I think the Zero W is going to have a very long shelf life.

Which is a short way of saying, at the risk of sounding flippant, one of the things you look for when deciding upon a microcomputer or microcontroller for your embedded system, is for it not to chuck out a shedload of heat. I’m struggling to imagine how one might integrate a Pi 4 without making an object which is basically one solid heatsink that’s literally too hot to handle.

But then again, the 4 does desktop, and server, so very, very well… Just a shame the Pibow Coupé can’t have etched-on GPIO numbers because of that massive heatsink.

Hopefully prices will fall such that the Zero W becomes as cheap as the base Zero, and everyone except for people who work in a round building in Cheltenham, will consider the W variant to be the go-to controller for anything that can cope with a 30-second boot time. (Remember the Pi 2 revision 2? Same CPU as the Pi 3 but no radio? Very handy bit of kit, that.)

Thanks, Roger, good Article. I have shared with Raspberry Pi-related product makers.

Hi Roger,

The Pi 3B/3B+/4 are very good platforms for testing out our Proof of Concept designs – price/performance is great.

However, viewing the Pi 3B as an evaluation platform and the Compute Module (CM3+) as the production/industrial platform. There is a large gap in design documentation to bridge the gap. Examples:

1) What WiFi/BT modules can be used with CM3+? Where is a schematic showing connections (SDIO/serial)? To add an external antenna requires FCC/IC certification in North America – this requires the ability to put the RF module into a test mode for WiFi and BT – can Pi Foundation offer guidance in this area? Or maybe a list of FCC-certified modules and external antennas are available? I think full schematics of CM3+ are needed for FCC/IC submission,correct? Will Pi Foundation release schematics under NDA?

2) The Pi 3B/3B+/4 do not offer u.FL connectors for external antennas. This is forcing us to migrate to CM3+. Otherwise Pi 3B/3B+/4 with our own HAT module is ideal. But the onboard chip/PCB cavity antennas do not give the range that an external antenna can – especially in a sealed metal enclosure.

3) What USB/Ethernet controller can be used with CM3+? The Microchip LAN7515 controller is sold exclusively to Pi Foundation. What are the alternates that will drop in and work with CM3+ and Raspbian drivers? How do I hook this up to CM3+ – need reference schematic.

4) Cameras that use the GPU port (CSI) are limited to OV5647 (obsolete) and Sony 8MP imx.219. Could other imager devices be supported with GPU? Why are full spec’s of Sony imx.219 proprietary to Pi Foundation? How am I supposed to design a camera for CM3+ without specifications? I understand that drivers use I2C to check for camera type – only OV5647 and IMX219 are supported.

5) The Broadcom SoC on these generate a lot of heat under full load (~ 3-4 Watts ?). Does Pi Foundation or other 3rd parties have recommendations on thermal mitigation – heat sinks, spreaders or heat pipes in a sealed enclosure?

Maybe the chip vendors (Microchip, Cypress, Broadcom) could publish design recommendations for CM3+? It would drive their peripheral chip sales for commercial applications of the Pi CM3+.

Hopefully most of this makes sense. Thanks for the opportunity to voice our concerns here.

I would hope the zero W will be replaced at some point, by something superior.
I’m thinking of a (future) pi zero, a pi pico at half the size of a zero, to be the main brains behind any kind of smart device.
A device with CPU, GPU and RAM integrated in one chip, on the bottom, and all interfaces (audio, wifi, USB) on the top. Having a 1 core, 2 thread CPU, of the same tdp, with perhaps a small bump in ram to 786MB.
The zeroW is my favorite by far of all their designs, with the 3B+ being second. The 4B being third due to thermal complaints.

But I don’t think the pi 0W is where it could be yet, and at some point will need replacing.

The pi organization can also possibly look into a replacement for the gpio pins, by using a connector instead (like a HDMI or pata interface), which is cheaper to implement, smaller, and easier to handle.

As far as new design usage, as main brains for 3D printing, coding, web design, movie theaters (media playback), smart fridges, smart TVs, car entertainment, …
The benefit of having a pi car radio, is that you’re not limited to a specific file format (eg: mp3), but can run anything from ogg, opus, wma, m4a, and even video!

Smart wear (clothing), or even modding a car’s fuel injection system with a Pi.
Get a Pi car diagnostic device,
A pi portable printer, for printing entry tickets, payment stubs or pricing labels on the go could go a long way!

The pi organization could earn most of it’s income from the industrial sector, by offering more affordable products than the competition, which could help keeping cost down for their user end products.

And a pi server… Still waiting for pi engineers to someday bring out 20 arm corex cores in one affordable board!

The PI four will work very well in environments that embedded x86 PCs already exist. There are large numbers of applications that simply require the product to be able to withstand 60°C. Industrial shop floor, machine control, production monitoring, etc. all benefit from a product that is much smaller much lower cost than an embedded PC. Having developed many products into such markets we are already working to come up with a standardized board carrying the pie for and some peripherals that we know we will need in a typical base level application.

We are very much looking forward to delivering the pie into markets that it currently is not in

Take the Pi4 and Pi3 and offer them without connectors. Same board just no connectors. This will allow better integration into some products.

Please make it easy for the Pi 4 to be plywood-mountable!

In Canada, all Plumbers and Electricians will normally mount all their parts to a simple plywood backing. Please make an official “tradesman” Pi case that is aluminium like the Flirc (and touches the SOC to act as a heatsink, allowing passive cooling for the Pi 4), BUT with with the innovation that all 4 corners have screw-holes large enough for standard plywood screws to comfortably pass through.
And Flirc might not steal away as many case sales from you. Their passive aluminium case beats the pants off the official, current plastic Pi 4 case, in the way of passive cooling.

Few points from my point of view
– Most of the industrial use cases will need more support & tuning on the OS level to match what can be done by different types of microcontrollers and Real-time Operating System (RTOS). Particularly, I would say providing well documentation on how to convert the lite version of the Raspbian to act as much close as possible to RTOS. There are plenty of articles on how to convert it to ReadonlyRoot, a lot of OS flavors and a lot of compact Linux distributions out there working (somehow) on Pi but yet there is NO official one from Raspberry Pi itself. I believe intensive documentation on how to do it (+ continuous updates) is what we need at this stage. this will boost the speed of implementing these nice boards in the industrial market.
– Another point I want to raise is being flexible in providing multiple options for simple things like official cases. Compute modules are suitable for companies to develop products, but many DIY projects can be also successful products using only the official hardware (you just to program it) e.g. CCTV using the Pi0W with official camera case. I believe providing may options for Pi4 cases to include many plugins can be useful e.g. waterproof cases, camera cases with PoE hat, case suitable with sense hat, camera action case for Pi0W…..

Hello,

I need to know how high are the chips on the Raspberry Pi 4 model B PCB. I’ve tried to get the precise heights, including min and max values since last month, but your web support supplied only a link to a known RPI4B outlines with the height only for the BCM chip (2.4mm).

I need to know, how high are the ETH PHY, USB HUB and DDR4 memory chips. I need the maximum and minimum values that can appear.

I’d also like to know the height of your PMIC chip.

A prototype of our heatsink which interfaces directly onto these chips has been left at your offices for your highest chief to check. I need to consult this asap, because that’s the last thing holding us back from serial production. The holidays are a miserable time for hardware developers, everybody’s on vacations :-(

FYI, interfacing the PMIC chip to the heatsink increases the stability of the RPI4 when overclocked. The BCM chip can heat more by about 5 degrees Celsius before rebooting.

Regarding DDR4 memory, we’re not sure, if there are any variances in the heights of various capacity chips (1/2/4GB versions). Including future chips. The problem is, if we make the heatsink tight to the memory chip, it may not interface directly with a future bit less high chip, or (the worst scenario), if the height of the memory chip increases in the future, it may lift up the heatsink and disconnect the BCM chip from the heatsink thermal-wise.

This is a very important decision to make – about the memory chip height. We can’t decide this without your engineering and manufacturing expertise and data. We don’t interface it directly now, there’s about 0.05mm distance between the memory and a heatsink (4GB version).

I’ve measured the respective heights using a caliper and rounded them a bit, so the presumed heights are now 0.9mm for ETH PHY and USB HUB, and 2.4mm for the BCM chip (also noted in your official documentation). Is it right? It cools very well, though…

The PMIC chip may be 0.8 or 0.9mm high. But this tiny 0.1mm difference can make a lot when interfacing to the heatsink… touching it or not.

Our thermal imagery shows that once the heatsink is installed, the hottest point is around the PMIC chip. And when overclocking, the BCM chip flies, but the PMIC struggles. For example, running at 1850MHz with over_voltage=4 leads to reboots soon after it heats up, but it’s super stable at over_voltage=2. If I remove the thermal bridge between the PMIC and the heatsink, it reboots sooner (at lower temperatures,over_voltage=4).

If I may ask one appeal as an engineer, please deliver the heatsink/case sample to your chief engineers to answer these question, so that we can make a presence in the next MagPi and have serial production running by then. A few of your people already work on this, but I’d appreciated if whomever influenced your top engineers to give it a priority. The heatsink samples are somewhere in your offices now :-)

Thank you very much and thank you for the Raspberry Pi – an excellent engineering feat!

Best Regards,

Lada

Hi,

Nice article!

I would say. Contact your partners RS and Farnell and ask for end customer data who bought Pi’s. Contacting the end customers directly will give you a lot of info I guess. I have worked for RS and I have had customers who used it for 3D printers, gambling machines etc. I am sure a lot of these customers are in for a chat with the Pi foundation.

Good luck developing for the future!

Regards Roger

Aggressively support UEFI (such as: https://github.com/tianocore/edk2-platforms/tree/master/Platform/RaspberryPi/RPi3 ). Instead of requiring every single OS to modify their software to support the Raspberry PI, they could instead use the common target of UEFI. Imagine SteamOS on the Raspberry PI via UEFI.

I absolutely love the Raspberry π! I use it all the time for various solutions. Sometimes when I need a temporary router, it works perfectly. A media server in the car to stream all of our movies (to keep the young ones happy and everyone else sane), it is a great solution! It’s low power use, meaning I can run it off one of my little battery banks for my phone. It’s tiny, meaning easy to transport and use in many situations. It’s also highly adaptable, with pretty decent hardware for the cost (I have the 3B+).

The only product that would be great to see, is a “Raspberry μπ” (Raspberry Micro-Pi). For development uses, remove everything unnecessary. No display ports, no USB, no ethernet jack, no ports at all really. Just the core components, and the GPIO. Of course, replace the GPIO with a female GPIO, since that saves space not having those pins stabbing out. A system designed to be ran completely headless. It could still run the normal old Raspbian OS in Lite. However, it could be absolutely tiny. Heck, maybe even develop an OS (if you can even call it that) that is pretty much just an Arduino. Seriously, an OS that simply just provides low-level frameworks, and allow the developer to write pretty bare-bones code for an optimal hardware-ish level product.

Having a tiny π would be great for this one project I’ve really wanted to work on for a while. I have an old Gameboy Advanced SP. I want to take out all the innards and replace it with modern hardware, while leaving the external components, such as the buttons. (I may replace the display if I have to, since I’m having trouble finding schematics for the pin layout).

Every model of π that exists is much too big for this project. I don’t need HDMI or USB or anything! Ideally, I would want WiFi and bluetooth, but I can deal without I guess. My plan was (when I originally wanted to use the π for this) to use the π as the central processor, and depending on the pins available and other restrictions, take an Atmega microprocessor (or similar) that has enough pins, and use it as a make-shift sort of GPU.

If there was a π that could handle this, I’d LOVE that. For a µπ, maybe have the purchase page have configuration options. Choosing to add/remove WiFi, Bluetooth, etc. Heck, it could even be an option to remove the SD card, instead using an integrated flash memory solution.

I want to close this up as to not write for hours, but to build on what I said for something else. In an upcoming π, adding the ability to have integrated storage would be amazing. Even if we still use the π as primary boot, maybe have the system also check (second, after SD) if the internal volume is bootable? Either way, it’d be awesome to simply store data. Possibly as some sort of persistent data system.

I used CM3 and design the carrier board for remote monitoring system, ex IoT-Energy, IoT-Agriculture application, here are my comments to add
1.) working environment temperature +85~-40,
2.) Support RS-485 x 2 Ports (4 Pins), [VCC, DATA(A), DATA(B), GND]
3.) RTC
4.) 4 Antennas, 4Gx2, GPSx1, BLEx1
5.) Dual Mini PCIe Sockets (1 for Communication Module, ex NB-IoT to WiSUN Mini PCIe Card, the other for Storage, ex USB to mSATA SSD)
6.) For power saving, all the modules can be power on/off separately

Will there be compute module 4? It would be very useful since 3x more power than cm3.

We are in the process of bringing to market a Personal Health Gateway that is based on the Raspberry pi 3b+. However, as we have attempted to move from pilot to production and secure the supply chain we have come into challenges in obtaining the 3b+.
In discussions with a supplier of a system on a module they indicated that the RPI boards would also be a supply side issue when deploying in scale. He asserted that Broadcom provides the core chips for RPIs at a discounted price but based on excesses capacity of manufacture only. The implication being that we could not be sure of having product.
Currently it has been impossible for us to obtain RPI 3b+ and we are being quoted lead times of 23 weeks.
Can you comment on the validity of the assertion, and on issues creating the current lack of RPI 3b+ in the market?