On Mon, 6 Mar 2017 23:38:01 -0500, rickman <gnuarm@gmail.com> wrote:
>On 3/6/2017 9:35 PM, eric.jacobsen@ieee.org wrote:
>> On Mon, 6 Mar 2017 18:00:21 -0500, rickman <gnuarm@gmail.com> wrote:
>>
>>> On 3/6/2017 5:21 PM, eric.jacobsen@ieee.org wrote:
>>>> On Mon, 6 Mar 2017 16:41:11 -0500, rickman <gnuarm@gmail.com> wrote:
>>>>
>>>>> On 3/6/2017 4:22 PM, eric.jacobsen@ieee.org wrote:
>>>>>> On Mon, 6 Mar 2017 15:58:09 -0500, rickman <gnuarm@gmail.com> wrote:
>>>>>>
>>>>>>> On 3/6/2017 3:18 PM, eric.jacobsen@ieee.org wrote:
>>>>>>>> On Mon, 6 Mar 2017 14:45:47 -0500, rickman <gnuarm@gmail.com> wrote:
>>>>>>>>
>>>>>>>>> On 3/6/2017 1:15 PM, Tim Wescott wrote:
>>>>>>>>>> On Mon, 06 Mar 2017 06:37:01 +0000, Steve Pope wrote:
>>>>>>>>>>
>>>>>>>>>>> Les Cargill <lcargill99@comcast.com> wrote:
>>>>>>>>>>>
>>>>>>>>>>>> Steve Pope wrote:
>>>>>>>>>>>
>>>>>>>>>>>>> I've recently seen in job listings, contract requirements etc.
>>>>>>>>>>>>> references to desiring engineers who are "experienced with Maker
>>>>>>>>>>>>> boards".
>>>>>>>>>>>>>
>>>>>>>>>>>>> What exactly is a Maker board? I gather it's not a brand name,
>>>>>>>>>>>>> but some general concept ... it sounds a little new-age or something.
>>>>>>>>>>>>> Is there a specific meaning?
>>>>>>>>>>>
>>>>>>>>>>>> Arduinos and Raspberry Pi are market leaders in "maker" market space.
>>>>>>>>>>>> "Maker" is a sort of movement where kids learn electronics in a club
>>>>>>>>>>>> environment using these and small breadboard-style peripherals.
>>>>>>>>>>>
>>>>>>>>>>> Sounds great.
>>>>>>>>>>>
>>>>>>>>>>> Maybe my concern is a prospective cutomer will take the stance that, a
>>>>>>>>>>> consultant would usually charge $80K to do a certain design,
>>>>>>>>>>> but with a "Maker Board" they can do it for $10K.
>>>>>>>>>>>
>>>>>>>>>>> My larger concern is .. they might be right. :--)
>>>>>>>>>>
>>>>>>>>>> Then we consultants should learn to do one-offs with maker boards!
>>>>>>>>>
>>>>>>>>> Who doesn't? Why would anyone *not* use a production board for a design
>>>>>>>>> if the volume isn't thousands? But if the term "maker board" is limited
>>>>>>>>> to the low cost units made for the hobbyist market, then I say there is
>>>>>>>>> a much smaller market for commercial use than boards made by a
>>>>>>>>> commercial vendor. The Beagle Board and rPi seem to be stable
>>>>>>>>> platforms, but you can't count on compatible units remaining in
>>>>>>>>> production for a long time.
>>>>>>>>>
>>>>>>>>> The cost of having to port software even between an rPi and an rPi 3 is
>>>>>>>>> *much* more than the savings of not using a commercial unit available
>>>>>>>>> for 10 years. It all depends on the project requirements.
>>>>>>>>
>>>>>>>> A large number of the cheapie boards run Linux on ARM cores, e.g.,
>>>>>>>> Beagle Bone, RPi, CHIP, Orange Pi, are essentially all software
>>>>>>>> compatible. i.e., executable developed on one will run on any of
>>>>>>>> them.
>>>>>>>>
>>>>>>>> Even if they're not ARM, if it's a reasonable Linux distribution it's
>>>>>>>> often just a recompile to port to a different platform. This assumes
>>>>>>>> that you've been careful about library usage, etc.
>>>>>>>
>>>>>>> It also assumes you are doing work that doesn't actually *use* the
>>>>>>> features of the board such as I/O. The TI chip the Beagle uses has
>>>>>>> auxiliary processors. You would need a cloned board to port that code.
>>>>>>> There's often a lot more to embedded apps than network calls.
>>>>>>
>>>>>> That was essentially the caveat on libraries that I made.
>>>>>>
>>>>>>>>>> It's also good to educate customers on how, while our bit looks
>>>>>>>>>> impressive as a line item, it's just one part of a larger whole, and we
>>>>>>>>>> (hopefully) know how to do our work to minimize the rest of the life-
>>>>>>>>>> cycle costs.
>>>>>>>>>
>>>>>>>>> Mostly that comes from not buying low priced, but short lived hobbyist
>>>>>>>>> hardware.
>>>>>>>>
>>>>>>>> This is a concern, but not a big one. Even the mainstream
>>>>>>>> development hardware is often a fast-moving target with short life
>>>>>>>> cycles.
>>>>>>>
>>>>>>> And the short life cycle products are mostly the higher volume products.
>>>>>>> Short life cycles don't mesh well with high NRE like custom
>>>>>>> development. Most projects with limited production and significant
>>>>>>> software development will either be long life cycle products or if short
>>>>>>> life cycle have high profitability which we all know is not so easy to
>>>>>>> find. These are relatively infrequent or we here would all be
>>>>>>> millionaires.
>>>>>>
>>>>>>>> I think the "maker" and hobbyist hooks in the cheapie dev
>>>>>>>> platforms actually make them live longer. For example, Arduinos have
>>>>>>>> been around for how long now? Way longer than most professional
>>>>>>>> development platforms. The Beaglebone has been around for a long
>>>>>>>> time and was one of the first low-cost Linux-ARM platforms and still
>>>>>>>> enjoys a lot of use, I'm sure patly because the hardware is totally
>>>>>>>> open-source. There are still people making special-purpose stuff
>>>>>>>> that is pretty much just a Beaglebone board with a few hacks to place
>>>>>>>> some specialized hardware on the board.
>>>>>>>
>>>>>>> Your definition of "a long time" is different from mine. The BeagleBone
>>>>>>> (BB) was introduced less than six years ago and is not totally
>>>>>>> compatible with the BeagleBone Black (Black) which was introduced not
>>>>>>> quite four years ago.
>>>>>>
>>>>>> Cycles move quickly these days. Not keeping up means early
>>>>>> obsolescence to competitors that do.
>>>>>
>>>>> Low volume designs can't afford to redesign every couple of years. High
>>>>> volume designs don't need maker boards.
>>>>
>>>> There are many low-volume applications that can afford frequent spins.
>>>> There are a number of markets where the electronics are the "cheap"
>>>> part of the system or the margins are high enough that the design
>>>> refreshes are worthwhile.
>>>
>>> You aren't making sense. It isn't the hardware that we are talking
>>> about getting redone, it is reworking the software to match the
>>> hardware. If it is a low volume app, it is hard to pay for NRE. Jobs
>>> where the margins are high are few and far between because they get
>>> grabbed quickly and if competitive the margins drop quickly.
>>
>> And like I mentioned before, many of these little cheap boards have
>> compatible ARM cores and all run Linux, so if you're careful about
>> your development process and interface modulatarity, you maximize
>> portability. Of course, planning for portability can be a good thing
>> for any project, but it is greatly eased in this case by common CPUs
>> and OSs. If you do it right, you don't even need a recompile.
>>
>> This was discussed already.
>
>Sure, but only if you use none of the features of the boards. This was
>discussed already.
No, you use the features through portable software interfaces, e.g.,
be careful with the libraries, or abstract the interfaces with careful
modular architecture, as previously discussed.
I move I2C and USB devices across these platforms with no change in
the software. Can even wiggle pins and use other hardware
resources, sometimes which are unique to one platform, but just detect
and gracefully adapt. It's actually been very simple to do and not
particularly time consuming. Same exact executable runs across
several platforms.
>>> High margin jobs are usually low volume and are limited by how much can
>>> be spent on the NRE. If they weren't high margin they wouldn't get done
>>> at all.
>>
>> There are all kinds of economic models for projects. That's one.
>> ;)
>
>It's actually basic arithmetic. To pay for NRE you either need large
>volume, high margin or both. When jobs are high margin they get grabbed.
Sometimes the funding does not come from unit sales. As I said,
there are many, many economic models for projects.
>>>> Low-cost, low-margin applications can benefit even more from the
>>>> cheapie "maker" board, because they update frequently. So from that
>>>> perspective it can be easy for a low-volume application to update for
>>>> cost savings or other reasons.
>>>
>>> Low cost low margin only makes sense with high volume. Then custom
>>> hardware does *exactly* what is needed without paying for extra
>>> features. This is basic economics.
>>
>> There are other times when low cost, low margin makes sense, and I
>> already mentioned one, where the electronics is not the main part of
>> the system. There are other models as well. The simple economic
>> models are good for generalities, but often real projects go in other
>> directions.
>
>You mean where the margin is in a part of the system that isn't the
>electronics. If the volume is low and the margin is low, there just
>isn't profit to bother with. There has to be profit somewhere or why do
>the job?
Sometimes there are reasons beyond money or sales income.
>>>>>>> Before the Black was introduced I inquired about the future availability
>>>>>>> of the BB and was told to *not* design it into any products unless I was
>>>>>>> willing to build my own hardware. They were not interested in
>>>>>>> supporting commercial apps meaning, they wanted to be able to make revs
>>>>>>> to the design without worrying about maintaining compatibility.
>>>>>>>
>>>>>>> I haven't inquired about the Black in this same way, so I don't know if
>>>>>>> they support commercial use or not.
>>>>>>
>>>>>> It's open-source, and TI doesn't care. There really isn't a "they"
>>>>>> any more. Using the design means populating TI parts, so they're
>>>>>> happy for people to run with it.
>>>>>
>>>>> You are confused. My comments were about *using* the Beagle boards in
>>>>> commercial work, not copying them. They recommend you do your own
>>>>> production for commercial work. The design is open source, but for a
>>>>> low volume project, why would you want to pay for ramping production,
>>>>> debugging, etc???
>>>>
>>>> Using one or not will just depend on the requirements for the
>>>> particular system. I still don't know who "they" is, but there's
>>>> nothing stopping anybody from sticking a COTS board in a box to do a
>>>> job and this is done more than some people might expect.
>>>
>>> No one said you can't use a Beagle board. But Beagle (I specifically
>>> conversed with the guy behind it all) may change their boards at their
>>> whim and have. They have removed some features. That's why they say
>>> *don't use these boards for commercial work*. The big name (and often
>>> big buck) vendors don't do this without advance notice at least if at
>>> all. Even then they only make mods when required by a bug usually.
>>
>> Most vendors can change their products on a whim, or drop them
>> entirely. With a BBB or Orange Pi you at least have the option of
>> ramping it up yourself or taking it to somebody else to make, but you
>> probably won't need to because the existing ecosystem may pick up the
>> slack on its own, since it can. That can't happen with proprietary
>> commercial products.
>>
>> I feel like we already had this discussion, too.
>
>Yes, because you keep repeating the same "most vendors" stuff.
>Reputable embedded hardware vendors make a commitment to their
>customers. If you choose someone else you are looking to be burnt.
HAHAHA! Seriously, crap happens to companies, to their management, to
the economic environment they live in, to their banks, to their
investors, and promises can't be kept. Boards change direction or
companies pivot and exit markets and drop entire divisions. Been
there too many times. Promises and commitments are only that. If
you bank on reputations and promises you may get burned. Get burned
enough times and you figure out that it's just part of the reality of
business.
>The utility of taking over production of an obsolete product is not very
>useful, especially when that product is obsolete because of EOL
>materials. You can't build a board you can't get parts for.
I've no idea what your point is here. It doesn't seem relevant to
the discussion.
>I've already stated this before too.
>>>>>>> THAT is the big difference between "maker" boards and commercial
>>>>>>> hardware. With commercial hardware your app will continue to run for
>>>>>>> the life of the product. That is why companies like BittWare, Signatec,
>>>>>>> RTD Embedded Technologies and many others exist.
>>>>>>
>>>>>> Actually, the "maker" boards generally *are* commercial hardware. And
>>>>>> my experience hasn't matched yours on "pro" platforms supporting
>>>>>> legacy users. You're always at risk to market forces, and if there's
>>>>>> not continued money in supporting your chosen platform, don't be
>>>>>> surprised to become an orphan.
>>>>>
>>>>> Which of the above companies ditched you? I've had chips obsoleted, but
>>>>> that was after more than 13 years in production. I can still buy them
>>>> >from a 1st tier disti who bought around 100k of them. Commercial boards
>>>>> are typically not removed from production until a chip is EOL'd. Once a
>>>>> board is in production the costs of maintaining that is minimal. The
>>>>> price may go up as the production volumes go down, but why kill a cash cow?
>>>>
>>>> I wish I could remember all of them to count how many times over my
>>>> career redesigns or development path changes had to happen because a
>>>> supplier or expected supplier unexpectedly dropped support of
>>>> something. This happens with big or small suppliers.
>>>
>>> Sure, products have a finite lifetime. But when you say "something",
>>> that is a big area. Embedded boards from a reputable vendor just don't
>>> get EOL'd without notice just like chips. I get advance notices from
>>> Lattice semi when they make a change in the way they make or package a
>>> chip!
>>
>> Good for you. Buy more stuff from them.
>
>I do. If you don't buy from someone who treats you the same way you are
>looking to get burnt and to burn your customers.
>>>>>>> If your application really only needs a few years of production or can
>>>>>>> afford a porting effort, then low cost units suffice. But if low cost
>>>>>>> matters, you likely aren't in high volume and low price isn't a real
>>>>>>> factor.
>>>>>>>
>>>>>>>
>>>>>>>> So I think many of the cheap hobbyist hardware platforms actually wind
>>>>>>>> up with longer life cycles than pro stuff.
>>>>>>>
>>>>>>> Only if you don't mind paying for the software over and over again.
>>>>>>
>>>>>> What software? One big reason to use the Linux platforms is that
>>>>>> much of the support and tools are open-source and free, not to mention
>>>>>> the OS itself.
>>>>>
>>>>> The software that is written for the app! If it is anything other than
>>>>> simple networking code, it uses features of the board that may be
>>>>> different in new versions. Even the BeagleBone and BBB are not 100%
>>>>> compatible in I/O etc.
>>>>
>>>> Again, that just depends on how careful you are on modularizing
>>>> dependencies on libraries or various interfaces. These days it's even
>>>> easier to make things modular or platform portable, e.g., use a USB
>>>> interface (or any standard interface with library support) to
>>>> application-specific hardware. The demodulator I linked was
>>>> developed on a BBB, but so far has ported with no changes to a wide
>>>> range of Linux/ARM platforms, and just recompiles to other platforms.
>>>> A little bit of planning helps.
>>>
>>> Modularize? Sure, decision hiding. But you still have to regression
>>> test the entire system when you make a change as large as replacing the
>>> hardware with an update. That costs money. Low margin or low volume
>>> apps have a hard time paying for this. With commercial hardware you
>>> will be given notice and can plan for it. With "maker" boards you find
>>> out when your last shipment starts failing in the field.
>>
>> In my experience in your last two sentences you can transpose
>> "commercial hardware" and '"maker" boards' and still be right.
>
>Then you are buying the WRONG commercial hardware. You also need to
>study the units you design in and ask questions.
I think most practicing engineers would love for it to be the paradise
you describe, but the reality of several decades of experience has
indicated otherwise. Clearly it is wise to pick vendors with the
highest probability of reliability, but nobody can deliver 100%.
This is why the really serious projects often require second sourcing
or lifetime buys. It doesn't matter what people promise, reality
often gets in the way. If you recognize the limitations you're in a
much better position to react when it goes wrong. If you rely too
heavily on promises or previous reputation or whatever, you can get
burned.
>> Back in ancient times we spent a lot of time and money trying to
>> figure out why existing software running on AT&T 1610 DSPs suddenly
>> started exhibiting a lot of very nasty bugs. Turned out a mask
>> change in the die had significantly changed the behavior of a few
>> instructions. Were we notified? Hell no. We had to figure it it
>> out on our own, even identifying the mask ID and which batch the
>> change happened, which wasn't easy. Were they apologetic? No. This
>> caused a huge problem.
>
>I guess that's why ATT doesn't make DSP anymore. Burn people once and
>they tend to not come back.
>> We stopped using their parts after that.
>
>Sometimes all you can do is lick your wounds.
Ah, BUT THE PROMISES! Such a reputable vendor! No large industrial
supplier would do such a thing!
Don't be fooled. This can happen with any vendor. Other very large
silicon suppliers have done the same. Remember the Pentium bug? The
list of products and companies that stuff like this has happened with
is very long, some no longer exist, some do. It's just the reality
of doing business.
>> I have many more stories about large, commercial vendors dropping
>> products without notice, making undocumented changes, etc. Companies
>> get bought, merged, sold, and the new management can change the
>> product direction or close entire divisions on short notice. It's
>> just how stuff works and you have to be prepared for it.
>
>Yep, that happens sometimes. At least you don't have the vendor telling
>you to not use their products in a commercial design because they want
>to be able to make changes without worrying about past customers usage.
>Make that mistake and you can only blame yourself.
>>>>>>>>>> My biggest concern about designing such a thing into a project is the
>>>>>>>>>> quality and product lifetime of the board -- I'd certainly brief a
>>>>>>>>>> prospective customer on said concerns if that's how they wanted to go.
>>>>>>>>>
>>>>>>>>> Yes, exactly. But there are applications with short product lives or
>>>>>>>>> that are not very hardware dependent. This is a factor that needs
>>>>>>>>> careful consideration.
>>>>>>>>
>>>>>>>> For something like the BBB (Beaglebone Black), the hardware (board)
>>>>>>>> design is open-source and free. Even if the multiple vendors all
>>>>>>>> stop making it, you can ramp it yourself as much as needed.
>>>>>>>
>>>>>>> Avoiding making a board is the whole point to using a COTS product. I
>>>>>>> made a lot of money supplying a custom board to a major networking
>>>>>>> company, mostly because they didn't want to incur the cost of ramping up
>>>>>>> production. There's a lot more to building a board than having the BOM
>>>>>>> and artwork. BTW, they already owned the design, they paid me to do
>>>>>>> that so it was theirs to begin with.
>>>>>>
>>>>>> Good for you, but there are many reasons to elect to use or not use
>>>>>> any COTS or custom platform, and there are many reasons why even a
>>>>>> tiny company may elect to do custom hardware rather than COTS, for
>>>>>> large or small quantities. It's a big world out there with lots of
>>>>>> crazy stuff going on, so the solution spaces vary greatly. The
>>>>>> cheapie "maker" or "embedded" platforms are just one potential
>>>>>> direction among many, but for some applications they're an easy way to
>>>>>> get stuff done quickly and cheaply.
>>>>>
>>>>> Yep, there is a market for "maker" boards, but that is limited to
>>>>> products at lower volumes and a short life span. The basic tradeoff is
>>>>> NRE vs. recurring costs. Low volume jobs need to minimize NRE and so
>>>>> can use "maker" boards as long as there is no need for a long life span.
>>>>> That usually means not much profit so again, a squeeze on NRE.
>>>>>
>>>>> If it isn't worth spending some money on, there isn't much money to get
>>>> >from it.
>>>>
>>>> It's just another tool in the box that can be very useful.
>>>
>>> Another tool for sure. The issue is just how useful. I prefer to hang
>>> my hat on commercial hardware where I have rapport with the vendor. The
>>> jobs where that isn't practical due to cost are few.
>>
>> I'm seeing more and more where leveraging low cost tools, platforms,
>> and test equipment has made a big difference, sometimes between having
>> a feasible project or not. Being able to offer that advantage to
>> clients can be valuable.
>
>That's great as long as they are aware of the potential problems it can
>create. If you had used the rPi model A in a design in 2012 you might
>have problems with the newer model A+. To start with the mechanical
>design is different. Connectors have moved and changed. They don't
>expect anyone to be using these boards in embedded systems and you can
>expect various changes and product deletions.
Clearly risk discussions should happen in any project. That's part
of the tradeoff analyses.
---
This email has been checked for viruses by Avast antivirus software.
https://www.avast.com/antivirus