Shop OBEX P1 Docs P2 Docs Learn Events
Ruminations while awaiting an FPGA image (was "Hello...... Anyone out there?") - Page 17 — Parallax Forums

Ruminations while awaiting an FPGA image (was "Hello...... Anyone out there?")

11213141517

Comments

  • SRLMSRLM Posts: 5,045
    edited 2014-08-07 14:22
    Delus wrote: »
    Just noticed the rad hard comment, if I get a chance I'll run the design through our synthesis tools in the IC process I work with (Unofficially rad-hard). I think I can safely post power estimates and area requirements if anyone is interested without violating our NDA (Though I should probably double check that one).

    How does this work? How does the chip architecture affect the radiation hardening?*

    * Ok, besides the error correcting logic. I'm thinking more along the lines of radiation hardening tries to avoid having any errors in the first place.
  • DelusDelus Posts: 79
    edited 2014-08-07 14:51
    Sorry if this ends up being a duplicate post (my last post didn't seem to work). Some processes are large enough and have thick enough layers to largely avoid radiation effects without adding additional error correction logic. The process I work with is one of those processes but is not officially branded as such. I mostly want to see how much area and power would be used/wasted on a uC design in the DALSA IC process. The rad-hard aspect just struck me as something other people might be interested in.
  • Bill HenningBill Henning Posts: 6,445
    edited 2014-08-07 15:05
    I'd be interested in power estimate and area requirements... all data is appreciated :)
    Delus wrote: »
    Just noticed the rad hard comment, if I get a chance I'll run the design through our synthesis tools in the IC process I work with (Unofficially rad-hard). I think I can safely post power estimates and area requirements if anyone is interested without violating our NDA (Though I should probably double check that one).
  • Heater.Heater. Posts: 21,230
    edited 2014-08-07 15:15
    Every schoolboy knows that propellers don't work in space.

    Perhaps Delus is about to discover that the Propeller MCU is actually a really good design choice for the harsh radiation environment of space!
  • ElectrodudeElectrodude Posts: 1,652
    edited 2014-08-07 20:45
    What does RealRandom.spin do on an FPGA prop 1? How not random is it?
  • cgraceycgracey Posts: 14,134
    edited 2014-08-07 22:23
    What does RealRandom.spin do on an FPGA prop 1? How not random is it?

    It would behave like a pseudo-random sequencer.
  • GenetixGenetix Posts: 1,754
    edited 2014-08-09 02:03
    Heater, a Propeller based pico-satellite.

    Chip, what changes have you made to the counters and video?
  • cgraceycgracey Posts: 14,134
    edited 2014-08-10 12:46
    Genetix wrote: »
    Heater, a Propeller based pico-satellite.

    Chip, what changes have you made to the counters and video?


    I've got it all unified under one peripheral which transfers pins to and from hub memory, transfers hub memory to DACs, and performs a DDS/Goertzel operation. It will be a week or so before I'll have an image to post.
  • mindrobotsmindrobots Posts: 6,506
    edited 2014-08-10 13:25
    cgracey wrote: »
    I've got it all unified under one peripheral which transfers pins to and from hub memory, transfers hub memory to DACs, and performs a DDS/Goertzel operation. It will be a week or so before I'll have an image to post.

    Haha!! Take your time....we're all busy with the P1 verilog now!! :lol:
  • Cluso99Cluso99 Posts: 18,069
    edited 2014-08-10 15:45
    I am surprised anyone is reading the P2 forum... too busy playing P1 FPGA ;)
  • David BetzDavid Betz Posts: 14,516
    edited 2014-08-10 16:39
    Cluso99 wrote: »
    I am surprised anyone is reading the P2 forum... too busy playing P1 FPGA ;)
    I wonder if the activity will shift back here once Chip releases a P2 FPGA image? I'll certainly start looking at what is required to update the PropGCC code generator for the new design.
  • mindrobotsmindrobots Posts: 6,506
    edited 2014-08-10 18:06
    David Betz wrote: »
    I wonder if the activity will shift back here once Chip releases a P2 FPGA image? I'll certainly start looking at what is required to update the PropGCC code generator for the new design.

    Sure it will because most of us have some form of attention deficit disorder! :smile:
  • TubularTubular Posts: 4,693
    edited 2014-08-10 18:11
    cgracey wrote: »
    I've got it all unified under one peripheral which transfers pins to and from hub memory, transfers hub memory to DACs, and performs a DDS/Goertzel operation. It will be a week or so before I'll have an image to post.

    That all sounds very exciting
  • Peter JakackiPeter Jakacki Posts: 10,193
    edited 2014-08-10 18:43
    David Betz wrote: »
    I wonder if the activity will shift back here once Chip releases a P2 FPGA image? I'll certainly start looking at what is required to update the PropGCC code generator for the new design.

    Yes, of course it most certainly will but we will only have the P2 image whereas we have P1 design files that we may play with to see if something else might be possible that could be incorporated into P2 perhaps? But I imagine that most of the work will be on testing P2 and developing our tools and software. I will look at running my own macro-assembler at least which will enable me to simplify my VM source and lay it out much better than the Spin tool can. All this is so exciting!!!
  • David BetzDavid Betz Posts: 14,516
    edited 2014-08-10 18:54
    Yes, of course it most certainly will but we will only have the P2 image whereas we have P1 design files that we may play with to see if something else might be possible that could be incorporated into P2 perhaps? But I imagine that most of the work will be on testing P2 and developing our tools and software. I will look at running my own macro-assembler at least which will enable me to simplify my VM source and lay it out much better than the Spin tool can. All this is so exciting!!!
    I've been waiting for you to announce that you plan to implement the Tachyon VM in hardware on the P1. :-)
  • Peter JakackiPeter Jakacki Posts: 10,193
    edited 2014-08-10 21:02
    David Betz wrote: »
    I've been waiting for you to announce that you plan to implement the Tachyon VM in hardware on the P1. :-)

    In that case I'd have to start from scratch I guess :) and it wouldn't be much use unless it was in production form! No, I'd be happy with the P2 and I'm sure that all of us together will be able to really utilize the P2 in a much shorter time than the P1 took, for a whole lot of reasons. So this would mean that we would need to have a P2 OBEX setup too. With the various languages and environments and tools this should be pretty awesome in a short space of time, perhaps even before the P2 is available in silicon. What a boost.
  • KeithEKeithE Posts: 957
    edited 2014-08-11 19:27
    In that case I'd have to start from scratch I guess :) and it wouldn't be much use unless it was in production form! No, I'd be happy with the P2 and I'm sure that all of us together will be able to really utilize the P2 in a much shorter time than the P1 took, for a whole lot of reasons. So this would mean that we would need to have a P2 OBEX setup too. With the various languages and environments and tools this should be pretty awesome in a short space of time, perhaps even before the P2 is available in silicon. What a boost.

    I thought it might be interesting to have a minimal just good enough prop 1 with your forth and a master APB interface for FPGA development. It would probably be relatively simple to do - if only I had more time… What would be the minimal cog count - just 1? I'm assuming it would just have forth in RAM at reset time so no need for the Parallax loader etcetera, but that would just be a question of how the "ROM" is initialized.
  • Bill HenningBill Henning Posts: 6,445
    edited 2014-08-11 19:36
    I got that covered..

    - DE2-115 + 2x DE0-Nano's are for P2 development
    - 2x BEMICRO CV's + 1x DE0-Nano will be for P1V development
    David Betz wrote: »
    I wonder if the activity will shift back here once Chip releases a P2 FPGA image? I'll certainly start looking at what is required to update the PropGCC code generator for the new design.
  • ozpropdevozpropdev Posts: 2,792
    edited 2014-08-11 20:37
    Now that we don't have multi-tasking on P2 I have to do it myself! :lol:
    (a) DE2 for P2 development
    (b) DE0 for P1V development.
    Which one will I play with today? I Can't decide so it looks like both! :lol::)
    Fun...fun...fun!
  • GenetixGenetix Posts: 1,754
    edited 2014-08-12 02:47
    Do any know what new counter or video features there are?
  • RaymanRayman Posts: 14,566
    edited 2014-08-12 11:32
    I've lost track of what's going on with counters and video...

    I think that maybe the old counters are gone and some functions moved to the new "smart pins".

    Analog video will now be supported by direct transfer from HUB memory to groups of 4 pins for RGB (VGA) video at HD resolution.
    I think Chip mentioned support for things like 24-bit digital video output too.
    But, I could be wrong...
  • jmgjmg Posts: 15,171
    edited 2014-08-12 17:46
    Rayman wrote: »
    I've lost track of what's going on with counters and video...

    I think that maybe the old counters are gone and some functions moved to the new "smart pins".

    Analog video will now be supported by direct transfer from HUB memory to groups of 4 pins for RGB (VGA) video at HD resolution.
    I think Chip mentioned support for things like 24-bit digital video output too.
    But, I could be wrong...

    I think that's right.
    The last plan I heard was to move the old counters into the Smart Pins, which expand with things like True PWM, Quadrature, and capture. (and serial ? hopefully QuadSPI ?)
    Undefined yet is if those smart-cells are per pin, or per pairs of pins.
    The new HUB access will include an in line CLUT at fSys/N ( iirc Chip has NCO (option?) here ).

    Once all that can stream wide video to DACs at fSys, allowing LCD style digital out is a logical extension.

    That may not be any-pin mapped ?
  • Dave HeinDave Hein Posts: 6,347
    edited 2014-08-13 08:39
    Does anyone recall when Chip said the P2 FPGA image would be available? I recall that he posted a week or two ago that it would available soon. I tried to find his latest post about the P2, but all his recent posts have been about the P1 FPGA.
  • mindrobotsmindrobots Posts: 6,506
    edited 2014-08-13 08:49
    You mean here, where on 8/10/14 he says,

    I've got it all unified under one peripheral which transfers pins to and from hub memory, transfers hub memory to DACs, and performs a DDS/Goertzel operation. It will be a week or so before I'll have an image to post.

    Silly me told him to take his time since we were all busy with the P1 verilog!
  • potatoheadpotatohead Posts: 10,261
    edited 2014-08-13 09:16
    Week after next. He was close to a week away, when Ken announced he would be answering questions about P1 this week. Since next week, he's supposed to get back on P2, maybe the week after is a reasonable, early expectation.
  • Dave HeinDave Hein Posts: 6,347
    edited 2014-08-13 10:09
    mindrobots wrote: »
    You mean here, where on 8/10/14 he says,

    Silly me told him to take his time since we were all busy with the P1 verilog!
    Yes, that's probably the one I'm thinking about. The P1 verilog is nice, but I would much rather have seen the P2 image by Monday. So now we have to wait 2 more weeks for the P2? These delays are becoming almost unbearable.
  • cgraceycgracey Posts: 14,134
    edited 2014-08-13 10:16
    Things like PWM and ADC accumulation are moving into the smart pins. Inside each cog there is an NCO-driven shifter/DDS circuit that is good for transferring pins to and from the hub quickly, moving data through DACs (video), and performing DDS/Goertzel operations.

    The DDS/Goertzel circuit outputs a sine (and cosine, if you want) and then correlates the sine and cosine with an incoming ADC bitstream to ultimately perform I-Q demodulation. This is good for making all kinds of measurements involving sine waves traveling through the real world. Sine waves are to nature what square waves are to logic. There's no need to do anything with the I-Q data. You could just use the DDS circuit to output a custom waveform of any frequency, up to Fclk/2, or 100MHz in the case of a 200MHz clock.

    Here are the notes I've been making for the implementation:
    XFRI	D/#,S/#		'init, reset phase, delay ~4 clocks, start
    XFRZ	D/#,S/#		'continue, reset phase (normalizes jitter on repetitive scan lines)
    XFR	D/#,S/#		'continue
    
    D/#:
    
    mode	dacs	pins	base	cnt						dac/pin output
    ----	----	----	----	---						--------------
    
    0000	dddd	ep--	----	xxxxxxxxxxxxxxxx  *	32-bit immediate	$xxxxxxxx
    0001	dddd	eppp	----	xxxxxxxxxxxxxxxx	 8-bit RFBYTE		$000000xx
    0010	dddd	epp-	----	xxxxxxxxxxxxxxxx	16-bit RFWORD		$0000xxxx
    0011	dddd	ep--	----	xxxxxxxxxxxxxxxx	32-bit RFLONG		$xxxxxxxx
    0100	dddd	ep--	bbbb	xxxxxxxxxxxxxxxx	 1-bit RFLONG LUT	$xxxxxxxx
    0101	dddd	ep--	bbbb	xxxxxxxxxxxxxxxx	 2-bit RFLONG LUT	$xxxxxxxx
    0110	dddd	ep--	bbbb	xxxxxxxxxxxxxxxx	 4-bit RFLONG LUT	$xxxxxxxx
    0111	dddd	ep--	bbbb	xxxxxxxxxxxxxxxx	 8-bit RFLONG LUT	$xxxxxxxx
    1000	dddd	ep--	bbbb	xxxxxxxxxxxxxxxx  *	 1-bit immediate LUT	$xxxxxxxx
    1001	dddd	ep--	bbbb	xxxxxxxxxxxxxxxx  *	 2-bit immediate LUT	$xxxxxxxx
    1010	dddd	ep--	bbbb	xxxxxxxxxxxxxxxx  *	 4-bit immediate LUT	$xxxxxxxx
    1011	dddd	ep--	bbbb	xxxxxxxxxxxxxxxx  *	 8-bit immediate LUT	$xxxxxxxx
    1100	dddd	--nn	nnnn	xxxxxxxxxxxxxxxx	DDS/Goertzel LUT	$xxxxxxxx
    1101	----	-ppp	----	xxxxxxxxxxxxxxxx	 8-bit WFBYTE		-
    1110	----	-pp-	----	xxxxxxxxxxxxxxxx	16-bit WFWORD		-
    1111	----	-p--	----	xxxxxxxxxxxxxxxx	32-bit WFLONG		-
    
    * uses S/# for immediate
    
    
    dddd
    ----
    0000	 -  -  -  -
    0001	 0  0  0  0
    0010	 -  -  0  0
    0011	 0  0  -  -
    0100	 -  -  -  0
    0101	 -  -  0  -
    0110	 -  0  -  -
    0111	 0  -  -  -
    1000	!0  0 !0  0
    1001	 -  - !0  0
    1010	!0  0  -  -
    1011	 1  0  1  0
    1100	 -  -  1  0
    1101	 1  0  -  -
    1110	!1  1 !0  0
    1111	 3  2  1  0
    
    ep-- = e:enable output, p:32-bit port
    epp- = e:enable output, pp:16-bit port
    eppp = e:enable output, ppp:8-bit port
    
    bbbb = %bbbb0000 base in LUT
    
    nnnnnn = ADC input pin
    
  • Bob Lawrence (VE1RLL)Bob Lawrence (VE1RLL) Posts: 1,720
    edited 2014-08-13 13:12
    cgracey wrote: »
    Things like PWM and ADC accumulation are moving into the smart pins. Inside each cog there is an NCO-driven shifter/DDS circuit that is good for transferring pins to and from the hub quickly, moving data through DACs (video), and performing DDS/Goertzel operations.

    The DDS/Goertzel circuit outputs a sine (and cosine, if you want) and then correlates the sine and cosine with an incoming ADC bitstream to ultimately perform I-Q demodulation. This is good for making all kinds of measurements involving sine waves traveling through the real world. Sine waves are to nature what square waves are to logic. There's no need to do anything with the I-Q data. You could just use the DDS circuit to output a custom waveform of any frequency, up to Fclk/2, or 100MHz in the case of a 200MHz clock.

    Here are the notes I've been making for the implementation:
    XFRI    D/#,S/#        'init, reset phase, delay ~4 clocks, start
    XFRZ    D/#,S/#        'continue, reset phase (normalizes jitter on repetitive scan lines)
    XFR    D/#,S/#        'continue
    
    D/#:
    
    mode    dacs    pins    base    cnt                        dac/pin output
    ----    ----    ----    ----    ---                        --------------
    
    0000    dddd    ep--    ----    xxxxxxxxxxxxxxxx  *    32-bit immediate    $xxxxxxxx
    0001    dddd    eppp    ----    xxxxxxxxxxxxxxxx     8-bit RFBYTE        $000000xx
    0010    dddd    epp-    ----    xxxxxxxxxxxxxxxx    16-bit RFWORD        $0000xxxx
    0011    dddd    ep--    ----    xxxxxxxxxxxxxxxx    32-bit RFLONG        $xxxxxxxx
    0100    dddd    ep--    bbbb    xxxxxxxxxxxxxxxx     1-bit RFLONG LUT    $xxxxxxxx
    0101    dddd    ep--    bbbb    xxxxxxxxxxxxxxxx     2-bit RFLONG LUT    $xxxxxxxx
    0110    dddd    ep--    bbbb    xxxxxxxxxxxxxxxx     4-bit RFLONG LUT    $xxxxxxxx
    0111    dddd    ep--    bbbb    xxxxxxxxxxxxxxxx     8-bit RFLONG LUT    $xxxxxxxx
    1000    dddd    ep--    bbbb    xxxxxxxxxxxxxxxx  *     1-bit immediate LUT    $xxxxxxxx
    1001    dddd    ep--    bbbb    xxxxxxxxxxxxxxxx  *     2-bit immediate LUT    $xxxxxxxx
    1010    dddd    ep--    bbbb    xxxxxxxxxxxxxxxx  *     4-bit immediate LUT    $xxxxxxxx
    1011    dddd    ep--    bbbb    xxxxxxxxxxxxxxxx  *     8-bit immediate LUT    $xxxxxxxx
    1100    dddd    --nn    nnnn    xxxxxxxxxxxxxxxx    DDS/Goertzel LUT    $xxxxxxxx
    1101    ----    -ppp    ----    xxxxxxxxxxxxxxxx     8-bit WFBYTE        -
    1110    ----    -pp-    ----    xxxxxxxxxxxxxxxx    16-bit WFWORD        -
    1111    ----    -p--    ----    xxxxxxxxxxxxxxxx    32-bit WFLONG        -
    
    * uses S/# for immediate
    
    
    dddd
    ----
    0000     -  -  -  -
    0001     0  0  0  0
    0010     -  -  0  0
    0011     0  0  -  -
    0100     -  -  -  0
    0101     -  -  0  -
    0110     -  0  -  -
    0111     0  -  -  -
    1000    !0  0 !0  0
    1001     -  - !0  0
    1010    !0  0  -  -
    1011     1  0  1  0
    1100     -  -  1  0
    1101     1  0  -  -
    1110    !1  1 !0  0
    1111     3  2  1  0
    
    ep-- = e:enable output, p:32-bit port
    epp- = e:enable output, pp:16-bit port
    eppp = e:enable output, ppp:8-bit port
    
    bbbb = %bbbb0000 base in LUT
    
    nnnnnn = ADC input pin
    


    With the DDS/Goertzel circuit I could probably make a audio pitch detector by out putting the desired signal on one cog and inputting the test signal on another cog and comparing the two. :cool:
  • ozpropdevozpropdev Posts: 2,792
    edited 2014-08-13 19:10
    Looks good Chip! :)
  • User NameUser Name Posts: 1,451
    edited 2014-08-15 13:15
    cgracey wrote: »
    The DDS/Goertzel circuit outputs a sine (and cosine, if you want) and then correlates the sine and cosine with an incoming ADC bitstream to ultimately perform I-Q demodulation.

    This strikes very close to home... Sounds absolutely fantastic!
Sign In or Register to comment.