Skip to contentSkip to conversation

kaffeine 0.8.8 USALS/GotoX problem

PATCH Hans Werner 1 participant 1 message Jul 14, 2009 Media (V4L/DVB)

lore
Share & tools
lore

Hans Werner

author
#0 lore raw
Quoted 14 lines

There is a bug in DvbStream::gotoX in Kaffeine 0.8.8, which is fixed by the
attached patch.

Look for example at what happens when azimuth = 15.951. The motor should be
driven to 16.0 degrees, but when 15.951 is rounded up with if (USALS>0.5) ++rd; the carries into the upper nibble of CMD2 and the lower nibble of CMD1 are not handled properly. The result is CMD1 represents 0 instead of 16
degrees and CMD2 is undefined, being set by the non-existent 11th element of
DecimalLookup.

Quoted 5 lines

The patch fixes it, and I also took the opportunity to fix the strange division by while loops.

Regards,
Hans

-------- Original-Nachricht --------

Quoted 24 lines

Datum: Tue, 14 Jul 2009 16:17:30 +0200
Von: "Jesko Schwarzer" <[email protected]>
An: "\'Hans Werner\'" <[email protected]>, [email protected]
Betreff: AW: [BUG][PATCH] kaffeine 0.8.8 USALS/GotoX problem

Hello Hans,

I looked into the patch - just to be informed what's going on and got confused by such code.
I reimplemented it to test it against side effects and still do not understand why it is written so complicated. Sure, I don't know the environment of the code, but does it matter in this case ?

If you like try these few lines (the last 3 lines) as a replacement:

	unsigned long USALS;
	int CMD1;
	int CMD2;

if ( azimuth>0.0 )

		CMD1 = 0xE0;    // East
 	else
		CMD1 = 0xD0;    // West
Quoted 8 lines

/*-----------------------------------------------*/

	USALS = ( fabs( azimuth ) * 16.0 + 0.5 );
	CMD1 |= ( USALS & 0xf00 ) >> 8;
	CMD2  = ( USALS & 0x0ff );

/*-----------------------------------------------*/

Regards
/Jesko

Hi Jesko,

thanks for looking at it. I see the main difference in your version is that there is no DecimalLookup for the bottom 4 bits.

I took a look at the Eutelsat specs here:
http://www.eutelsat.com/satellites/pdf/Diseqc/associated%20docs/positioner_appli_notice.pdf

The relevant section is 3.10.

Reading that I see that the lower four bits should never have been coded with DecimalLookup at all !! The 12-bit value should be a true binary fraction (not any coded form of decimal) representing the angle to a precision of 1/16th of a degree.

Whoever originally wrote the routine may have misinterpreted Table 2 which shows a lookup table converting the lower 4 bits of CMD2 to decimal, just for people who want to make STB displays and data entry limited to just 0.1 degree accuracy.

So I think your version of the routine is correct. Mine differs sometimes by 1/16th of a degree (not perceptible with my equipment), and is the same otherwise.

I have attached a new version of the patch. I compiled and tested with my rotor and confirmed it works.

Regards,
Hans

Signature · 4 lines
Release early, release often.

Jetzt kostenlos herunterladen: Internet Explorer 8 und Mozilla Firefox 3 -
sicherer, schneller und einfacher! http://portal.gmx.net/de/go/atbrowser
dvbstream.cpp renamed from dvbstream.cpp.orig +7 -18
@@ -758,31 +758,20 @@
 
 void DvbStream::gotoX( double azimuth )
 {
-	double USALS=0.0;
+	unsigned long USALS;
 	int CMD1=0x00, CMD2=0x00;
-	int DecimalLookup[10] = { 0x00, 0x02, 0x03, 0x05, 0x06, 0x08, 0x0A, 0x0B, 0x0D, 0x0E };
 
+	//rotation direction
 	if ( azimuth>0.0 )
 		CMD1 = 0xE0;    // East
 	else
 		CMD1 = 0xD0;      // West
 
-	USALS = fabs( azimuth );
-
-	while (USALS > 16) {
-		CMD1++;
-		USALS-= 16;
-	}
-	while (USALS >= 1.0) {
-		CMD2+=0x10;
-		USALS--;
-	}
-	USALS*= 10.0;
-	int rd = (int)USALS;
-	USALS-= rd;
-	if ( USALS>0.5 )
-		++rd;
-	CMD2+= DecimalLookup[rd];
+	//angle : 12 bits with a precision of 1/16th of a degree
+	//a true binary fraction, NOT binary coded decimal
+	USALS = ( fabs(azimuth)*16 + 0.5 );
+	CMD1 |= (USALS & 0xf00) >> 8;
+	CMD2  = (USALS & 0x0ff);
 
 	rotorCommand( 12, CMD1, CMD2 );
 }