Decoupling Capacitors

Humanoido

I'm designing a circuit that uses both the surface mount and dip propeller chips. The dip will receive the usual two decoupling capacitors, one on each side, as the chip has two voltage and ground pins. However, on the surface mount chip, it appears to have four pin pairs of Vss and Vdd. Will the surface mount chip require double the number of decoupling capacitors. What will be the likely result with only one decoupler? (with a standard 80MHz clock)

Leon

Each pair must have a decoupling capacitor mounted close to the chip, with wide tracks to them! Insufficient decoupling will result in unreliable operation and can even damage the PLL.

Humanoido

It's interesting the surface mount prop has four Vss/Vdd pairs and the DIP has only two. It also requires soldering four connectors instead of one on the Proto Board and adding four capacitors instead of two. A connector has 20 solder points (double row) so four per board is 80 solder points times 20 boards and so one thousand six hundred solder connections should keep me a little too busy.

LoopyByteloose

It is even more interesting that the Prop Manual shows only one pair of the DIP version connected - not a good idea for durable design. Nonetheless, the reason for less pins is likely larger wires feeding the power to the right locations on the chip.

Roy Eltham

Looking at the Propeller Proto USB board, it looks like they only have one decoupling capacitor. I've never had any issues with the Proto boards, so there must be something more to this. The way the Proto boards have the power and ground setup, that one decoupling capacitor isn't far from each VDD/VSS pair (it's right next to one set, and just across the chip from the others), that must be enough.

The Propeller Demo Board appears to only have 1 decoupler, also.
The Propeller backpack board has 2 decouplers on the Propeller power pins
The MSR1 Robot Controller board has 3 decouplers on the Propeller power pins.
The Propeller Servo Controller USB has 3 decouplers on the Propeller power pins.
The Propeller Platform USB has 3 decouplers on the Propeller power pins.
The Spinneret has all 4 decouplers on the Propeller power pins.

So, I dunno what is needed and what's not. Maybe it depend a lot on what others parts are on/near those power lines on the board?

Phil Pilgrim (PhiPi)

It's mainly a question of what's recommended, what you have room for, and what you can get by with. A lot depends on the layout. On the Propeller Backpack, for example (which I designed), there is a power plane beneath the Prop chip and a groundplane on the other side of the board. The decoupling caps (low-ESR multi-layer ceramics) are on opposite sides of the chip and have a value of 1uF each. One of the two decoupling caps is on the side where the crystal terminals are which, according to Sapieha, is very important.

I would consider connecting all four Vdd/Vss pairs, with 0.22uF ceramic bypass caps on pins 5/8 and 27/30, and a tantalum or large-value ceramic filter cap near the 3.3V regulator output (depending on regulator specs) to be a minimum requirement.

-Phil

Humanoido

Thanks very much for the data, research, and recommendations. I have installed four 10nF decoupling capacitors in one board and two in other boards for tests. So far, program performance is identical with hours of testing and blocks of tens of thousands of data bits communicated. With design considerations for top level performance, I agree that 4x is the way to go with the surface mounted prop on Proto Boards, inclusive of good filtering on the power source.

Cluso99

humanoido: There has been lots of discussions about this. You just need to search the older threads.

The Prop ProtoBoard indeed only has a single bypass cap at the prop pins. The circuit states it is a 1uf ceramic. There is a ground plane around the prop on the underside and a 3v3 plane encompassed within that under the prop. I suspect there is a ground plane under the prop on the top side. This is pretty much what I have done on my RamBlade.

However, the protoboard and other boards are not being pushed to the limits like I do on the RamBlades and TriBlades, and my future boards.

Just by running programs through the prop will not adequately test the design and decoupling properly. From tests Sapieha has done and understanding what he has found, I hypothesise that starting and stopping cogs (by waitxxx) instructions and with the counters exercised, with all cogs running, will cause the most errors. This is due to the start/stop currents. A good bulk cap such as a 10uF Tantalum under the prop and 0.1uF (100nF) X7R ceramics for each set of pins is a good choice. An extra 10nF under the prop is an added bonus. There are of course other ways of doing this too.

Somewhere I posted my RamBlade prop pcb section. I run it at 104MHz and ship 108MHz as well. It has run at 8x14.3MHz but not tested enough to be confident. It will not run at 8x15MHz=120MHz like the DIP in my TriBlade leaving me to suspect the DIP package may be a little more imune.

Sapieha

Hi Cluso.


It is not that It is Yours PCB that not can run 15MHz.

I have tested it more extensively AND it is ALL my propellers can RUN 14.318.180 BUT only 3 of them can run 15MHz.

Cluso99 wrote: »

humanoido: There has been lots of discussions about this. You just need to search the older threads.

The Prop ProtoBoard indeed only has a single bypass cap at the prop pins. The circuit states it is a 1uf ceramic. There is a ground plane around the prop on the underside and a 3v3 plane encompassed within that under the prop. I suspect there is a ground plane under the prop on the top side. This is pretty much what I have done on my RamBlade.

However, the protoboard and other boards are not being pushed to the limits like I do on the RamBlades and TriBlades, and my future boards.

Just by running programs through the prop will not adequately test the design and decoupling properly. From tests Sapieha has done and understanding what he has found, I hypothesise that starting and stopping cogs (by waitxxx) instructions and with the counters exercised, with all cogs running, will cause the most errors. This is due to the start/stop currents. A good bulk cap such as a 10uF Tantalum under the prop and 0.1uF (100nF) X7R ceramics for each set of pins is a good choice. An extra 10nF under the prop is an added bonus. There are of course other ways of doing this too.

Somewhere I posted my RamBlade prop pcb section. I run it at 104MHz and ship 108MHz as well. It has run at 8x14.3MHz but not tested enough to be confident. It will not run at 8x15MHz=120MHz like the DIP in my TriBlade leaving me to suspect the DIP package may be a little more imune.

Cluso99

Thanks Sapieha. IIRC Chip said he expected the PLL to top out at about 8x 14.318...MHz.