100msec interval. Step 1 light on, step 2 light off.
My light does start flickering at some speed, but is not flickering in the 100 msec rythm, at all. Looks more like some random timings.

I'm concerned about this. In a way, I know I can't work like this in real live situations. Worked with plenty other real hardware DMX controllers, did'nt have these problems, so the DMX technique itself is not the limitation for this.
Then it leaves few options where the source of this problem(s) maybe:
-FreeStyler
-My computer is to slow*
-My Velleman interface / driver can't respond fast enough.
(* don't think it is, CPU remains below 5% usage...)

Anywone here who can confirm something?

Tonight, I'm gonna write a tiny C++ program wich controls the Velleman interface directly though its driver. I'm gonna do that for testing if the Velleman interface is capable of responding fast enough to these actions. If Velleman responds fast enough, I'll know the problem is somewhere else, like FreeStyler.
If the problem (to slow reactions) persists then I'll know the Velleman is not suited for me, and going to find an interface that can handle what I want.
More thoughts about this?

[Röschi]

Hi,

i had this problem in a same way when I created a strobe sequence for my PAR56 LED (Step 1 value 255 red channel, step 2 value 0 red channel). I tried several scene times from 100ms up to 500ms and I noticed the same. The strobes sometimes came in the chosen period but about every 5th strobe was generated a way to early or got completely lost.

I think this problem comes from FreeStyler itself, because any other freeware DMX software doesn't show this problem with my interface (uDMX).

[MarSafe]

Hi, Remco,
don't worry about the interface. I'm using it for over 2 years now, and I have noticed this behaviour. A few months ago I used FS on a P3 733 with 512 MB and found that (for example) the Bezier-shapes were faster on the laptop than my scanners could actually follow. I moved up now from this laptop to a AMD Athlon 2500+ with 2 GB and the scanners are about as quick as the screen is (in 2.96j, that is. The dot doesn't move in 3.01). For a flash I use the flash-button of DMX400, and it respond very well. So I think the PC is the cause of it, although I haven't got a clue what you are using now. Can you let us know?

Regards,
Martin.

[remco_k]

So I think the PC is the cause of it, although I haven't got a clue what you are using now. Can you let us know?

Ofcourse;
Laptop, Compaq/HP NX6110
Intel celeron m @ 1.4 GHz
1 GB DDR SDRAM
Still I think that my laptop is fast enough for this to work good.

I'm quite happy reading the reaction of Röschi, he describes exact the same problem when using sequences with very short timing, using another interface.
So, that may leave us 2 options for the possible source of the problem:
-FreeStyler
-My computer is to slow

Either way, this evenening I'm going to try out a little C++ program, just to be sure. (Will be on a way faster desktop computer, so I'll test with FreeStyler on there too).
In the meantime I have written an e-mail to Raph (the author of FreeStyler) and asked him to think along in this topic. I hope he does, and just maybe if the problem is in FreeStyler, he can have a look at this.

[Röschi]

Ok, I just ran my sequence on my laptop, which is pretty faster than my desktop pc (1.83 GHz dual core). I actually noticed a great performance difference. Of course a LED Parcan isn't build for running strobes, but with my laptop the update lacks are quiet gone. Actually, now I can display a white strobe (3 channel, step 0 - 255), which is totally impossible on my pc!

[remco_k]

Well there is something!
Seems it -has- to do with the speed of the computer. (Although I'm still thinking that responding to a key and setting few bytes from 0-255 and back is not that much of a deal. A 300 MHz computer could do that 1000 times in a second, so why not we? Thats the reason I hope Raph is reading and thinking along.)

I just had a tiny brain wave, my wife's (work) laptop is way-way faster than mine. Poor old me.
When I get home, I'll steal it from her and do the same test on there to. Only disadvantage is when I go out for a light job, she can't work.

[aquas]

Hi,

I 've the same thing on live, because i use a old Presario 600Mhz, with...192Mb RAM, so when i try to increase speed to the moving heads...nothing hapen, but if i decrease speed, they react. And when i run a seq the speed decrease, if i run 2 seq decrease more, and 3 seq more...

Some day i've try to use the desktop... totaly diferent, much more speed than in the laptop, and the desktop is a old Athlon 1500+ with 512 RAM. I've never tried a New Dual Core....but i think that runs perfectly.

greats

[remco_k]

As promised earlier:

Tonight, I'm gonna write a tiny C++ program wich controls the Velleman interface directly though its driver. I'm gonna do that for testing if the Velleman interface is capable of responding fast enough to these actions. If Velleman responds fast enough, I'll know the problem is somewhere else, like FreeStyler.
If the problem (to slow reactions) persists then I'll know the Velleman is not suited for me, and going to find an interface that can handle what I want.
More thoughts about this?

Description:
Did that test on my 1.4 GHz laptop and 2.8GHz desktop computer, made (or better said, refurbished the Velleman SDK sample a bit) a tiny test tool wich puts out 4 DMX dimmer channels, freely assignable.
Also the maximum transmitted channel can be set. (wich speeds up the response time of theDMX universe if set lower than 512).

The results:
With my test tool I can get a very quick response from flash button to light.
If the max transmitted channel is set to 512 then it all gets slower, but not as slow in FreeStyler.
If the max transmitted channel is set lower, say 256 channels then it is way faster (about twice as fast), fast enough for working with it.
The lower the transmitted channels, the faster my DMX universe responds. Actually that complies with the way the DMX512 protocol works.
It does'nt have to transmit all 512 channels. So decreasing the max channel number in fact increases the response time of the DMX universe.
Need some theory? Look here: http://www.dmx512-online.com/packt.html
So it has nothing to do with FreeStyler, nor the Velleman interface. Its just the way DMX512 works.

The problem:
Now, back to my problem; I found that the Flash button responses are to slow in FreeStyler, but also in my test tool when max channels is set to 512.
Considering the above, I thought back to a topic I made earlier: FS always outputs 512 channels?
Where I'm asking around if FreeStyler does that. I think I know the answer by now; yes.
That is also the (or part of) source of the problem described in this topic; if FreeStyler could be freely configured to do max output channel X, then FS users can benefit greatly from it, because their DMX universe is much much more responsive.
Ofcourse, if users have a DMX universe where DMX channel 511 is in use, they can't lower that 'new' setting, otherwise that unit won't work anymore.

A little conclusion:
It seems like it partly is a DMX 512 protocol limitation and I learned that the hardware consoles I use also automaticly change the max transmitted channel to the lowest channel possible.
Reason: it's more responsive.
Some numbers:
If the max channel count is 512 then sending an DMX packet on the DMX universe takes up to about 24 msec in total.
If the max channel count is 256 then sending an DMX packet on the DMX universe takes up to about 12 msec in total.
If the max channel count is 128 then sending an DMX packet on the DMX universe takes up to about 6 msec in total.
... and so on. (Sort of...)
So, the lesser channels are transmitted, the faster the DMX universe responds.
Take in consideration that the Velleman driver could have some overhead, so sending 512 channels could take up to 50 msec in total...
Also take in consideration that FreeStyler takes some time to put out the values to the Velleman adding some more msecs to the total time.
Then flashing with 100 msec intervals, becomes nasty...

The question to Raph:
Can FreeStyler have a new option to set the maximum transmitted channel, so that (among others) the Velleman interface sends only the DMX channels below that number.(use the function SetChannelCount(long Count) in the dll).
I truly think that this solves my problem (and maybe that of others) because I use DMX universes with a maximum of 180 channels. That will speed up things greatly.
So this is actually a feature request now.
Maybe you can check if there's some improvement to be made while reacting to the buttons.

For others:
I have put the little test tool online, so you can see for yourselves:
http://www.digiplay.nl/download/K8062_Demo.exe
First assign the channels below the faders you like.
You can use the F1, F2, F3 and F4 keys as flash buttons.
With Max transmitted channel you can set the channel count that is sended along your DMX universe. Lower = faster.
Try 512 first. Flash fast with F1, see how 'slow' your DMX unit responds or even misses some flashes.
Now set the max channel to 4 and try again. Notice how fast it is now...

[djSupport]

couldn't the quality of you cable affect the speed as well??? also wouldnt u just go out and buy an enttec usb dmx pro with built in microproccesor so it should be damn fast!

[remco_k]

couldn't the quality of you cable affect the speed as well???

It could, but it does'nt. Using quality data cable for my testing.
Just a few meters, and terminated the DMX line. (tried the same tests without termination, no change)

also wouldnt u just go out and buy an enttec usb dmx pro with built in microproccesor so it should be damn fast!

Then still the DMX512 packet takes 24 msec to go and add FreeStylers 'delay' to that, no interface can make that faster.
The Enttec might be a little faster than the Velleman kit but I think still not fast enough.
The big change still remains in an option to configure FreeStyler to put out only (say) 180 channels. There is a lot time to win (8 ms instead of 24). It has to be considering it theoraticly, but in practice I have proved that. Real DMX consoles do this to, so why not implement a (great!) feature like this in FS?
Next, if Raph also can optimize FreeStyler a little to react some faster, then again some time has won.

Then it is damn fast, with the Enttec, but also with the Velleman and many (if not all) other interfaces.

[djSupport]

I think with the enntec pro you can reduce or speed up the timings.... fs only has to output somethiung then the pro deals with the dmx bit

[remco_k]

I think with the enntec pro you can reduce or speed up the timings.... fs only has to output somethiung then the pro deals with the dmx bit

The same goes for the Velleman, only its driver deals with the DMX bit.

In the meantime I got an e-mail from Raph, he's looking around in FS. I think we get there...

[remco_k]

... I think we get there...

And we did!

I e-mailed a few times with Raph, now I have a beta version of FS, where I can set the max number of DMX channels for my Velleman VM116 -and- I can set the FreeStylers refresh rate somewhat faster. Normally its hardcoded 50msec, meaning that every 50msec FS sends the new DMX data to the driver of your selected interface. Now it is configurable and that 50msec can be lowered to 20msec.

Seperate from eachother these 2 things already make improvements on the flashbutton reaction. But when I optimize both, then its lots and lots faster than in the 'old' FS. Now I can press my flash button, and the light goes on simultaniously*. It doesn't take a full day anymore...
I have set the max num output channels to 256 and the refresh rate to 20msec for my testing. This is a big, big, big step ahead!
Not just for me, I'm sure about that.
Ofcourse people could set the max num channels to 96, then it will be more and more faster...

Raph; thank you so much for thinking along and for your work!

* Ofcourse its not real simultaniously, but is sure feels that way now. And thats good!

[djSupport]

That sounds GREAT!!!

[remco_k]

That sounds GREAT!!!

It actually is...
The past half hour I've done nothing else than flashing my light on and off with the flashbutton... So fast... Wow!