EP#428: Repairing the Retro Future

Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Parker Dillmann (00:10):
Welcome to circuit break from MacroFab, a
weekly show about all thingsengineering, DIY projects,
manufacturing, industry news,and long lived designs and
complaining about Apple like wenormally do. We're your hosts,
electrical engineers ParkerDillmann. And Stephen Kraig.
This is episode 428 and andbefore we get into the podcast
proper, we have a platformupdate for MacroFab. We launched

(00:34):
the enhanced inventorymanagement system.
We've always done energyinventory management in the
platform, but it's we've gotbasically have a new display for
it, and it handles overage, andit handles, like, how you let's
say you up you you sent us somecomponents for your inventory so
we can store them, and you canbuild out of them. Well, what

(00:56):
now is when you place orders, itwill show, like, which parts of
that inventory is reserved forbuilds and that kind of stuff.
And it will actually tell youconsumption reporting and burn
down reports and that kind ofstuff. And then there's, like,
an improved inbound shipmentmanagement too, so you can see,
like, where your component is atin terms of when it's being

(01:20):
received in to the, HQ building.

Stephen Kraig (01:23):
So can you see when your part is picked for a
job and when it's gonna go outonto the floor?

It's yeah. So it has, like, a reserved and
then a work in progress section.Cool. So you can say, oh, you
know, let's say you place anorder and you need 20 of these
components, it will reserve 20of those for your order. That
way you can, like, queue upmultiple orders at once, And
then the moment it gets kitted,it goes into work in progress.
And kitting is when all thecomponents come together for

(01:53):
your order before it goes out tothe floor or get shipped out to
a partner facility.

I I can tell you from from my side, things like
that are absolutely wonderful.Given the job I work now, the
parts that I deal with areconsiderably more expensive than
previous jobs I've worked at.And so so a lot of times, I'll
get asked by my project manageror by my manager or other
people, hey. Where's x y z part?Can you give me a count on

(02:20):
what's at your manufacturer?
Blah blah blah. You know, whenyou have a reel of parts and
every single part is $1200, youyou wanna know the exact number
of parts and you wanna knowexactly where they are. And a
lot of times, when I'm askedthat question, it's, oh, let me
send an email to my Centimetersand have them count it and tell
me. But to be able to just lookat my inventory and be like,

(02:41):
yeah. I know they have thismany, and they're in this stage
kind of thing.
That's really helpful.

Yeah. That way you don't have to call us up or
email, you know, Jerry.

I do think that's funny because a lot of macro
fab, is I don't think you wereintending for this to be the
case, but a lot of macro fab isset up for people who don't
necessarily want to talk topeople, or it's set up where if
you don't want to talk topeople, you don't have to.
Right? It's all in the platformright there. I think that's kind
of funny how it's, like,introverts, Centimeters. Right?

It was planned that way. Yeah. It was. Wanted
part of the initial was, like,never had to pick up the phone
to figure out where your build'sat or even to start your build.
Mhmm.
What's really interesting is Isaw there's it's been this new
term on social media in this inthe hardware start or just the
startup space called hard tech,and people are like, we should

(03:37):
have a I I saw some discussionabout we should have a pizza
tracker for PCBA, and I'm like Ijust posted a screenshot. I'm
like, we've had this for 6 yearsnow. Yeah. We have a pizza
tracker for your PCBA order, soyou know exactly what's going
on.

Yeah. You know what stage it's any one point in
time. Yeah.

Yeah. So this basically is the same thing. It
gives you more visibility intoyour inventory that you have at
MacroFab.

And and you know what's funny? For basically no
work from your side, you beingthe engineer at
xyzdesigncompanyorwhatever, itmakes you look awesome. It, and
I'm not just trying to sellhere. What what what I'm saying
is say that project managercomes up and says, I need, I
need you to run a report of howmuch money we have in inventory

(04:25):
at our contract manufacturer.You do it in 5 minutes and send
it back to them.
You look like a rock star, andall you really did was go to
Macrobat and just say, hey. Giveme the list of parts that you
have.

See, we actually do

Yeah.

Is you just take 45 minutes to do it. Oh.
You get the information in 3minutes from the platform Yeah.
And then you go and take lunch.

But you still look like a rock star because 45
minutes is still really fast forthat.

Yeah. It's pretty good. Yeah. So when your
boss comes in and you say,that's gonna take a day, and you
come back 45 minutes later withthe answer.

Man. Okay. Quick side channel. I absolutely hate
estimating time to do a task ora project or or whatever. I I I
don't know a lot of people whodo, like, really enjoy that.
I'm I'm sure there's plenty ofbusiness guys out there that
really get their jollies out ofestimating time on projects and
things like that. But I I I wasdoing some recently, and I we

(05:25):
were kind of looking atestimating time based on hours
in a project. And a lot of timeswhen I think about estimating
time, I'm looking at, okay, 8hours is a day's worth of work,
but you're not constantlyworking on that project. So 8
hours like, if you say 8 hourson a spreadsheet somewhere, that
doesn't necessarily mean I comein Monday, and I leave with that

(05:48):
project done. Right?
That 8 hours could be spreadover 2 weeks because you're
doing all kinds of other things.I don't know. I'm go I'm ranting
a little bit here just becauseI've it's extended period of
time is going to take,especially with testing. Even if

(06:10):
you write a full test plan andyou kind of have an idea of this
probe goes here and I press thisbutton on this device, and I do
this kind of thing, it's stillincredibly difficult to get
anywhere near accurate.

Yeah. Oh, yeah. I agree. You kinda have to just
do it. You can get better as youget more experienced.
Sure. And, like, platforms likeTrello or Jira that can track,
like, track time and statusesand stuff, but that's that's
only as good as the data yougive it. So you have to be
diligent about, oh, like, foryour example, 8 hours over 2

(06:43):
weeks, you have to be diligentabout not just leaving, let's
say, a task in progressovernight or the task, like,
just in progress for the entire2 weeks. Mhmm.

Yeah.

How we kinda did that was we in our this is
in engineering, we had a ourlike, what everyone is working
on board, kinda just like yourcurrent task. So it was a way so
we could start measuring, like,workload and that kind of stuff.
Mhmm. We had 2 well, actually,like, more than 2. We had a

(07:16):
couple different, like, inprogress statuses, and wanted to
be, like, in process, like, inprogress.
I'm actively working on thisthing right now. Then there's
also in process, like, it's likeon the back burner because it's
like, I'm waiting on something.And we had different statuses
for, like, you're waiting on acustomer or you're waiting on

(07:38):
someone else at Macrofab orsomething like that. And that
got us pretty good, but it alsoyou have to be you know, your
team has to be diligent in doingit. And it's not something from
micromanaging.
It's just being like, okay.Every engineer's got 20
different projects. How do youknow that what's the process of

(07:59):
all 20 of those without havingto bug that engineer?

Yeah. Yeah.

And have a, you know, 10 minute meeting every
day. It was a way, basically, toget rid of stand ups and get rid
of meetings and just be able toget rid of all that kind of just
waste of time. Mhmm. And ithelped a lot. Yeah.
But, again, you have to bediligent, and it takes a while
to get off the ground and getpeople used to, you know, when

(08:27):
they task switch, they go andtell the system that they are
task switching.

Yeah. Feels a little micromanage y. But

You could say that, but it got it it opened up
3 hours every week for peoplebecause there was no more
meetings Sure. And that kind ofstuff.

You know, another thing is that kind of a system,
I think, works if whoever ismanaging the system itself
understands the type of workthat's going into it. I I once
worked under somebody who wasvery managy very micromanagy,
And they they ran a Trello esqueboard and had very explicit

(09:11):
dates for every little task thatyou do, but didn't understand
how engineering workednecessarily. So so there was
like, okay. This project kicksoff. You have 4 days to create
your schematic.
So design your bore not board.Sorry. Design your project, make
the schematic. And on the 4thday, the expectation was the

(09:33):
schematic is 100% done, and thenyou move into layout. And then
you have 2 weeks or whatever todo layout.
And, you know, at the end of 4days, I had the schematic, but
they were like, is this done?I'm like, well, it's done, but
I'm not gonna tell you it's donebecause that's not how this
works. And then by the time andYeah. Whenever I finished

(09:56):
layout, they were like, okay.Are you done with that schematic
and that layout?
And I was like, well, I have togo make my drawings for all the
manufacturing files. So, no, I'mnot gonna call both of those
done until they're done. Andthen after that, I had to get
all of it approved. So I'm notgonna say anything's done until
it's until everything wasapproved and then it all just in

(10:18):
magically in a second it allwrapped up and was all of the
tasks were complete. But themanager didn't understand that
you can call something done, butit's still open in a way.
During your layout portion, yourschematic is done, but you may
need to go back and makeadjustments to it. So it doesn't
just fit cleanly into this binof, oh, it's done, and I will

(10:40):
never need to see it again. It'sand so I think engineering has a
lot of wrinkles with that, andhaving a manager that
understands it is important.

Yeah. That's why we never really track
statuses like that, likeindividual tasks inside of a
project. It would just be likethe total project that the
engineer is working on.

Yeah.

So because each engineer has, you know, 20
different customers. Which one'sbehind, which one's our head.

Which one's on fire?

Yeah. Which one's on fire? And you don't so
that way, you can look at theboard, and we basically just had
a notification system thatwould've been like, hey, if a
project it's been in progressbut been blocked by let's say it
was blocked by an internalcustomer, which would be at
MacroFab Right. For longer than2 days, then the manager could
go and go, hey. I'm gonna gotalk to that one person that's

(11:38):
the blocker and figure out whythis is blocked and how do I
unblock this so we can getprogressing on this project
again.
Right. Right. Right. Whereas theonly other way to know that is
to have a stand up. And ifyou're doing a stand up every
week, once a week, let's say ona Monday or or on a Friday, you
won't know you you just havethis gap 7 day gap of

(12:00):
information.
So it works really well, but,again, it it you do have to be
diligent. I just it's betterwhen you don't have to
micromanage the tasks. It's justthe overall what are you working
on, and what's the status ofthat one project. Mhmm. So
keeping it like a pro like aprogress, like, product progress
tracker or just status of it,but, yeah, we're not tracking,

(12:24):
like, is the layout done?
Is the schematic done? Is that,you know, did you verify the the
document, like, the testdocument? That stuff wasn't in
there. It's just hideable stuff.Not even that.
It's just the let's say it was,like, the test building the test
documentation and implementationand all that stuff for, let's

(12:47):
say, Steven Craig amps. BecauseSteven Craig amps were is let's
say you're we're building yourstuff, and, there would just be
a top level ticket in that boardand you'd be assigned like an
NPI engineer and you'd beassigned a test engineer And
then but that test engineer isgonna have that ticket and be
like, okay. I wanna spendprobably an hour reviewing

(13:09):
Steven Craig's documentationthat they gave us. And so he'll
move it into in progressactually in progress. I'm
working on it right now Mhmm.
And work on it. And when he'sdone, you know, if it's actually
and okay. Well, the next step iswe're waiting on actually
getting in the equipment fromSteven Craig amps. So they'll
take the ticket and go, okay.I'm gonna put this in progress,

(13:30):
but it's in it's waiting on thecustomer to get us some stuff.
So they'll just move it intowaiting on customer, and they'll
just type in a note. This iswhere I'm, like, I'm waiting on
the actual equipment because wedon't have it yet. That's it.

Yeah. So it's

just it kinda just a mass like, the ticket
just kinda grows and collectsall this information. Yeah. Just
a history document. If you're anengineer you should be doing
that anyways is like recordingwhat you're doing on a project
that's kind of what that does.

Yeah. Yeah. I like that.

Yeah. Is there better systems? Probably. But
does the development team usesJIRA here and it was like, well,
that's just free for engineeringto use. So we that's what we
used.
Sure. Sure.

Yeah. I think when it comes to a project, the
way I I kinda prefer it, and youand I are talking about slightly
different methods and andindustries here. But but I I
like the idea that, say, acontract for a customer is
written, or corporates come justdown and says, I want a brake
controller for this particularcar. I want you to design that.

(14:37):
And they say, you know, in thecontract, we say, I want the
first PCBs in my hand in 3months.
And the engineering team says,okay. And that's as much as
needed because then theengineers that the engineering
team including your manager sitsdown and breaks that 3 months up

(14:57):
and they figure out how whenthey need to have schematics
done, when they need to havelayout done. I don't like it
when the higher level says youhave 2 days to do schematic and
you have 5 days to do thelayout. Blah blah blah. Like,
that never seems to work.
Just give me a date when I needto have a product or a PCB in
your hand, and then I will workwith my team to establish all

(15:19):
the intermediate dates.

Oh, yeah. I agree on that side. Yeah. Yeah.
100%.
That's a higher level. Let's saythe the document drafter. That's
how we can call that person.Yeah. Or the product.
Let's hope if I have another pword. The product person. The
producter. Yeah. I mean, becausethey're not engineer, but they

(15:43):
all they care about is they getthe PCB or the device by a
certain date.

Well and okay. And, yes, give me what that date
is that you have to have thatPCB or the product or whatever.
What's the deliverable? Give methat date, and I will come to
you on a very regular basis andtell you it's going well or it's
not. And if you need moregranular data about if it how
it's going, I will tell you themore granular.

(16:08):
But if you're okay with notgetting the granular, I I won't
feed you that unless you need toknow it. And you don't need to
necessarily feed super granulardates down to me, especially if
you don't understand generalengineering practices for double
lease of and that's not just me,like, being hoity toity or
anything like that. That's it'sjust, like, compartmentalize

(16:32):
each department and let them runtheir thing because they know
that they're the experts at howthe process flows through their
department. Let them do that.

Mhmm. No. I agree. You do have to have a
way, which it's fine, exceptthat you you still have in there
that if that person needs moregranular updates, let's just say
it's progressing well or notwell. K?
Yeah. Like, a a a thumbs up orthumbs down on you hitting your

(17:03):
dates. K? How do they knowwithout asking you?

Okay. So I honestly think that is one of
the largest jobs of what amanager does. So a manager is a
middleman between the individualcontributors and whatever other
department is there. So thegranularity for a manager is day
to day. And in some cases, hourto to hour.

(17:30):
It is walking to the team anddiscussing what are you doing
right now, how is this going,what is your progress, and then
distilling that into largerpackages of okay. Over 5 days,
Steven or Parker got this chunkof work done and distill that
and pack in a nice way thatmakes sense to whatever

(17:52):
department needs to know whatthe progress is. If I laid out a
switch mode power supply and Iand I got all my ground planes
really nice, blah blah blah, Myproject manager doesn't need to
know that. But my project but mymanager might say to my project
manager, we are 20% closer tobeing done. That's something
they do need to know.
It's all about who is youraudience, what are you trying to

(18:14):
convey to them, and to whatlevel of detail do they need.
And 99% of the time, they don'tneed as much detail as engineers
are willing to give or wantingto give. So knowing how to break
that down is a skill for sure.

Now what if your customer isn't in your
company, and it's because youare doing something for an
external customer?

Yeah.

And now they're tying up your support and your
and your and your and your saleslines trying to get information.

Well, that's just a bad customer. Once again, I
think

That's what that's what we're trying to do
with the Mac event platform isso you don't have to the more we
can record stuff and and make iteasier to record that
information and then tricklethat information to the customer
so they know how well or badtheir build is going, that
that's, you know, even better.

You know? Okay. So I haven't I've been on the
Macrofab platform a bit, but Ihaven't explored recently. Was
it do you have the ability toset up different dashboards for
people?

No. Not yet. We have a dashboard, but it's not
like a customizable dashboard.

Yeah. Okay. So the you know, just an idea. Say,
I have a login to my MacFabaccount, and I can see all of
this information. But the clientthat I'm making a board for
wants to see some of thatinformation as well.
I could make them a dashboardand send them a link, and that's
just something that they couldview. And they don't have any

(19:55):
editable things, but, you know,maybe the customer, the client
sent me some really expensiveparts or something and I sent
them to Macrofab and theclients, hey, I want to be able
to see them at Macrofab. I couldsend them a link and it shows
their parts that they sent to methat you guys now have. I don't
know. It's just a thought.
I that could be something that'scool.

You could do that with other services and
just using the MacFab API,basically, to get

that information. Build that.

Yeah. You could build that with something like
Retool, which is a it's like anonline app builder Right. That
you can basically glue a wholebunch of, APIs together and
databases. So you could hookthat up to the MacVeth platform
with your authentication andpull that information in and
just show a inventory dashboardfor one component for that

(20:43):
customer.

You can always do it that way. I I know what I was
just saying. There's veryextremely specific. Out of, you
know, all customers, there'sprobably only a very few number
that that would apply to. It'sstill I don't know.
It's it'd be a cool idea. Icould see that being a cool idea
splitting up based off of yourrole at a company. You know?

(21:06):
Purchasing sees this.Engineering sees that.
Management sees this.

Kinda have that. Yeah. And I remember when
I

was working there, you guys were setting
that up.

Yeah. We have different views, so to speak,
for procurement and well,procurement slash purchasing and
engineers on in organizations,but not nothing like what you're
talking about, though.

Yeah. And until recently, I can't see I haven't
had a need for that kind ofstuff. Most of the time that
I've dealt with customers, a a,you know, a a regular meeting
with them and a PowerPoint slidesaying, here's how we've managed
your stuff It's usually goodenough.

Yeah. It's just I I see a lot of people talking
about wanting the informationright away and not wanting to
talk to people.

Well and to kinda tie it all back, that's what
your inventory system allows youto do. Right?

Exactly. And I was just thinking about this,
but because there there are alot of customers that are old
school in in terms of they wantto call and talk to someone or
email in. What have I justconnected that to? I wonder if
you used 11 Labs which is a AIservice that you can basically
clone your own voice. That'swhat I used for the the charity

(22:23):
stream last year where it wasDarth Vader was reading all the
call outs from people.
I basically made a clone ofDarth Vader's voice.

Mhmm.

And so you you I could put a clone of my voice,
and it could be me talking toyou over the phone reading
basically what the back of ourplatform says.

I I wonder if you did that. I wonder if people
would be okay with that or ifthey'd be upset.

I think they wouldn't know. And then the
moment they figured they knew itwas an AI, they would be upset.

Yeah. I agree. So so there's this concept that if
you call a company and an actualhuman being answers the phone,
then, like, you get excited andthey feel like old school good
customer service when thathappens. Right? And you kinda
got to the point where youexpect if you call a large ish
company, you're gonna get you'regonna get whatever telephone

(23:16):
service they have, and then yougotta deal with that forever.
But is that the next step ofpissing people off? It's just
you do get a human, but it's anAI human. And I don't know. That
might actually work. That mightnot actually piss me off because
I can interact with it like ahuman and get what I want as

(23:39):
long as that all works.
I don't necessarily care.

Yeah. I don't care either because all I care
about is just gettinginformation I'm looking for as
fast as I can.

Yeah. The the under Yeah.

That's like what I call, let's say, AT and T
because my Internet is down ormessed up or something's wrong
with it. The first thing I dowhen I call these big companies
is I just mash the zero button abunch. Yeah. So I try to get a
actual human being on the phone.Sure.
Because I know the going throughthis the pressing all the
buttons and typing all theinformation in, it's gonna take

(24:11):
way longer than when I actuallyget a human on the phone. So if
I could skip to so I'm justtrying to get to the end where
the answer is. I do see an a ana hooked up to it would probably
not be good for data retention,though.

Do you have a because

think about well, think about how the
biggest breaches of security isin just social engineering is
the number one way of findingout yeah. Packing is mostly
social engineering nowadays.

Or you just convince the AI to give you data
that it shouldn't.

Yes. Exactly. Yeah. Yeah. So it's if if people
are gullible is not the rightword, but if people are
susceptible

Yeah.

To social engineering, I bet you an AI is
even more susceptible to becauseit doesn't it can't infer, I
guess. It it does it doesn't getsocial cues on, like, the phone
call and that kind of stuff.Right. Right. So I would say
that was probably your biggestrisk there.
Be interesting to try out,though. Maybe if I wasn't in

(25:19):
marketing now, I I would buildsomething like that just to try
it out.

In some ways, I feel like marketing is actually
an okay department to be in forthat.

I guess. Yeah. Because we don't have any AI
chat bots or anything like thaton our website. Yeah. Some of
those are some of them arereally good, like the Amazon
one.
When you talk to like, when youdo their chat thing that's on
their website Yeah. And that'san AI bot. There's no one on the
back end of that. Yeah. But someof them are pretty there was I
think it was GM slash Chevyimplemented a chat gbt bot for

(25:54):
their website, and you could getit to compile and run Python
code on their website.
Really? That was pretty yeah. Soit was pretty funny. You can
get, like, the chatbot toexecute code. That was pretty
good.

Yeah. I was not aware of that.

Yeah. That made the rounds on that. I think it
was, like, earlier this year. Ithink it was in January. But,
yeah, some interesting stuff forsure.
Yeah. So back back to that is,like, I I wouldn't care. As long
as the information is that's ifthat's the fastest way that
company can get me theinformation I need, so be it.
Don't let me put don't put me onI'd rather not be put on a

(26:34):
waiting list to finally getahold of someone.

Yeah. I've been on I've been on hold for an hour

and a half.

Yeah. Right. Right. AI would be so helpful
for answering all the reallydumb, simple questions that
probably burn tons of time.Okay.
We're we're going on all kindsof fun tangents here. But but
okay. Think of this. When when Igo to a website for for, an

(27:04):
establishment, let's say, arestaurant. Right?
If I go to a website for arestaurant,

the the the thing that

should be on the front page of a of a restaurant
is what are your hours andwhere's your location? Right?
Because 99% of the time, that'swhat

people Maybe also a website for food.

Well, okay. That's what all I'm saying is
the absolute most critical itemsthat are on there. Now if you
wanna add your menu or if youwanna add, you know, your
schedule of events or whateveryou're doing, like, if I don't
know if you have bands that playat your restaurant or whatever.
But what's the absolute mostcritical thing when you are a
restaurant or something likethat? Put that number on there.

(27:44):
I bet you restaurants I I I itwould be really fun to know how
many man hours are burned everyyear from people calling a
restaurant just being like, areyou open? I bet you it's
staggering how many hours areeven for a 10 second phone
conversation that somebody'sjust calling to see, are you
open kind of thing? AI couldtake care of those kinds of

(28:05):
things. Right?

Yeah. And especially, let's say, like, all
these restaurants, if you couldGoogle that restaurant, it has
its hours Yes. And all thatstuff right there. So that
information is really readilyavailable, and people still call
in and ask, what's your hours?

Exactly. Well but okay. So so there are times when
it's not necessarily superaccurate. It may be a holiday,
and you're not sure or andGoogle's still saying or Google
says holiday hours may bedifferent or something like
that, an AI could just be like,yep. We're actually closed.
Sorry.

It's hooked up to the on sign on your,

on your Yeah. For the neon sign?

Yeah. The the on neon sign or open neon sign.
Yeah. Yeah. Yeah.
It just hooked up to that.

So let a human say hard questions. Let an AI
answer the dumb ones.

That's a product right now. Okay. That's
is an on or an open sign. Idon't know why I keep saying on
sign. An open sign that's hookedup to a that's, like, an IoT
open sign.
K? Yeah. And it's hooked intoyour your POS system and for the

(29:21):
call system.

Yeah. And if the open sign is on, the chatbot
will say, yeah. Come on in.We're open.

Yeah. The phone tree will say we're open.

Right. Right.

Right. Right. That is a product. Yeah. I I
wonder if that has to exist, butmake it simple like that where
if the on oh, geez.
I'll just

say it again.

If the open sign is on Yep. Then it says
that we're open. And if it'soff, it says that we're we're
closed.

See, we're solving the real problems in the
world here.

But making it super simple, so it's not like
some other system that you haveto log into or, like, a button
they have to press, like, on thecomputer or whatever. It's
literally the sign turn it offand on.

Yeah. And, you know okay. A little bit further
than that, if you call have youwhen was the last time you
called, like, a Target or aWalmart and asked them if they
have something in stock? It'sbeen a while. Right?

I would say I can't remember because usually I
go onto their website and see ifthey have it in stock or not.

Yeah. But an AI could do that as well. Right?
Yeah. There's nothing sayingthat

I I don't know. It's more useful for grocery
stores because let's see. You'relooking for one particular
sauce. Yeah. And and it mightbe, like I mean, you know how
big HEBs can get grocery storesin here in Texas.
And onto the HEB website andgoing, okay. I'm looking for
this sauce. What aisle is it in?

Well, okay. So anytime you call one of those
big stores, you always have todo the whole dance of, can you
please connect me to theelectronics department? You you
you know, you always have to dothat because the first person
you talk is 100% guaranteed tonot be able to answer your
question. Right? But AI couldjust do that.

(31:08):
Right? It just knows theinventory, and they could just
spin it out

for you.

Just along the website, but

We should get on to the real topics, but I'm
wondering this what we justsaid. I wonder how the people
who hate AI just because AI isgoing to defend that idea Yeah.
Or attack that idea. Becausethere are some people out there
that just AI, I hate it. So Justbut just arbitrarily.

(31:35):
Yeah. Arbitrarily. Right. Why ispeople are so, like, black and
white nowadays?

Yeah. I don't know.

Polarized. Right? Yeah. Polarizing. Yeah.
People need to have more thanone bit of color.

Gray is a nice color.

Alright. So 36 minutes in. So the our first
topic that's actually on ourlist to talk about today, not
our dive into AI chat bots andMac fab platform updates is this
is a really cool article I foundon Hackaday, but it's from the
auto autopian? Autopian.Autopian?

(32:21):
You're right. That's what it is.It's about a, a person named
John who basically rebuilds ashort lived add on or option for
GM Toronados from, like, I thinkit's, like, 1989 to 92. 3 years,

(32:47):
you can get a what was called aVIC, a visual information center
inside your GM or Ultimate BealToronado, And this predates
those tablets that are, like,glued to the dash of every
single car nowadays by 2decades. Okay.

Yeah.

And, basically, it's a CRT touchscreen that has
controls for HVAC, has a tripcomputer compass. They also
integrated the car phone becausethis is 15 years before a cell
phone was a nothing, let alone,like, everyone having one. But
this is like an article aboutrepairing them, and what, John

(33:27):
does to repair them. Mostly it'sit's CRT repair because most
time it's just like the controlcircuitry goes out and they have
to replace some components. Mostof it's like electrolytic
capacitors, a lot like amplifierrepair.
Most time you're just replacingyou shotgun all the capacitors,
resolder everything, and itcomes back to life.

I I I love the the the old CRT stuff,
especially with electrolyticcaps because you can see the
caps go bad. Because the screenjust starts to wobble and
flicker, and you could just lookat it and be like, yep. I need
to replace caps. It's time.Yeah.

But it's a really cool I I knew they had
CRT. GM did this this I wentdown like the rabbit hole
reading about the VIC system,but what's even cooler is they
even had a short term pilotprogram with GPS navigation down
in Orlando, Florida. And they'vebeen there only 90 Yes. And 92.

(34:24):
Wow.
I I think they built, like, 200of these cars, and, like, Avis
Avis had a hundred of them thatyou could actually just rent,
and they had this GPS systeminto it. And it was like a
combination of using, like, thecar phone for data and the GPS
system because it you could alsoget just like a modern, like,

(34:46):
Android Auto setup or a modernnavigation. You get traffic
updates, and it would know thatyou were, like, off route. And
if you needed help, it was kindalike pioneer, and they're also
they're, OnStar service. So ifyou needed help, you could just
click the help button on yourdash, and it would call AAA for

(35:08):
you, and they they would knowwhere your location was and let
you get and help you get back onthe on track, basically.
Right. Right. Right.

I wonder what the resolution of it was. Because,
GPS is kinda crazy now. GPS iswhat? 6 feet now or whatever?

It's less than it's less than that.

Well, I mean, correct. It is less than that,
but but they don't necessarilygive you that much resolution if
you go to Google and whatnot.But, but regardless, I wonder
what it was at this time.

Yeah. I I think it was definitely probably,
like, 30 feet or so. But if youlook at the roads that's being
displayed on this CRT monitor,they're not particularly super
high risk. It looks like avector screen.

Yeah. Yeah.

I mean, it is a yeah. I don't know if they were
drawing them a lot as a vectordisplay, or are they rasterizing
it? I don't know that. Would becool if it's a vector.

That would be really neat.

But they're colored displays, so that

And they're touch screen too.

Yeah. They're touch they had colored vector
displays, so I don't know ifthey were vector or not, but
really cool, system. Everyonethat's interested in, like,
weird automobile stuff,definitely go check this out.
But this brought up ainteresting idea, or topic I
wanna talk about is if so thisis 30 year 34 years down the

(36:37):
road, right, that these CRTs arestarting to fail in these cars,
which, honestly, given theenvironment in a car with how
many, like, heat cycles you gothrough in the day, that's
actually kind of impressiveMhmm. For a TV.
But if you were trying to designsomething and that I was saying

(37:00):
like if you were designing theCRT monitor that's going in this
car, like, how long should thelifespan be?

Oh, that's a good that's a good question.

Because I mean, back way back in the day in the,
you know, forties, fifties, andsixties, you know, cars were
really only designed for 3, 4years max. If you got, like,
50,000 miles on a car, that wasa lot of miles for a car back
then.

Yeah.

And engines didn't last as long. In the
early days, engines didn't haveoil filters, lot of stuff. And
nowadays, if a car doesn't hit a100000 miles, it's a bad car,
bad design.

Oh, yeah.

And you can even say double that. If it
doesn't hit 200 1,000 miles,it's not even worth buying. Now
and most people are keepingtheir cars for 7 7 to 8 years
now for the first owner of thatcar, and it's gonna live 20 more
years later in the secondhandmarket. So but back then, it was

(38:02):
still like, do you did you planfor a 30, like, how would you
even gauge that? Like,especially when, like, early on
was, like, a low volume option.
But on the other side of thatis, let's say you want to design
something, like, an end of theworld electronic device where it
will still work until the sunexpands into a red giant and

(38:26):
consumes the earth. How wouldyou go about designing something
like that? I mean, first of all,electrolytic capacitors are you
can't use those.

No. No. Okay. So it's funny. I've actually dealt
a little bit of with thisrecently.
So, basically, what you'retalking about is reliability and
how do you guarantee reliabilityfor a long period of time. And
on top of that, what you'rereally getting at is how do you
calculate that you can guaranteereliability for a long time? And

(38:57):
there's a handful of, frankly,not that difficult e equations
that you could bang out thatsays, you know, at the end of
the day, what is yourprobability of failure at x
years given these inputconditions? Like, how many
temperature cycles it goesthrough every day and how many
what's its duty cycle of on? Acar is not a 100 duty percent

(39:20):
duty cycle.
You know, on average, it'sprobably the duty cycle of a car
is actually probably pretty lowbecause you drive to work, you
turn it off, and then you driveback home. Right? So all of
these are factors that can gointo these equations. But if
you're looking at just a singlepoint or a single, I don't know,
circuit, the reliability, it ittrails off at a certain point.

(39:43):
You you you just can't makesomething more reliable.
You can only spend so muchmoney, and that's when you start
saying, okay. Now we get intoredundancy. And redundancy, like
a dual redundant circuit,doesn't reliability. And so then

(40:04):
you have to start asking, whatabout triple or quadruple
redundancy? And so you build areally robust circuit.
You test the snot out of it, andthen you just multiply it. And,
frankly, to my knowledge, that'sone of the easier ways to get
more life out of something ofthis sort. And when it comes to
the driver let let let's say wewere actually trying to make a

(40:25):
CRT that last 50 years. Thefunny thing is the CRT itself
may not degrade, might not bethe the the key factor that
degrades. It may just be, youknow, the the flyback
transformer or the I'm sorry,the flyback circuit that does
the high voltage to it andthings like that.
So, you know, identify your keycircuits, test the snot out of
them, and then add a boatload ofredundancy would potentially get

(40:48):
you there. But the thing is,it's still only probability. And
and the way that you kinda workit out is you say you give a
probability number with a timeinput to your circuit or to your
equation. So you say somethinglike, at age 20 years, I want

(41:08):
95% of all of my product tostill be functioning. So you
have to allow for 5% of yourstuff to fail at at 20 years
because there is no such thingas a 100% success at 20.
You just can't guarantee that,but you can keep doing all these
little tricks to bump thatnumber up. And, eventually, you

(41:28):
have to just assume theremainder is risk of failure.

So yeah. Because we we talked about that
with, James Lewis Right. On aprevious podcast. I think it
was, like, entropy rules theworld or something like that is
the title of that podcast.

I like that.

But I'll rephrase the question.

Yeah.

If you were building an amplifier

Yeah.

That you wanted to pass down to your son Yeah.
What components would you pick?

Yeah. So so what I first said is more of, like,
the design equation side, andnow you wanna know more, like,
exact nitty gritties kind ofthings. So so, obviously, you
pick components that you knoware less temperature to have
less of a temperature dependenceand have nothing that can dry up

(42:19):
and have really good agingparameters. So film capacitors
as much as possible, things thatare potted. If you if if you
have something that is embeddedin urethane, a lot of times that
has a longer lifespan, Thingsthat, resist vibration, things
that resist temperature swing

Keeping the design cool. Keeping the design
Yeah. So, like, runningundersized traces or underrate
even not even, like, underrated,but just at rate resistors.

Yeah. Also, a circuit that that is used
regularly because because oneargument you could say, I wanted
to make an amplifier that Iwould give to my son in 30
years. I wouldn't just make anamplifier. Let it sit for 30
years, and then turn it onbecause that's not reliable
either. So something that getsregular use.
Right?

Yeah. It's an interesting thing to talk think
about because when because I'mgonna start going more into
automotive OEM, like,aftermarket designs for my
electronics moving away frompinball, and that's one thing
I'm thinking about. I'm like,well, if I design, let's say,
this motor controller and in in30 years, I want that motor

(43:33):
controller to still work. Mhmm.You know?
How do I because that's, like,the worst thing that can happen
is if I build let's say I build,like, a 100 of these motor
controllers and sell them, andin 3rd and I'm using some of
them. Right? In 30 years and ifthey if mine breaks, I can't
replace it because it's like a,you know, spokespoke motor

(43:54):
controller. Right? You know?
So I would like it, from mystandpoint, for the last
forever. And especially if you,like, make that as part of your
sales argument too, why it costsmore. It's you know, it's more
expensive as we are using thesecomponents for a longer life.
You can't say it can lastforever because there's you're
talking about those odds thatsomething will fail or

Can stack the odds in your favor.

You can stack the odds in your favor. Yeah. I
think that's very interesting tothink about because my computer
for video gaming at home, wewere talking about this last
year, beginning of last year,was it was starting that
computer was starting to fail.Mhmm. It was having weird issues
with this hard drive, and wecouldn't and I think it's been,

(44:41):
like, a month of podcastepisodes where, like, I was
bringing it on that floor home.
Yeah. It was sitting on thefloor, like, with all its guts
hanging out because it workedthat way. Yeah. And then the
moment you reassembled it thenormal way, it was it would
exhibit the same problem. Andthen that being, like, the south
bridge failed, so all everythingthat's that runs off the south

(45:02):
bridge on that motherboard isjust dead.
And, And, but there's some SATAports that run on the north
bridge. And if you plug intothat, it works just fine. It
doesn't care. And if you use thelower PCI Express lines, doesn't
work. If you use the upper ones,totally works fine.

That's weird.

And so I reassembled it in a different
configuration, still play gameson it every single day. But
what's interesting is thatmotherboard, I bought it. Jeez.
9 years is it almost 9 years oldnow? And one of the it was it's
an Asus TUF, and the sellingpoint was it doesn't have

(45:41):
electrolytic capacitors on it.
Mhmm. Because I was bitten bylong time ago, capacitor plague
was, like, a big thing inmotherboards. Probably still is
now. I just not in that newssphere anymore. I'm enthusiast
computer technology nowadays.
But anyways, back in the day,electrolytic capacitors, there
would always be, like, a badbatch, and motherboards would

(46:02):
just fail. And so I bought thistough motherboard because it it
it quoted what was the quote?Solid state capacitors. Yep.
Which means What?
Probably tantalum?

Those were more expensive, so I really doubt it.
It probably just Yeah. Tantalum.Yeah.

And I don't know what caused the southbridge
to fail on the board, but it didlast 9 years or 8 years before
it started exhibiting problems,which is the longest I've ever
had a motherboard last. Mhmm. Sothere's that. But definitely did
did I it definitely was not anheirloom motherboard, though.

Yeah. I don't know. It starts to get
difficult, especially withcomputers because they are way
more of a higher duty cycleproduct. They are they remain on
a lot longer. So they're in, youknow, they're functioning all
the time, and that just leads towhere.
Right? So 9 years soundsactually like a pretty long time

(47:00):
for something like that tocontinuously function. Like I
said, I think I don't know howmuch better you're gonna get
than that without doingredundancy.

Yeah. And so that's gonna be interesting
looking at I mean, I yeah. Iagree on the motherboard side.
Yeah. It's just the automotiveside.
I think what I'll do is I'llI'll make up my first design,
and we'll rip it apart andfigure out how to make it last
longer without,

Go ahead. Simplicity helps so much because
the fewer items you have inthere, your your probability of
survival goes up. The one of themain drivers in reliability is
is part count. The more partsyou have because because if you
think of everything as a system,you ask the question, what is
the probability of a failure?For every part that you add in

(47:51):
there, you add probability offailure.
Now you have to define what afailure means in that sense.
Okay. Let's say you have a TVwith a 1000000 LEDs on it or a
1000000000 LEDs on it. If 1 LEDfails, is that considered a
failure?

Oh, yeah. The whole dead pixel argument.

Right. Right. Is that considered a failure? If
so, I bet you if you run thereliability calculation on TVs,
you get abysmal numbers justbecause of the part count. But
their reliability is actually alot better than that.
So so the calculations actuallydo start to fall apart when you
have trouble defining whatfailure means. But in an

(48:31):
automotive sense, there's youend up having a lot fewer
functions that it does. Right?It's not like you're not doing a
1,000,000 things. You're doing15 or or whatever.
So what's nice is in your case,you can rely a lot on those
calculations because it is waymore stripped down and you can
really you can define what afailure means. If this output

(48:53):
just stops doing the output,that's a failure. Right? Yep.
And I'm assuming a lot of thestuff that you wanting to design
probably doesn't have a lot ofmaybe I'm wrong here, but
probably doesn't have a lot ofanalog control.
Right? It's probably a lot moredigital or just binary on off
kind of stuff.

Yeah. Well, the outputs yeah. The inputs are
mostly analog, though.

Well, yeah. You have sensor readings and things
like that.

Yeah. You're writing reading sensors that are
usually, like, 0 to 5 voltsensors.

Right. So not, one thing you can do to
increase, reliability is dowhat's called an EPSA,
electronics parts stressanalysis, where you take every
single part and you say, here'sthe max rated from the customer
or not sorry, the manufacturer.Here's a resistor. It says that

(49:47):
it can handle 1 watt. What am Irunning this resistor at?
Am I running it at 1 watt? Is itat its maximum rated? Well, it
it will have a degraded lifecompared to if you're running at
700 milliwatts or if you'rerunning at half a watt. And so
what you can do is do an EPSAwhere you arbitrarily pick

(50:08):
derated values. So, say, aceramic capacitor.
Let's say, you want a ceramiccapacitor. I'm just gonna pick a
random number. Let's say, a capthat's on a 24 volt line. That
cap has to be rated for at least24 volts, so you could pick a 25
volt cap. Right?
Because that technically meetsderating or not derating. That
technically meets the maximumrated. But it might make more

(50:31):
sense to pick a higher rated capor, you know, pick any
parameter, derate it by acertain amount, and you can
actually extend the life of thatcomponent. And so if you want to
get further reliability numbersor if you wanna get further life
out there, pick arbitrarily or,you know, go do some research on
on, you know, industry, whattheir derating values are. A lot

(50:55):
of times with this stuff, it'sjust 50%, 50 or 60%.
A lot of times with with thingslike tantalum capacitors, it's
actually 30%. You derate all theway down to 30% of them of the
part, and then you get extendedlife out of it. So there's a lot
of little tricks like that wherethe circuit doesn't change. You
just pick an appropriatecomponent based off of what kind

(51:20):
of lifespan you're trying to getout of it. So there's really
fun.
If if you really wanna putyourself to sleep, there's fun
documents out there that just gopart by part that says ceramic
capacitors, here's theirderating values. Bipolar
junction transistors, here'stheir derating values. Blah blah
blah. And just just hundreds ofpages of if you got this, run

(51:42):
this test, do this derating. Ifyou got this, you know, I could
I could show you those if youreally wanna fall asleep.
But, actually, you know what?That that would be a fun thing
to do. If you come up with adesign, it would be fun for us
to do a an EPSA analysis.

Yeah. On that. Yeah. We can do it as a as a,
yeah, exercise. I don't know ifthat would make great podcast
content if

You know you you know what might make fun? Doing
it live would be awful content,but what might be fun is doing a
first stab at a design, and thenyou and I do the analysis and
see how close did you get toderating in a high reliability
application with your first gutcheck on parts, and then come

(52:30):
back and say, we had to changethis and this. Now it's really
difficult to say because wechange these things, you'll get
an extra 5 years or 10 years oflifespan. It's really difficult
to put a number on that, but,you can't say that you at least
analyzed it for higherreliability.

Yeah. Yeah. Well, looks like I need to start
a design then.

Yeah. I'm excited now. Let's do this.

Yeah. It's gonna be a lot of fun. I don't
know if there's any breakersstill listening, but would you
want us to do a podcast episodeabout that? Let us know in the
discussion, form.macrev.com. Sowe have this other topic.
We can just go right into it.It'd be a long episode.

Why don't you know what? Why don't we
actually, I'm gonna I'm gonnasay, because we're at an hour,
let's put this let's push thisone. Let's go ahead and leave
this for another time or yeah. Ithink That'd

be the first thing we'll talk about next
week.

I think that that

works. Yeah. It's the Justice Department sues
Apple from monopolizingsmartphone markets. So we'll
talk about that first firstthing. We won't get sidetracked
for 30 minutes talking about

Project manager.

AI. Yeah. Project management with AI. So
thank you, everyone, forlistening to circuit break for
MacFab. We are your hosts, PerkyDolan.

And Steven Craig.

Later, everyone.

Take it easy.

Thank you, Yes You Breaker, for downloading our
podcast. Tell your friends andcoworkers about circuit break
the podcast for Macofab. Please,we need more listeners.

Also, go give us a review.

Yeah give us a review I think there's a link we
have a new like email newsletternow I don't think you can
immediately sign up for it yetwhich is not good, but it
exists. And I bet you somelisteners here got that email on
Monday, but there's a link inthat email that we'll send you
to Apple to review us. I don'tknow if there's other places

(54:29):
where you can review us, butplease review us. We need more
listeners. Everyone out therethat listens right now is very
loyal y'all listen to everysingle word that we say even
this word right here So if youhave a cool idea project or
topic you want us to discuss orthe community to discuss let us

(54:50):
know It's form.macfab.com, wherewe talk about personal projects,
discussions about the podcast,engineering topics, and news.
That's form.macrofab.com. And Ijust fixed it. It used to if you
just went to there withoutlogging in, it would only show
you half the content there.Fixed it. So now you see
everything.
If you even if you you can gothere without even logging in

(55:11):
and just you can view. You can,like, peep into the window.

All Episodes

Episode Transcript

Popular Podcasts

On Purpose with Jay Shetty

Stuff You Should Know

Dateline NBC

.css-15opob5{left:0;position:absolute;top:0.8rem;} All Episodes

.css-14f5ked{margin:0;word-break:break-word;display:-webkit-box;-webkit-box-orient:vertical;box-orient:vertical;-webkit-line-clamp:2;overflow:hidden;}EP#428: Repairing the Retro Future

Episode Transcript

Popular Podcasts

.css-r6mb8g{margin:0;word-break:break-word;display:-webkit-box;-webkit-box-orient:vertical;box-orient:vertical;-webkit-line-clamp:1;overflow:hidden;}On Purpose with Jay Shetty

Stuff You Should Know

Dateline NBC

All Episodes

EP#428: Repairing the Retro Future

On Purpose with Jay Shetty