@blinken@hachyderm.io
@blinken@hachyderm.io avatar

blinken

@blinken@hachyderm.io

SRE/electronics/infosec/data nerd, + paragliders.

Open hardware design & RF consulting: https://blinkenlight.co.uk

M7BLK

This profile is from a federated server and may be incomplete. Browse more on the original instance.

blinken, to animals
@blinken@hachyderm.io avatar

On a course in Oludeniz, Turkey. We were hanging out in the hotel lobby and one of the friendly and very pregnant hotel cats was cuddling next to us in the sun. All of a sudden, she was like MEEEEEEEEEEEEE and a kitten popped out, all over the nice hotel lounge. The bartender was not super impressed.

Congratulations to the new mum! We left her with some cat snacks and water.

A mother cat washes herself as she lies on a nice hotel couch with a just-born kitten in her lap.
Three fluffy, 6-hour-old kittens are suckling from a sleepy mum (wide photo)

blinken,
@blinken@hachyderm.io avatar

Also my clean black t-shirt is now covered in cat

lethalbit, to random
@lethalbit@chaos.social avatar

Something that probably only I would find useful, but would be nice in KiCad is the ability to assign a nested sheet a schematic symbol, and have the pins match up to the sheet pins.

Would be nice for a few things i'm bapping at.

blinken,
@blinken@hachyderm.io avatar

@lethalbit i agree it'd be nice if at least the exposed hierarchical labels (and their positions) were saved with the sub-sheet, which would make a sheet more symbol-like. It would be nice to make re-use/DRY easier. Currently I have a folder of sub-sheet modules I maintain, but the workflow feels a bit hacky in places.

blinken, to Electronics
@blinken@hachyderm.io avatar

Not to scale. But is it trustworthy? :S

jonny, to random
@jonny@neuromatch.social avatar

everyone i have figured it out, cats just have unchecked integer operations

blinken,
@blinken@hachyderm.io avatar

@jonny I must have the 32-bit model

jacqueline, to random
@jacqueline@chaos.social avatar

i like how all business-related software really wants to suck up to me and tell me how special and cool i am in its ad copy. it's very effective advertising on me.

blinken,
@blinken@hachyderm.io avatar

@jacqueline right, they have really weird and arbitrary limitations around tracking stock which i can only assume are because of terrible design decisions on the backend. But I still love 'em so much

glyph, to random
@glyph@mastodon.social avatar

psst. hardware hacking friends. come here. no, closer. just between you and me. soldering is fake, right? it's some kind of special effect that mechanical keyboard influencers use to juice the numbers on their tiktoks? nobody actually does this. like you're expecting me to believe that I can't run a unit test for a javascript function that adds two numbers together without ten git pushes and 200 lines of YAML but somehow you can attach a glob of glowing-hot metal to a 2mm target on the first try

blinken,
@blinken@hachyderm.io avatar

@InsertUser @glyph I think it's more correct to say that most engineering doesn't involve any unit testing

blinken, (edited ) to Electronics
@blinken@hachyderm.io avatar

While it looks like something from @NanoRaptor I promise you that this piece of modern medical equipment does indeed feature a mini-USB to double-USB-A cable for ... reasons that I'm sure made sense to the designers. I would love to meet them I have so many questions?

... there's even moulded tabs to make sure it goes in right way around!!!!!

I can't even

https://www.medshop.com.au/products/welch-allyn-gs777-diagnostic-set-range

blinken,
@blinken@hachyderm.io avatar

@maya_b @NanoRaptor I thought so too, but I believe that the double A end plugs into their proprietary wall-wart power supply.

So they control both sides of that connection (they have to for ISO60601 I suspect) and so they could pick literally any connector. Literally. Anything.

(the miniUSB end is permanently attached inside the equipment I think, someone sent me this so I'm not in front of it to confirm)

blinken, (edited )
@blinken@hachyderm.io avatar

update: the proprietary power brick included with the product (and containing the female side of the double-USB-A connector) is rated for a total output of 5V 0.5A.

blinken, to Electronics
@blinken@hachyderm.io avatar

Me: hey ChatGPT v4, show me your new v4 tech by generating a nice diagram explaining the the difference between SMA and RP-SMA connectors.

ChatGPT: Hold. My. Beer.

tubetime, to random
@tubetime@mastodon.social avatar

here goes nothing.

blinken,
@blinken@hachyderm.io avatar

@tubetime negative-DMA?

root42, to Electronics
@root42@chaos.social avatar

So the modern switching regulator replacements for the 7805 and LM323 produce around 300mV peak to peak ripple. The question is: how relevant is this? My drive runs just fine, but what I don’t know is what side effects this can cause?

blinken,
@blinken@hachyderm.io avatar

@azonenberg @whitequark @tubetime @philpem @root42 I got a ways - take a look at https://github.com/blinken/power-rail-probe + https://hachyderm.io/@blinken/111445303231432400

The probe works well and I use it day-to-day; the main issue before I 'release" it is a ~2dB dip in the frequency response around ~1MHz. I'd like to understand why that's happening. Opinions welcome!

Next, I plan to test the frequency response via a bias tee (given I lack a low-impedance AC source) to see if the dip is just because I'm testing with a 50ohm source.

blinken,
@blinken@hachyderm.io avatar

@azonenberg @whitequark @tubetime @philpem @root42 ...but if anyone would like one to test, I'd happily make & post you one for the cost of parts + shipping (London, UK). Just send me a DM.

blinken, to random
@blinken@hachyderm.io avatar

An old Google account was about to expire this morning, so I reset the password and logged in. It was created on 5th September 2004, and it's not been logged in to, ever - I guess I signed up for the invite waitlist and forgot about it.

It's a completely pristine time capsule of Google Privacy Policy updates.

The welcome email from the Gmail team 19 years ago is charming. Impressively, the links to ancient CGI help pages (eg, "http://gmail.google.com/support/bin/answer.py?answer=8301") still redirect to the right place.

A screenshot of an empty spam folder. No spam in 19 years.
First off, welcome. And thanks for agreeing to help us test Gmail. By now you probably know the key ways in which Gmail differs from traditional webmail services. Searching instead of filing. A free gigabyte of storage. Messages displayed in context as conversations. So what else is new? Gmail has many other special features that will become apparent as you use your account. To help you get started, we encourage you to visit our Help Center, there you can browse frequently asked questions, read our Getting Started guide, or contact the Gmail User Support Team. You'll also find information in the Help Center on such topics as: Importing your contacts from Yahoo! Mail, Outlook, and others to Gmail Using address auto-complete Setting up filters for incoming mail Using advanced search options You may also have noticed some text ads or related links to the right of this message. They're placed there in the same way that ads are placed alongside Google search results and, through our AdSense program, on content pages across the web. The matching of ads to content in your Gmail messages is performed entirely by computers; never by people. Because the ads and links are matched to information that is of interest to you, we hope you'll find them relevant and useful. We're working hard during our limited test to improve Gmail and make it the best webmail service around. Thanks for taking the plunge with us. We hope you'll enjoy Google's approach to email. Thanks, The Gmail Team

blinken, to random
@blinken@hachyderm.io avatar

Lest it be said I never finish a project, I finally installed the switch panel for my van this morning. It's only been three years.

This is a JLCPCB aluminium with some classy C&K switches. Some current-limited DC-DC converter modules from eBay (https://www.ebay.co.uk/itm/154400482539) provide the LED dimming for the "low" settings.

A stack of aluminium panels with lettering on a wooden desk.

blinken,
@blinken@hachyderm.io avatar

...and if you're wondering why there are five switches and six holes, I purchased SIX SWITCHES from Farnell and they sent FIVE SWITCHES AND AN INDUSTRIAL JOYSTICK. The sixth switch, it turns out, is permanently discontinued.

Look, I finish most of my projects. This one is totally on Farnell.

blinken, (edited ) to Electronics
@blinken@hachyderm.io avatar

Power rail probes are a handy accessory, because they let you apply a voltage offset before the signal hits the scope. This allows you to look at ripple on low-impedance, highish-voltage power rails.

Commercial versions are expensive (https://www.keysight.com/us/en/product/N7020A/power-rail-probe-2-ghz.html). It turns out they're not wildly complicated, so here's an version.

Naming things is difficult, so I called it 'power rail probe'. Lots still to do: feedback welcome.
https://github.com/blinken/power-rail-probe

blinken,
@blinken@hachyderm.io avatar

Haven't had much time to work on this, but the laser cut end panels arrived today & they look fantastic.

Some minor issues with fit and finish.

I couldn't work out why the light pipe wouldn't fit, but I got my full bodyweight behind it and it popped right in.

The other side of the metal enclosure, marked "remove cable before measurement" and featuring a USB-C connector and a power jack
The underside of a blue PCB with a plastic light pipe forcing one component off the board

blinken, (edited )
@blinken@hachyderm.io avatar

Alright, the power rail probe saga continues. Where did we leave off? I was stumped about the frequency response, which should have been flat but instead had an irritating ~3.7dB dip at about 1MHz. I was like, "oh maybe the problem was the potentiometer"

I replaced the potentiometer. The problem was not the potentiometer. The problem was thorny and involved my least favourite thing, differential equations, and caused me to spend months doubting my sanity.

A thread:

blinken,
@blinken@hachyderm.io avatar

I spent a lot of time trying to figure out the frequency response, and through
sheer trial and error ended up finding a configuration that was ... better by reducing the opamp gain (R3 -> 45k), and tuning the output resistors R1 & R7.

And by "finding a configuration", I don't mean simulation, it was days with a hot air gun and a VNA. But, the best it got was about -2.5dB, and I didn't fully understand why.

blinken,
@blinken@hachyderm.io avatar

I'd been thinking of the system as having a separate high- and
low-frequency paths - in theory, L1 filters high-frequency components, but of course it's more complicated. C1 forms a notch filter with L1, with a knee frequency around 800kHz.

The opamps (esp. U2) also add a frequency dependent response: they have bandwidth out past 20MHz. If we look at the SPICE simulation for a simplification of the circuit, there's a dip @ ~45MHz (possibly a zero in the transfer function).

A screenshot of a simplified schematic - a capacitor along the top path, and an opamp in inverting configuration along the bottom.

blinken,
@blinken@hachyderm.io avatar

(I am not going to compute the transfer function here, because, as I mentioned, I hate differential equations. My mechatronics degree was traumatic)

blinken,
@blinken@hachyderm.io avatar

The bottom line is, [hands waving] the opamps are not optimally stable in this configuration. What we need is to simplify things, and add some compensation to the design.

To begin with, let's drop the notch filter from the original design and replace with a simple LC high pass filter (with some damping resistors), tuned around 50kHz (marked in red). Let's then add some compensation to U2 to aggressively reduce the feedback frequency response (marked in blue).

blinken,
@blinken@hachyderm.io avatar

Component values were tuned using [hands waving] empirical methods and here's the result. U2 is not wildly happy about the amount of compensation in the
loop, but the overall result is actually pretty good: flat within 0.5dB, plus that area of oscillation at about 73kHz, but it's somewhat damped.

It's probably possible to improve on this, but I'm keen to respin the board and check the results of this in the real world. This yak has been well and truly shaved 🐃

blinken,
@blinken@hachyderm.io avatar

I mention yak shaving, because the v1.0 power rail probe is actually pretty handy tool - yes, it's got a -3.7dB dip @ 1.2MHz, but is otherwise within 1.3dB to 500MHz. Usually I'm looking for an order of magnitude on PSU ripple, for which it is extremely useful.

So, I wrote it up on GitHub: https://github.com/blinken/power-rail-probe

If you would like to buy one (it's <10% of the cost of a Tektronix TPR1000!), I have 8 of the original boards left, and it would help immensely with development!

blinken,
@blinken@hachyderm.io avatar

https://hachyderm.io/@blinken/111280654984096788

I'd also like to point out - my new proposal for RF debugging, smooshing your fingers all over the board - identified the problem, I was just too slow to figure it out.

D3 and D4 had the biggest effect on the frequency response when I touched nearby. I removed them, assuming the issue was the diodes themselves, but what I should have realised was that they're at the input to U2, which was critically damped/not stable.

Sadly, I wasn't paying attention.

  • All
  • Subscribed
  • Moderated
  • Favorites
  • •
  • megavids
  • thenastyranch
  • magazineikmin
  • everett
  • InstantRegret
  • rosin
  • Youngstown
  • slotface
  • love
  • khanakhh
  • kavyap
  • tacticalgear
  • GTA5RPClips
  • DreamBathrooms
  • provamag3
  • modclub
  • mdbf
  • normalnudes
  • Durango
  • ethstaker
  • osvaldo12
  • cubers
  • ngwrru68w68
  • tester
  • anitta
  • cisconetworking
  • Leos
  • JUstTest
  • All magazines
    •