21
\$\begingroup\$

I am the chief engineer at a small company. Our product popularity is growing rapidly, and now we are learning about painful lead times for microcontrollers. The owner is demanding product overhauls to avoid distributor lead times. My question is:

What is the norm in electronic manufacturing?

  • should I convince him to get a loan and keep 4-6 months of chips in stock at all times?
  • or do I just keep switching parts until I find things more readily available.
\$\endgroup\$
6
  • 16
    \$\begingroup\$ All large distributors a) have lead times on their website, and b) show the stock. Call your electronic part distributor! Ask them how much it'd cost if they guaranteed a certain stock. Call your mcu manufacturer's sale office! Make it clear that you'll switch to another device (of another manufacturer) for your obviously strongly growing demand, if they can't help you. You'll be surprised – they might come up with a very similar but more expensive chip that might suddenly be available cheaper, or hook you up with an AFE that can help you transition to another model that is more common. \$\endgroup\$ Dec 13, 2017 at 23:43
  • 7
    \$\begingroup\$ And: make sure there's really no ICs on the market anymore; octopart.com is a nice search engine, with which you can quickly check stocked amount at multiple distributors (use the "in stock only" option in the "availability" drop down menu). Other than that, yes, that is more of a financial then an engineering question. \$\endgroup\$ Dec 13, 2017 at 23:47
  • 2
    \$\begingroup\$ 'other actors in the market' opens a door with counterfeiters on the other side. \$\endgroup\$
    – Whit3rd
    Dec 14, 2017 at 1:21
  • 1
    \$\begingroup\$ mk20dx series from NXP, out of stock since October and none anywhere until April 1018. I have called mouser and avnet, no real help. I'm less than 10K units/year right now, but could turn into 100K soon, but not if the owner makes me redesign. \$\endgroup\$ Dec 14, 2017 at 16:18
  • 2
    \$\begingroup\$ @ZacharyScottDickerson Woah, the supplier's actually making you travel back in time a thousand years to get your micros??? ;) \$\endgroup\$
    – DerStrom8
    Dec 15, 2017 at 2:51

8 Answers 8

23
\$\begingroup\$

Overhauling the product to avoid lead times is absolutely the wrong approach. You will not win this battle, stuff will go out of stock, because life happens. You could very well switch to a new chip only to find out that in six months that part has gone out of stock, and the original part is back in stock. In the meantime, you will have burned a bunch of NRE designing and qualifying the new component.

The only way to guarantee that you have parts when you need them is to order parts early and hold stock either yourself or in a specialty warehouse.

I would only design a part out for supply issues if there was a persistent issue with getting parts (ie: it happened more than once or twice) and/or the vendor's delivery estimates were longer than they estimated. If TI tells me that a part's out of stock and it'll take 16 weeks for them to cut new wafers, well, that's life. As long as they actually deliver in 16 weeks I can't well complain, can I?

The smart, prudent thing is to forecast your production volumes well in advance. If you know you'll be trying to sell 10k units in Q2 2018, then you should be queuing up your contract manufacturer now in order to get all the purchase orders in place so you can build the units on-time.

In tandem with this, you can rely on credit from your contract manufacturer, especially if you have firm orders in the pipeline - so if you can't afford the parts now, your CM can absorb the cost and let you pay later. You might also be able to negotiate a deal with your contract manufacturer where they buy and store the parts and then sell them to you when you make a production run, especially if it's something that multiple customers use like passives.

At the end of the day, making hardware means spending money well in advance of selling a product. The Bolt blog has a great article on the hardware startup valley of death: it happens when you start selling product, because there's a long lead time between ordering a production run, executing on that production run, and sell-through in retail.

\$\endgroup\$
6
  • 1
    \$\begingroup\$ A lot of times you don't have a choice, you have to redesign the product to keep it on the shelf. If the choice is between not producing and selling the product or redesign, most people will redesign. Yes, you can select manufactures with better stocking plans, but this will probably come at cost. The best route is to stock components yourself (or make sure the assembly house does) \$\endgroup\$
    – Voltage Spike
    Dec 14, 2017 at 3:08
  • \$\begingroup\$ @laptop2d Depends on the nature of the product though. High volume consumer product? It'll probably take longer to re-design and push it through requisite third party testing than it would just to wait for the part. Low-volume, high-value product like a piece of specialty test equipment? You can probably afford to redesign, but really, if you're in that space, there's no excuse to not bank a reel or two of parts ahead of time - the component cost is usually a small part of the overall product cost on those, making reserve stock feasible. \$\endgroup\$ Dec 14, 2017 at 4:13
  • \$\begingroup\$ Yes, it's a new consumer product, and it's very expensive to keep raw materials in stock for 5K units, as it's a premium product. It will take much longer to re-design then to just wait for stock, but the owner is worried it will happen again. I previously warned about long lead times, but they did not have enough money to do so. \$\endgroup\$ Dec 14, 2017 at 16:28
  • \$\begingroup\$ What a great QA ! \$\endgroup\$
    – Fattie
    Dec 14, 2017 at 16:45
  • \$\begingroup\$ @ZacharyScottDickerson In that case, If I were in your position I would consider either credit or outside investment to improve cash flow. You clearly have a product that people want, so you should be able to get money to help you scale. If you already have outside investors, you could ask them for a bridge loan (in this case, with firm orders and a definite return, they'd probably say yes). Alternatively, you could look at a business loan from a bank, which they might be interested in if you carefully articulated your cash flow issues and show them you'll be able to repay. \$\endgroup\$ Dec 14, 2017 at 17:13
12
\$\begingroup\$

This is a problem that is best addressed in the design phase. There is one manufacturer I won't use (unless forced to) because of their actions decades ago. Another has gotten acquired by a company with a better reputation so their stock has gone up (figuratively speaking) in my opinion. There's no point getting some fancy feature or saving a bit of money if you can't scale your operation or get parts to keep customers in the worst case. In general terms, suppliers who rely heavily on large and volatile customers (eg. cell phones, which can go from feast to famine as the companies shift suppliers) might be worth avoiding.

One somewhat extreme measure is to do two complete concurrent designs to support two disparate and hopefully unrelated-in-the-future suppliers. Just go down the usual selection tree and pick the best two that are both fairly good and unrelated to each other. It won't double the design costs, maybe increase them by 20%. Interestingly, Samsung has done this in cell phones with the same base model numbers S8/S8+ using different processors for Asian and North American markets ( Exynos 9 Octa 8895M vs. Qualcomm Snapdragon 835).

Stuff still happens- companies with good product (eg. Luminary) get bought up and the product discontinued, for example. A fire or earthquake affecting a fab in Tainan on Taiwan may affect supplies from many companies if they are a foundry. A fire at a chemical plant in Japan in '93 (in a town few non-Japanese would recognize) disrupted supplies of epoxy resin, affecting chip makers worldwide. For that you need some stock, but that's more of a business decision- a bit like buying insurance. It looks like wasted money, until it doesn't.

What you probably don't want to do for products of any complexity is to run around putting out fires trying to redesign product under excessively tight time schedules once the stuff hits the fan. Gray market suppliers are a bit like loan sharks when this happens- they will overcharge and sometimes you will get parts that are not usable. The best case is that you pay a lot of money to get full-spec parts to keep your customers from canceling.

Incidentally, Microchip's CEO Steve Sanghi apologized late last summer that lead times were increasing and explained that the market has been fairly soft for several years, resulting in a lack of investment by the various players. As day follows night, the upswing is resulting in shortages, even for the best of them.

\$\endgroup\$
4
  • 2
    \$\begingroup\$ Come on, we all know who you're talking about! \$\endgroup\$
    – pipe
    Dec 14, 2017 at 14:53
  • \$\begingroup\$ LOL @pipe no, not in this case, but only because they don't make MCUs. It reinforces the point that there are some companies with better reputations in serving small customers than others. \$\endgroup\$ Dec 14, 2017 at 15:11
  • \$\begingroup\$ They do make MCUs. I have never used one from them, though. Said company does have a good reputation in terms of product quality, but horrible reputation when it comes to delivery times and prices. \$\endgroup\$
    – Lundin
    Dec 14, 2017 at 15:31
  • 4
    \$\begingroup\$ Yup, I've got my never again and preferred lists too. Microchip is high on the latter. That's not as much a lead time thing as it is about obsolescence. You can still buy a PIC 16C54 from the 1990s. Paying $5 for what you could today replace with a $0.50 part is worth it in many cases to not have to redesign, requalify, get new regulatory approvals, etc. \$\endgroup\$ Dec 14, 2017 at 16:19
8
\$\begingroup\$

Yeah and several sources have come to me lately saying lead times are blowing out too. This happens from time to time.

There are lots of things you can do. For instance if you are getting to the point where you would need a loan to buy material then a disti like arrow or someone will likely buy and hold parts for you and let you draw down against that stock. You can work all kinds of deals like that. They also have a thing where they'll be your cm and manage all part issues but I've never tried it.

I'm a huge fan of trying not to use anything I can't buy on digi or mouser so I always have them as a backup. They do run out of things too though.

Push your supplier or your local rep for that supplier. A lot of times if you issue a PO and make a lot of noise even if you are small 6wks can become 2wks.

You can go to brokers like flexible components. There is supply chain risk here though so best to know what you're doing. There is the risk of buying counterfeit, old or defective parts but I have used it to bail me out on occasion. I use flexible components they've been good to me.

Octopart is great for quick disti searches. I also keep some small cm contacts in Asia. Don't know where you are located but there is often overseas stock if there's non in the US.

I hope you have already tried to talk with your rep directly or failing that the vendor you're having trouble with. They can often get you parts that distribution can't

I've even been known to buy a reel of parts on my own dime that I kept in a drawer for emergencies.

\$\endgroup\$
9
  • 3
    \$\begingroup\$ especially in a small company, be a bit careful with recommending to buy things out of your own pocket – emplyees are often already heavily invested in that company, and doing such "backup investments" might mask risk and true cost to management. Gotta ask yourself: If I buy this myself now, isn't that just like giving a purpose-bound credit to the company? Shouldn't I be discussing this with the owner rather than doing this on my own? Shouldn't I be getting interest? Am I a bank or an engineer? I think OP's approach of thinking about recommending a loan is saner. \$\endgroup\$ Dec 14, 2017 at 0:40
  • 3
    \$\begingroup\$ Case by case basis I guess. I didn't mean buy $10k worth of parts. But $50 of oscillators in a drawer could save your butt or enable you to sell $250k worth of product and keep the company afloat. To each his own. Plus not everyone works for sane managers who understand forecasting and inventory. \$\endgroup\$ Dec 14, 2017 at 1:23
  • 3
    \$\begingroup\$ Where academic digital communications engineers have the meme of "let's assume our receiver is (magically) synced in time and frequency", the practical side of us (especially the side in me that sells my consulting powers to people) of course says "let's assume management is competent and sane". \$\endgroup\$ Dec 14, 2017 at 1:27
  • \$\begingroup\$ It's been many years since I've had that luxury :). Curse of the startup I guess. \$\endgroup\$ Dec 14, 2017 at 1:28
  • \$\begingroup\$ Point being: Management of my consulting business isn't sane, either. I'm my own boss. \$\endgroup\$ Dec 14, 2017 at 1:32
6
\$\begingroup\$

As mentioned by others, this really needs to be dealt with at the design phase (and therefore is most definitely an engineering activity), but also needs to be regularly reviewed; I have been bitten by a supplier calling up and stating they cannot find a part anywhere in distribution.

I ended up having to choose another part and do wire modifications to the board.

The key to that is understanding just when a part may go obsolete or have long lead times (yes, this is a bit like gazing at a crystal ball, but there are steps we can take to minimise issues).

Here is my list of things to watch out for:

1 The datasheet is 3 (or more) years old and has 'Preliminary' or 'Draft' stamped over it.

Run; here is an example.

2 When at the design stage, get the standard MOQ and lead time for the part; do not ever expect faster lead times than you get at this point. You may get a faster lead time, but there are no guarantees.

3 Some processor and FPGA families have multiple devices that are pin compatible.

You may not need extra functionality, but you have an escape route if you choose something from this line (I understand this is not how life works in general, but it is something to consider). As an example, the Silicon Labs Giant Gecko and Wonder Gecko differ only by processor core (the former has an M3, the latter an M4).

There are numerous micro-controller families that offer this sort drop-in functionality including footprints. Try and use items from these families if possible.

Specifying a part that has no chance of being replaced on the same footprint is asking for disaster by either going obsolete, or just as bad, on allocation.

4 Beware of large silicon parts (such as micros) that have very large MOQs or silicon vendors that have a small base of very large customers. Both of these are indicators of potential supply issues down the line.

5 Subscribe to the PCN (product change notice) list for each supplier / component that is critical to your design; this can yield useful information about the lifecycle of a component.

For obsolescence intelligence, use one of the various tools; these have a traffic light system and can often predict when a part will go obsolete; one of them even plugs directly into the Altium design environment.

Another example is here

These tools cost money, but how much is it costing if you cannot build, ship and bill?

The design process does have to be a bit longer to do the above checks, but the most expensive component is the one you cannot buy.

I understand your situation is not helped by the above, but perhaps your next project / product can benefit.

\$\endgroup\$
4
\$\begingroup\$

This is not an engineering function. Predicting future demand for your product is a sales function; and ordering parts for production to meet the predicted future demand is a production (or perhaps purchasing) function. Unless you wear those hats as well in the organisation (which isn't too unusual for a small company), this is not a problem with engineering - it is a problem with sales or production (or both) screwing up.

Granted, engineering may be able to help by finding a different chip, but that's going to take a ton of work to evaluate, qualify, price and redesign. Your owner needs to think very hard about this kind of thing.

Have you genuinely looked around all distributors? It's not unknown for devices to be completely unavailable, but it's relatively rare.

Some microcontrollers are available in pin-compatible versions from different suppliers - the Infineon C167 and STMicro ST10, for instance. This can give you protection against single-sourcing. In practise though you would normally qualify and stick to a single processor, from a reputable manufacturer. Embedded software is always dependent to a greater or lesser extent on the hardware you use - you can design to minimise that, but there is always some impact - so changes to the platform will usually take significant time to work through. It's never impossible to port to a new platform, but it does require some care and attention, and the time (and hence cost!) to do it should not be underestimated.

Note that peripheral devices - ADCs, DACs, etc. - could fairly easily be swapped out if they have standard interfaces such as a parallel bus or SPI. Whilst the behaviour of the device will be different, of course, a good driver design can isolate the rest of the firmware from the I/O details. The PCB can then have spaces for two devices to carry out the same I/O, with one fitted and the other not fitted, as device supply allows. There may well be performance impacts depending on which device is used (noise, slew rate, accuracy, etc.) but it should be possible to select two devices which are both adequate to meet requirements.

\$\endgroup\$
4
  • 6
    \$\begingroup\$ It can definitely be an engineering function, if you as an engineer didn't do the preliminary research and then picked a unicorn chip made of unobtanium, because the local shop had 10 available as "new-old-stock". \$\endgroup\$
    – pipe
    Dec 14, 2017 at 14:48
  • \$\begingroup\$ @pipe Fair point, +1 for that. \$\endgroup\$
    – Graham
    Dec 14, 2017 at 15:59
  • \$\begingroup\$ I feel like NXP MK20DX series was not a unicorn chip, but perhaps I'm wrong, as It's been out of stock since October everywhere until the following April. Unable to find anywhere in the world and it will be painful to re-design now that the product is taking off. I wear many hats, and lead many contractors, but we are growing the company very fast. \$\endgroup\$ Dec 14, 2017 at 16:41
  • 1
    \$\begingroup\$ @ZacharyScottDickerson A quick search on Google shows Mouser as selling them, and Mouser claim to have stock of various versions of it. No idea whether your particular variant is on that list, of course. Still, that's another part of a good design - choosing a device with various pin-compatible versions with different memory/speed so that you can upgrade/downgrade as needed. And that can help with these kind of problems too - better to buy the next price up than not be able to make it at all. \$\endgroup\$
    – Graham
    Dec 14, 2017 at 18:25
4
\$\begingroup\$

What you should do as an electronics designer is to ask the distributor:

  • Is this a popular chip that you keep in stock in larger volumes?
  • In case this is a chip which you do not keep in stock, then why am I even talking to you? I might as well buy it straight from the manufacturer in that case.

If your distributor does not keep parts in stock, get rid of them. That's their sole purpose. Now of course, some manufacturers refuse to sell directly to end customers. In other cases, the distributor can do nothing, since the problem is with the manufacturer not spitting out enough parts.

Over time, you learn which manufacturers that aren't trustworthy, which ones that have poor delivery times and which ones that love to EOL 50% of their products every year.

If your company is large enough to have a dedicated purchased, then your purchaser should be the one with this experience. So before you pick a certain part, check with the distributor and in-house purchasing both.

When you have done all of this, there's not much else you can do. There's so many new parts coming out all the time, that there is simply no way that you can have the latest MCUs in your products at all times, or you would do nothing else but porting code and making new PCBs 24/7.

Overall, electronics production needs to be planned at least half a year ahead. That is not the job of the designer, but of the purchaser or production manager. MCUs are very often critical key components and you probably want to keep a safety buffer of them in stock at all times.

\$\endgroup\$
2
  • \$\begingroup\$ I find almost NO MCU that has 10K units in stock at any distributor... I am fairly new to this side of the business, but am I doing something wrong? I called manufacturer and they bounced me back to distributors, which don't carry more than a few thousand at time of anything. I think in that case I will always be hit with lead times, meaning we must start stocking a 16 week supply of everything? \$\endgroup\$ Dec 14, 2017 at 16:47
  • \$\begingroup\$ I think a good distributor can make a ton of difference. They often will provide stuff in 10K units, so long as you're working with them. Pick your favorite, and call them up. Chat the procurement people up, and see if you can arrange a good deal. This way, you get the insight of someone who buys chips all day long, every day. \$\endgroup\$ Dec 14, 2017 at 18:39
4
\$\begingroup\$

Ah yes, "Gone on 27 week lead time" the bane of my life...

Firstly, when preparing BOMs as far as possible (and it is not always possible) list at least three alternative parts, and them make sure your code works with all of them (Flash memory devices, looking at you!), if an opamp does not NEED to be the TI version then also list the ones from a few other vendors. Does that cap have to be Panasonic, or would Nippon Chemi or CDW do just as well?

In the cases where it is not possible to list easy alternatives, maybe look to see if the same part is available on a different reel size, or in industrial rather then commercial temperature grade and list that as the alternate. Sometimes you can find pin compatible substitutes in an 'almost' the same part, so for example a PIC 32MX795L instead of a 695L, maybe the difference will make no difference in your application.

Then there are the vendors who are known to be 'interesting' this way, Maxim springs to mind as soon as you go off the completely jellybean parts, they have some cool stuff and you can get samples, but you try getting a few hundred....

Finally there is the stuff where you just have to suck it up, the ADV7511 HDMI chip did this to me recently, 19 weeks, even from AD, not amused.

\$\endgroup\$
1
\$\begingroup\$

Other things for consideration:

  • design the PCB to accommodate multi-footprint where possible
  • if components are single source, make it known. This is a risk to the business and the decision can be made to hold stock of these parts. This is generally a business decision
  • certain suppliers have 'A' and 'B' lists of components. 'A' list are new and they get pushed into new designs as preferred parts. They are likely to be around for longer
  • where components are unavailable, an alternative is the same part in a different footprint (using an adapter PCB). Leave space for this on the PCB if possible
  • consider souring from other countries eg. lcsc.com (JLCPCB uses them for their sourcing). You have to vet these suppliers well. There are lots of sharks out there.
  • if you're using a highly specialised part, form a very close relationship with the supplier/manufacturer to de-risk the business. If possible, 'contractualise' any formal requirements
  • plan ahead. If suppliers know your forecast, the easier it is for them to deliver
  • if you do your certification, qualification etc - try to get as much in as possible the first time around so you don't have to recertify with new components
  • avoid using discrete components (resistors, capacitors etc) in special geometries if you can
\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.