Another one of those decoupling capacitor questions.
AndrejaKo
Posts: 24
I recently ordered some DIP40 Propellers and plan to make a circuit on a perfboard similar to one pictured here: http://en.wikipedia.org/wiki/File:CopperCladPerfboard_1.png. I tried manufacturing my own PCBs several times, but I still don't have enough experience to make a complicated PCB myself, so it's either the perfboard or breadboard and from what I've heard, perfboards have lower negative influence on the circuit.
I don't have much experience with microcontrollers, but I did do some work with AVR ATmega 162 in the past.
After reading many threads discussing decoupling capacitors, they usually mention that near each pair of power pins 10 nF and 100 nF ceramic capacitors should be placed and that 10 μF tantalum capacitors should be used on power supply for each propeller. Few other threads mentioned 100 pF capacitors and 1 μF capacitors too.
Also the crystal pins are very close to the power pins on the DIP40 version.
This raised some concerns.
As far as I know, the Propeller doesn't require external capacitors near the crystal and the crystal should be directly connected to the XI and XO pins. I also heard from various sources that it's generally good idea to have crystal as close as possible to the crystal pins in order to get best signal.
Another problem is the thickness of capacitors, since I'm planning to use the through-hole components.
If I place all capacitors which I've seen recommanded, from 100 pF to 1 μF, their thickness would increase the distance between the capacitors and the chip itself. I think that I may be able to place first and maybe even second row of capacitors under the socket in order to shorten the distance between the capacitors and the pins. Would that be a good idea or not? Should I in this case skip some capacitor sizes? I think that I may be able to stack them all so that the distance between each row is the minimal 2.54 mm, but even that would make the capacitors of the last row (1 μF) very far away (5.08 mm if I place two capacitors under the socket or entire 10.16 mm if I don't) from the chip itself.
Any advice on that would be helpful.
Another concern is the crystal. It is in such position that it would interfere with capacitor placement. So how far from the chip it should be? I guess that I could skip a row of capacitors for the crystal itself, but that would place even greater distance between the capacitors and the chip.
Yet another point are component leads. The capacitors I have available have 5.08 mm between the leads, but the distance between the power pins on the propeller is (if my calculations are correct) 7.62 mm. So I'd have to bend the leads a bit in order to make a good connection. How big problem would that be?
And finally, a question which isn't about capacitors themselves. In the AVR Hardware Design Considerations, a inductor is placed between the capacitors close to the power source and the capacitors close to the chip itself. The idea is to prevent the traces form behaving as antennas by resisting the current change. I plan to have one 220 μF tantalum capacitor near the power connector for the board and have wires independently connect the power pins of the chips on the boards to that capacitor. It would seem logical to place the inductor after the capacitors and before those long wires. On the other hand I haven't seen any mention of inductors used in such way in Propeller documentation and here on the forum, so would its use be a good idea? If yes, what inductivity should I place there? On the ATmega, I used 47 nH, but I got the value from the design considerations document.
I don't have much experience with microcontrollers, but I did do some work with AVR ATmega 162 in the past.
After reading many threads discussing decoupling capacitors, they usually mention that near each pair of power pins 10 nF and 100 nF ceramic capacitors should be placed and that 10 μF tantalum capacitors should be used on power supply for each propeller. Few other threads mentioned 100 pF capacitors and 1 μF capacitors too.
Also the crystal pins are very close to the power pins on the DIP40 version.
This raised some concerns.
As far as I know, the Propeller doesn't require external capacitors near the crystal and the crystal should be directly connected to the XI and XO pins. I also heard from various sources that it's generally good idea to have crystal as close as possible to the crystal pins in order to get best signal.
Another problem is the thickness of capacitors, since I'm planning to use the through-hole components.
If I place all capacitors which I've seen recommanded, from 100 pF to 1 μF, their thickness would increase the distance between the capacitors and the chip itself. I think that I may be able to place first and maybe even second row of capacitors under the socket in order to shorten the distance between the capacitors and the pins. Would that be a good idea or not? Should I in this case skip some capacitor sizes? I think that I may be able to stack them all so that the distance between each row is the minimal 2.54 mm, but even that would make the capacitors of the last row (1 μF) very far away (5.08 mm if I place two capacitors under the socket or entire 10.16 mm if I don't) from the chip itself.
Any advice on that would be helpful.
Another concern is the crystal. It is in such position that it would interfere with capacitor placement. So how far from the chip it should be? I guess that I could skip a row of capacitors for the crystal itself, but that would place even greater distance between the capacitors and the chip.
Yet another point are component leads. The capacitors I have available have 5.08 mm between the leads, but the distance between the power pins on the propeller is (if my calculations are correct) 7.62 mm. So I'd have to bend the leads a bit in order to make a good connection. How big problem would that be?
And finally, a question which isn't about capacitors themselves. In the AVR Hardware Design Considerations, a inductor is placed between the capacitors close to the power source and the capacitors close to the chip itself. The idea is to prevent the traces form behaving as antennas by resisting the current change. I plan to have one 220 μF tantalum capacitor near the power connector for the board and have wires independently connect the power pins of the chips on the boards to that capacitor. It would seem logical to place the inductor after the capacitors and before those long wires. On the other hand I haven't seen any mention of inductors used in such way in Propeller documentation and here on the forum, so would its use be a good idea? If yes, what inductivity should I place there? On the ATmega, I used 47 nH, but I got the value from the design considerations document.
Comments
The idea is that the 100nF capacitors provide near-instantaneous power reserves at the chip level with little very inductance between the capacitor and the chip itself. The larger tantalum capacitor provides a slower power reserve because of longer leads and higher inductance between the board and the power supply mounted off board, possibly 30 to 100cm further away.
Mount the crystal within 2-3cm of the DIP package and keep the leads direct and reasonably short. A couple of mm either way won't make much of a difference. What gets people excited is when you leave 2-4cm long leads on the crystal or capacitors as well as another couple of cm of breadboard or perf. board between where the leads connect and the DIP package itself. That's a problem.
Also should I go for ceramic disk capacitors or multilayer capacitors? I know that I should go for the ones with lowest ESR, but I don't have ESR specifications for them.
Thanks for the update on the post. That solved the problem.
As decoupling capacitors use ceramic ones. Them need have fast response time to transients on Power lines to IC's/Propeller.
As Bulk one any Electrolytic/Tantalum - preferred Tantalum - Them are only to reserve some Power for Propeller in time it change Power consumption with start of NEW COG else restart working after waitpxx instructions.
As Mike said AS near as REASONABLE possible -
ONE thing to consider -- USE for Power lines bigger dimension of wire that normally uses for signal lines - That spare You much problems in Powering Propeller/Other IC's
I meant ceramic disk or ceramic multilayer capacitors, sorry for not being clear on that.
Also what type of wire should be thick enough? I normally use solid wire from Cat 5E Ethernet cables, so that should be 24 AWG. Should I go for thicker than that? The other option would be a 2x 0.35mm² cable.
@Mike Green
Then I'll use multilayer ceramic capacitors.
About the cables: I'm currently using 24 AWG cables from Cat 5E Ethernet cable. The other easily obtainable type of cable which would probably fit the board would be 2x0.35mm². Would the 24 AWG be enough or should I switch to the 0.35mm²?
As Mike said, a 100nF ceramic cap (use X7R not Z5U) close to both sets of prop power pins should be used. A 10uF tantalm closeby helps, as does a 10nF. Make sure you wire BOTH sets of power pins on the DIP40 (and 4 sets on the QFP). Wire the two sets of power pins together using as hort a connection as possible. I place my caps on the underside of the pcb under the prop. If you use a socket, putting them under the socket is good idea.
As you probably know, the power demands of CMOS chips is a chain of very high, but short, pulses of current (plus any output loadings). Your meter just shows the average of these.
I put 2 x 100nF SM caps under the prop, and one 10 - 47uF very near by (or in the DIP40 socket voids) and one 100nF for every additional chip. If there are not on board regs with their own caps then a 100+++uF goes on the board too.
Sorry as I must say that.
You have totally understood usage of Decoupling capacitors in range of 10nF to 100nF.
Them are not for Power Demands -- THEM are for shorting parasitic transients Frequencies that IC's produce ELSE can be disturbed OF on POWER Traces.
Ps. Different Capacitances are for different range of Frequencies
My past includes very high power radio, TV transmitters and microwave links, there they did use the whole group of caps etc (even if they had big bolts on each end).
Fortunately modern electronics seems to be not being pushed to the "cutting edge" now, and so calming things down is so much easier, and smaller
As for high performance FPGAs .. I couldn't afford one :-(
I got a nice answer explaining how to decouple those 20/40 chips here: http://electronics.stackexchange.com/questions/9888/how-to-connect-decoupling-capacitor-when-vcc-gnd-pins-arent-close/9892#9892
The choise of good quality components is important too, I have seen lots of those silver bodied SM caps (the ones that have about one third of the top blacked to indicate the negative connection) that are useless from new. On an ESR meter they read several Ohms. I often get rid of them and re-form the leads of a radial through-hole sort, tants are good, Sapieha always uses them and gets away with overclocking (in the extreem).
To that You need at.
As many Values IC needs - Depends on frequencies it shall work with.
That said, I'm a traditionalist (0.1uF MLCC near power/ground, extensive ground planes, segregated analog and digital grounds, star grounding arrangements, an occasional 10uF aluminum electrolytic cap located strategically, and a pi-section input filter with BIIIIG aluminum caps) and can state I haven't had a EMC issue. I tend to choose caps with a voltage rating at least 50% above Vcc or any "conceivable" transient.
I'm not a tantalum fan - had those dry out on me too many times. The only ones I'll use now are the Kemets with the internal fuses. These are nice too because the extra ESR from the fuse helps with the overall circuit damping.
QUICK HINT: IF you use a large value ceramic cap on the input of your circuit, beware! Let's suppose you power your circuit off of a wall-wart. You plug in your wall-wart and your long, inductive transmission line from the wart to your circuit is hot. When you plug it into your board, the low ESR of the ceramic cap will ring with that line inductance and give you an underdamped ring, the maximum voltage stress of which is double your wall-wart output voltage. If you have no TVS protection, or you've gone skimpy on your board's power supply design, life will become more difficult for you. An el-cheapo, high ESR aluminum electrolytic at this point can save you.
V/R
Mike