|
Artistic License claims to have "Invented" BAM. But nowhere have I found a claim that they have received or even applied for a patent.
Although I have found a dozen or so patents that would probably keep A.L. from getting one if they tried.
Not that it would matter very much, because this is not BAM.
BAM is flawed, and MIBAM represents a "Significant Improvement".
<br>
DT
|
BAM has been in public domain for over one year. That alone means it is not patentable.
|
....... BAM was published by AL *specifically* to place the idea & implementation in the public domain and keep companies from patenting it.Originally Posted by Darrel Taylor
Partly because there's been some ugliness in the professional lighting / theatre world by companies patenting (and then aggressively litigating) LED dimming and control techniques. These techniques have been obvious to those skilled in the trade since at least the 1970s, but the patents were granted nonetheless.
|
Excellent move by Artistic License.
Hopefully, by finding the Mirror Image improvement to BAM and releasing it to the public domain ...
RadikalQ3 and I have made it impossible for some squatter to patent a MIBAM equivalent too.
MIBAM - (Mirror Imaged Bit Angle Modulation)
http://www.picbasic.co.uk/forum/showthread.php?t=10564
<br>
DT
|
Hi Darrel,
May I pass along my late congrats' on some brilliant work and share a couple ideas?
Instead of using six instructions (6 cycles) per LED during each update interval you might consider updating one port at a time using an exclusive-or instruction (2 cycles) and data from an eight byte "toggle" array that's refreshed during that long 64T end-of-period interval. This reduces overhead from 288 cycles for 48 LEDs to just 12 cycles during each update interval. Using "toggle" data and an exclusive-or instruction lets us update MIBAM output bits without changing non-MIBAM bits on each port.
Since the three '1T' intervals seem to be a major bottleneck we might consider combining the center |2T|1T|1T|1T|2T| intervals into a single '7T' interval and use small in-line isochronous delays to effect precise timing between updates. This would practically eliminate "overhead" effects on the center three '1T' intervals.
Here's a pseudo C code driver example (sorry, I don't have PBP). I'm sure you realize that the "switch" function should be replaced with assembly language to reduce overhead and eliminate jitter and that the LEDs and "data benders" should probably be setup as macros as you did in your driver.
An assembly language version I wrote supports 48 LEDs at decent refresh rates at almost any clock (4-MHz, 177 Hz, 1T = 22-usecs).
Food for thought. Kind regards, Mike, K8LH
Code:unsigned char interval = 0; // ISR state machine var' unsigned char red = 44; unsigned char grn = 55; unsigned char blu = 66; volatile unsigned short ccpr1 @ 0xFBE; /* * * original intervals -> |64|32|16|8|4|2|1|1|1|2|4|8|16|32|64| * * modified intervals -> |64|32|16|8|4|----7----|4|8|16|32|64| * * */ void interrupt() // CCP1 "special event" interrupts { unsigned char Adat[8]; // porta "toggle" array unsigned char Bdat[8]; // portb "toggle" array pir1.CCP1IF = 0; // clear CCP1 interrupt flag switch(interval++) // { case 0: // 64T porta ^= Adat[7]; // portb ^= Bdat[7]; // ccpr1 = (64*tStep); // ccpr1 = half 2^7 (64T) break; // case 1: // 32T porta ^= Adat[6]; // portb ^= Bdat[6]; // ccpr1 >>= 1; // ccpr1 = half 2^6 (32T) break; // case 2: // 16T porta ^= Adat[5]; // portb ^= Bdat[5]; // ccpr1 >>= 1; // ccpr1 = half 2^5 (16T) break; // case 3: // 8T porta ^= Adat[4]; // portb ^= Bdat[4]; // ccpr1 >>= 1; // ccpr1 = half 2^4 (8T) break; // case 4: // 4T porta ^= Adat[3]; // portb ^= Bdat[3]; // ccpr1 >>= 1; // ccpr1 = half 2^3 (4T) break; // case 5: // 7T (2T+1T+1T+1T+2T) porta ^= Adat[2]; // portb ^= Bdat[2]; // ccpr1 = (7*tStep); // ccpr1 = 7*tStep (7T) delayCy(2*tStep-8); // 2T minus 8 cycles porta ^= Adat[1]; // portb ^= Bdat[1]; // delayCy(1*tStep-4); // 1T minus 4 cycles porta ^= Adat[0]; // portb ^= Bdat[0]; // delayCy(1*tStep-4); // 1T minus 4 cycles porta ^= Adat[0]; // portb ^= Bdat[0]; // delayCy(1*tStep-4); // 1T minus 4 cycles porta ^= Adat[1]; // portb ^= Bdat[1]; // break; // case 6: // 4T porta ^= Adat[2]; // portb ^= Bdat[2]; // ccpr1 = (4*tStep); // ccpr1 = half 2^3 (4T) break; // case 7: // 8T porta ^= Adat[3]; // portb ^= Bdat[3]; // ccpr1 <<= 1; // ccpr1 = half 2^4 (8T) break; // case 8: // 16T porta ^= Adat[4]; // portb ^= Bdat[4]; // ccpr1 <<= 1; // ccpr1 = half 2^5 (16T) break; // case 9: // 32T porta ^= Adat[5]; // portb ^= Bdat[5]; // ccpr1 <<= 1; // ccpr1 = half 2^6 (32T) break; // case 10: // 64T (end-of-period) porta ^= Adat[6]; // portb ^= Bdat[6]; // ccpr1 <<= 1; // ccpr1 = half 2^7 (64T) interval = 0; // prep for new period Adat[0] = 0; Adat[1] = 0; // clear Adat[] array Adat[2] = 0; Adat[3] = 0; // Adat[4] = 0; Adat[5] = 0; // Adat[6] = 0; Adat[7] = 0; // Bdat[0] = 0; Bdat[1] = 0; // clear Bdat[] array Bdat[2] = 0; Bdat[3] = 0; // Bdat[4] = 0; Bdat[5] = 0; // Bdat[6] = 0; Bdat[7] = 0; // /* * * use one "data bender" for each LED (16 cycles each) * * */ if(red.0) Bdat[0] |= 1; // 2^0 data RB0 pin if(red.1) Bdat[1] |= 1; // 2^1 data RB0 pin if(red.2) Bdat[2] |= 1; // 2^2 data RB0 pin if(red.3) Bdat[3] |= 1; // 2^3 data RB0 pin if(red.4) Bdat[4] |= 1; // 2^4 data RB0 pin if(red.5) Bdat[5] |= 1; // 2^5 data RB0 pin if(red.6) Bdat[6] |= 1; // 2^6 data RB0 pin if(red.7) Bdat[7] |= 1; // 2^7 data RB0 pin if(grn.0) Bdat[0] |= 2; // 2^0 data RB1 pin if(grn.1) Bdat[1] |= 2; // 2^1 data RB1 pin if(grn.2) Bdat[2] |= 2; // 2^2 data RB1 pin if(grn.3) Bdat[3] |= 2; // 2^3 data RB1 pin if(grn.4) Bdat[4] |= 2; // 2^4 data RB1 pin if(grn.5) Bdat[5] |= 2; // 2^5 data RB1 pin if(grn.6) Bdat[6] |= 2; // 2^6 data RB1 pin if(grn.7) Bdat[7] |= 2; // 2^7 data RB1 pin if(blu.0) Bdat[0] |= 4; // 2^0 data RB2 pin if(blu.1) Bdat[1] |= 4; // 2^1 data RB2 pin if(blu.2) Bdat[2] |= 4; // 2^2 data RB2 pin if(blu.3) Bdat[3] |= 4; // 2^3 data RB2 pin if(blu.4) Bdat[4] |= 4; // 2^4 data RB2 pin if(blu.5) Bdat[5] |= 4; // 2^5 data RB2 pin if(blu.6) Bdat[6] |= 4; // 2^6 data RB2 pin if(blu.7) Bdat[7] |= 4; // 2^7 data RB2 pin /* * * convert Bdat[] and Adat[] array "output" data to "toggle" data * * * * red 0x2C 00101100 (on RB0) * * grn 0x37 00110111 (on RB1) * * blu 0x42 01000010 (on RB2) * * * * Bdat 'output data' Bdat 'toggle data' Bdat 'toggle data' * * [0] 0x02 00000010 [0] 0x04 00000100 [0] 0x04 00000100 * * [1] 0x06 00000110 [1] 0x05 00000101 [1] 0x05 00000101 * * [2] 0x03 00000011 [2] 0x02 00000010 [2] 0x02 00000010 * * [3] 0x01 00000001 [3] 0x03 00000011 [3] 0x03 00000011 * * [4] 0x02 00000010 [4] 0x01 00000001 [4] 0x01 00000001 * * [5] 0x03 00000011 [5] 0x07 00000111 [5] 0x07 00000111 * * [6] 0x04 00000100 [6] 0x04 00000100 [6] 0x04 00000100 * * [7] 0x00 00000000 [7] 0x00 00000000 [7] 0x01 00000001 * * portb 0x00 00000000 portb 0x01 00000001 * * */ asm { movf _porta,W // get current PORTA bits andlw 0 // keep only MIBAM output bits xorwf _Adat+7,F // create 2^7 toggle bits xorwf _Adat+7,W // W simulates cumulative PORTA xorwf _Adat+6,F // create 2^6 toggle bits xorwf _Adat+6,W // xorwf _Adat+5,F // create 2^5 toggle bits xorwf _Adat+5,W // xorwf _Adat+4,F // create 2^4 toggle bits xorwf _Adat+4,W // xorwf _Adat+3,F // create 2^3 toggle bits xorwf _Adat+3,W // xorwf _Adat+2,F // create 2^2 toggle bits xorwf _Adat+2,W // xorwf _Adat+1,F // create 2^1 toggle bits xorwf _Adat+1,W // xorwf _Adat+0,F // create 2^0 toggle bits movf _portb,W // get current PORTB bits andlw 7 // keep only MIBAM output bits xorwf _Bdat+7,F // create 2^7 toggle bits xorwf _Bdat+7,W // W simulates cumulative PORTB xorwf _Bdat+6,F // create 2^6 toggle bits xorwf _Bdat+6,W // xorwf _Bdat+5,F // create 2^5 toggle bits xorwf _Bdat+5,W // xorwf _Bdat+4,F // create 2^4 toggle bits xorwf _Bdat+4,W // xorwf _Bdat+3,F // create 2^3 toggle bits xorwf _Bdat+3,W // xorwf _Bdat+2,F // create 2^2 toggle bits xorwf _Bdat+2,W // xorwf _Bdat+1,F // create 2^1 toggle bits xorwf _Bdat+1,W // xorwf _Bdat+0,F // create 2^0 toggle bits } break; // } }
Last edited by Mike, K8LH; - 26th July 2009 at 14:01.
|
WooHoo! Somebody's paying attention.
Hi Mike,
That's a VERY interesting idea!
Not sure at this point how I can let the user assign the pins at random, in any order, and any number of them, on any chip. But this sounds good enough to see if I can find a way.
For combining the smallest periods ... I already add the lsb from each side of the "mirror" into a single period to double the minimum interrupt time. Not sure where your 3rd one comes from. That would make the lsb worth 1.5.
I had also tried combining the 2 lsb's on both sides, but that reduced the resolution to 7-bit instead of 8, so the results weren't very good.
I'll see what I can do with the "Full PORT Press".
Thanks for the great idea.
Add: Hmmm, that would help minimze R-M-W issues too.
<br>
Last edited by Darrel Taylor; - 26th July 2009 at 16:56. Reason: RMW
DT
|
I'm not sure I follow you here. The duty cycle b0 bit output is still 1T long smack in the middle and the b1 bit still generates two 1T outputs on either side of the b0 1T output so we're not losing any resolution. I've just combined those |2T|1T|1T|1T|2T| outputs into a single 7T interval so that our minimum ISR interval is now 4T instead of 1T which gives us more headroom and much higher refresh intervals (even with the overhead of full ISR context save/restore)
I'm not exactly sure how to do it either and then we would still need to dynamically come up with Amask, Bmask, Cmask constants or variables for each port output-to-toggle routine but like you, I'm excited enough to try and find a way.Not sure at this point how I can let the user assign the pins at random, in any order, and any number of them, on any chip. But this sounds good enough to see if I can find a way.
With the new smaller 4T minimum ISR interval my assembly language test driver can achieve Kilohertz range refresh rates with some of the higher clock frequencies, though I'm not sure if that's really useful (grin).
Kind regards, Mike, K8LH
|
By combining those periods, you can't switch the LED for bit0, because it's combined in with bit1. That's what drops it down to 7-bit resolution. If it were a single output, you could combine them, but with multiple LED's, bit0 has to remain separate from bit1.
With MIBAM, it looks like this, with the red line being the "mirror".The 1T's on either side can be combined into a single period equal to 2T, but they can't be combined with the bit1 (2T's).Code:Bit- 1 0 0 1 |2T|1T|1T|2T|
I think I know how now, but welcome any more thoughts.I'm excited enough to try and find a way.
Just a matter of getting it done.
Thanks,
DT
Members who have read this thread : 1You do not have permission to view the list of names.
Posting Permissions
Bookmarks