The Z566 middle board connects the 40-pin ribbon cable from the Pi400 to the half-pitch 20-pin ribbon cable to the DUT. We can load code, debug code, check clocks and power supplies, and snoop code execution on a scope.
John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics
No.
John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics
As usual you snipped the rest of what what I had to say about that. Your own instant start oscillators deal with the problem (to the extent it needs dealing with) so your original claim doesn't make a lot of sense. Your instant start oscillators don't use particularly high Q resonant circuits, and you could mess with the Q between timing intervals if you wanted to.
Your flat negation of my claim is rhetorically neat, but (as usual) dishonest.
Sounds good, but what about before it goes into production? If someone crosses the streams and sends 50,000V up the RPi GPIO, how many hours/ days/weeks/months work is lost?
"No" isn't high rhetoric.
Wandering back on topic, I'd imagine that an on-chip BAW would be hard to start and stop very fast. Not only is it itself an acoustic device, but it will spray vibrations all around the parent chip, and they will bounce everywhere. I should look at the patents when I get time.
HP once sold a delay generator that started and stopped a quartz crystal oscillator. When I was working with Berkeley Nucleonics, I met a guy who had been on the HP design team. He was still suffering from PTSD.
John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics
The coupling between the mechanical resonance and the wires is quite weak, so the mass motion persists for some time.
I’ve occasionally used the ringdown of a crystal to calibrate successive-detection DLVAs, which have a nearly logarithmic response.
IIRC the free oscillation of an 80-MHz overtone crystal decays about 1dB per millisecond, which is super convenient for that.
Cheers
Phil Hobbs
github is not an editor, it a cloud platform for the version control system "git"
git was made by Linus for developing the Linux kernel
Quite astonishing that one guy build the foundation for the software that runs almost all the internet now
We rarely destroy first article boards, but we always build 5 or 6 so we could just pull up another.
John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics
Whatever they do, it's temporary until we release the production code. Then it is released as a formal revision, initially rev A.
I don't want shippable software floating off-site in some cloud.
John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics
It is as close as you can get.
Why would it "spray vibrations around the parent chip"? And why would it matter? It is an acoustic device, and you'd want the vibrating area to be confined be well defined, so as not to include areas with less well-defined mechanical properties. And on-chips devices aren't great microphones.
And good luck with understanding the patents if you could get at them.
As you'd know, if you had been able to appreciate the low distortion sine wave oscillator thread, starting and stopping a high Q oscillator is a slow process. You need variable gain to do it, and variable gain tends to be nonlinear, and feeds in harmonics, which you don't want, so you are stuck with very little variation in your gain and correspondingly slow start-up. Turn off can be quicker, but high Q means a lot of stored energy and you don't want to break the oscillator.
John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics
Lots of low level ripple there. Are you asking for help on ways to measure it accurately?
I'm thinking about the software, not the hardware.
Your picture showed the prototype/dev hardware attached to a Raspberry Pi 400, with the implication that the software was being developed on the '400. That's fine, and is pretty much what I've been doing (not with a '400, just a plain headless RPi). But one f*up with a bare wire or scope probe could kill the '400, and if that's where the only recent copy of the software lives you'll be having a bad day.
My solution is to have the source on a network share, mounted on the RPi, which is then nominally stateless. That share is backed up etc., so if the RPi is toast it can be quickly replaced, nothing lost.
Not criticizing, just curious if you have a different workflow that I could consider, if you're prepare to share it...
I'm also considering virtualising the RPi development environment*, so it can be preserved with zero space.
Am 08.07.26 um 05:36 schrieb john larkin:
The ripple located at the "right" location just 5 ns before the start edge tells me that the oscillator does not come from nirvana to work on command. It could be done with a gate or transmission gate.
40 years ago I did such start/stop oscillators with cross coupled AD9500 for an ultrasonics phased array @ ~5 MHz and electronic beam steering. The AD FAE must have been the happiest man on earth.Gerhard
The AD9500 programmable delay generator chip wasn't all that good.
I'd used the On-Semiconductor MC10198 ECL monostable with an external DAC do the much same job earlier, and while the idea of getting both parts in the same package was seductive, the performances wasn't as good.
The idea is to have an oscillator that goes from gate input to clock output in 3 ns and then clocks with crystal accuracy and picoseconds of jitter for days after trigger. And stops clean in under two cycles.
Given a giant sampling scope, plugged into the AC line for ground, a couple of feet from the oscillator, a mile from Sutro Tower, a bit of noise is understandable.
John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics
We display this poster prominently:
My software guy likes to type on a PC and move the stuff to the PI400 to load and debug the 2040 based products. He uses github too, I think. I'll ask for details, or if you want I can get you in contact.
I don't code much myself lately; I prefer to draw than type. I picked the RP2040 and PI400 and planned the overall scheme and designed the Z566 board between the Pi400 and the DUT. It lets us check power supplies and clocks and scope code execution and stuff.
The products have a tiny pushbutton. A short push is reset. A
10-second push starts the 2040 up in boot mode, where it looks like a USB memory stick. A user can then drag and drop a single file to upgrade the Pi and FPGA code. That is SO easy.The RP2040 is awesome. They did everything right. The state machines are cool.
The Pi people just announced that it's OK to run the 2040 at 200 MHz. People were overclocking it past 400 anyhow.
John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics
We can all dream.
Being careless about your testing environment isn't a great excuse for unimpressive results.
We're designing a new one now, eight channels, with the RP2040 cpu and PoE and a cheap Efinix FPGA.
I'm eliminating the xo phase-lock and just letting the LC oscillator rip open-loop this time. So long-delay jitter will be mediocre. Lots of people want delays but don't need picoscecond jitter, is the idea.
I could remote the sampling head on an extension cable, but why bother? What matters is the performance of the final product.
The extension cables cost a lot more than the SD-series sampling heads now.
John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics
I'd have to have an application to get interested enough to think about it. Stroboscopic electron microscopes (in England) and exciting electron spin resonances (in the Netherlands) did get me involved but that was forty and thirty years ago respectively.
The implication is that the crystal oscillator took longer to get your LC oscillator properly phase locked than you like to admit.
Only a few of your customers noticed, so you still had a product you could sell.
And lots of customer have very specific performance requirements and you now know which ones you can keep happy.
You can buy second hand SD-series sampling heads, but new extension cable have to be made from scratch?
Have something to add? Share your thoughts — no account required.
Ask the community — no account required