• Industrial Automation Engineer (unregistered)

    <Sigh.> Just... <Sigh.>

  • Foo AKA Fooo (unregistered)

    A "biotech company that [...] had a product, but they couldn't produce enough of it" -- where have I heard this recently ... well, probably just a deja-vu ...

  • (nodebb)

    The HW problem is common, and the solution is not a bad one... BUT if the sensor wire was hard driven (say 100mA of current at all times, just in different directions) in both states there would not be the opportunity for to "pick up" the noise from the PWM....

    ps : Worked on exactly this type of stuff back in the late 1960s through 1990.... Has everything that was learned been forgotten?????

  • rtlgrmpf (unregistered) in reply to Foo AKA Fooo

    I seriously hope it's another one. We are already f....d enough.

  • akozakie (unregistered)

    This seems a bit unfair towards Roy. Sure, there are some WTFs, but it looks like the guy generally delivered on the promise. You get what you pay for. If the fix was sufficient (the article doesn't say that)... Well, they got a mostly working solution that they could get to work well enough for a fraction of the cost, when a market-standard solution was too expensive for them. And they were able to get the constructor on the phone for debugging when needed, so even post-sale support actually exists. OF COURSE it's going to be of lower quality, less reliable and harder to maintain, that's why they could pay less.

    Sure, he could do it all better. Design better, code better, build better. Guess what - if he had that level of skill and knowledge and was willing to put this much work into his products, he would also price his work far closer to the market average.

  • LCrawford (unregistered)

    At least Roy's wiring job was neat!

  • (author) in reply to TheCPUWizard

    Has everything that was learned been forgotten?

    … yes? Like 90% of the technology industry is to get frustrated with the current standard way of doing things, invent a new way that doesn't have the frustrations, discover that the frustrating parts were necessary and then reinvent them badly, and then act like this is something entirely new.

  • 🤷 (unregistered) in reply to akozakie

    This seems a bit unfair towards Roy. [...] You get what you pay for.

    Exactly. Roy delivered sooner than promised and cheaper than promised, for the fraction of costs that any other solution. What did they expect? That Roy was some kind of magician who can deliver "better than market standard" products for 1/3 of the price?

  • Pabs (unregistered) in reply to Remy Porter

    Are you describing Windows 10 there Remy?!

  • Red Five (unregistered)

    I think that if the home position was sensed by triggering a low rather than a high, it would also have solved the problem. But the sensor wires still should be isolated and shielded from the stepper driver wires.

  • MiserableOldGit (unregistered) in reply to 🤷

    I agree, it must have been one of Roy's first attempts at this sort of thing or he would already have known about the issue. He certainly should have been able to work out, given enough time with an oscilloscope and a dummies guide to electronics. It's a cheap rush job, sounds like a bit of a success to me, normally attempting that kind of workaround ends up costing more for less.

  • MiserableOldGit (unregistered) in reply to Remy Porter

    Completely wrong ... has to be far higher than 90%

    I get kicked out of meetings for saying things like "Yes, it is a fresh new idea, not at all like the identically worded suggestions I've heard a dozen times before, but that stubborn old bastard Mr Basic Physics just won't get on board. "

  • (nodebb) in reply to 🤷

    Roy delivered sooner than promised and cheaper than promised, for the fraction of costs that any other solution. What did they expect? That Roy was some kind of magician who can deliver "better than market standard" products for 1/3 of the price?

    This isn't as far-fetched as it sounds. If you only need the product to do 1/10 of what a market standard product would do, then 1/3 of the price to build just that 1/10 at reasonable quality may be plausible. (Not sure how that dynamic played out here, but e.g. it's exactly what happens when a client already has a market standard product and then contracts me to add a custom report.)

  • Brian Boorman (google)

    It has a name. It's name is "Crosstalk".

  • Angela Anuszewski (google) in reply to TheCPUWizard

    Or they didn't know about it in the first place. Depends on their background.

    At my first real job, (the one making radios), they had a policy of hiring EEs as their software developers. I asked my manager about the reasons for that. He said that in his experience, you can teach an EE to write software, but there's no hope of teaching a CS or SE how electromagnetism works.

  • Duston (unregistered)

    BUT if the sensor wire was hard driven (say 100mA of current at all times, just in different directions)

    Yeah, but a current changing directions also induces a current in a parallel wire.

  • Foo AKA Fooo (unregistered) in reply to akozakie

    As I understand it, the tragic part is that he probably spent considerable time on the software work-around when a simple electric solution would do. TRWTF to me was when he boldly claimed the whole batch of controllers was defective rather than considering that his newly built application might be the problem. (If it worked before, and suddenly fails with a new batch, that's when you can begin to suspect the batch to be defective -- but still more likely, some detail has changed, probably still within specs, and your application wasn't prepared for it. Occam's razor ...)

  • Best Of 2021 (unregistered)

    To be fair to Roy this is the sort of thing that is only obvious in retrospect, it isn't wrong with his software, and probably didn't show up when he was wiring stuff up manually on his test rig. He looks stupid after you find the wires are too close together, but it's not at all obvious in advance that insulated wires within the same cluster are going to mess you up.

    Works fine with software mocks and with the physical test rig, but not with the production boards? Blaming the board is not unreasonable, and bodging a solution when working on a tight schedule and budget, seems pretty fair enough to me.

    If there's a WTF here it's that they didn't put a physical robotics guy with Roy to deal with physical problems.

  • (nodebb) in reply to Remy Porter

    Remy - unfortunately I have to agree with you. "Shiney" (or "Squirrel") is possibly the bigest plage in tech.

  • (nodebb)

    Duston - There needed to be a voltage wire (ie NOT ground) in the cable, and a ground wire. Now instead the "single pole" switch use "double pole", not the switch wire is NEVER floating and susceptible to crosstalk, or pickup)

  • Dave (unregistered) in reply to MiserableOldGit

    Funnily enough, I used to make a good living simply translating between people like you and PHBs. I was basically saying the same thing you would, but in their language. If you do it right, the boss kicks everyone else out of the meeting and the work gets done.

    Then I realised the same skills transfer to a field where if something needs banging against a wall, I get to use a hammer instead of my head.

  • Worf (unregistered)

    The trick is you don't have a switch indicate HIGH when it's hit, but LOW. Active-Low logic exists for a reason. Interference from other lines will induce voltages on the sensor line which can have an impedance driving it. (Using more current is just a way of saying "stronger driver" and gets you some noise immunity).

    Of course, the easiest way is to pull up to HIGH and then have the switch indicate home by pulling that line to ground. The interference will have to be exceptionally bad for it to bring the signal line low, at which point the switch debounce logic typically cleans it up.

  • (nodebb) in reply to Pabs

    Node.js, maybe?

    Addendum 2021-02-05 00:58: [TODO: Continue adding to this list until it includes 90% of existing technology.]

  • (nodebb) in reply to Pabs

    The only thing in Windows 10 I experience to NOT work smoothly is the index search, especially in the start menu. It used to be very powerful and useful in Windows 7, then was gimped to the point of uselessness in Windows 8's mobile-like design and only partially restored with Windows 10 - for instance you can'd find "weekly_worktable.xls" by searching for "worktab" anymore, while classic start menu gives the result despite accessing the same underlying index.

    Other than that, I don't really get Windows 10-bashing.

  • 516052 (unregistered)

    It's the constant forced and annoying updates that are the primary problem. Like just now there was an update that literally disabled flash player without any recourse for the end user. I don't want my OS provider removing software from my PC without permission.

  • MiserableOldGit (unregistered) in reply to 516052

    Are you sure that wasn't the Adobe updater? Flash player has been uninstalling itself since it was deprecated in December. For most of us, that's good-bloody-riddance-not-a-moment-too-soon, but I can see how it could be mightily annoying and unnecessary for some.

  • 516052 (unregistered)

    Yep it was a windows update. Just google "windows update removes flash player". Am at work now so can't dig around too much.

    And say what you want about flash but it was a platform that supported far too much software for anyone to reasonably be comfortable with the idea that a corporation can just decide to kill it on a whim. It'd be like if Steam shut down tomorrow and took down all your games with it. Not cool.

    And I can understand not wanting to support it any more. That's fine. But actually forcefully killing it without giving your users any choice on the matter is where I draw the line.

  • DanK (unregistered) in reply to TheCPUWizard

    Worked on exactly this type of stuff back in the late 1960s through 1990.... Has everything that was learned been forgotten?????

    No, but competent engineers cost more than the off-the-shelf solution.

  • MiserableOldGit (unregistered) in reply to 516052
    And I can understand not wanting to support it any more. That's fine. But actually forcefully killing it without giving your users any choice on the matter is where I draw the line.

    Can't argue with that ... I assumed that must be due to some compelling serious security vuln, but apparently not.

    Must just be, as the thing has had its day in terms of new dev, they want to retire it into a support model that leaves those with dependencies no choice but to allow Adobe/Harman get their hand deep in your pocket and squeeze your nuts until they bleed. I suppose MS, Apple and the various others would play along as that's largely their business model too.

  • holy shit now I'm actually commenting here... (unregistered) in reply to 516052

    To be fair Adobe did give two and a half years advance notice.

  • (nodebb)

    At least Roy wasn't a "Real" electronics engineer.

    The kind who will write thousands of lines of nested for / switch state machines, indented so far to the right so as to ensure the all important single point of return (which as we all know is really the hallmark of good code). The purpose of all this being to carefully and intricately time when the limit switch is read, or implement some sort of hysteresis such that it was effectively "tuned" to the capacitive and inductive coupling between the specific cable he used to build the thing.

    Thus creating a heisenbug that only emerged on particularly hot or cold days, or came into being as the properties of the conductor changed over many thousands of cycles, or when the temperature changed by a couple of degrees, or when the cable manufacturer made the next batch of multi-core stranded cable which has slightly different metallurgical properties.

  • (nodebb)

    Nice to know somebody is doing hardware-software systems, I started to believe I was the only one and because, you know SOFTWARE IS CHEAP and SOFTWARE CAN DO ALL ZE THINGS!!!1! It's clearly obvious fault of the software programmer if things don't work and cannot be fixed.

    Such as... just insist software can magically rewire circuit boards so +5V won't be poured into the poor 3.3v logic.

    Or: insist insist the software should find a way to let a open-collector to ground mosfet close to +5V when off...

    But my favourite: insist the software can be done in 4 days and the schedule to be completely realistic. What can possibly go wrong.

  • Officer Johnny Holzkopf (unregistered) in reply to Angela Anuszewski

    Since when is high school physics "background"? And what academic adyoocaisien is needed today if a job applicant is expected to be able to read, write, and understand his/her native language?

  • Officer Johnny Holzkopf (unregistered) in reply to 516052

    Simple thing: it is no longer your personal computer if someone else is in control...

  • (nodebb) in reply to UserK

    Gotta love Electronics Engineers ;)

  • (nodebb) in reply to R3D3

    I don't know how this is related to this thread, but OSX search works TOO well. By default instead of searching the folder you're currently in, it will go through every email and document on your machine and return you so many results you've got no hope finding what you're actually after. Yes, it's fixed with one click, but I really wonder how big that index file is.

    Addendum 2021-02-07 17:07: OK, I'm told by my co-worker it's called "Finder" and that it also looks in the cloud.

  • (nodebb) in reply to TheCPUWizard

    Yes, its all been forgotten..

    In a previous place of employment a couple of years ago I was referred to as 'the IT guy'* and wasn't allowed to touch 'the wires'. In a particular project the wires were carrying a simulated encoder output from a PLC high-speed counter card. The outputs of these cards were documented as being fairly high-Z on low and cannot sink much. The cabling also had a lot of capacitance over the run. The result - at high speeds the slew rates basically meant the device receiving the encoder signal never saw it go zero. It may have not helped much that the grounding terminal for the PLC backplane was floating in the breeze when I saw it to. It was nigh on impossible to convince the Elec dept to put some pull-down resistors and check the grounding terminals, but after months of trying to convince them that was the problem they finally relented and it worked perfectly (albeit the resistors were also hanging in the breeze off some wago connectors that they had spliced into the wiring loom. Quality job.

    • actually a B Eng Comp Sys
  • Gummy Gus (unregistered)

    This kind of thing goes way back. Way back in around 1970, our research lab bought a very expensive tape recorder. It turned out to be useless on days with low humidity, as just playing a tape would cause minor little sparks from the unreeling tape, and that would throw the control logic into conniptions, like full rewind, like every ten seconds.

    I looked over the logic and it was hopeless-- it was first-generation fairchild RTW logic, which had about 0.3 volts of noise margin. No amount of added capacitors to bypass noise helped, not at all.

    Now some people could get rTl to work-- Seymour Cray used about 70,000 RTL gates in his 6600 supercomputers and they ran just fine. But most folks gave up on RTL logic, it was just too flaky.

    This was a 1975 $3000 tape recorder, so not cheap, but totally unusable during the low humidity winter months.

    In this fine article, they probably just needed like a 1K pull-down resistor and a 0.01uF integrating capacitor across each input.

  • eric bloedow (unregistered)

    reminds me of "twisted pairs"...wires were twisted around each other in a helix pattern to avoid a similar problem.

  • Roy (unregistered) in reply to akozakie

    It's the old saw:

    "Good, fast, cheap - pick any two"

  • Some Ed (unregistered)

    Not every start up in every industry. Back in the day, I worked for a contracting startup, which went to making a profit so quickly that their financial founder was employee number 2, because they needed to have as an official employee the first person they sent to work on a contract, and that was before they finished getting their incorporation documentation finalized. I've heard of other contracting companies that experienced the same thing.

    Sure, I have picked a point in their existence when they didn't yet officially have costs, because they didn't officially exist yet. That having been said, in the course of their first five years, they mostly hired employees after they'd contracted them out to do work, which basically meant that all billable employees were generating billable hours on day one, until they reached the sophistication level of having new hire training. We still did sometimes, even after that - I was hired on in year six, about two weeks after they'd contracted me (or whoever else they could hire) to do work, and my new hire training happened six months late because they needed me on the job before my start date.

    Yes, they had overhead employees, and overhead costs, such as the office space rentals. But for those first six years, they were consistently late on hiring all of those. When the founder finally managed to sell the company off, the new owner did an audit, and found more money in unbilled work than the founder had been looking to make by selling off the company. Most of that unbilled work was simply from having not managed to hire enough accountants and billing department staff to keep up with all the work the company was doing.

    To be fair, that sort of thing is not endemic to the contracting industry, there are a lot of startups that do struggle there. Just not all of them. I suspect there are other industries where not all startups have the profit struggle as well, but I don't know what they would be.

Leave a comment on “Coming to Grips”

Log In or post as a guest

Replying to comment #:

« Return to Article