SBCs Other Than RasPi?

In this situation some Raspberry Pi users attach a POE hat and this allows the Raspberry Pi to be powered by a single Ethernet connection and to boot from a network image. You can’t get much better than that for Raspberry Pi’s that are deployed in hard to get to areas. The only thing that will require a visit to the Raspberry Pi is a failure of the Pi itself.

You could probably find a refurbished or off-lease slightly older PC for that price. I think the last PC I bought for an office was about $250 and it was a quad core i5 with a 128gb ssd. That was about 1yr ago and it’s been running fine since.

This is for a small form-factor medium-volume production item.

Maybe:

  • Intel NUC for example (like a mac mini)
  • Udoo (unknown for me never used it)
  • Raspberry Pi Comute Module 3+ 32GB eMMc flash disk (1GB memory)
  • Micro ITX (just like any pc only tiny)

Or search google for “tiny pc” “small pc” “tiny x86 sbc”

or look here; http://onlogic.com

Well, that would certainly rule out an off the shelf PC then, but it also opens up a whole new realm of concerns I would guess. A lot of the other boards beside raspberry pi are hard to get hold of at times and also seem to change fairly often. Not sure I would base a product on them unless I could get some kind of assurance from the manufacturer that I could get new boards made if needed. I guess it depends on the production needs of whoever is producing the product you are working on.

Personally if they want something like a pi I’d just stick an industrial rated sd card in it and call it good. Not sure you’ll do much better than that with anything else. Not to mention if the card does fail it will be cheap and easy to fix that problem, but unless you are writing data to the card heavily all the time it should last a few years easily. Just be sure to set the os up so it’s not writing things that you don’t need.

We use rPi 3B+ with the SanDisk “High Endurance Card” µSD cards and the units in our lab are beat to heck and back. We’ve just started using the 2GB RPi 4 units with the same card with similar success. On the high(er)-end, we use the RockPRO 64 from Pine64. It’s VERY robust, but does cost a few dollars more than the RPi 4. We use 64GB eMMC cards as system disks for those and we’ve seen zero issues in over 2 years.

From everything that my team has uncovered, the issues with the generic µSD cards is related more to misuse (unplugging while powered on being the biggest culprit - the RPi’s socket is non-buffered). If you put the units in a case that prevents access to the µSD card without removing power, I suspect that you’ll never see this issue.

[quote=482385:@Tim Jones]We use rPi 3B+ with the SanDisk “High Endurance Card” µSD cards and the units in our lab are beat to heck and back. We’ve just started using the 2GB RPi 4 units with the same card with similar success. On the high(er)-end, we use the RockPRO 64 from Pine64. It’s VERY robust, but does cost a few dollars more than the RPi 4. We use 64GB eMMC cards as system disks for those and we’ve seen zero issues in over 2 years.

From everything that my team has uncovered, the issues with the generic µSD cards is related more to misuse (unplugging while powered on being the biggest culprit - the RPi’s socket is non-buffered). If you put the units in a case that prevents access to the µSD card without removing power, I suspect that you’ll never see this issue.[/quote]
Thanks, Tim, the Pine RockPRO looks like it could be just the ticket, and it’s less costly than the Up Board we use now elsewhere in the product line. Back to the original question, can I develop and run Xojo programs on the RockPRO as easily as developing on the RPi? Will the remote debugger stub run on it?

I’ve had terrible luck with RPi SD cards, and I’ve taken every precaution. I’ve never removed an SD card while power was on. I’ve never powered down an RPi without first doing a software shutdown. I’ve never overclocked an RPi. I’ve tried various power supplies, on the assumption that a bad power supply may have been putting glitches on the power bus, but to no avail.

It should raise red flags that if you go to the official RPi website, you’ll find that they maintain an extremely long list of manufacturers and model numbers of SD cards that work or don’t work. That shouldn’t be necessary. They should just work. The fact that sometimes they work and sometimes they don’t, is indicative of a serious design error in the RPi. Yes, I know that the majority of people have never had a problem, but if you find yourself in the other group, you’ll never trust an RPi ever again. I would stay the hell away from RPi’s running off of SD cards. They may be a fun toy, but they’re not suitable for mission critical applications.

Anything that is tested and qualified appropriately for its use is suitable for mission critical applications.

I have used raspis as voip servers since the original model b. I had one running in my home for several years on a 4gb sd card and it never had a problem. I upgraded that one to raspi 3 to test for a customer who wanted one to run his business and home phones. It is setup to automatically back up to a second card and we’ve never had any issues with that one either. It’s been in service over 2 years now. In another year we’ll swap out the backup card for a new one and just keep on going. In fact I’ve only had one premature failure of a raspi I put in service and it was put in service with a new brand of sd card, knowing that we might have an issue.

The main problems people have with raspis are pretty easy to deal with. Buy storage and power supplies suitable for the use you intend. I would recommend that with any type of computer hardware put in service for mission critical usage.

I have it running on a banana pi, but it will only run on a couple of processors not all 8

I strongly disagree with this sentiment. We buy 100-150 SD cards every quarter and the number of counterfeit cards that we find when we have to go outside of normal channels is ridiculous. We see this mostly in the 16GB and 32GB size variants since those are the two most popular sizes. I have a box of 32GB cards that we’ve received from an otherwise reputable supplier that fail the tests in ChipGenius (get it if you don’t have it). We have now moved explicitly to using SanDisk “High Endurance Card” cards and our “fake” rate has dropped to 0%.

@Robert Weaver 's experiences mirror those of my client, despite their having purchased “approved” memory cards from supposedly reliable sources. Swapping in “backup cards” every two years as Kevin Windham suggests is not a viable scenario. We went with the Up Board from Aaeon because it uses eMMC which is somehow miraculously more reliable than SD even though they both use the same kind of flash AFAIK. The Up Board is too expensive and too big for this new project, though, and it also needs wi-fi, which the Up Board doesn’t have built-in. Definitely liking the Pine RockPro if I can develop Xojo apps for it.

It looks like this just reports common device properties like PID, VID, etc., same as Apple’s USB Prober and other utilities. How does that prevent counterfeits? Can’t the counterfeiters just use the real manufacturer’s info?

Not that we’ve found. The chipset is the key. It’s not about “preventing”, but rather identifying. Other tip-offs are the quality of the packaging and the labeling. We received one set back in January that was not even packaged - just the plastic cases rubber-banded together with a laser-printed paper tag that said - “OEM Bluk, no packaging” - yes, they even misspelled “bulk” …

Not trying to be flippant here, but your need of zero maintenance is a “Pi”-in-the-sky requirement for something as inexpensive as a RasPi. If you need industrial level performance and reliability, you’re not going to get it for $55. I’ve been in this game since 1984 and these types of designs really are a “you get what you pay for” design. There really is a reason that the government pays $3K for a Motorola marine radio when we could purchase a consumer grade radio for $495.

I should have also mentioned that I tried at least a dozen different name brand SD cards (mostly SanDisk, but also some Lexmark and a couple of others), and all from a reputable seller with trackable supply chain. All of the cards failed. I don’t believe that it was entirely the fault of the SD card. I believe that there’s a borderline timing issue with the RPi itself. If you get a Pi that’s just a little bit off spec, it’s going to corrupt SD cards no matter what kind you use. If you get a good Pi, then it will probably work fine with no issues. My biggest complaint is that whenever anyone complained about the problem on the official RPi site, they got shouted down with claims that they were:

  • Removing the card without powering down,
  • Powering down without first doing a software shutdown,
  • Overclocking the processor,
  • Using a non approved power supply,
    even when they were doing none of these things.

The fact that the designers remained in denial about the problem meant that we were never able to find the true cause of the problem. It also meant that if they wouldn’t admit to there being a problem, then they had no intention of fixing it. Or, if they did privately recognize there was a problem and then secretly fixed it, would they later come out and announce that “We fixed that problem that never existed”? Without ever logging it as a known issue and then later announcing that they tracked down the problem and fixed it, how can anyone have any faith that it’s now safe to use the thing?

Again, I do recognize that the majority of users have never experienced the problem. However the minority that has experienced this problem is a very large minority.

We are in agreement that for $55 you shouldn’t expect industrial grade reliability. On the other hand we have deployed Arduino based controllers (no SD cards to worry about) in some very harsh industrial environments, and they’ve outperformed some very expensive industrial grade hardware.

I totally missed this one. Yes, there are currently a number of 32bit ARMHF (hard float) options that will run the remote debugger and Xojo apps.

Also, I do see that the 2GB model is only $59. That’s quite a bargain considering it’s 6 core and Big-Little architecture.

Don’t be too dependent on the PCIe option as it so far only supports the Marvell/ASMedia SATA card 100%. I have successfully built a 24TB NAS using it and two 12TB Exos drives.

Arduino is a much simpler device, so making any type of comparison between the two device classes is like comparing a Lexus to a Schwinn. Far less can go wrong with an Arduino. It’s more resilient under even high temperature uses or areas where access is a problem. As a tradeoff, it’s also not a full computer like the RasPi and clones.

Be sure to qualify your eMMC choice as well. Those also fail in time like all flash memory does. If you need something that will last more than two years with minimal failures, be sure to buy something rated for the amount of writing you expect to need, and test it before purchasing in quantity.

BTW, the 3 yr rotation on the backup card, not the primary card, is peace of mind insurance for the customer. Like I said before my original PBX I had installed at my house was running as a voip server, with asterisk logging for several years on a 4GB card, sandisk IIRC, and never had any issues. All servers should have backups and the extra card for my customers backup just makes it very easy for them to fix a failed card. I expect it will fail eventually, maybe 5-7 years of service. All the customer needs to do if that happens is just unplug the pi and move the backup card to where the main card is and plug it back in. Simple.

Also, both of their previous PBX boxes that were run off actual PCs had hard drive failures. They weren’t equipped to deal with those and experienced about a day of downtime both times. Now if they have a failure it should only be maybe 10 minutes.

I recall there being some changes made to the software and some default boot settings changes as well to help with some of the issues you described. Part of the problems some people had were also related to the way the usb bus was set up. The way it was powered and the fact that the reader was also running on usb had something to do with it as well. I definitely remember struggling with power issues before all that information came out. I remember trying to run it off various USB ports and having intermittent issues. Once I started using higher voltage and amperage power supplies my particular problems went away. I also started adding up the power requirements of the stuff I plugged into USB so I’d know when I was in the danger zone. A basic mouse and keyboard were never an issue, but more than that could get sketchy.

We use a 4A supply for all of our implementations. In bulk, the cost is around $4 more than a 2.5A, but the result is so much more predictable.

One option that we used until we found a proper source was this Amazon Basics USB hub. We just threw the hub away (recycled, actually) :).
Powered 7 Port USB Hub