Wednesday, 7 February 2018

The frustrating bit...

The FB2900 is close, and one of the big things that is adding quite a bit of time, and a lot of cost, this time, is new safety standards. These were not in place for the last product.

Launching a product, especially an electronic one, has a lot of hurdles. It used to be the EMC testing (generating radio/electrical noise, and being susceptible to such) were the "biggies" that needed expensive external test houses. The "electrical safety" was a minor part of testing (can you poke your finger where it could get zapped?)... To be fair it was a bit more - things with mains power have to meet various rules for air gaps and insulation, but there are well known and tried and tested ways to design this in.

However, times have changed... The new safety standard (a little gem called BS EN 62368-1:2014) has a lot of things in it.

This is good!

I have seen too many things with power bricks melting and nearly causing fires. However, it is new, and this means some of the things you need to know when designing are not 100% clear. Unless/until the examiner at the test house makes a decision based on his reading of the spec, you cannot be sure. It is quite possible that something seemingly simple can catch you out and cause lots of delay.

I am pondering doing a more comprehensive blog some time, but here are just a couple of examples of the issues we have had to consider...

Holes matter!

The FB2700 had a grid of holes on the side for ventilation. The new FB2900 is way lower power (typically 5W) and does not need holes. We originally planned to leave the case design the same, but turns out the holes matter. Because they were more than 3mm this meant the case is not a "fire enclosure" (I think that is the term). This means things like deliberately overloading the power module to try and break it, and setting the damn thing on fire to see if any plastic comes out of the holes! Thankfully we do not need the holes any more, and the screw holes for rack mounts are small enough not to count. This meant a minor rework of the metalwork which added more delay.

Labels matter!

The front and back have screen printed polycarbonate labels. Obviously these are purely cosmetic, they could not possible matter, surely? After all the serial number label has all of the safety warning and CE marks and so on.

Wrong, they matter. Or might matter. Apparently there are factors of "touch temperature" - i.e. if the metal case is at max operating temperature, can you hurt yourself touching it while plugging in a cable? That sort of thing. I don't fully understand this one yet, but it meant the labels matter.

It gets more complex! There are LED holes that are over 3mm, and they are plugged by light pipes which are flame retardant and rated to the max operating temperature, so plugged holes. However the label provides an extra layer of flame retardant over the hole just to add extra safety. So err on side of safety and use flame retardant labels.

When we did the FB2700 the symbols and warnings on the labels were a simple checklist. Like power supply ratings, and CE mark, and class II (double insulated) mark, and so on. We simply put them on and we complied. Now it is part of the safety testing to check the label and check the associated paperwork (the quick start guide). So the production sample we send really has to be proper finished and paperwork and label and everything. That adds more delay.

The good news

This may sound depressing, but it is really good that things are tested this well. We are winning!

The production sample is off to the test house to finalise their testing and approval. Once we have that, we can stamp a CE mark on it and start shipping. Of course, something could come back and bite us and add more delay, but fingers crossed. We have a good hardware team on this and they do a damn good job, so should not be any issues.

The future

There is one thing that is not tested at all, and I can see that in 10 years time this may be very different. This is a network device. It is increasingly important that all network devices are sensible in the way they work - controlling access properly, protecting against attacks, and ensuring updates and patches as needed. None of this is subject to any formal standards or tests now. I wonder how that will change over the years. I wonder how easily a wince or linux box will pass? Fun times.

5 comments:

  1. As part of my work I recently came across the DO-326a spec which might be an indication of where things go...

    ReplyDelete
  2. So there already is a standard relating to "cyber security" (iso IEC 27001 iirc) that a number of industrial control system hardware is starting to be sold as compliant to. I feel that is more driven by a customer demand from businesses insisting that new hardware has cyber security without actually understanding what they are asking for. As such the standard is pretty weak, qualitative and entirely optional.

    And I also believe there is a specific standard that anything that plugs into a phone socket needs to comply with. But that's more for the protection of openreach's gear than for the benefit of consumers.


    So it's not totally impossible that things are going to start appearing saying they comply with a "cyber security" type of standard and that will eventually lead to it being a defacto requirement. After all, why would I buy your widget that doesn't comply when there are 100s of others that do.

    ReplyDelete
  3. And meanwhile, hoards of Chinese manufacturers just slap a CE mark on their gear and flog it on eBay and Amazon, with roughly 1% chance of any action being taken. 😠😠😠

    ReplyDelete
    Replies
    1. Just slapping a CE mark on gear that's compliant is legal. CE marking for almost all equipment (the rare exceptions are things like implanted medical devices, where if they go wrong somebody will definitely die) is self-certified. The manufacturer affixing the CE mark certifies that they've checked their product complies. They can use third parties if they want, and doubtless a small outfit like A&A would choose to, but if you make 50 000 units per day you can have a team that does your checks in house. There are probably hundreds of people in China who have responsibility for conformance to the exact standard RevK mentions at their employer.

      Self-certification is a sensible approach because it balances risk, these are relatively low risk products (compared to pacemaker, or a jet engine) and so spending far more to achieve a slightly safer product isn't worth it.

      My mother used to make stuffed toys. Not on an industrial scale, she probably sold a few hundred a year at most. But since they were for sale, and are clearly toys, even back then they needed a CE label. She was able to purchase the test kits needed (to check materials used won't hold a flame for example, and that eyes and noses can't be torn out by an angry toddler unless they have Hulk strength) and have fabric CE labels printed with her name and address on them, very affordably. If she'd needed an independent evaluation of each toy it'd have been far too costly to bother at all.

      Delete
    2. Nick, good point. I wouldn't mind so much if the gear actually met the standards, but far too often that's not the case. From US power plugs plus a Hong Kong deathdapter* to disconnected protective earth, via crappy voltage separation and terrible power factors, there are countless small manufacturers that have very little incentive to comply. It makes a mockery of the CE trust mark, and results in unfair competition.

      *—© Big Clive on YouTube

      Amazon and eBay might be complicit in this. Amazon in particular are arguably legally liable for those products that are Fulfilled by Amazon.

      For the avoidance of doubt, I'm talking about the smaller, scrappier manufacturers, typically. Not the established ones. Also, although the comment published says "Chinese" I edited it soon after to read "overseas," which reflects what I actually mean (for whatever reason, that edit hasn't been published).

      Delete