Near disaster with PCB design

Why not USB? A micro USB connector and an FTDI chip make that easy. Looks like a serial port to any PC.

Sounds like a lot of fun. These super-interdisciplinary things let you cross domains freely, and keep the competition down.

Digitally programmable RF switched attenuators are super wideband and cheap. Or analog multipliers. Or trimpots!

Next-gen benchtop digital delay generator. We've been working on that, on and off, for three years, constantly interrupted by other stuff. Which means that the technology keeps changing! The outputs are much better now, GaN instead of the phemts. And the triggered, phase-locked oscillator is much better and far more mind-boggling; I barely understand it myself. I think we'll finish it this time! I'll send you one.

I'm also doing a few small 1-guy projects, idea through PCB layout, during the lockdown. The dual-path GHz o/e, a tiny delay/pulse generator, a pretty high voltage pulse amp, a USB Mach-Zender driver with built-in pulse and bias generator, a few little gadgets like that that might sell in moderate volume over the years.

We are also doing an SFP-based data transceiver for a biggish aerospace company who won't tell us what it's for, except that the light goes through a rotary coupling, optical slip rings of some sort.

Lotta little stuff now. I like things that can be started and finished quickly, without a lot of software or FPGA stuff. I also enjoy doing an occasional PCB layout, at least small ones.

We use USB serial in some applications too. MODBUS is useful in instrument applications because you may be talking to many detectors, and it saves beaucoup wires and computer ports. USB serial also has some enumeration funnies on Windows, whereas with MODBUS you always know whom you're talking to. Simon has developed a super swoopy MODBUS stack, including code generators in Python for all the access methods and stuff. We're primarily using it in our digital IR sensors for the fire control system for harvesters.

Looking forward! I use my P400 quite a lot, especially for testing. At some point I'll try to pry that GaN trick out of you. Maybe over a Ron Zacapa. ;)

Never did get into doing my own PCBs--I just nag the draftsman. ;)

Magdalen, formerly known as Beautiful Layout Hunchback, is now doing test and demo board layout for Cirrus Logic in Austin, and really likes it. So I did have some good influence after all.

Cheers

Phil Hobbs

OK. The totem pole is obvious; the gate drivers are not.

Some stuff is too hard for me to explain; I just know it when I see it. Even when someone is doing a board for me, like the big one now, I like to do some critical-path place and route. You only have to know how to use part of the tool for that.

ElderBrat, thednageek, got her PhD and her husband in Austin. She really liked it too. Austin I mean. She worked her way through school as a BMW motorcycle mechanic. Both fe-brats are scary.

afaik you can get a device id from FTDI and have it enumerate as "PB's magic dothat" if you want

though these days it might involve getting a drivers signed by MS

Since we own the serial command language, it's simpler to add a command to identify the device attached to a given serial port, but that's a headache for the application programmer, and hence relatively likely to result in midnight phone calls. (I hate midnight phone calls.)

Cheers

Phil Hobbs

onsdag den 27. maj 2020 kl. 18.48.45 UTC+2 skrev snipped-for-privacy@highlandsniptechnology.com:

I often use the standard FTDI cable with a 3.5mm jack, all that is needed on the board is a 3.5mm jack socket, having the ftdi chip right at the pc also seems to be more reliable on noisy stuff

We used to do that, but as you say, MS just made life more and more difficult. (FTDI try their best, and provide as much tools as they can for easing the pain.)

I can suggest three ways around all this, that we have used.

1. Dump windows when you can - if you use Linux, you can use udev settings to give you consistent port names based on serial numbers, order of insertion, physical place on the tree of hubs, or whatever you like. I have a hub glued to my workbench - any FTDI serial device connected to ports 1 to 4 turns up as /dev/ttySerial_hub1 .. 4.

1. Ignore Windows comms ports, and access the device directly using the FTDI DLL's, or pyftdi. This lets you figure out serial numbers or other information about the devices, as well as accessing them as serial ports (with pyserial) - thus you can know which device you are talking to.

2. Don't put the FTDI chip on the board - put a 6-pin header with the TX and RX as 3.3v (or 1.8v) TTL signals, then use a serial cable from FTDI. This is the smallest and cheapest on the board, and ideal for serial ports that you only occasionally need.

for that I usually used the ftdi cable with a 3.5mm jack

Windows cheerfully and randomly names COMn ports every powerup. I usually scan them all to see who's where. Our products have an ID query for model/version/serial number.

We sometimes get power from the USB cable too. We haven't had much trouble putting an FTDI chip and USB connector on our board, into a uP uart. We prefer the big ole B connector when we have room. You can mate it under a table in the dark, and it seems to be indestructable.

...AND on a different drive (crash insurance).

Sure, if you want to throw all your work away. and give up any chance of even a manual merge. Pretty clever feature.

You obviously don't know git well at all. Checking out an old commit doesn't lose any information. You can also check it out into a new branch, e.g.

git checkout -b backOne HEAD~1

which doesn't disturb the master branch even slightly.

Reverting a commit does throw stuff away, but normally it's only used to fix extreme breakage.

Cheers

Phil Hobbs

Take a peek at LTM4663 maybe.

If the user is not wanting to use a standard Windows or Linux box to control it then they may prefer not to deal with the USB standard - it is a mess. Not surprisingly many USB implementations are flaky and can't be relied upon to stay working without reboots for months or years.

Galvanic isolation of USB is more awkward than many other standards.

Also there are distance limitations for USB that can be inconvenient away from an office location. I have been inconvenienced by this one.

It is possible to jam the USB B connector into an Ethernet socket on the same board, rather than the USB socket, when doing so without looking. I know from experience. Never underestimate the ability of users to screw things up - especially if the user is the board's designer!

be relied upon to stay working without reboots for months or years.

The harvester sensors talk to a PLC and can be reflashed over the serial link, which is a pretty good trick. (Simon figured it out, not I.)

We like the FTDI parts and have used them several times--I'm a fan of things that Just Work. We also use printer-style B connectors when possible, for the same reason as the printers.

The SiPM/MPPC gizmo just pins out the UART using a USB ESD protector with 604-ohm resistors either side. If the one on the connector side arcs over in an ESD event, the TVS will clamp the transient and the other resistor will protect the MCU. (Hopefully.)

Cheers

Phil Hobbs

Interesting part, thanks. 40 mV p-p is super noisy for a front end or diode laser, though. The alumina end faces have epsilon ~ 10, so all that stuff just comes roaring in. A class-H is a bit more complicated but much much quieter.

Cheers

Phil Hobbs

Oh, that's the old Crown (I think) audio trick. Quiet and reasonably efficient but lots of parts.

Is one side of the peltier closer to the sensitive stuff? Maybe you could ground that side and drive the other from a switcher. If you only cool, the drive would be unipolar and easier to filter than a bridge.

Unipolar class-H isn't hard, I guess, if you have the supplies. There's probably a clever way to eliminate the diode drops.

I guess you could design a slew-rate-limited switcher, give up a little efficiency for quiet.