July 8, 2025 • 21 mins
Unlocking the Power of Zephyr RTOS: Insights from Eli Hughes on Embedded Systems
This episode is a deep dive into the world of Zephyr with special guest Eli Hughes. Eli shares his journey from working with legacy systems to adopting Zephyr RTOS, detailing the challenges and triumphs he encountered along the way. Learn about the standout features of Zephyr, its capability to accelerate Industrial IoT projects, and the benefits of standardized development practices. Eli also shares valuable insights on how to smooth the transition to Zephyr and how the open-source community can aid in the journey.
Whether you're a software engineer, a developer, or someone curious about Embedded Systems, this episode is packed with valuable information that can help you leverage Zephyr for your next project.
Tune in to learn, grow, and get inspired!
Resources:
00:00 Introduction to the EdgeVerse Techcast
00:49 Meet Eli Hughes: Zephyr Expert
01:48 Zephyr Origin Story
02:52 First Impressions and Challenges with Zephyr
05:00 Navigating Zephyr's Learning Curve
11:12 Zephyr's Unique Features and Debugging Stories
15:59 The Power of the Zephyr Community
19:30 Future of Zephyr and Conclusion
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
(00:11):
Welcome back everyone to theEdgeVerse Techcast, where we explore
information that is vital foryour success in the frontiers of
Embedded Systems and Edge Computing.
Today we're heading back into thewild world of Real-time Operating
Systems and joining us as someonewho's lived it, Eli Hughes.
Whether you're migrating fromLegacy Code or just RTOS, curious.
(00:34):
Eli has been there, done thatand has probably debugged it too.
He's helped companies acrossthe industries adopt Zephyr, and
today he's here to share some hardearned lessons from the trenches.
Eli, welcome to the Techcast.
Can you kick us off with a quick intro?
Who you are?
What you do, and howZephyr entered your life?
(00:55):
Sure.
Well, it's good to be herewith you, Bridgette and Kyle.
So I have lots to talk about.
Just some background about me.
I kind of started life early on in theApple II era, like in middle school.
But you know, my first real exposureto Embedded and Microcontrollers
was the Motorola 68HC11, right?
(01:15):
Then went to Motorola DSP in whichmigrated into the Freescale years,
which I happened to be doing allwork with Phillips and NXP before
it became NXP as we know it today.
As well as I taught at PennState University: Audio and
Acoustics Microcontrollers.
Even participated in what'snow the NXP Cup with students.
(01:37):
Today I'm kind of a Zephyr fanboy, right?
And one of the things I do is I helpthrough the NXP Pro support program
to, kind of sell the dream as it were.
Well, we're so happy to have you Eli.
So I'm gonna kick us off here, justkind of the, the Zephyr origin story.
What is the backstory?
How did you find yourselfworking with Zephyr?
Was it love at first configurationor more of an arranged marriage?
(02:01):
Um, yeah, I, it was somethingthat I was totally random.
It wasn't planned.
I was working on a really cool project,kinda circa it started in 2017,
but the Zephyr piece came in 2019.
We were building this system to monitoringfermentation process for beer brewing.
(02:23):
And uh, I was engineering anultrasound probe to measure
yeast and bubble activity.
Well, we were building an IoT solutionand we were looking at a cellular
modem that kind of integrated amicrocontroller and the modem all in one.
And at the time, Zephyr was new.
But the only way it was supportedby, it was Nordic, was Zephyr.
So I kind of started when Zephyr wasfairly new and with a hard project.
(02:47):
It was a little rough, it took alittle time, but I became hooked.
All right.
So looking back on this firstexperience, what was the one
standout feature or golden arrowthat made you say, "Yep, this is it".
So, it came through thispath of like Industrial IoT.
So, you know, the project hadall these cool things about it.
(03:08):
Of course it was dealing with likebeer brewing, but in the chemistry.
Fundamentally as an industrial IOTproduct where we had a lot of Embedded
Networking and Secure Networking.
So we were doing a lot of work withlike Microsoft Azure and IoT Hub
and these security protocols, whichfor an embedded system can be tough.
And Zephyr was the first thing.
(03:30):
Once we got over the hump of learningit, that kind of just allowed us to
work in that world, and especiallyright on our PCs, to really dive
deep into cloud connectivity.
In that kind of golden arrow.
That was one piece.
The other piece came in was westarted, like with the Zephyr
piece in 2019, but everyone knowsin 2020 the world changed, right?
(03:51):
With COVID-19 and Supply Chainand everything was chaos.
Well, like every other startup,we were pivoting, right?
We had to change our productwith chips, we could get in
actually a different modem.
And that's where Zephyr, that piecereally came in that we could quickly
pivot to completely different platforms.
NXP being one of them.
Bringing in some NXP chips with havingthe port process be relatively simple.
(04:16):
And that for me waslike, "Yeah, this is it".
Okay, so before we talk about maybemigration pains or switching RTOSes,
did you have the chance to try the beer?
I didn't say this out loud.
So my partner's in crime,they were big into craft brew.
I, for the life of me, neverin my life actually liked beer.
(04:37):
I just never had the taste for it.
I went right to the whiskey, right?
So I, uh.
I got to see the process at brewery's,little, big and everything in
between, and it was just really neat.
So not a beer drinker, but they definitelyneeded you for the coding and the backend.
Yeah.
Well when we went into the distillers,you know, that was a different story.
(04:59):
Thanks for sharing that.
And you know, we know that switchingRTOSes rarely plug and play.
So what are some common bumpsyou've seen teams hit when moving
from legacy platforms to Zephyr?
I think just kind ofpersonalizing it is like.
When you go into it, you gottahave like really fresh eyes
(05:19):
and not make any assumptions.
So I made the mistake if I waslooking at it completely through the
lens of what I was used to 'cause Isaid, "Oh, it's embedded. It's like
every other embedded thing", right?
I had done a lot with BareMetal or FreeRTOS and things
with a lot less structure.
My first takeaway is like,"My God is this complicated".
(05:39):
Then I had a, a colleague, hesaid, Eli's like nothing in
this tech world is intuitive.
Like we don't, we're not bornwith intuition about RTOSes.
It's all about how we'reconditioned and what we know.
And so once I kind of picked up on someof the pieces and like there's kind of
five components of Zephyr and saw howthat they were kind of intertwined.
(06:00):
It got like really easy.
A lot of it came down to.
Some of the terminology.
It's kind of perceived thatZephyr is like Light Linux.
Which it's really not.
It's a Linux Foundation project and itborrows concepts, but at the end of the
day, you're doing the same thing, right?
You have a bunch of source code, you'rebuilding it, and you get this binary.
It's just that Zephyr has standardizationaround the front end of how you configure
(06:24):
things with KConfig and Devicetree.
But at the end of the day, it'sjust C files that it compiles.
Kind of the other piece, and I thinkthis hit some teams a little harder,
is that Zephyr is structured aroundlike modern Software Engineering
and architecture practices, whichis a little different than a lot of
embedded projects where it might be oneperson and files on their hard drive.
(06:47):
It encourages to useGit and version control.
And some of these softwarepractices you do see in other
realms outside of Embedded.
That can cause a little friction.
But once you're in and understandit you can see the benefit,
especially with distributed teams.
And for those navigating the same path.
What resources or tactics helpsmooth out the rough spots?
(07:11):
Well, I'm glad you asked that.
So there's one thing and it'swhat helped me get around
some pieces I was stuck with.
I realized when I was stuck onsomething, it wasn't Zephyr's fault.
I just was asking the wrong question.
I didn't pose my questioncorrectly to get a good answer.
And so one of the biggest reframeswhen you're looking for help.
(07:33):
Is how to reframe a question, and I startthis when I help train other people.
A lot of embedded whenyou're working with a chip.
Like say you're, you're just getting upto speed with the i.MX RT, like the 1064,
which is a really popular Crossover part.
You might ask a question like,"Well, how do I make ethernet work
on the RT1064 with Zephyr?" Well, andthat's actually the wrong question.
(07:57):
The reframe is simple.
"How do I make ethernet work inZephyr?" The RT1064 part comes
later, or actually never, you mightnot even have to deal with it the
way Zephyr deals with hardware.
Once I reframed it, of flipping itaround or just "How do I do this
in Zephyr?" You can look at samplesfor completely different parts and
(08:18):
get your solution because thingsare structured exactly the same way.
Now just as a future hint to yourlisteners, maybe get excited some of
these lessons learned we're puttingtogether with some training so we
can help accelerate you and also withan eye more on structure for teams.
Right?
How does a professional team get upand running with Zephyr, and what are
the shortcuts to get you going fast?
(08:41):
Yeah.
We'll have to keep our listeners up todate when you have that training ready?
Before I ask this nextquestion, you talking about
the team structure of Zephyr.
I have seen a lot of feedback fromcustomers that it's coming from the
top management for the software teams.
They are saying we need to run our companyin a more structured software manner.
(09:05):
And to your point, they lookat Linux and other open-source
software, and they say we needsomething a little bit more scalable.
There's not these gaps that are providedby suppliers, or you get something
here and you put it together with this.
You are engaging in something that wasbuilt to scale with large software teams.
(09:27):
You mentioned Git.
The tools of Zephyr based on West.
From the origin, Zephyr's beenbuilt to work with large software
teams with complex software.
Versus a lot of the existing developmenthas been coming from the other direction.
It's more Bare Metal, moreof a hardware environment.
So it's really great to seeZephyr updating the way embedded
(09:49):
software's being tackled.
Yeah, that's an excellent point.
And just to kind of add tothat, and it's something when I
show people where to find help.
I mentioned, Zephyr, is likefive pieces and one of those
pieces is the Test Framework.
Some of the objections of "Well, it'sopen-source, anyone can do whatever".
When you look at the model of howit works, no one can just go commit
(10:11):
something into main without a process,as well as automated tests that are
running like 24- 7 on everything.
Yeah.
So it's structured well, and I actuallyshow people, "Hey, if you wanna know how
the SD card interface works. Here's howit's actually tested in Zephyr and you
can run those." And that's actually thebest way to discover how to do new things
(10:34):
because it's literally the way it's testedacross thousands of platforms and stuff.
And that's something that NXP adopted.
Eli, you may be familiar.
We redid the architecture of the SDK.
We have the head of thetest branch of Zephyr.
He is the Manager of our testenvironment and they've adopted some
of these automated test requirementsfor submissions of code changing.
(10:58):
That allows our SDK to become more robust.
So yeah, Zephyr's making some hugeimprovements for the way people
develop software and deliver software.
All right.
Well, let me jump into the next question.
Sorry for the tangent there.
Let's talk about some of the cool things.
Why don't you talk about a feature thatyou found in Zephyr that made a big
difference that you think other softwareengineers should be notified about?
(11:22):
Yeah, and what I'm gonna talkabout, it's kind of speaks to not
only this distributed nature, but.
This idea that sometimes you don't havethe hardware before you get started.
You might have an EVK, but you don'thave your actual target hardware.
One of the things Zephyr can do isyou can take your Zephyr project and
there's all kinds of boards you cantarget, like the RT1064 EVK, but there's
(11:44):
boards for what's called POSIX Native.
And what that allows you to dois you can build your Zephyr code
so it runs on a PC under Linux.
And so I show this when I train people.
A great, way to use thatis, let's say you're doing a
product with an embedded GUI.
Well, when you run Zephyr on POSIXnative, you get a simulated LCD.
(12:06):
It's the same code you would run inyour target, but you're little LCD
window just pops up on your monitor.
So you could try differentsize and resolutions.
The person may be doing LVGLcan be developing on day one
in Zephyr code running with thePOSIX Native target, right away.
And it's not like they lose any workwhen they get their real hardware.
(12:27):
It's still the same Zephyr code.
But this test cycle of just (12:29):
Editing, Testing, Debugging is really quick.
I like the graphics example'cause it's easy to understand.
But that same port also simulatesSerial Ports and CAN Bus that
get reflected through your PC.
One I used is CAN Bus.
If I had a CAN adapterhooked up to my machine.
(12:51):
I write my Zephyr code.
I run it in Linux.
That CAN message comes rightout of my USB CAN adapter
'cause there was a Linux driver.
It allows you to develop yet anothertest or development framework to
work before you even have hardware.
Or maybe you haven't evenchosen your hardware yet.
I wish everyone knew aboutit 'cause it's just so neat.
Hmm.
That's why we have you here.
Thanks for sharing the insights andI know we're picking your brain.
(13:14):
So I wanna transition to debugging.
Is there any memorable in the trenchesstories you can share where Zephyr either
saved the day or made things interesting?
Well.
I have this, I don't know, youcould classify this either as a
superpower or a mental illness.
But, uh, when I wanna understand somethingI say, what if I didn't have anything?
(13:37):
Meaning no dev boards, nothing.
Right?
So I was starting to usethe i.MX RT 685, circa 2021.
It's a great Crossoverpart for Embedded Audio.
So I said I'm gonna start from scratch,meaning I'm gonna do a board design
everything and bring up a new Zephyrboard port from scratch with the RT685.
(13:59):
End-to-end, just to see the whole process.
I wrote about it and it's actuallyon the Linux Foundation website.
I think we published that in 2022.
So it's a few versions old, butit's the same fundamental material.
Now the interesting part is that like anybig project, Zephyr moves and evolves.
The APIs from three years ago aremostly the same, but things improve.
(14:24):
Zephyr makes it really easy onceyou're plugged in to switch versions.
Don't just switch a version'cause you see a new one's out.
They're well tested and everything, but itcan also introduce some friction as well.
'cause you're working with thisthing that's always moving, evolving.
The good news is Zephyr makesit easy to test and revert once
you're plugged into the tool.
(14:45):
All right.
Excellent.
So maybe don't bite off moredebug than required by upgrading
to an unnecessary version.
Or don't do.
What I do.
I'm on a project that we're gettingtowards the end of it, and I say, "Hey,
the bleeding edge in main, the latestthings out. Let me just use the latest
bleeding edge." when 4.1's released.
(15:06):
That's good to do on a test branch.
Right.
Don't necessarily jump in.
But the good news, it's allopen and you can see it, right?
Yep.
Yep.
And Zephyr does have long-term support.
They have an LTS branch, whichwould even be more secure and
stable, for those super cautious.
I think an important point here too.
And this is something thatI've had to help people with.
Just because you see Zephyr,you can go to Zephyr's GitHub.
(15:30):
If you're working in like evenan air-gapped environment, like
the way Zephyr is structured,everything can be completely
internal to your organization, right?
You can have your own copy of it at aspecific version, on your servers, in
your network that is 100% of your control.
It's not like you haveto go out to the web.
It supports all the differenttypes of workflows, which is
(15:53):
just a superpower in my opinion.
Super, super flexible, which bringsus to the next comment I have.
Eli, more and more companies just thismonth a few other large semiconductor
companies jumped on board with Zephyr.
Most of it is for this new technologythat you're describing, but I
think some of it is the good oldFear of Missing Out, or the FoMO.
(16:14):
So one of the otherstrengths is the community.
This open-source and allthe contributing members.
How has this Zephyr ecosystem helped you?
Do you think that this community withall of these contributors is what
is going to carry Zephyr into thefuture and really differentiate it
from some of the other alternatives?
(16:35):
I do, and I think for me.
What's really helped me, aside fromsome of these other cool features.
There is the boring part, right?
The boring part is ZephyrStandardizes how you build a project.
When you type "west build" and how youstructure these Lego blocks together.
It's the same for any part.
Because that's now standardized.
It allows a lot of peopleto be involved, right?
(16:59):
Even though that's a boring aspect,it means you can bring new people
into a team, into your company, andbring them up to speed with something
that has a standard framework.
One example is I was really excited.
There's a fairly recent part, the MCXA156.
It's just a really neat general purposemicrocontroller I mean, I had the board
and in five minutes of plugging it in.
(17:20):
No manual.
I could type "west build", and I was upand running on the FRDM Board doing some
sophisticated things with other piecesI had because of the standardization.
The other side of it, like youmentioned, because it is open
source with this structure.
I go to GitHub every week just to pokearound to see who's working on what.
What issues are there?
(17:41):
Sometimes your answers, if you're havinga problem just search the issue tracker
and someone solved it and it's toosolved two years ago, or at least someone
showed someone else how to do something.
Just being able to do thatand learn how to poke around,
it really accelerates a team.
Yeah and one other exciting thing,on the same topic, is the Zephyr
Development Summit, the ZDS, correct?
(18:04):
Mm-hmm.
Just the engagement withall the community members.
It's like a big team, right?
Even though they may wear a differentjersey, it's like everybody's working
towards the same goal, which is different.
It's just feels different.
You feel like you're gettingassistance from someone who
may have an ulterior goal.
Something that they're workingon, but they're in parallel, and
(18:26):
so you're assisting each other.
And that's I think a difference.
So, one thing I've seen as wellthat I think speaks to that.
It allows people to come intothe fold who might not have
otherwise been embedded developers.
There are a lot of people who are reallygood with code that is very much embedded.
(18:46):
So think about someone who's writingc code for a server backend in Linux.
They've never playedwith a microcontroller.
But Zephyr is like at this bridgebetween structured development
to your classic embedded.
That those people cancome from that world.
And you also have people who areused to Bare Metal microcontrollers
(19:06):
who now can bridge to the other sideinto these higher end processors,
like the RT family is a good example.
I think to your point, yeah, it'sbringing more people into embedded
that might not have been there.
Which is awesome.
It's gonna be exciting.
I think we're gonna have alot more episodes on Zephyr.
And I think, we're gonna have you back.
Correct?
When the training comes out oryou'll have some other topics that
(19:28):
we can dive into, that'll be great.
So one thing I'll leave withthat I'm really excited about
and NXP is really pushing.
You can see this just inthe board's directory.
We think about Zephyr asRTOS for microcontrollers.
Well, there's now support coming outfrom the NXP side who are running
Zephyr on i.MX Applications Processors.
(19:49):
Say like the i.MX 8 or the 93 whereyou're not even running Linux, you're
loading Zephyr onto this 2 GHz MPU,which is like totally mind blowing.
It's kind of a new area that we'reseeing customers adopting this.
It's certainly excitingtimes for the future.
All right, we'll load that upBridgette, on the next topic.
(20:09):
Thank you so much for joiningand sharing your journey today.
It's clear from our conversation thatZephyr has a lot to offer our listeners,
but like anything new and powerful,it comes down to a learning curve.
I agree.
Eli, this has been an absolute treat.
I'm so happy you joined usand are willing to come back.
You've given our listeners a greatunderstanding of why they should consider
(20:31):
following your path towards Zephyr.
All right, well it was nice having aconversation and I'll certainly be back.
And to our listeners, don't beafraid to get your hands dirty.
The trenches are where the magic happens.
You know Bridgette, sometimes when youget down into the trenches and you look
around, you might just realize thatit's a place where people are coming
together to work closely on a common goal.
(20:54):
In this case, that's theadvancement of Embedded Software.
So once again, thanks forstreaming this EdgeVerse Techcast.
Bridgette and I truly appreciateyou listening and until next
time, keep exploring at the Edge.
Welcome back everyone to theEdgeVerse Techcast, where we explore
information that is vital foryour success in the frontiers of
Embedded Systems and Edge Computing.
Today we're heading back into thewild world of Real-time Operating
Systems and joining us as someonewho's lived it, Eli Hughes.
Whether you're migrating fromLegacy Code or just RTOS, curious.
(00:34):
Eli has been there, done thatand has probably debugged it too.
He's helped companies acrossthe industries adopt Zephyr, and
today he's here to share some hardearned lessons from the trenches.
Eli, welcome to the Techcast.
Can you kick us off with a quick intro?
Who you are?
What you do, and howZephyr entered your life?
(00:55):
Sure.
Well, it's good to be herewith you, Bridgette and Kyle.
So I have lots to talk about.
Just some background about me.
I kind of started life early on in theApple II era, like in middle school.
But you know, my first real exposureto Embedded and Microcontrollers
was the Motorola 68HC11, right?
(01:15):
Then went to Motorola DSP in whichmigrated into the Freescale years,
which I happened to be doing allwork with Phillips and NXP before
it became NXP as we know it today.
As well as I taught at PennState University: Audio and
Acoustics Microcontrollers.
Even participated in what'snow the NXP Cup with students.
(01:37):
Today I'm kind of a Zephyr fanboy, right?
And one of the things I do is I helpthrough the NXP Pro support program
to, kind of sell the dream as it were.
Well, we're so happy to have you Eli.
So I'm gonna kick us off here, justkind of the, the Zephyr origin story.
What is the backstory?
How did you find yourselfworking with Zephyr?
Was it love at first configurationor more of an arranged marriage?
(02:01):
Um, yeah, I, it was somethingthat I was totally random.
It wasn't planned.
I was working on a really cool project,kinda circa it started in 2017,
but the Zephyr piece came in 2019.
We were building this system to monitoringfermentation process for beer brewing.
(02:23):
And uh, I was engineering anultrasound probe to measure
yeast and bubble activity.
Well, we were building an IoT solutionand we were looking at a cellular
modem that kind of integrated amicrocontroller and the modem all in one.
And at the time, Zephyr was new.
But the only way it was supportedby, it was Nordic, was Zephyr.
So I kind of started when Zephyr wasfairly new and with a hard project.
(02:47):
It was a little rough, it took alittle time, but I became hooked.
All right.
So looking back on this firstexperience, what was the one
standout feature or golden arrowthat made you say, "Yep, this is it".
So, it came through thispath of like Industrial IoT.
So, you know, the project hadall these cool things about it.
(03:08):
Of course it was dealing with likebeer brewing, but in the chemistry.
Fundamentally as an industrial IOTproduct where we had a lot of Embedded
Networking and Secure Networking.
So we were doing a lot of work withlike Microsoft Azure and IoT Hub
and these security protocols, whichfor an embedded system can be tough.
And Zephyr was the first thing.
(03:30):
Once we got over the hump of learningit, that kind of just allowed us to
work in that world, and especiallyright on our PCs, to really dive
deep into cloud connectivity.
In that kind of golden arrow.
That was one piece.
The other piece came in was westarted, like with the Zephyr
piece in 2019, but everyone knowsin 2020 the world changed, right?
(03:51):
With COVID-19 and Supply Chainand everything was chaos.
Well, like every other startup,we were pivoting, right?
We had to change our productwith chips, we could get in
actually a different modem.
And that's where Zephyr, that piecereally came in that we could quickly
pivot to completely different platforms.
NXP being one of them.
Bringing in some NXP chips with havingthe port process be relatively simple.
(04:16):
And that for me waslike, "Yeah, this is it".
Okay, so before we talk about maybemigration pains or switching RTOSes,
did you have the chance to try the beer?
I didn't say this out loud.
So my partner's in crime,they were big into craft brew.
I, for the life of me, neverin my life actually liked beer.
(04:37):
I just never had the taste for it.
I went right to the whiskey, right?
So I, uh.
I got to see the process at brewery's,little, big and everything in
between, and it was just really neat.
So not a beer drinker, but they definitelyneeded you for the coding and the backend.
Yeah.
Well when we went into the distillers,you know, that was a different story.
(04:59):
Thanks for sharing that.
And you know, we know that switchingRTOSes rarely plug and play.
So what are some common bumpsyou've seen teams hit when moving
from legacy platforms to Zephyr?
I think just kind ofpersonalizing it is like.
When you go into it, you gottahave like really fresh eyes
(05:19):
and not make any assumptions.
So I made the mistake if I waslooking at it completely through the
lens of what I was used to 'cause Isaid, "Oh, it's embedded. It's like
every other embedded thing", right?
I had done a lot with BareMetal or FreeRTOS and things
with a lot less structure.
My first takeaway is like,"My God is this complicated".
(05:39):
Then I had a, a colleague, hesaid, Eli's like nothing in
this tech world is intuitive.
Like we don't, we're not bornwith intuition about RTOSes.
It's all about how we'reconditioned and what we know.
And so once I kind of picked up on someof the pieces and like there's kind of
five components of Zephyr and saw howthat they were kind of intertwined.
(06:00):
It got like really easy.
A lot of it came down to.
Some of the terminology.
It's kind of perceived thatZephyr is like Light Linux.
Which it's really not.
It's a Linux Foundation project and itborrows concepts, but at the end of the
day, you're doing the same thing, right?
You have a bunch of source code, you'rebuilding it, and you get this binary.
It's just that Zephyr has standardizationaround the front end of how you configure
(06:24):
things with KConfig and Devicetree.
But at the end of the day, it'sjust C files that it compiles.
Kind of the other piece, and I thinkthis hit some teams a little harder,
is that Zephyr is structured aroundlike modern Software Engineering
and architecture practices, whichis a little different than a lot of
embedded projects where it might be oneperson and files on their hard drive.
(06:47):
It encourages to useGit and version control.
And some of these softwarepractices you do see in other
realms outside of Embedded.
That can cause a little friction.
But once you're in and understandit you can see the benefit,
especially with distributed teams.
And for those navigating the same path.
What resources or tactics helpsmooth out the rough spots?
(07:11):
Well, I'm glad you asked that.
So there's one thing and it'swhat helped me get around
some pieces I was stuck with.
I realized when I was stuck onsomething, it wasn't Zephyr's fault.
I just was asking the wrong question.
I didn't pose my questioncorrectly to get a good answer.
And so one of the biggest reframeswhen you're looking for help.
(07:33):
Is how to reframe a question, and I startthis when I help train other people.
A lot of embedded whenyou're working with a chip.
Like say you're, you're just getting upto speed with the i.MX RT, like the 1064,
which is a really popular Crossover part.
You might ask a question like,"Well, how do I make ethernet work
on the RT1064 with Zephyr?" Well, andthat's actually the wrong question.
(07:57):
The reframe is simple.
"How do I make ethernet work inZephyr?" The RT1064 part comes
later, or actually never, you mightnot even have to deal with it the
way Zephyr deals with hardware.
Once I reframed it, of flipping itaround or just "How do I do this
in Zephyr?" You can look at samplesfor completely different parts and
(08:18):
get your solution because thingsare structured exactly the same way.
Now just as a future hint to yourlisteners, maybe get excited some of
these lessons learned we're puttingtogether with some training so we
can help accelerate you and also withan eye more on structure for teams.
Right?
How does a professional team get upand running with Zephyr, and what are
the shortcuts to get you going fast?
(08:41):
Yeah.
We'll have to keep our listeners up todate when you have that training ready?
Before I ask this nextquestion, you talking about
the team structure of Zephyr.
I have seen a lot of feedback fromcustomers that it's coming from the
top management for the software teams.
They are saying we need to run our companyin a more structured software manner.
(09:05):
And to your point, they lookat Linux and other open-source
software, and they say we needsomething a little bit more scalable.
There's not these gaps that are providedby suppliers, or you get something
here and you put it together with this.
You are engaging in something that wasbuilt to scale with large software teams.
(09:27):
You mentioned Git.
The tools of Zephyr based on West.
From the origin, Zephyr's beenbuilt to work with large software
teams with complex software.
Versus a lot of the existing developmenthas been coming from the other direction.
It's more Bare Metal, moreof a hardware environment.
So it's really great to seeZephyr updating the way embedded
(09:49):
software's being tackled.
Yeah, that's an excellent point.
And just to kind of add tothat, and it's something when I
show people where to find help.
I mentioned, Zephyr, is likefive pieces and one of those
pieces is the Test Framework.
Some of the objections of "Well, it'sopen-source, anyone can do whatever".
When you look at the model of howit works, no one can just go commit
(10:11):
something into main without a process,as well as automated tests that are
running like 24- 7 on everything.
Yeah.
So it's structured well, and I actuallyshow people, "Hey, if you wanna know how
the SD card interface works. Here's howit's actually tested in Zephyr and you
can run those." And that's actually thebest way to discover how to do new things
(10:34):
because it's literally the way it's testedacross thousands of platforms and stuff.
And that's something that NXP adopted.
Eli, you may be familiar.
We redid the architecture of the SDK.
We have the head of thetest branch of Zephyr.
He is the Manager of our testenvironment and they've adopted some
of these automated test requirementsfor submissions of code changing.
(10:58):
That allows our SDK to become more robust.
So yeah, Zephyr's making some hugeimprovements for the way people
develop software and deliver software.
All right.
Well, let me jump into the next question.
Sorry for the tangent there.
Let's talk about some of the cool things.
Why don't you talk about a feature thatyou found in Zephyr that made a big
difference that you think other softwareengineers should be notified about?
(11:22):
Yeah, and what I'm gonna talkabout, it's kind of speaks to not
only this distributed nature, but.
This idea that sometimes you don't havethe hardware before you get started.
You might have an EVK, but you don'thave your actual target hardware.
One of the things Zephyr can do isyou can take your Zephyr project and
there's all kinds of boards you cantarget, like the RT1064 EVK, but there's
(11:44):
boards for what's called POSIX Native.
And what that allows you to dois you can build your Zephyr code
so it runs on a PC under Linux.
And so I show this when I train people.
A great, way to use thatis, let's say you're doing a
product with an embedded GUI.
Well, when you run Zephyr on POSIXnative, you get a simulated LCD.
(12:06):
It's the same code you would run inyour target, but you're little LCD
window just pops up on your monitor.
So you could try differentsize and resolutions.
The person may be doing LVGLcan be developing on day one
in Zephyr code running with thePOSIX Native target, right away.
And it's not like they lose any workwhen they get their real hardware.
(12:27):
It's still the same Zephyr code.
But this test cycle of just (12:29):
Editing, Testing, Debugging is really quick.
I like the graphics example'cause it's easy to understand.
But that same port also simulatesSerial Ports and CAN Bus that
get reflected through your PC.
One I used is CAN Bus.
If I had a CAN adapterhooked up to my machine.
(12:51):
I write my Zephyr code.
I run it in Linux.
That CAN message comes rightout of my USB CAN adapter
'cause there was a Linux driver.
It allows you to develop yet anothertest or development framework to
work before you even have hardware.
Or maybe you haven't evenchosen your hardware yet.
I wish everyone knew aboutit 'cause it's just so neat.
Hmm.
That's why we have you here.
Thanks for sharing the insights andI know we're picking your brain.
(13:14):
So I wanna transition to debugging.
Is there any memorable in the trenchesstories you can share where Zephyr either
saved the day or made things interesting?
Well.
I have this, I don't know, youcould classify this either as a
superpower or a mental illness.
But, uh, when I wanna understand somethingI say, what if I didn't have anything?
(13:37):
Meaning no dev boards, nothing.
Right?
So I was starting to usethe i.MX RT 685, circa 2021.
It's a great Crossoverpart for Embedded Audio.
So I said I'm gonna start from scratch,meaning I'm gonna do a board design
everything and bring up a new Zephyrboard port from scratch with the RT685.
(13:59):
End-to-end, just to see the whole process.
I wrote about it and it's actuallyon the Linux Foundation website.
I think we published that in 2022.
So it's a few versions old, butit's the same fundamental material.
Now the interesting part is that like anybig project, Zephyr moves and evolves.
The APIs from three years ago aremostly the same, but things improve.
(14:24):
Zephyr makes it really easy onceyou're plugged in to switch versions.
Don't just switch a version'cause you see a new one's out.
They're well tested and everything, but itcan also introduce some friction as well.
'cause you're working with thisthing that's always moving, evolving.
The good news is Zephyr makesit easy to test and revert once
you're plugged into the tool.
(14:45):
All right.
Excellent.
So maybe don't bite off moredebug than required by upgrading
to an unnecessary version.
Or don't do.
What I do.
I'm on a project that we're gettingtowards the end of it, and I say, "Hey,
the bleeding edge in main, the latestthings out. Let me just use the latest
bleeding edge." when 4.1's released.
(15:06):
That's good to do on a test branch.
Right.
Don't necessarily jump in.
But the good news, it's allopen and you can see it, right?
Yep.
Yep.
And Zephyr does have long-term support.
They have an LTS branch, whichwould even be more secure and
stable, for those super cautious.
I think an important point here too.
And this is something thatI've had to help people with.
Just because you see Zephyr,you can go to Zephyr's GitHub.
(15:30):
If you're working in like evenan air-gapped environment, like
the way Zephyr is structured,everything can be completely
internal to your organization, right?
You can have your own copy of it at aspecific version, on your servers, in
your network that is 100% of your control.
It's not like you haveto go out to the web.
It supports all the differenttypes of workflows, which is
(15:53):
just a superpower in my opinion.
Super, super flexible, which bringsus to the next comment I have.
Eli, more and more companies just thismonth a few other large semiconductor
companies jumped on board with Zephyr.
Most of it is for this new technologythat you're describing, but I
think some of it is the good oldFear of Missing Out, or the FoMO.
(16:14):
So one of the otherstrengths is the community.
This open-source and allthe contributing members.
How has this Zephyr ecosystem helped you?
Do you think that this community withall of these contributors is what
is going to carry Zephyr into thefuture and really differentiate it
from some of the other alternatives?
(16:35):
I do, and I think for me.
What's really helped me, aside fromsome of these other cool features.
There is the boring part, right?
The boring part is ZephyrStandardizes how you build a project.
When you type "west build" and how youstructure these Lego blocks together.
It's the same for any part.
Because that's now standardized.
It allows a lot of peopleto be involved, right?
(16:59):
Even though that's a boring aspect,it means you can bring new people
into a team, into your company, andbring them up to speed with something
that has a standard framework.
One example is I was really excited.
There's a fairly recent part, the MCXA156.
It's just a really neat general purposemicrocontroller I mean, I had the board
and in five minutes of plugging it in.
(17:20):
No manual.
I could type "west build", and I was upand running on the FRDM Board doing some
sophisticated things with other piecesI had because of the standardization.
The other side of it, like youmentioned, because it is open
source with this structure.
I go to GitHub every week just to pokearound to see who's working on what.
What issues are there?
(17:41):
Sometimes your answers, if you're havinga problem just search the issue tracker
and someone solved it and it's toosolved two years ago, or at least someone
showed someone else how to do something.
Just being able to do thatand learn how to poke around,
it really accelerates a team.
Yeah and one other exciting thing,on the same topic, is the Zephyr
Development Summit, the ZDS, correct?
(18:04):
Mm-hmm.
Just the engagement withall the community members.
It's like a big team, right?
Even though they may wear a differentjersey, it's like everybody's working
towards the same goal, which is different.
It's just feels different.
You feel like you're gettingassistance from someone who
may have an ulterior goal.
Something that they're workingon, but they're in parallel, and
(18:26):
so you're assisting each other.
And that's I think a difference.
So, one thing I've seen as wellthat I think speaks to that.
It allows people to come intothe fold who might not have
otherwise been embedded developers.
There are a lot of people who are reallygood with code that is very much embedded.
(18:46):
So think about someone who's writingc code for a server backend in Linux.
They've never playedwith a microcontroller.
But Zephyr is like at this bridgebetween structured development
to your classic embedded.
That those people cancome from that world.
And you also have people who areused to Bare Metal microcontrollers
(19:06):
who now can bridge to the other sideinto these higher end processors,
like the RT family is a good example.
I think to your point, yeah, it'sbringing more people into embedded
that might not have been there.
Which is awesome.
It's gonna be exciting.
I think we're gonna have alot more episodes on Zephyr.
And I think, we're gonna have you back.
Correct?
When the training comes out oryou'll have some other topics that
(19:28):
we can dive into, that'll be great.
So one thing I'll leave withthat I'm really excited about
and NXP is really pushing.
You can see this just inthe board's directory.
We think about Zephyr asRTOS for microcontrollers.
Well, there's now support coming outfrom the NXP side who are running
Zephyr on i.MX Applications Processors.
(19:49):
Say like the i.MX 8 or the 93 whereyou're not even running Linux, you're
loading Zephyr onto this 2 GHz MPU,which is like totally mind blowing.
It's kind of a new area that we'reseeing customers adopting this.
It's certainly excitingtimes for the future.
All right, we'll load that upBridgette, on the next topic.
(20:09):
Thank you so much for joiningand sharing your journey today.
It's clear from our conversation thatZephyr has a lot to offer our listeners,
but like anything new and powerful,it comes down to a learning curve.
I agree.
Eli, this has been an absolute treat.
I'm so happy you joined usand are willing to come back.
You've given our listeners a greatunderstanding of why they should consider
(20:31):
following your path towards Zephyr.
All right, well it was nice having aconversation and I'll certainly be back.
And to our listeners, don't beafraid to get your hands dirty.
The trenches are where the magic happens.
You know Bridgette, sometimes when youget down into the trenches and you look
around, you might just realize thatit's a place where people are coming
together to work closely on a common goal.
(20:54):
In this case, that's theadvancement of Embedded Software.
So once again, thanks forstreaming this EdgeVerse Techcast.
Bridgette and I truly appreciateyou listening and until next
time, keep exploring at the Edge.
Popular Podcasts
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com
The Herd with Colin Cowherd
The Herd with Colin Cowherd is a thought-provoking, opinionated, and topic-driven journey through the top sports stories of the day.