EEVblog #485 - Agilent TrueVolt 34461A Multimeter Teardown

[mikeselectricstuff]

... lamp ... proof of proper training on its use

 

How do people get to be in charge of making these rules? By banging their heads really hard on a wall like this? Whoever lasts the longest wins?

The problem is as soon as you put someone in charge of this sort of thing, they have to justify their existence and be seen to be doing something. Just like politicians, there are never enough rules/laws - if someone decided everything was OK they'd be out of a job.

[c4757p]

Whining that your multimeter has WinCE is like whining that the capacitors in your television have black plastic labels. "But I don't like black! They should be blue!" Who gives a damn, they work and you don't see them! Jesus, I've heard of wine snobs, but embedded OS snobs? Piss off...

Meant in the nicest of ways, of course.

[c4757p]

The problem is as soon as you put someone in charge of this sort of thing, they have to justify their existence and be seen to be doing something. Just like politicians, there are never enough rules/laws - if someone decided everything was OK they'd be out of a job.

Not necessarily true, though, or rule books would just be piling up everywhere around the world with the right way to lick an ice cream cone and which side of your head to part your hair on. Obviously it happens in some places, but most places aren't quite that draconian. There's definitely something more authoritarian and legalistic in some companies' cultures than in others.

[fpga]

The problem is as soon as you put someone in charge of this sort of thing, they have to justify their existence and be seen to be doing something. Just like politicians, there are never enough rules/laws - if someone decided everything was OK they'd be out of a job.

We strayed way off topic on this thread. I just hope you don't have policies or people like that where you work.

[jancumps]

Whining that your multimeter has WinCE is like whining that the capacitors in your television have black plastic labels. "But I don't like black! They should be blue!" ...

Wow, you worried me there for a moment. I had to check this:

[mikeselectricstuff]

The problem is as soon as you put someone in charge of this sort of thing, they have to justify their existence and be seen to be doing something. Just like politicians, there are never enough rules/laws - if someone decided everything was OK they'd be out of a job.

Not necessarily true, though, or rule books would just be piling up everywhere around the world with the right way to lick an ice cream cone and which side of your head to part your hair on.

You're not familiar with the EU then... they seem to do little else.

[ChadSeibert]

The problem is as soon as you put someone in charge of this sort of thing, they have to justify their existence and be seen to be doing something. Just like politicians, there are never enough rules/laws - if someone decided everything was OK they'd be out of a job.

Not necessarily true, though, or rule books would just be piling up everywhere around the world with the right way to lick an ice cream cone and which side of your head to part your hair on.

You're not familiar with the EU then... they seem to do little else.

Eh, there's always quibbling over existing laws. No need to argue about ice cream cones when you can just argue about the same thing over and over. 

[ejeffrey]

Ha ha, you've apparently never worked for large utilities, IT is worse than HR.

No, although I have worked for a large company.  Also I know that it is impossible for most companies that require non-trivial technology to avoid windows, at least if you count windows CE.  This meter is the tip of the iceberg.  I don't think you can even buy an oscilloscope that doesn't run some version of windows any more.  If the IT department tells the CTO that you can't use your $500,000 65 GHz oscilloscope, or your $10M photolithography system, or whatever, you know what happens?  The IT guy gets fired.

Seriously, these companies have the ability to set up private networks for running special purpose equipment.  If a regular person asks, they lie and say no because it is less work and gives them a power trip.  Once someone important finds out that they are impeding actually doing work, the solutions magically come out of the woodwork.

[SeanB]

At work I plug in my laptop, and have no problems with it, as all I am doing it for is to use a network printer direct. not going to connect to the Windows server just to print, so CUPS just works direct a lot faster. As well I am remote control for the IT provider, I often get a call to go and change tapes for the off site backup, as they either need to reinsert or the regular driver ( Jay) is running late, so I go and change tapes and leave them for him to collect the next day. Simple as there are 4 days of tapes there for use, just in case. The backups go off site, along with a hard drive kept elsewhere.

If somebody had to bring up the rule of not plugging in a phone charger probably most of the staff would be on strike the first day. All good and well, but if you do not actually have electricity at home and have a cellphone.......

[dr.diesel]

I don't think you can even buy an oscilloscope that doesn't run some version of windows any more.

Sure you can:

https://www.eevblog.com/forum/testgear/equipment-running-windoze/

The $500,000 65Ghz Scope is running a full blown version, you could load antivirus if you wanted - although not condoned I'm sure. 

But your correct, for that kind of money you could get the right people involved, but for $1000 meter, they'd tell you to piss off, especially when there are plenty of other alternatives.

[ejeffrey]

But your correct, for that kind of money you could get the right people involved, but for $1000 meter, they'd tell you to piss off, especially when there are plenty of other alternatives.

The point is, any company that has a significant investment in special purpose hardware has had to solve this problem already.  If you are one guy working on hardware in the corner at a large company that mostly just pushes paper around or does software development, then yes you have a problem, and not just for this meter.  If you work in an actual lab environment, this isn't a problem because it has already been solved.

Also, in my experience, if they won't let you connect something running WinCE to the network, they won't let you connect *anything* to the network except for desktop PCs and networked printers specifically purchased by the IT department.

[Alexei.Polkhanov]

Windows CE has been rock solid for 15 years.  The development environment is light years ahead of everyone else...Visual Studio is in it's own league and really has no competition.  Unix/Linux environments are much as they were 20 years ago...and 30 years ago...and 40 years ago.  It's like playing with tinker toys.  The threading model and libraries for Windows are more advanced and more usable than anything you'll find elsewhere.  Basically, it's a programmer's paradise if you approach it with an open mind.

Assuming you're going to use an OS at all, WinCE is the slam dunk, obvious first choice for a device like this, and by a wide margin.  Someone else mentioned cost.  I believe Core still costs <$5 per device.  It's not like I'm even a fan of Microsoft or Windows, but we're not talking about desktop environments.  It's about selecting the right tool for the right job.

Visual Studio is very good dev environment, I still consider it a nice 'editor' more than anything else and I would jump of it into WinDbg to debug. Choice of dev environment is purely personal preference. I may argue that when I use Emacs and compile on command line things seem to go faster. 

With regards to threading and other useful features (as in useful for EMBEDDED OS) I would strongly disagree. CE maybe somewhere between VxWorks and QNX since Windows CE has advantages for embedded UI apps, VxWorks is definitely better when code size, performance and hard real-time matters, followed by RTLinux which is free but has larger footprint, QNX has sexy, compact UI framework. Someone who used pSOS and other embedded OS probably will challenge my opinion here and they maybe right. For example Mars Pathfinder is running VxWorks, Cisco replaced (RT)Linux by VxWorks in WRT routers and saved money on memory even if VxWorks is not free.

Perhaps where CE has really failed them is longer than necessary boot times. I would say that anything more than 4 seconds is unacceptably long - this is just a fronted for DMM, not a space station. I have seen entire Ubuntu Linux booting faster on PC with flash drive!

[jancumps]

This thread now fully complies with the rule that if someone mentions an ide, someone else mentions vi or emacs. 

[SeanB]

Actually you really should not have to boot, if your hardware is fixed ( in a meter definitely) just have an image decompressed to ram on startup that is already running, in the init code you set up registers and such with boot code, uncompress image to ram, set all pointers and stack to a preset frozen state then the bootloader kills itself by jumping into an entry point, when the main starts running as if it never stopped from the initial state, and then sees the bootloader as free memory. Can be done in under 1s if you do proper image and compress it into rom and use a simple bootloader. If you want to do firmware updates have a simple link looked at by the bootloader code during startup which jumps to look for the new firmware and checksums it then writes it to flash then gives a message pass or fail and halts the processor. Reboot without the jumper and away you go..

[John Coloccia]

I like VxWorks.  I've used it and it's generally a nice environment to work in.  I certainly trust it for critical applications more than WindowsCE, and I suspect that's why you see it used frequently in space, military vehicles, etc.  It's missing some really great thread features that you find in Windows, though (it's generally Posix compliant).  For example, some of the most powerful features Windows brings to the table are things like Events, WaitForMultipleObjects(), and things like that.  Semaphores and Mutexes are nice, but the way windows handles thread synchronization is a thing of beauty.

But if it's going to run my life support system, VxWorks isn't a bad choice over Windows!.  I've used pSOS too.  Nothing really remarkable other than it's teeny tiny.  I guess Wind River bought it and subsequently killed it.  Didn't they do the same to RTLinux too?

[rsjsouza]

Actually you really should not have to boot, if your hardware is fixed ( in a meter definitely)

I don't disagree with your statement above at all, but in practice what I usually see in the OSes I worked with is they don't get it completely right. Over time a great number of suspend/resume switching ends up in unknown states or unexpected behaviour (peripherals don't get properly restored to their pre-suspend settings, external devices don't get properly configured, etc.). In this case the equipment manufacturer tends to go the safer way and simply restores the system to a known-good full reset configuration (apart from some minimal settings before turn off).

I also think the manufacturer does not spend too much resources in optimizing boot times for a bench equipment - from a marketing perspective this is not as significant as in a portable device.

[SeanB]

True, but as you know the hardware you set all the registers and other hardware registers on startup and force load them. Not going to take more than 1000 instructions to just write all the registers, less than 50mS to do it all at each reset whether from power on or soft reset. Not a probe, look up in table, decide on action and wait for timeouts from hardware to respond, just push it out to known stuff and boom you are ready to resume from an image ready to run.

[Alexei.Polkhanov]

Windows brings to the table are things like Events, WaitForMultipleObjects(), and things like that.  Semaphores and Mutexes are nice, but the way windows handles thread synchronization is a thing of beauty.

I see where you going, yes you absolutely right - WFMO is much more efficient than select() because you don't have to iterate over all handles to find one that triggered. I have been tracking this problem since 2001. I never had to solve it for embedded platforms, but for large servers on *NIX (like Linux) problem was known as C10K problem (http://www.kegel.com/c10k.html) because waiting on 10,000 sockets with select() is not very efficient. Back then in 2000 there were already several solutions like kqueue, then we got libevent, epoll etc. You can configure you kernel to use these or to enable them as option when you configure kernel for your platform. Problem is that you have to be aware of these options, pros and cons and so on. RTLinux kernel is not same as Linux. I have not heard any sell pitches for CE, but I can guess where they would try to attack RTLinux for example.

I think Linux kernel will always be ahead of any other OS in terms of features and supported platforms because it is available for everyone to change, and so every university and every developer can add/change whatever they want. And that is what they do. Latest experimental algorithms in OS design mostly come to Linux first, but because there are no single company behind it it is up to developers/users to discover and use them.

[John Coloccia]

I really thought RTLinux had kind of died with a whimper.  I should give it another look.  I didn't realize that it was still being actively maintained.  I guess I assumed that after Wind River gave up on it that it just became sort of an artifact of a more exciting time....the late 90's....LOL.

Oh, how fast things change around here

[Alexei.Polkhanov]

I really thought RTLinux had kind of died with a whimper.  I should give it another look.  I didn't realize that it was still being actively maintained.  I guess I assumed that after Wind River gave up on it that it just became sort of an artifact of a more exciting time....the late 90's....LOL.

Oh, how fast things change around here

Sorry for confusion I almost forgot about Wind River, what I was referring to by RTLinux is Real Time Linux, which is a Linux kernel with PREEMPT_RT patch. I think RedHat is maintaining it.  Whatever was done specifically for RTLinux now ended up in many other "Embedded Linux" distros.

Situation with Embedded Linux in general is very confusing. You can get lost in names and distros and never find your way out :-) For example I have Nokia phone, N900, that runs Maemo OS, then there was "MeeGo" which is a basis of current project called "Sailfish" developed by Jolla - company that was started by people ejected from Nokia after it signed its deal with Microsoft. Debian and Ubuntu have distros targeted for embedded/mobile as well.

Just look at list of Linux OS supported by Nios II (Soft core running on Altera FPGA):
...
Linux   Nios II   Timesys
Linux   Nios II   Wind River
Linux   Nios II   SLS
Linux   Nios II   CodeSourcery
Linux   Nios II   Open Source Community
...
µCLinux   Nios II   SLS
µCLinux   Nios II   Open Source Community
...

It seems that everyone has its own flavor of embedded Linux. It is a little bit of a Embedded Linux zoo.

[lewis]

Dave - When's the review coming? Looking forward to that one...

[EEVblog]

Dave - When's the review coming? Looking forward to that one...

It has been mostly shot. Just need a final few minutes of material and a summary.
It's already an hour long 
I'm going to have to leave out any PC based playing around for brevity.
So should be #489

[John Coloccia]

Windows brings to the table are things like Events, WaitForMultipleObjects(), and things like that.  Semaphores and Mutexes are nice, but the way windows handles thread synchronization is a thing of beauty.

I see where you going, yes you absolutely right - WFMO is much more efficient than select() because you don't have to iterate over all handles to find one that triggered. I have been tracking this problem since 2001. I never had to solve it for embedded platforms, but for large servers on *NIX (like Linux) problem was known as C10K problem (http://www.kegel.com/c10k.html) because waiting on 10,000 sockets with select() is not very efficient. Back then in 2000 there were already several solutions like kqueue, then we got libevent, epoll etc. You can configure you kernel to use these or to enable them as option when you configure kernel for your platform. Problem is that you have to be aware of these options, pros and cons and so on. RTLinux kernel is not same as Linux. I have not heard any sell pitches for CE, but I can guess where they would try to attack RTLinux for example.

I think Linux kernel will always be ahead of any other OS in terms of features and supported platforms because it is available for everyone to change, and so every university and every developer can add/change whatever they want. And that is what they do. Latest experimental algorithms in OS design mostly come to Linux first, but because there are no single company behind it it is up to developers/users to discover and use them.

At one point, I wrote a little library using posix semaphores and mutexes that simulated the way Windows handled events and things like that, and it was pretty efficient.  This was many years ago.  I forget the details of exactly what I did, but I think it was something along the lines of launching threads to wait on individual synch objects to get signaled, and then each caller actually waited on a different object that sort of signaled when something in their list was ready.  I guess it really ended up being kind of a thread synchronization mini-kernel that simulated what Windows was doing.

I think it's long gone, but I have some old stuff kicking around here.  If I happen to find it, and I think it's useful, I'll maybe post it here or start another thread on it somewhere else.

[lewis]

Dave - When's the review coming? Looking forward to that one...

It has been mostly shot. Just need a final few minutes of material and a summary.
It's already an hour long 
I'm going to have to leave out any PC based playing around for brevity.
So should be #489

Awesome. The longer the better!

[CChin254]

Here is the IC Identification of this device. The PDF (in the next post) has the cost distributions for the manufacturers and some other stuff.