Raspberry Pi Pico

[DavidAlfa]

I've looking at it since it came out... The dual core really caught my attention.
It's amazing how fast MCUs are getting, multicore, even GHz clocks.

How good are the documentation and libraries vs stm32?

[exe]

I've been wondering about the PIO for a while.
Can it be used for old-school parallel bus? Like 16 bit address 8 bit data? Or the 16 bit address + data multiplexed?

I'm no expert, but if you explain in more details may be I can help. PIO can pull/push in one go up to 32 pins.

[tszaboo]

I've been wondering about the PIO for a while.
Can it be used for old-school parallel bus? Like 16 bit address 8 bit data? Or the 16 bit address + data multiplexed?

I'm no expert, but if you explain in more details may be I can help. PIO can pull/push in one go up to 32 pins.

Well, I know that you can write more than 1 GPIO pin at the time. Imagine the bus of a 6502. 16 bit address 8 bit data. R/W. PHI2 , which is a clock input for the bus. Maybe it should be an output for our theoretical system. DBE for external drivers.
All this is possible to do with a 100MHz microcontroller, for a 1MHz speed bus. I'm just wondering if it is possible to move it "one level down" and do it with the PIO, freeing up resources.

[thm_w]

Saw a recent blog on measuring ADC performance of Pico: https://pico-adc.markomo.me/
Very well documented, DG4000 was used.

[Nominal Animal]

I've been wondering about the PIO for a while.
Can it be used for old-school parallel bus? Like 16 bit address 8 bit data? Or the 16 bit address + data multiplexed?

Multiplexed yes, although I only looked at the PIO details to see if I could experiment with PDM with it (no, the PIO units have no adder).  Do check the datasheet for yourself, though.  The PIO instruction set is on page 341, I recommend reading from page 330 (Chapter 3. PIO) onwards.

The IN and OUT PIO instructions can set/get the state of any number of pins (up to 32), but the set is contiguous and always starts at the same pin (unless the 2040 modifies the PINCTRL register for that state machine).  So, for 16-bit address and 8-bit data, I think you could do it if you push 32-bit words to the FIFO, consisting of the data byte in the least and most significant bytes, sandwiching the address bits.  Writing the address would always write the data bits too (set all 24 pins).

The multiplexed form seems much easier.  Essentially, the PIO "program" will pop off an address, emit it to the PIO pins, and toggle the additional clock/read/write pin.  A read program will then read the PIO pins and push it to the 2040 proper; a write program will pop off the data, and emit it to the PIO pins.  Each instruction has an optional delay (generally, 0-31).  The limit of 32 instruction words per PIO unit (four state machines) might be an issue, but you'll need at least two of the state machines (read and write operations) anyway.  Autopull and wrap will reduce the number of state machine instructions you'll need.

I believe you can also do "block" DMA to the multiplexed bus, with the PIO program autoincrementing (or decrementing) the address.  For example, you supply the PIO program with the length of the data, the initial address, and the data (bytes or words), and it will write the interleaved address-data to the pins as needed, without the DMA'd data having the addresses interleaved.

Obviously, I haven't tried this myself, as I don't have a use for it, so there might be gotchas I haven't seen.

[ale500]

Did someone try with segger's embedded Studio ? According to their website the latest version (5.42) should have a package for the RP2040. But I do not see this package in the list of available packages. (I clicked on the non commercial license, I have an edu jlink).

[SiliconWizard]

Saw a recent blog on measuring ADC performance of Pico: https://pico-adc.markomo.me/
Very well documented, DG4000 was used.

Thanks. It doesn't look that good, but I'm not very surprised.

[exe]

Btw, speaking of overclocking, according to [1], it seems most board can go up to 330MHz, and some go over 400MHz. Unfortunately, external flash stops working at much lower frequencies.

[1] https://www.raspberrypi.org/forums/viewtopic.php?t=301902

[Dubbie]

Wow that writeup on the ADC was really excellent.

[MK14]

Btw, speaking of overclocking, according to [1], it seems most board can go up to 330MHz, and some go over 400MHz. Unfortunately, external flash stops working at much lower frequencies.

[1] https://www.raspberrypi.org/forums/viewtopic.php?t=301902

That's very interesting. That could mean that future releases of the PI Picos, could be at higher maximum clock speeds. It also means that anyone who understands the risks/problems with overclocking, could experiment with overclocking these modules.

It is possible they started out at the current 133 MHz limit, until they could do more tests and optimize the MCU, enough to safely go to higher maximum clock frequencies. Maybe it is the peripherals and/or external flash chip (as already mentioned, faster ones may not be available, are too expensive and/or may be unsuitable for other reasons), which are limiting the maximum reliable clock frequency.

N.B. Overclocking is NOT recommended for serious/professional/production applications/use.

[guenthert]

RP2040 chips (not the Raspberry Pico board) now available for $1.  Well, some 40000 of them at least.  In autumn larger quantities are expected.

https://www.raspberrypi.org/blog/raspberry-pi-rp2040-on-sale/

[SiliconWizard]

RP2040 chips (not the Raspberry Pico board) now available for $1.  Well, some 40000 of them at least.  In autumn larger quantities are expected.

https://www.raspberrypi.org/blog/raspberry-pi-rp2040-on-sale/

Now that's pretty cheap. I wonder how many of them they'll have to sell before they break even. (I'm also wondering about the licensing deal they got from ARM. Probably a pretty nice deal here.)

[guenthert]

RP2040 chips (not the Raspberry Pico board) now available for $1.  Well, some 40000 of them at least.  In autumn larger quantities are expected.

https://www.raspberrypi.org/blog/raspberry-pi-rp2040-on-sale/

Now that's pretty cheap. I wonder how many of them they'll have to sell before they break even. (I'm also wondering about the licensing deal they got from ARM. Probably a pretty nice deal here.)

     I don't know the first thing about finances and have no insight into the details here, but my guess would be that the R. Pi Foundation as a non-profit organisation has written off the development costs and just needs to recoup OpEx.  And it's my (perhaps flawed) understanding, that chips, once developed, can be made fairly cheaply, if existing infrastructure can be utilized (no investment into better machinery needed).

[lucazader]

Another thing to note here:
They are $1 in qty 1. (and right now discount for volume)

They do expect the per chip price to drop when they start selling full reels later this year.
source: https://www.cnx-software.com/2021/06/01/buy-raspberry-pi-rp2040-one-dollar/
alasdair allan made a comment on this article (he is the Technical Documentation Manager at rpi)

[brucehoult]

     I don't know the first thing about finances and have no insight into the details here, but my guess would be that the R. Pi Foundation as a non-profit organisation has written off the development costs and just needs to recoup OpEx.  And it's my (perhaps flawed) understanding, that chips, once developed, can be made fairly cheaply, if existing infrastructure can be utilized (no investment into better machinery needed).

Once you have the full-reticule mask set made -- which costs $800,000 for the 40nm node the RP2040 uses according to 2019 figures I found -- that stays in the library at the fab and one or more wafers can be run off at any time in the future for according to the same 2019 source $2274 per 40nm wafer

The die is 2 sq mm so a 300 mm wafer would be around 35000 chips. The cost per bare chip is therefore something like 6.5 cents.

Testing and packaging will be the main part of the cost, not the chip, especially if you're not trying to recover NRE costs.

[lucazader]

The die is 2 sq mm so a 300 mm wafer would be around 35000 chips. The cost per bare chip is therefore something like 6.5 cents.

That makes it sound like for this first batch of chips that were produced, they only needed probably 2 wafers to get the 40k units that they are shipping, assuming a 60% ish yield.
That's pretty interesting.

How does the 2 sq-mm die size compare to similar chips from other manufacturers? (eg STM32G0/F0, ATMEL/PIC, NXP)

Edit: This article from 2014 seems to suggest around a 3.2 sq-mm die size for the KL03 line of chips
https://www.techdesignforums.com/blog/2014/02/27/freescale-kl03-mcu-design-tuning/
and the arm IP page give a bit more about the Die of the m0+ core without peripherals, around 0.006 sq-mm for the 40nm node
https://developer.arm.com/ip-products/processors/cortex-m/cortex-m0-plus

[brucehoult]

The 0.006 mm^2 for a core is of course just the actual size of useful stuff.

I *think* the 2 mm^2 for the chip includes the space around the actual circuitry for the wires to be attached ("pads"), and also the gap for the saw to munch up to cut each chip out from the wafer.

[Rasz]

They are $1 in qty 1. (and right now discount for volume)

where did you get that from?
They are $1 in qty 1 and there are NO volume discounts, in fact it sounded like they prohibited dealers from selling more than 1 per person.

[lucazader]

They are $1 in qty 1. (and right now discount for volume)

where did you get that from?
They are $1 in qty 1 and there are NO volume discounts, in fact it sounded like they prohibited dealers from selling more than 1 per person.

as menioned in the original post:
source: https://www.cnx-software.com/2021/06/01/buy-raspberry-pi-rp2040-one-dollar/
alasdair allan made a comment on this article (he is the Technical Documentation Manager at rpi)

[exe]

They are $1 in qty 1 and there are NO volume discounts, in fact it sounded like they prohibited dealers from selling more than 1 per person.

I've seen in one shop that they sell 1 piece max (which is weird), but here you can buy at least 10 if/when it's back in stock: https://www.okdo.com/us/p/raspberry-pi-rp2040-microcontroller-pack-of-10/?src=raspberrypi .

Unfortunately for me, it's not on sale yet in the Netherlands (or I didn't find it). I also bought a few rpi pico modules, so I'm probably still going to transplant ICs from them.

[tszaboo]

They are $1 in qty 1. (and right now discount for volume)

where did you get that from?
They are $1 in qty 1 and there are NO volume discounts, in fact it sounded like they prohibited dealers from selling more than 1 per person.

There might be no volume discount, but here is a reel at farnell that can be ordered.
https://nl.farnell.com/raspberry-pi/rp2040tr13/mcu-32bit-133mhz-qfn-56/dp/3766080?CMP=GRHB-OCTOPART

RPI foundation has their right to set up their pricing policy. If it is "flat 1 USD" than its like that.
I dont mind the pricing, it seems very reasonable. It also doesn't put small players at a disadvantage.

[gamalot]

I ordered 20 pieces of RP2040 at PiAustralia at a unit price of 1.50 Australian dollars.

[phil from seattle]

They are $1 in qty 1. (and right now discount for volume)

where did you get that from?
They are $1 in qty 1 and there are NO volume discounts, in fact it sounded like they prohibited dealers from selling more than 1 per person.

There might be no volume discount, but here is a reel at farnell that can be ordered.
https://nl.farnell.com/raspberry-pi/rp2040tr13/mcu-32bit-133mhz-qfn-56/dp/3766080?CMP=GRHB-OCTOPART

RPI foundation has their right to set up their pricing policy. If it is "flat 1 USD" than its like that.
I dont mind the pricing, it seems very reasonable. It also doesn't put small players at a disadvantage.

Hoping to get some clarity on volume soon.  I'm not sure how the flat rate thing will play out for distis.  In particular, for far east assemblers like JLCPCB. I think it would be awesome if I could have a RP2040 board assembled by them.

[Rasz]

They are $1 in qty 1 and there are NO volume discounts, in fact it sounded like they prohibited dealers from selling more than 1 per person.

I've seen in one shop that they sell 1 piece max (which is weird), but here you can buy at least 10 if/when it's back in stock: https://www.okdo.com/us/p/raspberry-pi-rp2040-microcontroller-pack-of-10/?src=raspberrypi .

so
-NO volume discount
-not selling more than 1 per person

There might be no volume discount, but here is a reel at farnell that can be ordered.
https://nl.farnell.com/raspberry-pi/rp2040tr13/mcu-32bit-133mhz-qfn-56/dp/3766080?CMP=GRHB-OCTOPART

>Note: this product is not in stock
>You can place a pre-order by clicking on the button, or indicate your interest by clicking on the link

RPI foundation has their right to set up their pricing policy. If it is "flat 1 USD" than its like that.

its 1 usd same way RTX 2080 MSRP is $699

I dont mind the pricing, it seems very reasonable. It also doesn't put small players at a disadvantage.

its the old Soviet trick - the pricing is always very reasonable when there is no merchandise in stock

[thm_w]

its the old Soviet trick - the pricing is always very reasonable when there is no merchandise in stock

Because they are still producing them 

"Available to back order for despatch 04/08/2021"
https://uk.rs-online.com/web/p/microcontrollers/2211610/