What is the expected lifespan of pedals using surface mount parts?

[SeanCostello]

Hi all:

I have been thinking about creating DSP based pedals for a few years now. As far as I know, there are no through-hole DSPs of significant power out there, at least for non-military apps. I had presumed that a DSP pedal would be surface mount only.

However...

I started thinking about the survivability of these pedals. I know that through-hole pedals can survive for decades - I have a late '70's Big Muff, and a late '60's Super Fuzz, and both work great. For that matter, guitar amps can survive for a long time as well - mine only date to the early '70's, but I know that there are plenty of people out there using amps from the 1950's. Part of the survivability is the fact that these pedals and amps can be repaired. I had the footswitch of my Super Fuzz go out in the late 80's, and paid to have it repaired in the early 90's, in the pre-Internet days, so no schematic, but they were still abe to trace things out. If a transistor, resistor, or cap goes, you can just switch it out.

If I make digital pedals, it won't really be for the money (I won't turn a lot of money down if it comes my way, but I am not expecting it). My goal would be to embed some of my algorithms in hardware, to create as permanent an artifact as I can. Computer plugins only last a few years before the hardware is obsolete, or the hosting environment changes. Open source software would be a good long-term plan for the survivability of the code, but a bad short-term plan from being able to sell the algorithms in other formats. I want something that guitarists could use 50 years from now, if possible.

So, what is the best solution? Will surface mount survive the ages? Or should I combine surface mount for the DSP with through-hole for the rest of the circuitry? How the heck would you assemble a mixed through-hole and SMT for mass production?

Thanks for any thoughts y'all may have on the subject.

Sean Costello

[iaresee]

What is it about surface mount that you think would make it break down faster? If anything I'd say the solder bonds are better than hand-built, via-based PCB designs because the surface has to be ultra-clean before you put it through the soldering process. As long as someone isn't poking at the parts there's no reason to expect surface mount to not last as long as any other PCB manufacturing method.

[Processaurus]

As far as repairablity you could do like in the old days, glue a schematic to the inside of the box.  Could even laminate it or print it on plastic or something.  You could label the parts with both a part number and their value.  You could hide a little bag of any hard to find IC's in the box (especially ones you program).  I've done that for stuff I made were I knew the people would be touring and wouldn't be shy about opening the pedal up if something broke.

I don't think reliability wise there is anything worse about surfacemount.  The 805 parts are doable to desolder and solder a few on (like you would with a repair) with a normal iron and good eyes or a magnifying glass.  You could even make the circuit easer to visually follow than a through hole board if you put all the tracks on top (weird idea!).

[Paul Perry (Frostwave)]

A couple of my products are mixed surface mount & through hole (mainly because, some key components were only available as surface - so you might as well have a bunch put in a the same time).
I just get the smt stuff done by a factory & then stick the big stuff & sockets etc in by hand.

[petemoore]

I don't think reliability wise there is anything worse about surfacemount.
  +1 ... why wouldn't they be...
  Maybe the reputation following the type of 'peripherals' construction [such as pots and jacks soldered into the PCB] follows SMT because it's often used with 'fast construction' techniques which may be a little less sturdy, able to be repaired, and reliable than...say...metal jacks with flexible connection wirings, mounted in a solid metal case...ie all the weak points made stronger, or flexible enough to withstand a jacknut coming loose without pulling board traces off.

[Ice-9]

I myself am also building a range of pedals that use surface mount components, although they are not dsp based at the moment. A range of dist/od and phaser flangers etc. I have been prototyping them at the moment so have etched my own boards and hand soldered the surface mount components.
Things i have noticed is that its easier and quicker to make (only holes to drill for the sockets switches etc. ) easier/quicker to mount the components and troubleshooting is easier as replacing components if necessary is simple. As for reliability most newer pedals are built with sm components now, i just bought an mxr evh phaser and took it to bits to see how they build them and that is 99% surface mount. Although i did drop it and broke the script/evh switch (if any one knows where i can get a replacement ?DPDT latching). anyway i have found my pedals are working with no problems at all but then again there only a few weeks old.  So time will tell

[iaresee]

I don't think reliability wise there is anything worse about surfacemount.
  +1 ... why wouldn't they be...
  Maybe the reputation following the type of 'peripherals' construction [such as pots and jacks soldered into the PCB] follows SMT because it's often used with 'fast construction' techniques which may be a little less sturdy, able to be repaired, and reliable than...say...metal jacks with flexible connection wirings, mounted in a solid metal case...ie all the weak points made stronger, or flexible enough to withstand a jacknut coming loose without pulling board traces off.

I've had this exact conversation on TGP recently and I contend that as long as your I/O ports are mounted to the case there's no need to worry about insertion and removal forces being transferred to the board/joint area. Of course, if you're surface mounting your I/O you need precision machined cases because you've got a very low tolerance on where those through holes in the case need to be in order to get the whole thing to fit. But it can really be a thing of beauty when down right. The MXR line up, case in point, looks gorgeous inside and the boards are all suspended elegantly but the surface mount jacks and pots. Really quite lovely packaging.

[R.G.]

That's an interesting and insightful line of thought, Sean. I would expect Mark to have some comments here.

My last job while still in the belly of the beast was in a third-level RAS (Reliability, Availability, and Serviceability) group. We lived and breathed Mean Time To Failure (MTTF), cost of service, parts stocking, estimated life, and the like.

Survivability is going to be a composite of the raw reliability (i.e., doesn't break down) and the service-abilty. The likelihood of it no longer being useful is the product of the probability of failure and the probability of NOT being able to repair it. You have control of the probability of failure. You have zero control of the probability of repair.

Surface mount parts are in general at least as reliable as through-hole parts or better, with one exception. They are highly sensitive to board flexing. They have no compliance in wire leads to flex with the board, and so board flex will crack solder joints and parts where it would not with a through hole circuit. The solution is to not let it flex. Thick boards, highly supported and decoupled from the outside of the box so temperature, handling, dropping, etc. will not couple through to the board and flex it. Conformal coating is a good idea for damping resonances - and preventing dampness. The military knows how to make electronics survivable. If you want to do this, just go get the military specs on how to make surface mount electronics and do it. After all, they do make processor based stuff which survives years of storage, then being shot out of a artillery gun.

But your question is about repairs. I think that your hope for repairs is in keeping the algorithms available for whatever computers then exist. Here's why.

The industry by and large does not repair SMD. Manufacturing is so cheap and repair people so expensive that the economics are against it. It can be done, of course, but that's not the industrial way. I suspect the most cost effective way to keep a batch of pedals running for 50 years is to make twice as many PCBs as you need, then salt the remainder away in a safe deposit box for ten years to get past the infant mortality. Once you're past infant mortality, the failures per year will be a tiny fraction of the total population, and a ready replacement board will keep the entire population going for a very long time. And a new board, even one stockpiled for a decade or two, will be far cheaper than a human to replace a part.

In the USA, a person cannot support themselves as a repair tech on less than about $50 per hour. There are cheaper techs, but they are not making a living at it with lower rates. I know the going rate is far higher than that in some place, Switzerland in particular. Could be different other places. If the cost of a total replacement board and mailing is less than the cost of bench time to repair, then it makes economic sense to just replace the board. If you combine this with a specialized repair depot where you apply manufacturing techniques to repair, you could recycle some of the failed boards. But it's likely to still be cheaper to make more boards in the first run.

The more specialized the part, the more likely it is to not be available. I probably ought to tell everyone the saga of the SAD1024s sometime. But a DSP chip is a very fungible industrial part. Industry has paid some very expensive MBAs a lot of money to help them perfect not doing anything at all that does not return a desired rate of return on investment RIGHT NOW. That means that although there is some stocking of a few parts as samples and prototypes, I suspect that most things like DSPs are made to order in elaborate just-in-time delivery schemes. The parts are not even cooked in the wafer fab line until they will be needed for a shipment to a buyer who has arranged for his manufacturing line to put them in boards. The ideal situation for both the part maker and the product maker is that the parts don't exist until the day they are put into the final PCBs. This is very different from the industrial concepts of the 50s, 60s, and 70s, and from China having decided to join the world economy.

So I think that replacement DSPs are not going to be available at all, except by accident. There will be DSPs, but their programming and even architecture will likely be different.
But a good programmer can trascode an algorithm into a new architecture. That's why so-called "dusty deck" Fortran programs are still run in the oil industry.

IMHO, put your effort into reliability and document the algorithms. Release the algorithms to the public domain once you're through with any money you make out of it.

[gez]

The only difference I've discovered so far is that the protection diodes on certain CMOS chips don't seem to be able to take as much current in surface mount form, despite what the data sheet says.  When I was doing the layout for one of my circuits the only way I could cram it all in was by omitting a current limiting resistor on the gate of a CMOS chip used for switching.  The pulse to the gate was very brief and I calculated that I should be able to get away with it.  Breadboarded it with DIL chips and it worked fine.  Totally fried the surface mount equivalent, though...and every replacement I put in (I'd forgotten about the omission so couldn't understand what was going wrong). 

That was a tough lesson to learn as I wasted a lot of time (hand soldered board). 

[Paul Perry (Frostwave)]

As always, wise words from RG.
I know from bitter experience what he speaks of. Time is what beats one on repairs - and, at least for me, it's faster to dive into through hole stuff & clip a part out, than to go to the surface mount. yes, one can repair smt - but one can't do it quickly! at least, I can't ! When I was pricing my smt products, i should have built in a % of 'wasted boards".
As it is, the entire profit has gone because of a few boards that aren't really fixable.
(that's also the downside of having a high component count product, as opposed to a fuzz with only a dozen!)