hardware question - propeller not spinning...

dougm

I built a sort of general IO board for the propeller but I can't seem to get it to connect. I based my design almost entirely on the datasheet for the propstick but using the LQFP.

At first the crystal wasn't spinning so I replaced it with an oscillator and have confirmed that 5Mhz is being driven into XI. Then I noticed that the ResN pin wasn't getting pulled up, so I added a 1K pull-up to that line. It does get pulled up when the software attempts to connect (using the DTR pin on the FT232).

The FT232 appears to be working fine now. I did have to fix it so that VCC was connected to +5 rather than +3.3, but it shows up as a COM port, and I see activity when the Propeller tool attempts to connect.

So I was wondering if you experts could take a quick look at my schematic and see if anything looks amiss?

Thank you,

DougM

LoopyByteloose

Schematic? Seems you omitted it.

Oh, now I see.. it is so faint. I will try to read it, but may not be successful.

AV_Tech

Hello DougM.

When I look at the datasheet of the PropStick i see an (big) difference at the circuit around the 32K Eprom in compare with your circuit. Why you have chance this good working section. When I build an circuit around de propeller chip by following the circuit (of the propstick) it's always working.

Greeting AV

LoopyByteloose

Well, it is very hard to read your schematic. But I can see that you have only one capacitor filtering all 4 power inputs. That is not appropriate. Each ground and 3.3 pin pair should have a capacitor as close as possible to the Propeller. Very risky to take a 40pin DIP package schematic and to adapt it to the other package as you miss this necessary detail.

I am not up to speed on USB interfaces. You could by pass that and use a RS232 to attempt to confirm that the Propeller is functional and the USB is your problem or vice versa.

There are other elements of the schematic that I don't understand. Maybe someone else will.

dougm

Thank you for your feedback, sorry for the hard-to-read schematic, I exported it out of Eagle at 400DPI but I'm not sure how well it translated when uploaded. I can email a cleaner copy if you PM me an email address.

A couple of notes:

The eeprom chip isn't populated yet so we can eliminate that.

5 and 3.3v are measuring correctly. I do have 100n decoupling caps on each active component (except the ADC oddly) as well as several caps on the 3.3V reg.

I will try talking to it directly using logic level serial by bypassing the FT232 and let you know if it works.

If you are willing to send me a schematic if your working design I would be happy to see if I can figure out where mine is different.

Thank you,

DougM

Rayman

That looks like Eagle...
Here's my PSM3 schematics for these items.
SSD1921_Test9c_FT232.pdf
SSD1921_Test9c_Prop.pdf

Couple differences:
I'm using the QFP44 instead of DIP, but that shouldn't matter.
I use FT232 with 5V I/O so that it can control 5V power, but that shouldn't matter too much either.

dougm

Thanks Ray,

I also found the schematic for RevB of the quickstart board and there are several differences in the USB circuitry, so I'll look at your schematic and that one and see if I can find the issue. I also note that as mentioned above each VDD line needs its own decoupling cap which unfortunately sounds like a new rev of the board.

Thanks for your help all, I've got some stuff to chew on.

DougM

LoopyByteloose

Personally, I don't have a working design.

But all of the Parallax Propeller production boards have schematics that are open-source for your reference. Designs vary quite a bit. I generally avoid USB entirely if I build something and use the 40pin DIP.

Rayman

Loopsy is very correct that every Vdd input should have it's own decoupling cap in ideal world.
For DIP version, I usually put a 10uF Ta cap on either side.

You should be able to solder those to your PCB pretty easily as is since VDD and GND pins are close together...

Tubular

The RESn thing is odd - if BOEn is connected to Vss (like normal), then I believe RESn should be weakly pulled up.

I'd be checking solder joints around the BOEn and RESn pins, the reset button itself, etc.

Peter Jakacki

dougm wrote: »

Thanks Ray,

I also found the schematic for RevB of the quickstart board and there are several differences in the USB circuitry, so I'll look at your schematic and that one and see if I can find the issue. I also note that as mentioned above each VDD line needs its own decoupling cap which unfortunately sounds like a new rev of the board.

Thanks for your help all, I've got some stuff to chew on.

DougM

I don't think that this decoupling cap on every pin should be sprouted like dogma, as it often is. The DIP version only has 2 pins for power so one on each pin only amounts to 2 caps yet it's the same chip. If you have good and tight power and ground planes or pours then you don't really need all these caps, sure, have them if you want, but you don't need them. If the layout is good then you have very low resistance and impedance between these pins and even with a single cap well placed it will perform better than many designs I see with wayward and badly placed decoupling caps snaking their way back to the pin.

BTW, I notice your EEPROM is set to write-protect, but I guess that is intentional then for when you do place it? The reset line should be reading high so make sure you have your 3.3V rail within range. 1K is an awfully low value to use for an optional reset pullup.

@Ray: 10uf tants make very lousy decouplers, one is normally included for "droop" and possible regulator stability issues, especially with LDOs but this type of cap is useless for any high frequency response.

EDIT: Here's a simple schematic that you can access from an online Prop itself:

Mark_T

The important point is each supply/ground pair needs a decoupling capacitor close to it via a
low-inductance path. This can be achieved with a single cap even if you have power and ground planes
under the chip and decoupling on the back. If you don't want to mount caps on the back of the board
then its hard to route low-inductance paths between pairs unless sets of I/O pins are unused.

dougm

Ok, now this is becoming a real mystery.

I removed the USB chip and connected directly to TX/RX (inverting tx with rx and vice versa along the way) with a standard TTL level serial converter. I see the TX LED blink when it tries to enumerate the propeller.
I re-checked the library to make sure the pinouts agree with the datasheet and the footprint.
I removed the transistor that was controlling the RESn line and checked to see that it floated it up (it does) and also manually pulled it up

So essentially the whole thing has been reduced to the chip, an oscillator and TX/RX lines and still it doesn't work.

Has anyone ever received a bad chip? I bought the un-marked #P8X32A-Q44-U

Is this chip super finicky about power conditioning?

Thanks,

DougM

Phil Pilgrim (PhiPi)

dougm wrote:

Is this chip super finicky about power conditioning?

Yes, that's what people have been trying to tell you. Without proper bypassing, Props have been known to have their PLLs fried. However, typically, it's still possible to load programs, since the loader does not use the PLL. In any event, though, proper power conditioning is an absolute must.

Has anyone ever received a bad chip?

I doubt it, since Parallax tests all of the chips before selling them.

-Phil

Peter Jakacki

Phil Pilgrim (PhiPi) wrote: »

Yes, that's what people have been trying to tell you. Without proper bypassing, Props have been known to have their PLLs fried. However, typically, it's still possible to load programs, since the loader does not use the PLL. In any event, though, proper power conditioning is an absolute must.


I doubt it, since Parallax tests all of the chips before selling them.

-Phil

But Phil, the fried PLL has been traced to bad grounding with heavy or inductive loads using the Prop's pins as not so happy go-betweens. The worst that not having a decoupler could do is make the Prop behave unreliably, but it can't fry it.

@dougm, I've never had a dud Prop, a dud board or dud assembly perhaps, but the Prop's are pretty hard to kill. On some boards I have an LED pulled-up on the SCL line, it's handy as a status indicator plus when the Prop starts up it attempts to boot from the EEPROM so you will see the LED flash for a moment. Try that but if it doesn't come up then check your supply volts, with a scope if possible to make sure the supply isn't drooping while the chip is trying to come out of reset. If you don't have a scope then put your meter on AC range which should read effectively zero, but if there is ripple it will show up. Other than that a photo or two will help.

Phil Pilgrim (PhiPi)

Peter Jakacki wrote:

But Phil, the fried PLL has been traced to bad grounding with heavy or inductive loads using the Prop's pins as not so happy go-betweens. The worst that not having a decoupler could do is make the Prop behave unreliably, but it can't fry it.

The general wisdom of the forum would indicate that it's somewhat easier than that. For example:

Mike Green (in another thread) wrote:

Not only do you have to wire Vss and Vdd to all of the points designated for them, but you need to have bypass capacitors (like 100nF) for Vdd/Vss on both sides of the DIP package close to the pins. The loading capacitors are internal to the chip, so you just have to have the crystal, also close to the chip with short leads.

The PLL just happens to be (by testing) the weakest link on the chip. It doesn't take much noise on the power supply or ground to go from 3.3V to over 4V where damage will occur. The additional Vdd/Vss pins are provided for a reason ... to distribute the supply (and ground) currents across the chip so the voltage drops across the chip can be minimized at high currents. If they're not supplied or not properly bypassed, induced noise may have a few more tenths of a volt to ride on and the margins are tight.

IOW, a noisy power supply and inadequate bypassing can be enough to cause problems. I'm sure I could find other references...

That said, though, I do not believe a bad PLL alone is the OP's problem.

-Phil

Peter Jakacki

Phil Pilgrim (PhiPi) wrote: »

The general wisdom of the forum would indicate that it's somewhat easier than that. For example:



IOW, a noisy power supply and inadequate bypassing can be enough to cause problems. I'm sure I could find other references...

That said, though, I do not believe a bad PLL alone is the OP's problem.

-Phil

The noise actually being a product of CMOS switching transients but the regulator is never fast enough to try and compensate (esp LDOs) so as to overshoot but that is why we have small ceramic decouplers there, they are fast enough. Supply spikes can only do damage when they have sufficient current which is not the case as that current has to come from somewhere and the clamping effect of a whole silicon substrate soon dampens any hopefuls. Typically the worst effects of inadequate decoupling is the operational reliability of the device especially at higher frequencies. Many times those "spikes" that are observed are improperly grounded scope probes etc, the test instrument is only as good as the operator.

But as you said too, I don't believe it's a bad PLL, because even when this is fried it is still possible to program and run the chip without a PLL. I don't know what the OP's problem is, I can't put my finger on it, but I'm sure he has

dougm

I'm persuaded that the issue is not power. I am now driving it with 4 C-cell batteries and a shunt reg, so ripple should be essentially non-existent and the AC measurement on the meter says 0.003v so that should be sufficiently clean, and that also eliminates the possibility of voltage droop at startup.

I have physically removed all components from the PCB except the chip, the 3.3v reg, the caps and the oscillator.

Peter's test of putting an LED on the SCL line indicates that the thing is working - the SCL line kicks maybe 50ms after I power on the chip.

I've triple-checked that the serial lines are going to the right pins.

Here is a piccie of the offending board. I can also post the Eagle .brd file if you'd like.

.

The only thing I haven't done is test with a different PC, but this PC connects to a PropStick without a problem.

Thank you for your help.

DougM

Ariba

dougm wrote: »

Ok, now this is becoming a real mystery.

I removed the USB chip and connected directly to TX/RX (inverting tx with rx and vice versa along the way) with a standard TTL level serial converter. I see the TX LED blink when it tries to enumerate the propeller.
I re-checked the library to make sure the pinouts agree with the datasheet and the footprint.
I removed the transistor that was controlling the RESn line and checked to see that it floated it up (it does) and also manually pulled it up

So essentially the whole thing has been reduced to the chip, an oscillator and TX/RX lines and still it doesn't work.

Has anyone ever received a bad chip? I bought the un-marked #P8X32A-Q44-U

Is this chip super finicky about power conditioning?

Thanks,

DougM

You also need to connect the DTR line, I hope your serial converter provides this signal. The TTL levels must be 3.3V (there are also converters with 5V levels which can damage the Propeller).
The DTR should be connected via the Transistor circuit. But for a first test I think it works also with a direct connection to RESn input or a simple capacitor (1..10nF) between DTR and RESn.

To connect to the Propeller no external oscillator is needed, the boot code uses the internal oscillator. It's also normal that you get not oscillation at the crystal pins until a code is downladed which enables the XTAL mode.

The Propeller is not picky about the supply voltage, good bypass caps are mainly important if you switch indcutive loads.

Andy

Genetix

Doug, are you by chance near any equipment that might be generating a lot of electrical noise?

Phil Pilgrim (PhiPi)

Try removing the oscillator. You don't need it for program loading or running programs with the internal RC oscillator.

-Phil

dougm

Andy, unfortunately my external USB --> TTL serial adapter is 5V logic and does not have the DTR line available, but I think I can fix both of those things. That said I may have been driving 5V into the RX line on the Propeller, though when I was using the onboard converter it would have been 3.3. So there's a chance that if it wasn't dead before it is now.

Genetix, I'm not near anything that generates significant electrical noise, and the PropStick works ok.

Thanks to the couple of people who noted that the oscillator does not run at first startup, that's one thing off my list. :-)

I'm going to have a beer and noodle on this for a while, will let you know what I get to when I get back to it.

Thank you!

DougM

Peter Jakacki

I have devices where I only have the rx/tx lines and no DTR available but I can still reprogram them just by releasing the reset as I tell the PC to load it. If you do an F10 to RAM then this should tell you whether it is a reset problem. The other thing is that I have also used a small cap (1 to 10nf) from the FT232 DTR directly to the Prop reset without the transistor, and that works every time.

BTW, the +5V directly into the Prop may not actually kill it as the current although higher than allowed is still low enough but I do know that this higher voltage will appear on the supply rail during reset when the Prop's current draw is low and I have observed that the Prop doesn't boot properly in these circumstances.

LoopyByteloose

Peter Jakacki wrote: »

I don't think that this decoupling cap on every pin should be sprouted like dogma, as it often is. The DIP version only has 2 pins for power so one on each pin only amounts to 2 caps yet it's the same chip. If you have good and tight power and ground planes or pours then you don't really need all these caps, sure, have them if you want, but you don't need them. If the layout is good then you have very low resistance and impedance between these pins and even with a single cap well placed it will perform better than many designs I see with wayward and badly placed decoupling caps snaking their way back to the pin.

@Ray: 10uf tants make very lousy decouplers, one is normally included for "droop" and possible regulator stability issues, especially with LDOs but this type of cap is useless for any high frequency response.

Well, there is a tendency to get a lot of different opinions from the Propeller Forum. I just prefer to go with what Parallax does in production as problems quickly get noticed, mentioned, and resolved.

Regarding the capacitors on each pair of power inputs... Dogma or not, at least provide SMD pads for all four pairs on the 44 pin board. You may get by with less, there was a thread years ago that explained the real value of following this well-accepted design.

Extra pads for bypass capacitors is often done in industrial production. They then populate the board minimally and observe performance. If it is good, they leave them empty. If there is a problem, they don't have to go to a second run of board productions. It is an excellent cost and time saving approach.


And 10uf by-pass caps are absurdly large for this situation. The concept is to trap transient spikes that are at a much higher pulse than a 10uf might resolve. 0.1uf or even 0.01uf are in the ball park... any value inbetween is okay as well as a starting point. Higher are lower values may need a scope to verify utility.

Some of us get into 'bigger is better' mode without actually understanding that capacitors need to fit the task that they are expected to do... and different tasks require a different approach to selection of size, voltage, and the quality of the capacitor.

Tantalum is not always the best solution, and certainly not always a necessary solution. The topic of which kind of capacitor to use is vast. If you want high quality .. often mylar or polypropylene are preferred as an alternative to tant. But by-pass capacitors tend to do nicely as ceramics.

Peter Jakacki

Loopy Byteloose wrote: »

Well, there is a tendency to get a lot of different opinions from the Propeller Forum. I just prefer to go with what Parallax does in production as problems quickly get noticed, mentioned, and resolved.

Regarding the capacitors on each pair of power inputs... Dogma or not, at least provide SMD pads for all four pairs on the 44 pin board. You may get by with less, there was a thread years ago that explained the real value of following this well-accepted design.

As much as I love Parallax I can't say that all of their designs could be held up as exemplars, I've seen 1,000uf caps "just to be sure" on the outputs of regulators and I've seen all four caps faithfully fitted but ineffective due to routing and power/ground planes. I would rather omit a cap and route tight to minimise the effective resistance and inductance of the path which is the whole point of decoupling due to shoot-through, core, and I/O current noise.

The other big thing that gets left on it's ownsome is the ridiculous HC49 gull-wing dipole antenna "SMD" crystal that ends up out away from the chip or has all those noisy I/O lines squeezing past it ready to inject their little spikes right into what is effectively a high-impedance amplifier that becomes the master clock for the whole shebang. Considering that there have been much better alternatives for many years I can't understand why anyone would stick with these parts. When I've used them in the past I'd shield them with ground planes and use guard tracks etc but they are just way too big otherwise. Now don't get me started on breadboard practices.... plug-in and otherwise....

LoopyByteloose

I guess it would be ideal to have exemplar designs published and done by peer review. But the truth is I have no idea of how to implement such a scheme.

I just tried to make the case for providing extra pads to eliminate the need to redesign and reorder a set of boards.

++++++++++
Sticking with old parts might just be a remaining inventory issue. I am only guessing, but it does get expensive to stay up-to-date with everything and buy smaller lots of components.

@Peter J
I do consider your opinions thoughtful and informative, but I tend to fall back on referring to the Parallax schematics and related Forum discussions.

Parallax has indeed gotten distracted at times and rushed to issue a board that was not fully sorted out. Maybe they should just have you as a paid consultant for pre-release review?

There are other vendors with boards; such as Gangster Gadget, Mikronauts, and so on. They might publish alternative schematic material of a superior design.