Sample & Hold, did they go out with the buggy whip?

Den onsdag den 19. august 2015 kl. 22.20.04 UTC+2 skrev John Larkin:

I just saw this one, he might be interested

-Lasse

Ok, but we have so many other things to do on this project that it would be much nicer if SPI worked out of the box. Just send data and it gets there. Not we have to bare-metal it :-(

And then mine became a Lutheran. Her late aunt: "It is really great that you guys found back to The Lord. But did it have to be a Lutheran church?"

An unusually large percentage of the really good engineers I worked with went to Jesuit High. They must be doing something right. I don't think that the average American public high school with unionized staff and all that will ever come close.

That looks complex. I think we'll do it the old, disgusting way, reload the whole web page maybe once a second. We can also load a modest block of canned javascript into the browser, and then let the browser ask us for or poke realtime data, using our defied telnet text commands.

We hope to do several web-enabled instruments, each with an ARM running bare metal out of its 256Kbyte internal memory.

The LPC3250 has a pretty fancy SPI controller, so all you have to do is set it up and poke data into it. It's so fancy that sometimes it's easier to bit-bang things.

Telnet isn't something I'd be using these days. Asking for trouble.

Cheers

Phil Hobbs

Well, the boxes will use ascii text commands for user control, whatever that is. Serious users won't want a web page interface; they'll want their own code to talk to the boxes. So we can have our javascript talk that same command syntax, on top of the static HTML that sets up the web page.

I think he means ASCII over TCP/IP, which indeed is not the same thing as Telnet even though the windows telnet program only supported that, most primitive, telnet mode.

There's plenty of things using that as a basis, HTTP for a start.

I'm quite happy to use plain unencrypted ASCII on a serial port or USB, but _not_ over TCP/IP. The former are like the boy in the bubble compared with anything connected to the Internet these days, and it's only going to get worse. Some of John's stuff runs super expensive and potentially dangerous hardware, which makes it an attractive target for crackers. Keeping the sys tems air-gapped is nice, but it's hard to guarantee what users do.

Cheers

Phil Hobbs

Huh?

TCP/IP is not even remotely equivalent to the Internet. After you've plugged about ten serial cables into something, a nice Ethernet port and TCP/IP starts to look a lot better.

Besides which , UDP is Vastly better.

Generally more bandwidth is better.

John is likely sufficiently obscure to be completely invisible to crackers.

This is a good thing.

Not that hard. The IT department really does not want test equipment on the same network the billing and admin people use anyway.

I know ours does not.

Gesundheit.

It stands for "Transaction control protocol / internet protocol".

And I didn't claim it was "equivalent to the Internet", just that some moron^H^H^H^H^H customer was bound to connect it to the Internet, which they will, as you well know. Somebody will want to be able to fix things from home rather than having to drag himself back to the lab at

More convenient, I agree. (I have a bunch of GPIB instruments controlled by a Prologix GPIB-Ethernet box, for instance.) "Better" depends on what other folks do, some of whom are emphatically not your friends. Sad but true.

Unless you care if your packets get there reliably. A dropped packet causing a jet engine rotor to disassemble itself would be A Bad Thing. UDP with a handshake is probably slower than TCP/IP.

More fun, anyway.

You hope. The Iranians thought the same about their uranium centrifuges, I bet. The thing about IP threats is that the toolkits don't forget stuff--they just keep getting nastier. Lamarckian evolution is a _lot_ faster than Darwinian. The vulnerability of SCADA equipment in this country is seriously scary.

What they want and what they get is often quite different.

The SCADA and IoT worlds are very very different from a corporate computer network: far more vulnerable, far more dangerous, and far less well supervised and patched.

Comp.risks is an excellent read on that point, as is Bruce Schneier's blog.

Cheers

Phil Hobbs

We need S-PORT capability on this product for 12 audio channels. But the main reason I could not use something like the LPC3250 is that we absolutely don't want a large BGA package. This gear will get banged around a lot. Since I started mountain biking again I found out what repeated shocks can do to electronics. Last week it shook my battery charge balancer to death.

My users don't connect their control networks to the world. So far, we've had no complaints, no requests for encryption.

We have had to certify (and sometimes modify products) that our boxes can't be used to physically transport any information from inside a secured area, out to the world. I suppose someone could write a secret message on a piece of paper and stuff it inside one of our boxes.

If someone did stumble onto one of our boxes that somehow had a public IP address (which my users generally don't allow) they could type HELP and get a summary of the box ID and the command set.

A serious uP needs a lot of pins. The LPC3250 has 296 balls, which would be awkward in a gullwing package.

We haven't done serious vibration testing, but we have found BGAs to be far more reliable than gullwing connections.

I found the opposite. Rarely any failure with gullwing and if so then it was usually a poor reflow job to begin with. Maybe the same with BGA but you can't easily inspect those.

The ADSP-CM408F has 176 gullwing pins. That's way more I/O than we need, especially since many modern buses are serial.

That's not a bug, it's a feature!

We have had zero ball connect failures, with millions of ball connections, new production and some FPGA replacement. We've had some problems with gullwings, mostly things that looked soldered but weren't, which are mostly found in inspection/test.

More and more parts are BGA these days; it's almost unavoidable. Laptops and tablets and phones are BGA, and are pretty rugged.

Our stuff isn't mounted in 19" racks and gets banged around a lot. Someone tugs on the cable and the box slides off a table crashing onto a tile floor. Or it is pole-mounted and someone tips over the whole pole cart ... *THWOCK* ... chingalingaling ... and they expect this not to break stuff.

I had a major tire blow-out on my mountain bike on Friday ... RAT-TAT-TAT ... *PHOOMP* . Among many other things this also destroyed the charge controller of the central battery which is located behind the seat post. One of those newfangled leadless packages seems to have let go.

Not really. The upside is that BGAs have created a whole new trade and job opportunities for people:

I don't trust BGA.