I'm back, just continue doing my multipropeller board, at this time, I've one master with twelve slaves connected working fine, and now I'm going for the last four slaves.
I'll be posting code, and more pictures, when I get all work perfect, hope could be usefull for somebody.
The boards are home-assembled by hand.
One pict shows many flat wires connected to test the slave data recieved with anothers boards not shown here.
Thanks so much Timothy.
They were assembled at home by hand, yes by me,.... but the PCB are from a company.
Talking about the weldings, I found that the poor problem, is to place the component in the correct position, then to weld them I use a gel flux (that i found here in my country) and it's so easy !!! like some video in another thread shows it, neither I've not used a special welder tool.
The hard work, was to place the 0603 capacitors.....[noparse]:([/noparse]
Yes the pocessing power was too much, like there are a lot of chips per slave, and I can't test it before to design the PCB's, and they will be running signals at high frequency, I prefeer to use one LM2937 each two boards, but really is too much for this. I hope that the supply current would be greater.
I hope to be usefull for somebody when I finished this, willl be one propeller master controlling sixteen slaves in a parallel bus, with three control lines.
I repeat, when I get it finished, I will post some code, at this momment I've twelve slaves running well, I suppose that next four will be not problem for me.
Due I can't get any oscillator chip to control all the clocks, I use one oscillator chip every four slaves... and one xtal with the master (because I can't get work TV driver with a "xinput" anf freq of 80_000_000), that's too bad too [noparse]:([/noparse] ... but it is working at the needed speed. Sure with a oscillator pack for all (like Chip told to me) could be get a higher speed of processing.
......think that the master propeller job is:
1- Doing the master job.
2- Reading a PC keyboard.
3- Using I2C to save some data at the EEPROM (same EEPROM that the propeller uses for program code).
4- Using RCTime to get a value from a pin.
5- Showing video at a TV.
6- Reading the data to send to the slaves from a SD card, at more or less 400Kbytes/sec (Rokicki's code).
Thanks so much !! Quattro..
Sorry... not clear for me what you mean..[noparse]:([/noparse]
But if understand well....not yet..I've still four slaves more to get work..... to get that [noparse]:)[/noparse] ....but I'm closer now.
What is this for? It looks cool and I think that a lot could be done with so many cogs communicating with each other.
Like the finish on the PCB's to.
Thank you
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Check out my robot using the propeller robot.tmcp.com.au
If you offer cheap PCB fabrication, perl programming or any other helpful services please email me at. anthonybmyatt@yahoo.com.au
Hi computer guy.
It's for a big video LED curtain.
The slaves has 6 cogs working each, (main, slave, R, G, B, Clocks).
I've three threads arround this project.
would love to see some pics of the finished display.
How large will it be?
Thank you
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Check out my robot using the propeller robot.tmcp.com.au
If you offer cheap PCB fabrication, perl programming or any other helpful services please email me at. anthonybmyatt@yahoo.com.au
computer guy.
It will be about 33ft by 21 ft (No exactly relation 4:3, neither 16:9)
Be patient, I'll post some code, and of course many picts of the job finished, I can't be.....
At this time we've done only 1/32 of the total....
We still have to buy some IC's, nobody has those in USA, we only found them in China, we bought last week 300 pcs to some guy, they sent to us 67 pcs fried... If somebody knows about one Chinesse honorable, please !! let me know !! It will be apreciated.
Seems like when somebody ask there for IC's, all answer saying... "They are new and unused parts"....we can't believe that !!! ..don't must it be obviously ?? or I'm living in another planet ?
What IC is it just out of interest 300 pcs is a lot.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Check out my robot using the propeller robot.tmcp.com.au
If you offer cheap PCB fabrication, perl programming or any other helpful services please email me at. anthonybmyatt@yahoo.com.au
PS: I want to leave it clear now....Sorry... I don't want to offend Chinesse people...I'm sure I wasn't not lucky with that man, I must not to generalize with all Chinesse people. Only that sellers seems to be suspects.
I joined the forum only some weeks ago and got aware of this thread just by Albero's recent request.
Sure, not every one of you is an EE, but even the educated hoppyist need not take Chip's verdict for granted
But why is he right?
(1) Whenever you want to transmit signals, you have some limt to their upper frequency f:
1/f = 2*R*C (for minor distortion you should even use a factor of 2.pi!)
But what is C and what is R?
(2) A CMOS input generally has a capacitance of around 10 pF (it can be less, it can be more)
When we connect N inputs the capacitance to be driven will be N*10pF
(3) We output a voltage swing of around 3 V, and we - but not safely! - can do it at 30 mA, so easily compute 100 Ohm.
(4) Putting it all together:
1/f = 2*N*10pF*100Ohm = 2 ns
i.e. we can use 500 MHz with ONE connected Pin, and 25 MHz with 20
(5) However... If we transmit at this frequency continously we shall need 240 mA (for all 8 lines) which is not rercommendable for the Propeller...
So the basic signal clock should be FAR below 25MHz when 20 other chips are connected..
---
And always keep in mind : 80% of electronics is nothing but the clever application of
U = RI
P = UI
Q = CU
I = dQ/dt
Edit: Well, I missed all the postings in this thread since April 1st somehow.... So this sounds a little bit out of place now, as Alberte has made phantastic advances. But I should like to leave it here as it might help someone else in the future....
Not trying to be anail, and please correct me if i'm wrong,
You mentioned you'd need around 240 mA· for 8 driven pins right? I think your calculations assume a full phase/continuous draw and not a half wave - which is what digital signal is - which·should be around half that calculated if you consider duty cycle that is.
And i thought the formula for freq was fc = 1/ (2 * pi (R C)) as foundhere..·on page 3 of 33
N maybe ima bit ol school here,
······· /\
···· /· E· \·
· /
\
/· I··· |·· R·· \
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔ E3= Thought
http://folding.stanford.edu/·- Donating some CPU/GPU downtime just might lead to a cure for cancer! The average PC while browsing the internet typically uses less than 30% of it's potential, why not donate a portion of the rest for cancer resaerch?
Nearly correct It is always tricky to mix the correct "sinusodial" approach of Fourier Analysis with our digital world of constant Highs and Lows and a FAST transition in between.
A "fast transisition" sent through a low pass filter will be "distorted", the main "digital" effect beeing that its "slope" will decrease, thus rectangled signals become triangled You loose when the signal has no longer enough time to reach its high levels at all.
It depends of the usage (especially on the signal conditioning in the receivers, e.g. Schmitt-Triggers) what you are willing to accept. The factor 2*pi is generally used for analogue applications, meaning a drop down of 3 dB for the corresponding sine wave. I stated this in my original posting, I think....
I think however "2" will be fine for the really upper limit of digital signals, though that could be "marginal design" - but you have a chance
Current flows during loading and unloading of the capacitors which truely is an exponential effect and has to be averaged (but never forget the short time max current!); the matter is not whether your signal is "High", but how long you stay "high" or "low" in relation to the rise and fall of the signal. I have neither looked it up nor measured directly.
I think the value I gave FOR THE MARGINAL FREQUENCY (!) is correct. The factor of 2 you assumed is correct for half the frequency, which is the safe case anyway.
As an afterthought: Anyone can easily and safely experiment, as this is not a matter of "high" frequencies but of the relationship between C and f, so a very cheap oscilloscop will do:
Connect one of the Prop pins with a 100 Ohm resistor and a "low inductance" 100 nF capacitor (e.g. a WIMA MKS2 or FKS2) to ground and generate f = 10 kHz, 50 kHz, 100 kHz, 200 kHz, and 1 MHz and watch the signals; current can be deduced from the voltage at the 100 Ohms resistor. You probably have to "integrate" yourself, as the value your general purpose multi-meter shows will be questionable for AC...
( A good value would be given by an old fashioned coil-based instrument - BUT DON'T INTRODUCE ANY INDUCTANCY)
The external 100 Ohms resistor adds to the (presumable) also 100 Ohm internal impedance, so the max current will always be limited to save 15 to at most 30 mA.
Braves at heart can add a second experiment using a 10 Ohms resistor afterwards, but don't go up to 1 MHz in that case...
----
And maybe I should add a disclaimer
deSilva said...
As an afterthought: Anyone can easily and safely experiment, as this is not a matter of "high" frequencies but of the relationship between C and f, so a very cheap oscilloscop will do:
Connect one of the Prop pins with a 100 Ohm resistor and a "low inductance" 100 nF capacitor (e.g. a WIMA MKS2 or FKS2) to ground and generate f = 10 kHz, 50 kHz, 100 kHz, 200 kHz, and 1 MHz and watch the signals; current can be deduced from the voltage at the 100 Ohms resistor. You probably have to "integrate" yourself, as the value your general purpose multi-meter shows will be questionable for AC...
( A good value would be given by an old fashioned coil-based instrument - BUT DON'T INTRODUCE ANY INDUCTANCY)
As far as adding a resistor from a pin to a junction and a capacitor (of any size) from that junction to ground or to VCC/VDD esentially creates an integrator and will generate a sawtooth pattern at a given freq determined by
I think:
f = 1/ (2 * pi * R * C)
I include the parallax serach linkfor your use in serching through the forums for answers. Don't use the forums search button, it's dotnetBB and it's junk! I've seen the boys (primarily Beau) do some amazing things with the prop like A/D conversion with a feedback loop in the nS range with two SMD resistors solidered right onto two pins - i looked for 20min but couldn't find it. Another thread i read there was a post where (noname) took a prop and overclocked it to 160MHz!
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔ E3= Thought
http://folding.stanford.edu/·- Donating some CPU/GPU downtime just might lead to a cure for cancer! The average PC while browsing the internet typically uses less than 30% of it's potential, why not donate a portion of the rest for cancer resaerch?
I'm agree with you in some of your calculus.. but...
What this mean ?
deSilva said...
(5) However... If we transmit at this frequency continously we shall need 240 mA (for all 8 lines) which is not rercommendable for the Propeller...
So the basic signal clock should be FAR below 25MHz when 20 other chips are connected..
When we will need 240mA ? I'm transmiting from a master (out) to a high impedance input (in)..at one time each slave ...the rest of the slaves will be deactivated..
Also in my case, I assume that I will have 6pf each propeller, ·by 17/bus... so about 102 pF maximunn capacitance in the line. (each pin). And using your calculus, it's give me a top freq of about 3Mhz in this bus.
Well Alberto, this is fine. Early postings talked of 20 Props, and 10 pF will be on the safe side
I wanted to (again) stress the fact, that all the current you need has to do - and only! - with the capacities you are loading and unloading and this is directly proportional to the top frequency you can use. As the Prop CAN deliver 30 mA I wanted to show up the limits for that!
As you are working in the sub-Mhz area there will be no problems! Except maybe for the clock input (5 MHz?) , but as I understand your concept, you connect only 4 Props to one oscillator.
So this is all really fine.
My intention was just to give some background, WHY it is fine
Thanks so much deSilva !
Yes, at the end I used only 16 slaves....my first question about 20 props, was to "generalyze" the question.
I really apreciate that calculus ...it could be usefull for me, in next times too.
And correct, I used an 80Mhz oscillator each four slaves.....like all of them (I can found), support only a maximun of 30pF load, maybe we could consider another oscillator chip, to get all props with the same clock phase... "Quattro" ( had suggested some for me one...but I did can't get it. )
I think, could be usefull too.... to talk about "The PCB lines large"..... I did it very very short, as short I can. but my question is: How to impacts that parameter in the final speed (capacitance) ??
Could be really interesting .....because I must deal with some board assembly problems, due the "size" of the boards. They are so close one to another, and that give me many problems with the wired, from them to the controller "Output" connectors.
Maybe, when I finished this design..I will try to do some similar but, "digitizing" real time video signal (instead getting data from SD card), and coudd be great, to do the future slave boards, not so close one to another, and with the output connectors included in the slave PCB's.
Obviously, I think to use A/D converters for the "digitize" job, and I only need to take "96x64" RGB samples per frame at more or less 20fps.
Very interesting project and nice construction.· Just food for thought.
It is possible to stack 20 Propeller Protoboards as another prototype.·
Since each of the I/O pins have two holes on the board, One row of Pins·0 to Pin 7 could provide an 8pin·socket [noparse][[/noparse]pointing up] and the other row could have a row of the same I/O pointing down.
I suspect that you won't get much more compact by building 20 Propellers on one surface.· Vertical integration is likely a better fit.
And since each has it's own power regulation and programing port, the Prop Clip can easily move around for programming.· If anything fails, merely replace the board with another.
Of course, the sync of the oscillators remains in question.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
"Everything in the world is purchased by labour; and our passions are the only causes of labor." -- David·Hume (1711-76)········
···················· Tropically,····· G. Herzog [noparse][[/noparse]·黃鶴 ]·in Taiwan
I wonder if you could use a 74***4050Hex buffer to distribute the clock to all the chips evenlly?
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔ E3= Thought
http://folding.stanford.edu/·- Donating some CPU/GPU downtime just might lead to a cure for cancer! The average PC while browsing the internet typically uses less than 30% of it's potential, why not donate a portion of the rest for cancer resaerch?
RinksCustoms said...
I wonder if you could use a 74***4050Hex buffer to distribute the clock to all the chips evenlly?
Why do you "wonder" and not check it? You do not need 20 Props but just a tiny 200 pF capacitor
And why do you think 50 mA is so much better than 30 mA?
Can you use them in parallel (180 mA ? 0,6 W?)? The spec says 300 to 500 mW are enough ..
I assume Alberto is only using 5 MHz rather than 80 MHz as he said...
Hi all.
Kramer:
I think perhaps 20 props could be working fine...(but with a lowest comunication speed )
I never think to use protoboards, due I need to have near the prop chip, 9 LVDS chips more...
But the connector idea is very interesting for my next project !!! I will try it ....thanks !!
I thought about all props in the same board, but I find that they would be more closely using my vertical integration, and easy to place the LVDS chips too in the PCB design.
Bob:
I use 16 props for drive each one a LED screen...and then, all of them (16), makes a big LED video screen
Each slave uses 5 COGS each, one for Red channel, one for Green channel, one for Blue channel, one for clocks, and one for slave comunication.
RinksCustoms.
I think it is the same as use one oscillator chip for all propellers ?? I must put in parallel all "4050" buffer inputs, so I will have a charge in the oscillator about a lot of pF..... The problem to drive all clock inputs with one chip is not the "drive capacity" but the capacitance charging the oscillator chip,·what I have,·lets me only·30 pF max.
Then If I would connect the out of the oscillator to a only one buffer, and then, drive with it all the XINs....I will never have a 80Mhz signals on all clocks, due the capacitance of the 16 props charging the buffer with 96 pF. (I think that)
deSilva:
I dont understand very well what you mean...sorry
But, I'm using one 80Mhz oscillator pack, each 4 propellers, going the "clk signal", into the "four" XIN propellers pins in parallel....
usig this:
_clkmode = xinput
_xinfreq = 80_000_000
For all:
In this thread, "Qauattro" suggest me a special oscillator chip for these types of applications....maybe it could do the job easily. He told me about the CY27EE16 Take a look at this.
BTX,
Thanks, I was wondering if I was just sounding foolish.
I don't think I will ever stack 20 Proto Boards, but maybe 4 or 5 is in my future.·
I certainly am one of those that is having trouble filling all 8 cogs.· But one never knows.
At least there is an easy option to start experimenting [noparse][[/noparse]I already bought a 5 pak of the Proto boards]
I suspect that coordinating timing in software would work, but be a bit slower.· Other I/O pins could signal bus status to smooth out communications.
I am very interested to see what your· LED display will look like and how complex you expect it to be.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
"Everything in the world is purchased by labour; and our passions are the only causes of labor." -- David·Hume (1711-76)········
···················· Tropically,····· G. Herzog [noparse][[/noparse]·黃鶴 ]·in Taiwan
BTX said...
deSilva:
I dont understand very well what you mean...sorry
Never mind, Alberto, I am used to that
(1) I see absolutely no reason why you don't use the PLLs, so you can reduce the external clock to comfortable 5 or 10 MHz. You will have A LOT of EMS pollution with your approach.
It should not be a matter of clock phase locking!
(2) If - and I think it is not! - phase locking is no problem, there is no need for a "broadcast approach" , you can easily "chain" the clock from place to place. I want to say:
Rather than this:
-> #1
OSC -> BUFFER -> #2
-> #16
Do this:
OSC -> #1 -> BUFFER -> #2 -> BUFFER -> .... #16
There are intermediate solutions of course.
I think one should avoid under all circimstances to output 80 MHz with a 500 mA buffer.. This will make a great broadband VHF-transmitter
Yes Kramer.
I'm so anxious too to see finished...but still a lot lot of work to do [noparse]:([/noparse]
deSilva.
Hahahah...OK...but sometimes I feel really bad, when somebody talks about something, and I answer another.....please sorry
Yes I really don't know ...why I did it, maybe because Chip said it to me... and I didn't think anymore in that.
I really think I would prefeer t ouse a 5Mhz clock instead 80Mhz [noparse]:)[/noparse], but better could be that Chip answer this.... why he told about use 80Mhz clock knowing the bus characteristics ?.
About you suggest of: OSC -> #1 -> BUFFER -> #2 -> BUFFER -> .... #16
What will happaned with the "delay" introduced by so many buffers .... I thought in that...but I thought too, that differences between first slave board and the last one, would be a problem.
Using "analogue lingo" you would say the "phase" is shifting - it seems that no one is interested in shift-locked clocking, so forget it...
In "digital lingo" you would call it "delay", a typical delay is 100 ns for a buffer at low voltage; this improves much when going up to 5 V!
This indeed is a lot compared to the instruction execution time of 50 ns. But again, I don't see a reason why this should lead to problems....
When doing software synchronization by sending messages through your bus system you are off by many instructions.... Synchronization for better than - say - 1 mys will need a specific timing signal sent to ALL Props phase un-shifted!
This can be accomplished by a pulse each - say! - milli second, where a COG inside every Prop can wait for. When detecting this pulse the cog can compute the slippage of this Prop by it's internal knowlege of the local time phase of the main loop at this moment. It then calibrates an internal delay the main loop must run through regularly, so that the difference will decrease. This is a "soft" methode to avoid breaks. It is not "setting", but "regulating" the clock.
You will need (and most likely have already) something of this kind, when your Props are not excatly "instruction locked".I presume even the time from reset to start of first user instruction might vary...
Comments
I'll read the datasheet carefully, but it seems that is designed for this purpose...
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Regards.
Alberto.
Quattro
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
'Necessity is the mother of invention'
I'm back, just continue doing my multipropeller board, at this time, I've one master with twelve slaves connected working fine, and now I'm going for the last four slaves.
I'll be posting code, and more pictures, when I get all work perfect, hope could be usefull for somebody.
The boards are home-assembled by hand.
One pict shows many flat wires connected to test the slave data recieved with anothers boards not shown here.
These are some picts.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Regards.
Alberto.
Also, why so much processing power?
This is in an interesting thread.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Timothy D. Swieter
tdswieter.com
One little spark is all it takes for an idea to explode
They were assembled at home by hand, yes by me,.... but the PCB are from a company.
Talking about the weldings, I found that the poor problem, is to place the component in the correct position, then to weld them I use a gel flux (that i found here in my country) and it's so easy !!! like some video in another thread shows it, neither I've not used a special welder tool.
The hard work, was to place the 0603 capacitors.....[noparse]:([/noparse]
Yes the pocessing power was too much, like there are a lot of chips per slave, and I can't test it before to design the PCB's, and they will be running signals at high frequency, I prefeer to use one LM2937 each two boards, but really is too much for this. I hope that the supply current would be greater.
I hope to be usefull for somebody when I finished this, willl be one propeller master controlling sixteen slaves in a parallel bus, with three control lines.
I repeat, when I get it finished, I will post some code, at this momment I've twelve slaves running well, I suppose that next four will be not problem for me.
Due I can't get any oscillator chip to control all the clocks, I use one oscillator chip every four slaves... and one xtal with the master (because I can't get work TV driver with a "xinput" anf freq of 80_000_000), that's too bad too [noparse]:([/noparse] ... but it is working at the needed speed. Sure with a oscillator pack for all (like Chip told to me) could be get a higher speed of processing.
......think that the master propeller job is:
1- Doing the master job.
2- Reading a PC keyboard.
3- Using I2C to save some data at the EEPROM (same EEPROM that the propeller uses for program code).
4- Using RCTime to get a value from a pin.
5- Showing video at a TV.
6- Reading the data to send to the slaves from a SD card, at more or less 400Kbytes/sec (Rokicki's code).
I'm wondering with the propeller !!!!!!!!!!!!
Here's another view.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Regards.
Alberto.
Quattro..
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
'Necessity is the mother of invention'
Sorry... not clear for me what you mean..[noparse]:([/noparse]
But if understand well....not yet..I've still four slaves more to get work..... to get that [noparse]:)[/noparse] ....but I'm closer now.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Regards.
Alberto.
What is this for? It looks cool and I think that a lot could be done with so many cogs communicating with each other.
Like the finish on the PCB's to.
Thank you
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Check out my robot using the propeller robot.tmcp.com.au
If you offer cheap PCB fabrication, perl programming or any other helpful services please email me at.
anthonybmyatt@yahoo.com.au
It's for a big video LED curtain.
The slaves has 6 cogs working each, (main, slave, R, G, B, Clocks).
I've three threads arround this project.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Regards.
Alberto.
would love to see some pics of the finished display.
How large will it be?
Thank you
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Check out my robot using the propeller robot.tmcp.com.au
If you offer cheap PCB fabrication, perl programming or any other helpful services please email me at.
anthonybmyatt@yahoo.com.au
It will be about 33ft by 21 ft (No exactly relation 4:3, neither 16:9)
Be patient, I'll post some code, and of course many picts of the job finished, I can't be.....
At this time we've done only 1/32 of the total....
We still have to buy some IC's, nobody has those in USA, we only found them in China, we bought last week 300 pcs to some guy, they sent to us 67 pcs fried..
Seems like when somebody ask there for IC's, all answer saying... "They are new and unused parts"....we can't believe that !!! ..don't must it be obviously ?? or I'm living in another planet ?
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Regards.
Alberto.
What IC is it just out of interest 300 pcs is a lot.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Check out my robot using the propeller robot.tmcp.com.au
If you offer cheap PCB fabrication, perl programming or any other helpful services please email me at.
anthonybmyatt@yahoo.com.au
TLC5940NT, I need 1000 pcs more.
PS: I want to leave it clear now....Sorry... I don't want to offend Chinesse people...I'm sure I wasn't not lucky with that man, I must not to generalize with all Chinesse people. Only that sellers seems to be suspects.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Regards.
Alberto.
Sure, not every one of you is an EE, but even the educated hoppyist need not take Chip's verdict for granted
But why is he right?
(1) Whenever you want to transmit signals, you have some limt to their upper frequency f:
1/f = 2*R*C (for minor distortion you should even use a factor of 2.pi!)
But what is C and what is R?
(2) A CMOS input generally has a capacitance of around 10 pF (it can be less, it can be more)
When we connect N inputs the capacitance to be driven will be N*10pF
(3) We output a voltage swing of around 3 V, and we - but not safely! - can do it at 30 mA, so easily compute 100 Ohm.
(4) Putting it all together:
1/f = 2*N*10pF*100Ohm = 2 ns
i.e. we can use 500 MHz with ONE connected Pin, and 25 MHz with 20
(5) However... If we transmit at this frequency continously we shall need 240 mA (for all 8 lines) which is not rercommendable for the Propeller...
So the basic signal clock should be FAR below 25MHz when 20 other chips are connected..
---
And always keep in mind : 80% of electronics is nothing but the clever application of
U = RI
P = UI
Q = CU
I = dQ/dt
Edit: Well, I missed all the postings in this thread since April 1st somehow.... So this sounds a little bit out of place now, as Alberte has made phantastic advances. But I should like to leave it here as it might help someone else in the future....
Post Edited (deSilva) : 7/8/2007 12:22:32 AM GMT
Not trying to be anail, and please correct me if i'm wrong,
You mentioned you'd need around 240 mA· for 8 driven pins right? I think your calculations assume a full phase/continuous draw and not a half wave - which is what digital signal is - which·should be around half that calculated if you consider duty cycle that is.
And i thought the formula for freq was fc = 1/ (2 * pi (R C)) as found here..·on page 3 of 33
N maybe ima bit ol school here,
······· /\
···· /· E· \·
· /
\
/· I··· |·· R·· \
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
E3 = Thought
http://folding.stanford.edu/·- Donating some CPU/GPU downtime just might lead to a cure for cancer! The average PC while browsing the internet typically uses less than 30% of it's potential, why not donate a portion of the rest for cancer resaerch?
A "fast transisition" sent through a low pass filter will be "distorted", the main "digital" effect beeing that its "slope" will decrease, thus rectangled signals become triangled
It depends of the usage (especially on the signal conditioning in the receivers, e.g. Schmitt-Triggers) what you are willing to accept. The factor 2*pi is generally used for analogue applications, meaning a drop down of 3 dB for the corresponding sine wave. I stated this in my original posting, I think....
I think however "2" will be fine for the really upper limit of digital signals, though that could be "marginal design" - but you have a chance
Current flows during loading and unloading of the capacitors which truely is an exponential effect and has to be averaged (but never forget the short time max current!); the matter is not whether your signal is "High", but how long you stay "high" or "low" in relation to the rise and fall of the signal. I have neither looked it up nor measured directly.
I think the value I gave FOR THE MARGINAL FREQUENCY (!) is correct. The factor of 2 you assumed is correct for half the frequency, which is the safe case anyway.
Post Edited (deSilva) : 7/9/2007 9:44:13 AM GMT
Connect one of the Prop pins with a 100 Ohm resistor and a "low inductance" 100 nF capacitor (e.g. a WIMA MKS2 or FKS2) to ground and generate f = 10 kHz, 50 kHz, 100 kHz, 200 kHz, and 1 MHz and watch the signals; current can be deduced from the voltage at the 100 Ohms resistor. You probably have to "integrate" yourself, as the value your general purpose multi-meter shows will be questionable for AC...
( A good value would be given by an old fashioned coil-based instrument - BUT DON'T INTRODUCE ANY INDUCTANCY)
The external 100 Ohms resistor adds to the (presumable) also 100 Ohm internal impedance, so the max current will always be limited to save 15 to at most 30 mA.
Braves at heart can add a second experiment using a 10 Ohms resistor afterwards, but don't go up to 1 MHz in that case...
----
And maybe I should add a disclaimer
Post Edited (deSilva) : 7/9/2007 9:36:30 AM GMT
I think:
f = 1/ (2 * pi * R * C)
I include the parallax serach link for your use in serching through the forums for answers. Don't use the forums search button, it's dotnetBB and it's junk! I've seen the boys (primarily Beau) do some amazing things with the prop like A/D conversion with a feedback loop in the nS range with two SMD resistors solidered right onto two pins - i looked for 20min but couldn't find it. Another thread i read there was a post where (noname) took a prop and overclocked it to 160MHz!
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
E3 = Thought
http://folding.stanford.edu/·- Donating some CPU/GPU downtime just might lead to a cure for cancer! The average PC while browsing the internet typically uses less than 30% of it's potential, why not donate a portion of the rest for cancer resaerch?
I'm agree with you in some of your calculus.. but...
What this mean ?
When we will need 240mA ? I'm transmiting from a master (out) to a high impedance input (in)..at one time each slave ...the rest of the slaves will be deactivated..
Also in my case, I assume that I will have 6pf each propeller, ·by 17/bus... so about 102 pF maximunn capacitance in the line. (each pin). And using your calculus, it's give me a top freq of about 3Mhz in this bus.
Which is enough for my application.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Regards.
Alberto.
I wanted to (again) stress the fact, that all the current you need has to do - and only! - with the capacities you are loading and unloading and this is directly proportional to the top frequency you can use. As the Prop CAN deliver 30 mA I wanted to show up the limits for that!
As you are working in the sub-Mhz area there will be no problems! Except maybe for the clock input (5 MHz?) , but as I understand your concept, you connect only 4 Props to one oscillator.
So this is all really fine.
My intention was just to give some background, WHY it is fine
Yes, at the end I used only 16 slaves....my first question about 20 props, was to "generalyze" the question.
I really apreciate that calculus ...it could be usefull for me, in next times too.
And correct, I used an 80Mhz oscillator each four slaves.....like all of them (I can found), support only a maximun of 30pF load, maybe we could consider another oscillator chip, to get all props with the same clock phase... "Quattro" ( had suggested some for me one...but I did can't get it. )
I think, could be usefull too.... to talk about "The PCB lines large"..... I did it very very short, as short I can. but my question is: How to impacts that parameter in the final speed (capacitance) ??
Could be really interesting .....because I must deal with some board assembly problems, due the "size" of the boards. They are so close one to another, and that give me many problems with the wired, from them to the controller "Output" connectors.
Maybe, when I finished this design..I will try to do some similar but, "digitizing" real time video signal (instead getting data from SD card), and coudd be great, to do the future slave boards, not so close one to another, and with the output connectors included in the slave PCB's.
Obviously, I think to use A/D converters for the "digitize" job, and I only need to take "96x64" RGB samples per frame at more or less 20fps.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Regards.
Alberto.
It is possible to stack 20 Propeller Protoboards as another prototype.·
Since each of the I/O pins have two holes on the board, One row of Pins·0 to Pin 7 could provide an 8pin·socket [noparse][[/noparse]pointing up] and the other row could have a row of the same I/O pointing down.
I suspect that you won't get much more compact by building 20 Propellers on one surface.· Vertical integration is likely a better fit.
And since each has it's own power regulation and programing port, the Prop Clip can easily move around for programming.· If anything fails, merely replace the board with another.
Of course, the sync of the oscillators remains in question.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
"Everything in the world is purchased by labour; and our passions are the only causes of labor." -- David·Hume (1711-76)········
Post Edited (Kramer) : 7/10/2007 2:10:20 PM GMT
P.S are you going to create a higher resolution video driver or higher color video driver? or just settle with 400 or 256?
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Realize that I am really a mad scientist··· and
Don't forget it!
http://raydillon.com/Images/Illustration/GameArt/WildIsle/WildIsle-Ink-ScientistClose.jpg
·
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
E3 = Thought
http://folding.stanford.edu/·- Donating some CPU/GPU downtime just might lead to a cure for cancer! The average PC while browsing the internet typically uses less than 30% of it's potential, why not donate a portion of the rest for cancer resaerch?
And why do you think 50 mA is so much better than 30 mA?
Can you use them in parallel (180 mA ? 0,6 W?)? The spec says 300 to 500 mW are enough ..
I assume Alberto is only using 5 MHz rather than 80 MHz as he said...
Post Edited (deSilva) : 7/10/2007 11:35:12 PM GMT
Kramer:
I think perhaps 20 props could be working fine...(but with a lowest comunication speed )
I never think to use protoboards, due I need to have near the prop chip, 9 LVDS chips more...
But the connector idea is very interesting for my next project !!! I will try it ....thanks !!
I thought about all props in the same board, but I find that they would be more closely using my vertical integration, and easy to place the LVDS chips too in the PCB design.
Bob:
I use 16 props for drive each one a LED screen...and then, all of them (16), makes a big LED video screen
Each slave uses 5 COGS each, one for Red channel, one for Green channel, one for Blue channel, one for clocks, and one for slave comunication.
RinksCustoms.
I think it is the same as use one oscillator chip for all propellers ?? I must put in parallel all "4050" buffer inputs, so I will have a charge in the oscillator about a lot of pF..... The problem to drive all clock inputs with one chip is not the "drive capacity" but the capacitance charging the oscillator chip,·what I have,·lets me only·30 pF max.
Then If I would connect the out of the oscillator to a only one buffer, and then, drive with it all the XINs....I will never have a 80Mhz signals on all clocks, due the capacitance of the 16 props charging the buffer with 96 pF. (I think that)
deSilva:
I dont understand very well what you mean...sorry
But, I'm using one 80Mhz oscillator pack, each 4 propellers, going the "clk signal", into the "four" XIN propellers pins in parallel....
usig this:
_clkmode = xinput _xinfreq = 80_000_000For all:
In this thread, "Qauattro" suggest me a special oscillator chip for these types of applications....maybe it could do the job easily. He told me about the CY27EE16 Take a look at this.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Regards.
Alberto.
Thanks, I was wondering if I was just sounding foolish.
I don't think I will ever stack 20 Proto Boards, but maybe 4 or 5 is in my future.·
I certainly am one of those that is having trouble filling all 8 cogs.· But one never knows.
At least there is an easy option to start experimenting [noparse][[/noparse]I already bought a 5 pak of the Proto boards]
I suspect that coordinating timing in software would work, but be a bit slower.· Other I/O pins could signal bus status to smooth out communications.
I am very interested to see what your· LED display will look like and how complex you expect it to be.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
"Everything in the world is purchased by labour; and our passions are the only causes of labor." -- David·Hume (1711-76)········
Never mind, Alberto, I am used to that
(1) I see absolutely no reason why you don't use the PLLs, so you can reduce the external clock to comfortable 5 or 10 MHz. You will have A LOT of EMS pollution with your approach.
It should not be a matter of clock phase locking!
(2) If - and I think it is not! - phase locking is no problem, there is no need for a "broadcast approach" , you can easily "chain" the clock from place to place. I want to say:
Rather than this: -> #1 OSC -> BUFFER -> #2 -> #16 Do this: OSC -> #1 -> BUFFER -> #2 -> BUFFER -> .... #16There are intermediate solutions of course.
I think one should avoid under all circimstances to output 80 MHz with a 500 mA buffer.. This will make a great broadband VHF-transmitter
Post Edited (deSilva) : 7/11/2007 11:27:28 AM GMT
I'm so anxious too to see finished...but still a lot lot of work to do [noparse]:([/noparse]
deSilva.
Hahahah...OK...but sometimes I feel really bad, when somebody talks about something, and I answer another.....please sorry
Yes I really don't know ...why I did it, maybe because Chip said it to me... and I didn't think anymore in that.
I really think I would prefeer t ouse a 5Mhz clock instead 80Mhz [noparse]:)[/noparse], but better could be that Chip answer this.... why he told about use 80Mhz clock knowing the bus characteristics ?.
About you suggest of: OSC -> #1 -> BUFFER -> #2 -> BUFFER -> .... #16
What will happaned with the "delay" introduced by so many buffers .... I thought in that...but I thought too, that differences between first slave board and the last one, would be a problem.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Regards.
Alberto.
In "digital lingo" you would call it "delay", a typical delay is 100 ns for a buffer at low voltage; this improves much when going up to 5 V!
This indeed is a lot compared to the instruction execution time of 50 ns. But again, I don't see a reason why this should lead to problems....
When doing software synchronization by sending messages through your bus system you are off by many instructions.... Synchronization for better than - say - 1 mys will need a specific timing signal sent to ALL Props phase un-shifted!
This can be accomplished by a pulse each - say! - milli second, where a COG inside every Prop can wait for. When detecting this pulse the cog can compute the slippage of this Prop by it's internal knowlege of the local time phase of the main loop at this moment. It then calibrates an internal delay the main loop must run through regularly, so that the difference will decrease. This is a "soft" methode to avoid breaks. It is not "setting", but "regulating" the clock.
You will need (and most likely have already) something of this kind, when your Props are not excatly "instruction locked".I presume even the time from reset to start of first user instruction might vary...
Post Edited (deSilva) : 7/11/2007 1:00:09 PM GMT