The Mysterious 50 Ohm Impedance: Where It Came from and Why We Use It (2021) | Hacker News Reader

The Mysterious 50 Ohm Impedance: Where It Came from and Why We Use It (2021)

resources.altium.com

222 pointsamelius78 comments

Just a quick comment on the product, since this it effectively an ad for Altium.

They typically present themselves as the most popular solution, but they also very clearly out-price any hobbyist at $10k for a perpetual licence. Their "hobby" version CircuitStudio lacks critical features and has 0 support and 0 updates and the forums are just crickets. But KiCad is free, open source, and looks similar enough that I had a great time following the Altium tutorials by Rick Hartley with it.

In my opinion, that also invalidates their claim that Altium is the most commonly used PCB design software, because there likely are 100x more hobbyists using KiCad than people able to spend $10k on a hobby.

It seems to me like Altium is developing like Eagle. They used to court the makers and hobbyists and then greatly profited when those people started working. But now both of them are mostly in the business of milking companies who have existing data in their proprietary file format.

What has your experience with Altium been like?

It’s the best layout software I’ve used. Nevertheless, I’ve started preferring KiCAD for personal projects for the reasons mentioned above.

Kicad also did a GIGANTIC leap forward in the past two years.

Yep. Altium started content marketing like mad right around the time KiCad became good enough and got the glitches / crashing under control. KiCad version 6 was a huge jump in quality.

I don't mind the content marketing, I just wish they had a plan that I could rationalize paying for as a hobbyist and they simply don't.

It's hard to stay in business selling something for $0. Meanwhile, there are companies with more money than $Deity that look at $10K as a rounding error compared to their high-end engineer's salary. So Altium follows the money. As a huge open-source nerd, I used to think something like $10K for Altium was a borderline crime. Now I sell robots for the same amount. (And people in the industry still say I'm charging too little.) Long story long, if you think $10K is too much for something, you're not their target market. (And as long as there are viable open-source alternatives, I don't have a problem with that.)

Wait what kind of robotics? After ChatGPT, that's the next big thing.

It's in my bio, but I after my "internet fame" with Selenium and Appium, I now make physical robots that do the same thing: tap buttons on phone screens for end-to-end testing. (There are lots of little reasons why a physical robot would be preferable over an emulator or a USB-tethered device, but perhaps, that's another thread for another day...)

Oh I was hoping it was secret TidyBot competitor. I've always dreamed of building a phone screen tester, and then you could sell it as a cloud bare metal cellphone

$10k is kind of a crime for individuals. For businesses, absolutely, high prices for critical tools with support contracts is a no brainer. I just about coughed up a lung when I heard how much my company pays for 2 floating licenses of specialized engineering software (six figures per year). But in terms of ROI, it's easy to justify.

Your robot could easily sell at 20-30k. If you add some kind of calibration/support service to guarantee whatever operating parameters, you could charge a subscription on top of that. Test equipment is normally fairly expensive and specialized, and no one who runs tests in a commercial environment wants to be in a situation where they are their own checks/balances. So even though they could easily build/maintain test equipment themselves, companies will avoid it if possible. It's why businesses will buy the $5k name brand multimeters, test probes, etc instead of the $200 no name one with the same specs.

I used to work at a small company where the most expensive licence was $10k. One of our big customers basically asked us to make a more expensive licence so they could give us more money to try and make sure we didn't go away.

Right but small company making two-sided boards with a few dozen chips don't need the 80% of extra features it comes with.

With that kind of licensing it just makes possible future users go to competition.

I switched from Eagle to Altium for a few reasons

- Autodesk took my 3-years subscription money and then ditched Eagle, did no work, updates or improvements on the product I'd actually subscribed to, and tried to force me to move to a cloud-based app. No thanks. Never again, Autodesk.

- Altium is a world away from Eagle, with its push/shove routing, rules, field-solver for impedance matching, dynamic lookup for components, just, well, everything basically. It's a steeper learning curve than Eagle, but it's far and away just better as well.

- A perpetual license is what makes it worthwhile. Altium is already way more than I need, as a hobbyist, which gives me a lot of room to grow into it. I'd been using Eagle for ~20 years before switching, and I'd spent a lot on that over the years. If I get the same time-scale out of Altium that I did out of Eagle, it's not so bad averaged out.

Altium ended perpetual licenses in 2021. I have a perpetual license of Altium 21, but when my salesperson came back for Altium 22, they had stopped the perpetual licensing scheme. To put numbers to it, it used to be $5000 for a license and $2000/year for a support contract on that license ($7000 in the first year covering the license and the support), and it's now something like $3000 for an annual license with no perpetual option.

KiCad has been slowly getting toward feature parity with Altium 21, and I have had more fun with KiCad for hobby projects anyway. For those who tried it years ago, when I first tried KiCad in 2014, it was pretty crappy, but at some point between then and now they got their act in gear. The modern version is pretty close to a professional tool, and they are essentially taking feature ideas from Altium, so it's a pretty easy switch.

Can’t be true. I bought a perpetual license just over 12 months ago (as in 5 days over 12 months ago). It still works fine, and it says it’s perpetual.

Edit: and isn’t. They offer a perpetual license here [1]

[1]: https://www.altium.com/altium-designer/licensing

Oh, they brought it back and doubled the price. How lovely.

We managed to negotiate a couple of grand off (including waiving the first year of maintenance) when we needed another license a few weeks ago but yeah it's a bit scummy.

> and they are essentially taking feature ideas from Altium, so it's a pretty easy switch.

I wish this was more of an attitude in the software space, especially the open source software space.

You don’t need to completely clone something, but paid software often tests UI with focus groups, A/B testing etc; Might as well profit off of that and steal the good parts. An example of how it’s not done is Photoshop vs GIMP/Krita. Both programs have extreme NIH syndrome, where they go out of their way to not implement something in the way Photoshop has, leading to -at times a very- convoluted UI and UX.

https://www.gimp.org/docs/userfaq.html#i-dont-like-gimps-use...

One of the arguments I've heard against this is that it makes software feel like a "cheap clone". If most features are here, but there are a few missing or slightly different, you will resent their absence, or notice the difference, and conclude the clone is "incomplete" and inferior.

And indeed, it would be very hard to make a perfect clone, but it may not be desirable either. There may be better ways/tools to achieve the same results, and if you just look for the features you are used to, you'll never find them.

I haven't used Photoshop in years, but regularly use GIMP. It was not always intuitive, but now that I'm used to it I find the workflow quite effective.

Lmao, coming from gimp that's rich.

After the big UI change that put everything in one window instead of million small ones I don't really have problems with GIMP UI.

Since we are talking about cloning, a good example is the clone brush / stamp. In Photoshop it’s it’s own tool. In Krita (and perhaps GIMP?), it’s just another brush you have to dig for and find, and adjusting the clone settings is very annoying.

This is a common occurrence, where you will google for “where is X in Krita / GIMP”, where X is a very regular tool that for some opaque reason has been stuffed away in a very strange place. Photoshop stores their glassware in the cupboard? We better keep it in the fridge!

I hope that at some point Krita or GIMP or maybe even a unified Paint.net+Pinta gets a Blender moment, where a few bigger companies are tired of paying gobs of money to (mainly) Adobe and start to pay for development.

This is indeed a good example.

I think what you are looking for in GIMP is the Heal tool[1]. It sounds a bit like the xyproblem[2,3]: what you want is a way to erase part of the picture, not "the clone/stamp brush in GIMP". It may also be related to the "curse of knowledge" cognitive bias, but you have to unlearn some of what you already know (and doesn't apply here): having a different interface is supposed to help a bit here.

I understand why you'd just want "photoshop but free", but there are a lot of reasons why it would be complicated to obtain, if achievable, not to mention... desirable?

Regarding the heal tool, on my UI (2.10.34) I have to long-press the "clone" button, next to the eraser, to get it. Now that I mention it... the "clone" tool [4] may be what you are looking for? I tend to prefer the heal tool though. The "clone" tool was here since ~2004-2007 (timespan of the 2.2 version) according to the manual[5], which makes me question your memory, though.

Krita is more of a painting program than an image editor, so this tool is less needed. I am also less comfortable with it, but I found the "smart patch tool" [6] in under 30s of looking up "krita remove part image" in DuckDuckGo. I also found references to the clone brush, so I looked for it, that sounds more like what you describe[7].

Anyway, I don't want to attack you or anything, I would just like to point out that in general, when switching tools, it's better to look for the idiomatic way of achieving the same result, than a 1:1 equivalent of every step.

[1]: https://docs.gimp.org/en/gimp-tool-heal.html

[2]: https://en.wikipedia.org/wiki/XY_problem

[3]: https://xyproblem.info/

[4]: https://docs.gimp.org/en/gimp-tool-clone.html

[5]: https://docs.gimp.org/2.2/en/gimp-tool-clone.html

[6]: https://docs.krita.org/en/reference_manual/tools/smart_patch...

[7]: https://graphicdesign.stackexchange.com/a/124754

I got burned on “perpetual” licenses from Avid with Pro Tools. I hate licensing games these days

I wouldn't believe anything marketed as "perpetual" or "unlimited." The real limit is either written in the fine print, or you're just begging for the rug to be pulled out from under you later when they do change the fine print. These words are nonsense marketing-speak, and it's surprising people fall for them.

Among hobbyists, Altium is probably just as pirated as Photoshop.

I thought you couldn't pirate Photoshop anymore?

Kicad is getting better and better but in a few ways it is still basically a toy compared to the big boy packages.

That being said altium is pretty rotten internally it seems, just looks nice.

KiCad is getting better and better. At this point Altium is copying features _from_ KiCad, so you know the competition is real :).

Altium is still a Delphi-ish program, though with lots of things now just compiling to .Net. The core layout package has always been solid with a very small core developer group running it which is really more what its known for. The library management parts with collaboration are actually getting a lot better (ECAD<->MCAD, etc).

Altium gained tons of weird stuff with a super weird push to "FPGAs for everything", a giant off-shore (not Australian or Carlsbad) development team, and... all of that crash and burned and no actual customer actually understood what was happening. Much of that legacy is still in parts, but most of the features have been removed by now.

If you need it, the field solver for high-speed diff pairs in Altium is head and shoulders above the approximation techniques in the online tools - which never agree with each other anyway. Kicad is no better than the online ones.

The price of a standalone field solver (which would be more capable than Altium’s, for sure) is more than the price of Altium…

Given that an FPGA can output a lot of diff pairs at pretty high speeds these days, even us hobbyists can have a need for this sort of tech.

99% of the time, I’m routing CMOS logic, i2c, SPI, maybe a CAN network here or there. I have maybe 10 projects where I worked with anything so critical that it required the most precise diff pair length matching. I think SRAM was the only that actually was affected by lackadaisically ran diff pairs.

Even then, KiCAD has successfully guided me to routing a number of signals that are capable of data transmission well into the megahertz range.

Designed gigabit Ethernet board and one with FPGA+DDR3 using KiCad 5 and it was ok. Don’t see benefit in OrCAD or Altium here. Maybe for 60 GHz there might be some benefit. Bet that’s not my working frequency range.

Yeah. If you need it. Which in most times is resounding "well it would save me some time but not really, I can get by".

I always laugh at their ads/sponsorships on YouTube videos. I am sure Altium Designer is great, but there is no way you can afford it if you don't have significant income directly derived from the things that it saves time with.

The shit that other engineering fields go through to practice their craft is always depressing. If you're a mechanical engineer, half your salary is Solidworks. Electronics engineer, Altium. Software engineer? Everything is free. Occasionally someone shows up with a fancy IDE and wants $75 a year for it, and that's as high as it goes. Weird stuff.

Depends on how good the alternative are, $75 can easily be justified.

IntelliJ was far more productive than its peer in its first 5 years or so

back in 2004 I remember coming across "understand for c++", which cost something like $1000/seat but which was absolutely the best thing I have seen for navigating and working with a large inherited codebase.

Understand is still around, and probably still that expensive. I’m not aware of a better tool, but it has its weaknesses. Often we would make understand-only changes to help it “see” better, since we were taking diagrams straight from Understand and pasting them into our paperwork.

It's a professional tool, they're probably just looking at the professional market for the claim (i.e. it would be from sales data so Kicad isn't going to even show up)

For hobbiests, yes. In industry, for organizations with many engineers who need to collaborate together on many boards making up a complex system, Altium is really the only choice.

Also check out diptrace, I'm thinking of going with it for my next fee boards because it seems to have a very intuitive focus on its UI and also costs less than allium (they have a free tier), I just don't bees the power of Altium, but am tired if the archaic UI and workflows in eagle, kicad, etc.

Altium is far from the only choice for professionals, but if you have to ask, you can't afford them.

Kicad is so amazing. Extremely powerful tool with no imposed limits that is completely free and widely used.

I just finished the third rev of my open source low cost high power brushless motor controller in kicad. Open formats means Thea Flowers was able to make a beautiful web viewer that takes a GitHub link to a kicad board or schematic file and displays it for ya. You can see my new board in your browser right here:

https://kicanvas.org/?github=https://github.com/Twisted-Fiel...

I also recently designed a PCB for a friends art project where I needed to place a bunch of LEDs in concentric rings. I used a nice python library for kicad board files to generate a board with all the LEDs placed, then imported the rest of my components and continued with layout as normal.

But beyond the fact that it’s a high quality tool, and has easy to manipulate and read open formats, and loads of easy to use third party tools, is the fact that kicad is and always will be free and open source. That means that when I make my boards open source (and all of my work is open source), I know anyone anywhere in the world can download kicad and use the designs.

I used altium to design some boards at google. It’s clearly a powerful tool. If that’s already what you know, you don’t want to share your designs or collaborate with others who don’t have altium, and you have a license, then go ahead and use it. Or if you want to get a job at a large firm that likely uses altium, then go learn the tool. But if you want to learn a PCB tool for 99% of hobbyist and a good deal of professional work, kicad is likely a better choice.

Last time I got a quotation it was more than 10k USD per seat per unit time (1 year or 2 years, not sure).

I don't think they do perpetual licenses (anymore?).

TLDR: it's worse than 10k perpetual :p

> In my opinion, that also invalidates their claim that Altium is the most commonly used PCB design software, because there likely are 100x more hobbyists using KiCad than people able to spend $10k on a hobby.

I guess that depends whether you look at per-user or per-PCB basis

Altium is an enterprise level software that is just in a different league as kicad. Both are great products but it’s no sense comparing two very different products. I’ve never personally seen a company seriously use Kicad. I’m sure it’s possible but it’s not a tool you want your company to depend on. I find it hard to believe altiums ‘most commonly used’ claim. Maybe cadence just isn’t as worried about marketing? Haha.

The article is not that good, but whatever.

I am new to electronics myself and quite fascinated how has the art progressed in the past hundred or so years. There are people remembering "we were doing it wrong". Incredible.

Anyway, if you are interested in signals traveling inside those cute green multi-GHz boards you make your living off, go watch Rick Hartley.

https://www.youtube.com/watch?v=ySuUZEjARPY (How to Achieve Proper Grounding)

https://www.youtube.com/watch?v=QG0Apol-oj0 (What your Differential Pairs Wish You Knew)

Electronics has and likely always will feel like arcane magic to me

My favorite bit of "electronics magic":

https://www.youtube.com/watch?v=PD7DzTIFJdU

(this has more to do with TFA and impedance-in-general than might meet the eye at first glance)

Get this book:

https://www.opencircuitsbook.com/

It’s a book featuring macro photographs of cut-aways of electronic components. It’s the first book that helped me really think of electronic components as a physical things whose function followed from physical principles, rather than an arcane collection of various bits of black magic strung together.

Windell Oskay & Eric Schlaepfer's book has brilliant photographs and insightful text ... a coffee table book to delight any techie. They slice through connectors, semiconductors, and components ... showing the wonderful microscopic world of everyday electronics. A real joy!

I agree! I kinda know nothing about electronics since "my first electronic kit" as a 8 year old, and that book of cutaways shocked me as how many things in my computer aren't nanoscale magic!

It’s a beautiful demonstration of how increasing one parameter (frequency) can successively invalidate whatever model you’re using for the system, where negligible errors in the model eventually become functional circuit components.

frequency exposes the efficiency of transmission through a material.

That's why one of the best books on high speed digital electronics is titled High Speed Digital Design: A Handbook of Black Magic [1]

[1] https://www.amazon.com/High-Speed-Digital-Design-Handbook/dp...

Especially when you think about what electronics are made of: things mined from the earth. They made rocks compute for us!

Where else would they come from, the moon? It's to be expected that for large scale products like computers to be made from common materials.

High-speed digital and RF are more "magical" than a lot of other forms of electronics. It is quite a bit more approachable if you limit yourself to ~200 MHz signals at most, then you don't really have to worry about the RF properties of your circuits as long as you keep the wires pretty short.

I like Rick. At least they got that it was the original coax cables that drove it. Then once you have lots of measurement equipment and cables, it's hard to change (even SMA is 50ohm).

You can still get 75ohm BNC cables (close to 77) and 93ohms (and I think some 150/300ohm antenna cables?). Once you get over a few GHz all the inputs are balanced differential anyway and you have T networks to optimize power reflection.

It doesn't mention at all that free space impedance that matters a lot to actual aerial antenna design is 377ohms. I don't know enough about it, but assume they're using transformers to balance loads (about 300ohms?).

The impedance of a folded dipole is 300 Ohms.

And the impedance of a 1/2 wave dipole is 73.1 Ohms.

However there are many other types of antennas

see https://en.wikipedia.org/wiki/Dipole_antenna

Many antennas are designed for a given gain and polarization, and you get the impedance you get from the materials and geometry used. Almost all of them have a matching network at the entrance to the antenna, and usually one set up based on actual S-parameter measurements. If you're really fancy, you can get custom coax cables made to match your antenna at any arbitrary impedance, and then put a much better matching circuit in your rack instead of using one on the antenna.

High-speed digital signals haven't entirely settled on 50 ohm (100 ohm differential) impedance, though. SSTL, HSTL, and POD (used by different types of memory and high-speed DACs/ADCs) often use 35-40 ohm impedance instead of 50. It's also not uncommon to see drivers having lower impedance and receivers having higher impedance (eg having a 32 ohm driver and a 48 ohm receiver on a 40 ohm wire) because a little bit of reflection can actually help signal integrity here.

Long digital wires like twinax Ethernet are all 100 ohm differential signals, though, since reflection is not beneficial there at all.

All the medium speed stuff I see (not 64Gb Serdes or even Thunderbolt) is differential current drive/receive though they may be voltage limited on the output. Those still seem to be designed as 50ohm, but they have some fairly high resistance connection's. Even at 1Gb getting a clean eye requires lots of equalization and careful routing, if you want to go 10+ inches.

> usually one set up based on actual S-parameter measurements

My brother-in-law made this funny remark couple years back. He's into mechanical engineering. "Having a simulator is awesome, but sometimes it's just faster to let nature do the calculations for you."

It's funny how you can't even measure things at high frequencies without parasitics screwing everything up.

But people are coping: https://electronics.stackexchange.com/questions/560225/where...

> quite fascinated how has the art progressed in the past hundred or so years. There are people remembering "we were doing it wrong". Incredible

If you think about it, that's practically a given in any engineering field that stuffers from high iteration / experimentation cost.

300 ohms is also very common in applications which use https://en.wikipedia.org/wiki/Twin-lead

For even lower loss, ladder-line is available in 450 ohms.

Which is the only replaceman open-wire feeder at 600 ohms. They don’t make it any longer, you’re on your own to construct it or make-do with ladder line.

Yes you can definitely build your own ladder line. The insulators are available. But it's probably easier to use window line [0] instead.

[0] https://www.trueladderline.com/blog/window-line-open-wire-fe...

The most common impedance in actual practice for the sort of transmission line you have in mind is 75 ohms, not 50. Most of the cable TV network, and thus cable Internet, runs on this. That's why RG-6 and RG-11 grows on trees.

There is no big secret behind all of this. Transmission lines by definition have some impedance. Practitioners long ago adopted a handful of standard values that suit the common use cases given prevailing constraints, mostly driven by cost.

For things that transmit to the air (antenna feeds), 50Hz is the standard, not 75.

My radiotech teacher told me it (as in half-wave antenna impedance) was just like pi or c, I didn't believe him.

Turns out it's a result of a weird spacial integral. Go figure, I gave up radio.

Took me right back to my education and I never really thought about impedance values and why they were chosen.

We used to build 10base2 BNC terminators ourselves with empty connectors, a soldering iron, crimp tool and some hot glue using 47ohm resistors albeit with low tolerance values.

According to my boss they outperformed the industrially produced ones.

Back then, the network problems I was facing was either on the software/configuration side (wrong IPX/SPX network id) or simply faulty pc power supplies (I.e. getting shocked by mains voltage when touching ground, frequently happening after thunderstorms; also the terminators would be hot to the touch in such cases).

I wouldn't be surprised if this article got posted because of the video "Why doesn't a 75 Ohm cable measure 75 Ohms?" [0] which the recommendation algorithm has picked up some days ago.

[0] https://www.youtube.com/watch?v=a-8CaGOmWDU

I just recently started watching this guy "Why doesn't a 75 Ohm cable measure 75 Ohms?" https://www.youtube.com/watch?v=a-8CaGOmWDU

Great work to learn from.. thanks for this series.. please do more follow up on such tool layout of complete antenna design!! I would recommend some sort of time lapse capture of antenna design in background with timestamp to jump to the explanation..

Interesting, I always thought 50 Hz was just picked for the sake of it at some point in the distance past and then just continued as legacy.

The same way with mains voltage of 50Hz (Europe) or 60Hz (US). Higher frequencies would be much better for transformer efficency (though transmission radiation loss is higher too) and is also much safer for humans. For the former reason they use 400Hz on airplanes. But early mains generators could not reach that kind of frequency.

The article is about the 50 Ohm impedance of coaxial cables, not the 50 Hz frequency of the power grid in many countries.

Of some coaxial cables. Back in the 90s I just accepted every terminator was 50hz, and the only coax I really dealt with was ethernet. In the 00s I got into television and almost every terminator was 75Hz, with 50hz being specially marked.

I can't remember the last time I saw a terminator. I still deal with plenty of coax, but it's from one piece of active equipment to another and they self terminate.

(thinking about it I do see some specialised terminators on N type connectors on spectrum analysers, but I don't really do satellite so never deal with them)

I know that, I'm a HAM radio operator.

I was just using the 50 Hz as an example of something that was chosen pretty arbitrarily. I always thought the 50 Ohm was also chosen like that. I had no idea there were serious technical considerations in choosing it.

I should have written 50 Ohm, not Hz in the first sentence of my post though, that was a mixup. But I'm very aware of it.

There is some 1930s bell labs paper on coaxial cables that considered available materials and concluded 30 ohms was best for power handling, 77 ohms was best for low loss, and 50 ohms was selected as a manufacturable compromise. The publication I'm thinking of would have been by Espenschied most likely.

Edit: ah, I see the fine article also says this-- in any case the name might be helpful to someone looking up the citation.

Reply on news.ycombinator.com