iSixSigma

Digital design project opportunities

Six Sigma – iSixSigma Forums Old Forums General Digital design project opportunities

Viewing 9 posts - 1 through 9 (of 9 total)
  • Author
    Posts
  • #43812

    digitalman
    Participant

    One area that my company struggles in is finding opportunity for Digital Hardware Design Engineers to utilize Six Sigma processes in the design phase.  This seems to be due to the limited parametric data that digital designs offer.  Are there some concrete examples that people have of using the various tools on their design projects? 
    I’m hoping for something more parametric and “advanced” than VOC, fishbone and the like. Maybe some RSM, HW/SW integration Monte Carlo, more math-based.
    Thanks.

    0
    #139474

    Heebeegeebee BB
    Participant

    Not a fan of VoC/QFD, huh?
     

    0
    #139475

    digitalman
    Participant

    Well, actually I think they’re fine, they just seem to be much more applicable to a different aspect of design than the board-level work that the majority of designers spend their time at.
    They are a lot of work, if done formally, and (I think) more appropriate at higher project scopes.  I’m hoping for some examples of 6sigma tools used on the outputs of VOC that trickle down to the implementation groups.

    0
    #139509

    Semiconductors Senior
    Member

    Hi digitalman,
    May I ask you: what kind of digital hardware do your engineers/colleagues design? Can you please give examples? Where do you see limitations in parametric data and why? Why do you (or your environment) think “Six Sigma is hard to apply for us” ?
    I admit the digital-view tends to hide opportunities. So I’d like to understand your specific situation better, if that’s possible in the public.

    0
    #139512

    Saherngu
    Participant

    We’ve been struggling years coming up with plain old metrics for digital/analog/layout (Hardware) design. We’ve thought about design complexity/effort for a productivity metric but haven’t done anything further than think about it.  Need a process with measurements to even begin to apply six-sigma for quality. We’ve started counting board respins but reasons could be poor requirements instead of poor design. We’re trying to come up with measurements on Change Orders due to design error. But is that going to help get the “HW design process” under control? A thought i have is to record the design defects caught at Critical Design Reviews. 

    0
    #139516

    digitalman
    Participant

    Seems like James and I are doing the same kind of work.  I’m in digital hardware design for medium to large-scale board designs (small amounts of analog along with lots of high speed digital including large FPGA designs). The items James brings up are valid applications, but I’m sensing that the “qualitative”-ness limits acceptance and widespread use.  Since a lot of 6sigma is culture change, when engineers are involved they have to be able to apply parametrics to a problem before they trust a solution. That would certainly be a hurdle here, but only because the “digital” area seems so “non-normal”.
    I’ve been focusing my thinking toward the FPGA aspect, as that is typically a well-contained, detailed design. Metrics gathering could be automated to some extent, and it is a microcosm of DMADV to some extent.
    Since our 6sigma projects require showing dollar amount savings, I haven’t been able to correlate design review or board design defects to actual cost, perhaps for the same reasons as James.
    It’s very difficult with the allocated resources to dedicate the time necessary to demonstrate effectiveness of a 6sigma process change/improvement, so there has to be a clear (almost immediate) benefit in order to get funding.
    It’s a tough one.

    0
    #139641

    Semiconductors Senior
    Member

    Hello James, digitalman,Thank you for your information. Please let me think out loud to share my thoughts with you. I welcome your feedback.
    The first thing I would suggest is to get rid of the mind-blocking terms, like “digital”, “analog”, “PCB”, “FPGA”, “code review”, “correlations” and similar. The reason is that our creative mind locks eagerly on those specific terms and their associated burned-in imaginations – which is an effective way to stop any kind of progress in reach. (I know this effect very well from my last search where the object was just in front of me or just to my right, and it took me days to find out, because it was so obvious to turn left for the search. Or the last time I searched a book in my private library.)
    Instead I would suggest to use generic terms, like a “product”, or a “requested result”, a “thingy” or similar. Instead I would suggest to take a helicopter view. Let’s climb …… and ask ourselves: “assume we can have everything under the sun (and beyond): if so, what would be the ideal process here?”
    I can’t give the answer, only you can. Perhaps the most ideal situation would be: “it costs 0 $ to generate the thingy, it takes 0 hours and the receiver can just use the thingy right away.”I would suggest to rethink this provocative Ideal Final Result carefully: don’t accept too little. I would suggest: Don’t hinder yourself by historical hurdles, things unimaginable, or defects which have been around ever since. I would suggest to abandon thinking of defects during this helicopter trip entirely. Instead, I would suggest, to shape out, to reveal, to pinpoint the purpose, the very best this process can produce, the reason why it’s there. It serves an ultimate purpose (in society and economy) and this purpose is …… (your words here).
    After a while it’s time for a landing.Things appear different down to earth, here. But we’ve seen the horizon: “It’s there, into that direction, and it’s just 10 miles to go …” (This I would expect hear you saying now).Down to earth.digitalman, you ask for clear, immediate benefit. I can think of two extremes. A) your process differs from the Ideal Final Process we’ve just witnessed above: then the next 6 sigma steps should be clear enough and the bottom line contribution might be estimated. B) your process is already quite close to the ideal and it’s time to take the next step. [Cost by the way is a critical issue: waste has to be reduced, investment has to be made as it generates results.]What can be the next step?There are many characteristics well managed businesses exhibit. One is: market position – there are more. The right market position helps to bring in the right customers, which result in a nice financial situation.Market position is more than a share (which needs to be defined [e.g by region, by customer, by delivery channels or similar ?]). What are the shares of substituting products, i.e. same products from competitors, or different products which serve the same purpose for your customers? Which substitutions can become available tomorrow? What is the perceived quality level of what you do today and what is the perceived customer benefit today? What should it be?These kind of considerations are outward-looking to give inward-focus.
    These kind of considerations are strategic considerations. They help you to shape the future of your business by looking ahead, rather than being punished by market responses (which just react on past decisions, i.e. react far too late).And it’s at least one order of magnitude beyond the pure defect-oriented perception of your situation. Of course, I think, defects now can be defined here as well, and business process can be adjusted and improved: on a more important level, I suppose.
    Having thought all this what would I suggest?I would suggest: backtrack a route now, from the horizon, from the rational market position you should be in, backwards to the problems to be solved first (e.g. by 6 sigma), backwards to the few vital DMAICs or DMADVs to do to first steps out towards the horizon.Then, I would expect hardly any problem in defining the appropriate metrics and the appropriate parametrics – and the time frames and the benefits and … . At least you will know, why, how and how much you should reshape your FPGA-business, your PCB-business etc.
    Do my thoughts help you? Please let me know.
    Semiconductors Senior

    0
    #139770

    Electra
    Participant

    Hi Digitalman,
     
    It has been a few years since I have been working in the same area as you, and in the meanwhile I’ve learned to enjoy the 6Sigma practices.
     
    A have a few questions about your concerns. Do you want to be able to theoretically predict the ideal/correct board design linked to a certain SW design? Or do you just want to have a more predictive design process?
    I think in the end I think they are the same but have some different way of approaching the problem.
    Some 6Sigma tools I can think of (and yes they logically follow the 6Sigma cycle):
    Translate requirements from customers, development, testing, process, customer support (you should complete the list) into clear design rules/requirements.
    Do intensive FMEA’s on your design, forgetting the border between HW and SW
    Take into account parallel paths, add testing possibilities in the design you can omit in a later stage, or leave them in for repair facilities at a customer support center.
    Maybe can make a specific board to develop the SW, having all the options for SW debugging on board without some of the other HW requirements who can hinder you when designing and testing initially the SW.
    Follow up the status of FMEA’s
    So in the end try to avoid redesigns and errors, and increase the ability to correct errors or modify requirements.
    This might already be the way you are working.
    Try to gather facts : design/debug lead times vs
          Re-used building blocks
          Number of nodes in the design
    Design modifications + costs vs hours spending FMEA’s etc
    Facts can be displayed in graphs
    Entering a planning in MS Project for example AND following it up can predict to certain extent the actual expected time needed (S-curve!) in an visual way.
    In my precious job we entered all (SW + HW) problems in a database (this doesn’t have to be complex). This means you have statistic data you can use to predict the end of the debugging phase (again S-curve) and data that tells you how long it takes to find the root cause, the solution, the implementation of the solution, the re-testing and success rate of the solutions. Very valuable data I think!
    Facts can help you convince others of a certain way of working : taking time in the beginning of the process for having a more predictive ending. Facts also prove improvement, facts you can relate to working hours, scrap PCB’s due to design errors… So costs…. So having a real cost 6Sigma saving…. (I experience the same push on costs, on hard savings so I understand your need to translate improvements to money).
    Another option might be to describe the digital design in an analog way. In the end everything is analog.
     
    Some thoughts…
     
    Electra
     

    0
    #139814

    Jonathon Andell
    Participant

    This doesn’t apply in every digital situation, but there are some cases where it makes sense:Digital signals are transients of volatge in time. In some situations, those transients (derivatives, integrals, min/max values, times spans between events, etc.) may contain crucial information. You could investigate whether send/recieve signals are synchronized well, or how much power is being used, or whether peak voltages are too high or too low.Here’s another: if you are mixing and mathcing components or placements at the board level or above, you could do a DOE or RSM, with time or speed as one response, power draw as another, and heat dissipation/buildup as a third.I even did it once on design of computers – choice of bus type, caches size, etc. – and used programs drawing on different kinds of processing (data transfer, number crunching, etc.) as a kind of “outer array.” The response variable was a bit complex, and remains confidential, but it was a big success.Coming up with a good response variable can be a fair chunk of work in any of these scenarios, but the reuslting analysis and information can be quite valuable.

    0
Viewing 9 posts - 1 through 9 (of 9 total)

The forum ‘General’ is closed to new topics and replies.