• Home
  • Help
  • Search
  • Login
  • Register
Pages: [1]
Author Topic: Should a dev kit be unreliable?  (Read 1351 times)
Full Member

Karma: 15
Posts: 141

View Profile
« on: August 30, 2010, 01:29:52 PM »


I cannot stay out of this debate any longer. There are too many apologists on this forum for Marvell and Globalscale.  When people complain about the problems with the  Sheeva Plug, the apologists immediately start singing the refrain "Its only a dev kit, dev kits are not reliable" 

This is not the case in the Sheevaplug and the Guruplug.  These devices fail due to shortcomings in the design process.  I have been an electronic design engineer for nearly 50 years and I have seen (and committed)  lots of sins in that time.

When a product is mooted, the marketing people produce a set of goals, what we used to call "design for's" . This goals were prioritised. and would include time to market, production cost,  packaging, performance, environment (operating and non operating) and reliability. These goals were handed to the designers, who went away and scribbled on a fag packet,  they then returned their list to the product manager of what was really possible within the two normally non negotiable parameters which are cost and time to market. By a iteritive process the design, specification etc of the product would reach the design specification stage, to which the design engineer would toil. Bear in mind, if a feature was not in the design spec then the design would not blink an eye, He has enough to do with what is on the list.

When I examine my Sheevaplug (v1.3) it becomes very apparent that the design goals for this product were 1. packaging, 2.  performance  3. cost.  ( not necessarily in that order).

The packaging meets the design goal, A smart "wall wart", nice plastic moulding, clever interchangeability of mains plug types for various countries.

The main pcb is a very nice design and all the features work to specification. It is well engineered and has no ECO wires or apparent changes.

I think that when the designers had acheived the above items , they had either run out of budget or time to finish the job properly.

The first design problem with the Sheeva was the quality of the bought in mains power supply.  Just visually comparing the quality of the main pcb with the power supply pcb reveals the power supply is very cheaply designed and constructed. The pcb is cheap biscuit (FR2) not glass (FR4).  The components are also cheap. They are standard commercial "entertainment" grade devices with a nominal operating temp spec of 0 deg C to  85 deg C. The construction  (components and wires),  positioning and soldering is awful. None of the capacitors in the power supply are rated for use in a switching reg, i.e. they are cheap aluminium electrolytics and not low ESR types as they should have been. These capacitors will fail very quickly.
The problems of the power supply life is further compounded by the high operating temperature the devices operate in.  I have measured the temperature in the power supply at the surface of the filter cap when operating at over 70 deg C.  The life expectancy tables of  cheap aluminium caps at this temperature show around a 200 day life (on 24/7). That is not taking into account additional failures due to high ripple current  on an non "low ESR" capacitor.
It appears that this was the cheapest power supply that could be obtained. WHY? Was there no budget available for a better one or indifference on the part of the product manager?

It is also obvious that there were no proper environmental testing of the product in development, otherwise the designers would have realised that the device does not have enough cooling in the small form factor chosen by the marketeers.   To any designer with any experience, just looking at the devices would raise some doubts, a couple of hours with a thermocouple meter would have proved it.

Why was the product allowed to ship with these defects?  Time to market or "Its only a dev kit, if it burns out it doesn't  matter" philosophy?

In that case why is the main pcb design so very good?

What I am trying to question, is that if Globalscale had spent the same amount of effort on the power supply and on the environmental budget for the devices as they had on the design of the main pcb then the product would have been so much more reliable,  that there would have been no one complaining about Globalscale.

The Sheeva plug is very nearly a quality product, some bad decisions let it down,  but not I believe the fact that it is a dev kit has caused these design failures!

I have not had the opportunity to examine a Guruplug, but reading all the problem posts on this forum, would suggest the device has an extreme heat budget problem, due to the extra features.  The remarks regarding the Sheevaplug could equally apply to the Guruplug.





Karma: 2
Posts: 8

View Profile
« Reply #1 on: August 30, 2010, 05:46:06 PM »

         Let be the first to say it:   "Well said".   You are 100% correct in your evaluation.  Having talked to Henry (from GlobalScale) at the PlugIN event, it sounds like there was a large dropoff in quality from his vendor (for the power supply).  OK, I can deal with that.   However, since there are so few other options (of vendors), that winds up affecting a very large percentage of the Sheeva/GuruPlugs out there.  And now most of our units are failing for the exact reasons that you discussed.

Without Marvell pushing the quality of the product and looking at the longer term (rather than how many ARM chipsets did we sell this quarter), this will take a long time to come to fruition.  And BTW, the Intel/Atom chipset IS catching up with regarding to power consumption Wink  To say otherwise is just naive...   The Asus EEE box only takes 15-16 watts at full chat and that is with a disk drive, wifi, output for a real display, etc.  (I know, I have one that I measured with my handy KillAWatt).  And that is with a (old, first gen I think) Atom N270.


Jr. Member

Karma: 1
Posts: 83

View Profile
« Reply #2 on: August 31, 2010, 12:50:16 AM »

I have not had the opportunity to examine a Guruplug, but reading all the problem posts on this forum, would suggest the device has an extreme heat budget problem, due to the extra features.  The remarks regarding the Sheevaplug could equally apply to the Guruplug.

Your remarks about the quality of the PCBs apply to GuruPlug as well. The main PCBs (two of them) look nicely designed. But the power supply PCB reminds me of those hand-made amateurish boards I used to make for a hobby in the old days.

Lack of knowledge is not such a big problem, unwillingness to learn is.

Pages: [1]
Jump to: