DSM2 work

[pmackenzie]

I thought of one other thing that might be useful - getting the beeper to go off when in range check mode, and perhaps some text on the screen?
Nothing too obnoxious as far as the beeper, just the occasional short beep to remind you.

Pat MacKenzie

[Clivew]

Are we at a point where Pat's latest code update should be put on as the latest revision of er9x? It seems like it is quite reliable. It can always be tweaked further later.

Mike.

I would vote yes to that, since all functions appear to work perfectly!

Clive

[Kayle]

Hi,

Are we at a point where Pat's latest code update should be put on as the latest revision of er9x? It seems like it is quite reliable. It can always be tweaked further later.

Mike.

Yeah that would be great. I need the FrSky version

regards
Kayle

[cre8tiveleo]

Here is some info from rcgroups,
http://www.rcgroups.com/forums/showthread.php?t=1482559

Spektrum DM9 module and it has a fixed mode of operation (9 channels, 10 bit resolution, no 2 way binding)

The channel order of the incoming serial data is as follows:

Byte 1 = 0x00
byte 2 = 0x00
byte 3 = Throttle High Byte
byte 4 = Throttle Low Byte
byte 5 = Ail High Byte
byte 6 = Ail Low Byte
byte 7 = Elev High Byte
byte 8 = Elev Low Byte
byte 9 = Rudder High Byte
byte 10 = Rudder Low Byte
byte 11 = Flaps High Byte
byte 12 = Flaps Low Byte
byte 13 = Gear High Byte
byte 14 = Gear Low Byte
byte 15 = Aux1 High Byte
byte 16 = Aux1 Low Byte
byte 17 = Aux2 High Byte
byte 18 = Aux2 Low Byte
byte 19 = Aux3 High Byte
byte 20 = Aux3 Low Byte
etc... up to 14 channels. Again, this part is not fully tested as the RX with the most channels I have is the AR7000. I have tested the serial input mode successfully with 6 channels so far.

Also, there is no need to send 14 channels worth of data. You only send as many channels as you need to the module. However, they must be sent in a regularly occurring burst every 22ms just like a PPM frame. The timing of the DSM2 transmissions is based on the start of the incoming PPM frame or first received serial character.

--------------------------------------------------------------------------------

So I'm thinking , since you're sending 6 channels of data (1-6) and if the ar9000 is expecting at least 9 channels of data (1-9) that the data packet is being read twice by the ar9000 (two consecutive sends to make up one packet)

ar 9000 wants : 1 2 3 4 5 6 7 8 9 (up to 14) in 22ms
dsm2 from er9x : 1 2 3 4 5 6

what ar 9000 translates from what it gets from the er9x tx.

1 2 3 4 5 6 1 2 3
4 5 6 1 2 3 4 5 6 * Jitter from rudder channel (4) onto throttle channel (1)
1 2 3 4 5 6 1 2 3
4 5 6 1 2 3 4 5 6 * Jitter from rudder channel (4) onto throttle channel (1)
.
.
.

To a laymen type compared to what you all can do, this makes total sense to me.

[pmackenzie]

Interesting link, thanks for posting it.

It must be something like that, but the module should be taking care of all this.
I tried sending 8 channels to the LP4 module and the system did not work at all, but that was with the old ISR. I will try again with the new one.

The frame width is a bit shorter than the 22 msec they talk about and that might be it, but when I tried to make it longer the MCPx did not work.

I will try to work up a version where you can set the frame length in the menus, so you can experiment with making it longer to see if it helps the AR9000.


Pat MacKenzie

[MikeB]

The final item I still want to look at is latency. It is likely not that good right now (or at least not as good as it could be), since the outputs for the next frame are locked in ~20 msec before it is sent.
The call to setupPulsesDsm2() would ideally be done just before the beginning of the frame, however the timing is important since it seems like the modules are quite sensitive to the frame width.
Pat MacKenzie

I thought about that, in the OCRB compare interrupt, (set to 200), when it interrupts at 200, set the compare to (say) 35000 and return. When you get this second interrupt, set the compare back to 200 and call setupPulsesDsm2. This should reduce the latency significantly.

If you increase the number of channels, make sure you increase the size of the pulses union.

Mike.

[MikeB]

ar 9000 wants : 1 2 3 4 5 6 7 8 9 (up to 14) in 22ms
dsm2 from er9x : 1 2 3 4 5 6

what ar 9000 translates from what it gets from the er9x tx.

1 2 3 4 5 6 1 2 3
4 5 6 1 2 3 4 5 6 * Jitter from rudder channel (4) onto throttle channel (1)
To a laymen type compared to what you all can do, this makes total sense to me.

Quite possibly, and it may just use the channel 4 byte of the second frame, and then realise it doesn't want any more so ignore the rest so only channel 1 is funny. What happens if you put a servo in channel 7, is it a copy of channel 1?. If so you have a quick, short term solution, throttle on channel 7 output.

Mike.

[pmackenzie]

Here is a version that lets you set the frame length by using the PPM Frame length adjustment in the setup screen.
I changed it so you can make the adjustments in 0.1 msec increments instead of 0.5 msec.
AR6110 works from about 19 to about 23 msec.
MCPx doesn't like it much over 22 msec, but looks to work fine right down to 12.5 msec!

I wonder if this means that we can get 11msec performance out of the faster Spektrum receivers

[cre8tiveleo]

...What happens if you put a servo in channel 7, is it a copy of channel 1?. If so you have a quick, short term solution, throttle on channel 7 output.

Mike.

For the time being (since the jitter issue startes) I have the throttle mapped on channel 6 (love er9x for that function alone) and keep my 5th channel for remote switch, the rest are the same. If it were indeed a copy on channel 7 of 1, and 2 on 8 and 3 on 9... that would be uber cool, two elevator and two aileron ports... automagically... (I'll fill the thing with servos and see what happens)

I think I tried a servo on channel 7 (but my mind is slipping and I could of just went to 6 thinking that's all the tx was capable of sending), and nothing happened, I will try it again tonight when I get home.

[cre8tiveleo]

...I wonder if this means that we can get 11msec performance out of the faster Spektrum receivers

Awesome, I will try it out once I get home, that was quick! Thanks Pat!

Cheers

[Rob Thomson]

I thought of one other thing that might be useful - getting the beeper to go off when in range check mode, and perhaps some text on the screen?
Nothing too obnoxious as far as the beeper, just the occasional short beep to remind you.

Pat MacKenzie

That should be easy to do . Just a call to the audio class!


Sent from my iPhone using Tapatalk

[pmackenzie]

I will look into that, but it will have to come from the 10msec timer section I think.
Not much of an object oriented programmer I am afraid, the whole "class" thing just adds a level of fog for me.
But I am working on it, one of the reasons I jumped into the ER9X thing in the first place.

I messed around with the frame length (easy to do with the version I posted a couple of posts back) , and with the 6115 DSMX receiver it does reduce the frames length going to the servos as you reduce the frame in 9X.
I will have to devise a test to see if it is outputting new data every time, or just doubling up on the previous frame (Like FrSky does in its useless "High Speed" digital servo mode)
AFAIK, 6115 is not advertised as being capable of high frame rates.

6100 and 6110 keep their frame rate fixed. They will work at 11msec incoming, but not 12. But there is no advantage since the servos still get ~20 msec frames.

Pat MacKenzie

[Rob Thomson]

Lol

We all have our weak points and strong points.

I am amazed at how you manage to work out the hardware interfacing stuff - yet find classes easy!

to make your life easy - just use this wrapper function

audioDefevent(uint8_t e);

e.g. audioDefevent(AUDIO_WARNING1);

That should do the trick!

Once you have it working we can finetune the sound to suit.

Rob

[cre8tiveleo]

Here is a version that lets you set the frame length by using the PPM Frame length...

Well, the frame length does something, but it doesn't fix the issue with the ar9000.

I'm going to play with the settings and other dsm modes to explain it better... but basically...

frame length of 12.5 jitter is hyper fast... as the frame length increases, the jitter slows. at a point the bind lights flash, slowly, cycle, then then quicken as the frame length increase, kinda cool.. I'll make a video of that and post it. At a certain frame rate, the servo does a fancy dance and just freaks out, not just jitter... it's also kinda cool to watch.. the cats love it.

Channel 7 and above seem not to be affected at all. there isn't anything there, no flickers, or anything.

Will report more later. Any chance of sending more 'packets' to fill in the missing 3 channel data? so it's 123456789 instead of 123456... 123456...

Cheers

[cre8tiveleo]

...
frame length of 12.5 jitter is hyper fast... as the frame length increases, the jitter slows. at a point the bind lights flash, slowly, cycle, then then quicken as the frame length increase, kinda cool.. I'll make a video of that and post it. At a certain frame rate, the servo does a fancy dance and just freaks out, not just jitter... it's also kinda cool to watch.. the cats love it....

And, here is the video of the funky servo dance. (the Hitec is on ch1, the one blue hk servo is on channel 6 mapped to 1, and isn't affected when I do rudder *TAER* and the other sero *right side* is on channel 3, yeah, no servo on 4.. my bad)



just another vid...



one more, cause I was bored.



sorry about the video quality, borrowed the from the wife.

[pmackenzie]

Model Match works
There was a post on RCG in the DSM2 thread:
http://www.rcgroups.com/forums/showpost ... tcount=333

I had tried the same thing earlier, but it did nothing. Based on his post I gave it another try, and it works with my DX4e module. I simple changed the second byte to 1+ the current model's number.

Works perfectly

The problem with this method is that if you move the models around the number will change, so perhaps something like adding up all the characters in the model name and using the remainder would be better? (and not conflict with Spektrum's intellectual property)

Code: Select all

static uint8_t *Dsm2_pulsePtr = pulses2MHz.pbyte ;
void setupPulsesDsm2(uint8_t chns)
{
    static uint8_t dsmDat[2+8*2]={0xFF,0x00,  0x00,0xAA,  0x05,0xFF,  0x09,0xFF,  0x0D,0xFF,  0x13,0x54,  0x14,0xAA};
	uint8_t counter ;
	CSwData &cs = g_model.customSw[NUM_CSW-1];

    pulses2MHzptr = pulses2MHz.pbyte ;
    
		// If more channels needed make sure the pulses union/array is large enough
    if (dsmDat[0]&BadData)  //first time through, setup header
       {
	   switch(g_model.ppmNCH)
	   {
	   case LPXDSM2:
	     dsmDat[0]= 0x80;
	     break;
	   case DSM2only:
	     dsmDat[0]=0x90;
	     break;
	   default:
	     dsmDat[0]=0x98;  //dsmx, bind mode
	     break;
      }
    }
   if((dsmDat[0]&BindBit)&&(!keyState(SW_Trainer)))  dsmDat[0]&=~BindBit;		//clear bind bit if trainer not pulled 
   if ((!(dsmDat[0]&BindBit))&&getSwitch(MAX_DRSWITCH-1,0,0)) dsmDat[0]|=RangeCheckBit;   //range check function
   else dsmDat[0]&=~RangeCheckBit;
	dsmDat[1]=g_eeGeneral.currModel+1;  //DSM2_Header second byte for "model protect"
    for(uint8_t i=0; i<chns; i++)
		{
      uint16_t pulse = limit(0, (g_chans512[i]>>1)+512,1023);
      dsmDat[2+2*i] = (i<<2) | ((pulse>>8)&0x03);
      dsmDat[3+2*i] = pulse & 0xff;
    }

		for ( counter = 0 ; counter < 14 ; counter += 1 )
		{
    	sendByteDsm2(dsmDat[counter]);
		}
    pulses2MHzptr-=1 ; //remove last stopbits and
    _send_1( 255 ) ;	 //prolong them
    _send_1(0);        //end of pulse stream
		Dsm2_pulsePtr = pulses2MHz.pbyte ;
}

Clive - I am working on a version to send more channels.

Pat MacKenzie

[Clivew]

Hey, What a hero!
Brilliant work Pat, I think we all owe you big time for all your efforts!
Do you ever sleep?
Best regards,
Clive

[pmackenzie]

Sleep - not enough

Ok, here is a version that puts out 9 channels of data. In LP mode it still only puts out 6, because my LP modules won't work with more than 6.
This way it retains compatibility.

Mike - I setup the Union as

Code: Select all

union p2mhz_t 
{
	uint16_t pword[92] ;   //72
	uint8_t pbyte[184] ;   //144
} pulses2MHz ;

184 = (2+Number of channels*2) * 10 +4

Also made the DSMDat array larger (2*9+2)

And there was a loop in the routine that used the constant "14" that I changed to:
for ( counter = 0 ; counter < 2*chns+2 ; counter += 1 )

Works fine in 9 channel mode with my AR6100 and DX4e module. Anxious to hear if it solves Clive's AR9000 problem.
Since this also has the Model Protect feature you should have to rebind.

Can't try it with my MCPx. Had a mid-air Monday night, and the board won't connect or bind anymore.

Pat MacKenzie

[cre8tiveleo]

Clive has an Ar9000 too?

Man, you need some sleep!

Cheers, I'll check this out shortly.

[cre8tiveleo]

okay...



lpxdms will bind with everything... the other two modes , don't bind with anything.

cleared all my memory of models, just did one model, tried each mode... only the first bound and was the same. the others, nada, bind light on rx's kept flashing, I tried 4 different rx's. ar9000, ar6200 6100 and 6110e

This was the 9ch version I tried.

[cre8tiveleo]

tried the version just before the 9ch, and everything binds, and keeps bound, even when switching dsm2 modes.

[pmackenzie]

What module are you using? More than 6 channels only worked for me with the DX4 module.

I can get it to bind with my AR6100/AR6110 in 9 channel DSM2/DSMX mode, but it won't bind with the 6115.
I can force a bind to happen in the 6115 by doing it in LP mode. Two problems, when I switch to 9 ch mode the receiver is not happy, and the bind will be in DSM2 not DSMx.

Can you try to do a bind with the AR9000 in LP mode, then switch to DSM2/DSMX mode?

If that works then I should be able to limit the number of channels sent during bind to 6. And also figure out a way to limit the number of channels.
It would be easy, but the # of channels in the model data is being used to hold the operating mode. Hmm.

I also need to test this with my AR7000 to see if the data is actually going out, but now I have to get some sleep

Did you confirm the "model protection" is working, assuming you are not using an auto bind module?

Pat MacKenzie

[cre8tiveleo]

...
Can you try to do a bind with the AR9000 in LP mode, then switch to DSM2/DSMX mode?
...

That's the only mode it binds in. It's the only mode my ar6100, 6200, 6110e 6100e bind in. Once they are bound, and I switch to dsm2 only or dsm2/dsmx the bind lights go off.. and they won't rebind, unless I go into lpx mode.

That's with the 9ch version. the other version, it binds in all modes. I haven't checked the model thingy, but I will before I hit the sac, which is soon too, early day tomorrow. LAUNDRY DAY!

Cheers, and thanks again for all the great work so far!

Ahhhhh

[cre8tiveleo]

Oh yeah, Model match... WORKS!!!! My god that is awesome! Beers sir! I owe you some beers!

Bound 4 different rx's to 4 different models, (all ar6000 series family) I used the hex file before the 9chdsm2 to test. (will test the 9chdsm2 hex more tomorrow, when my eyes can focus)

Went through each model 1- 4, bound each rx. left all the rx's on (yes, since they aren't attached to a plane it's safe) then just switched models. and only the rx that was bound to the model functioned, the others just watched as their brother controlled servos. very nice!



I go sleep now too... cheers!

[Clivew]

Here is a version that lets you set the frame length by using the PPM Frame length adjustment in the setup screen.
I changed it so you can make the adjustments in 0.1 msec increments instead of 0.5 msec.
AR6110 works from about 19 to about 23 msec.
MCPx doesn't like it much over 22 msec, but looks to work fine right down to 12.5 msec!

I wonder if this means that we can get 11msec performance out of the faster Spektrum receivers

Watch this space!
I have an AR8000 coming, I wll report back.
Everything else continues to work perfectly, including "model protection" and range check
on both lp4 and dx4e modules plus dsmx on the latter.
Shame about your mcpx Pat!
Clive

[Rob Thomson]

For what it's worth...

Had a good flying session with my LP4DSM module and Micro Sukhoi today.

No glitches or issues - just works!

i guess these issues are related to specific tx modules.

Rob

[MikeB]

Pat, my er9x source files have the DSM2 changes I put in, plus some other fixes. I'd like to commit these other fixes. Are we at a point where the DSM2 stuff is good enough to commit? We can commit other improvements later. If we are, could you let me have the source files you have changed please.

Mike.

[cre8tiveleo]

Don't do the 9ch stuff yet, everything else works great though.

Cheers.

[pmackenzie]

Pat, my er9x source files have the DSM2 changes I put in, plus some other fixes. I'd like to commit these other fixes. Are we at a point where the DSM2 stuff is good enough to commit? We can commit other improvements later. If we are, could you let me have the source files you have changed please.

Mike.

I will do a bit more testing today, but other than the >6ch stuff I think everything is OK.

One thing I am not quite happy with is the menu item where you select the DSM mode.
There can be a tiny glitch when you switch to DSM2 from PPM, the first value you see for the sub mode selection is invalid because the range checking on the variable is not done yet.
(The valid range is different between DSM2 and PPM, and I am borrowing the PPM variable to maintain the same model data structure)

So for example if you have the PPM mode set to 6CH and then switch to DSM2 you will see " msec0" in the next field, garbage from another string.
As soon as you scroll over and select it the call to CHECK_INCDEC_H_MODELVAR with the valid DSM2 range will occur and you get a valid string.

The invalid value does not cause a problem for setupPulses because I used the "default" case, but it does not look very professional.

Pat MacKenzie

[Kayle]

Hi,

i put a AMTX11 from a DX4e in my 9x an it just works great. I use a seperate 3,3V regulator and a levelshifter with 2 resistors ( 10k and 20k ). I set my Model to dsm2 and ....... wow it woks

Thanks for all to make this possible.

Kayle