Maxim Semi Deliveries

I've found that it depends on who you are. If you're a small company, avoid them like the plague. If you buy in the millions, they really want your business and are quite responsive (prices are good, too).

Think of it this way, if you're small, they have great specs. ;-)

I've heard a lot of talk like in the comment here, and my impression is it's still going on:

But I can't say because I do not use Maxim parts in designs. Ever. I do design them out a lot, usually because they can't deliver quantities.

There is some nice Maxim stuff though :-)

It sounds like no change, then.

LTC is what Maxim should have been. Superb parts, but available and with great support.

Yeah, but she is over-dressed for the situation.

If a spec is in a datasheet with min and max limits, then it is tested to be within those limits, generally over temp and voltage, though you need to read the test conditions. The test program and datasheet limits are 100% in sync. There are other tests that you don't see that may cause a part to be rejected, but every part that ends up going to a customer will meet published specifications. There are additional tests for process parameters at the wafer level that will keep product outside the average process from reaching the end user.

As I have stated many times, take a black magic marker and blot out any spec without limits. Your device should perform based on those limits. If you need more, you need to pay for a custom flow with screening. Or you need to contract for a custom part. Graphs are just bench tests of a typical part at the time the product was released. You don't want to put too much faith in a graph.

As fabs are retired and replaced/acquired over the years, there will be minor variations in product performance. Tooling is replaced. Raw wafer vendors changed. The process itself is just a statistical goal. This is real life. However a chip is more likely to meet the design goal on a volume basis than some half assed home brew attempt at performing the same function.

Maxim does about $600 million per quarter in revenue. Clearly they are shipping product.

It is always good practice to make business decisions on the advice of one post on usenet.

Maxim's problem was creating too many parts. Gifford thought it would make the company hard to second source because the competition would have to figure out what to second source. Problem was the competition wasn't that stupid. They just second sourced the winners.

At this point, the leadership of the Maxim should be replaced with people that know what they are doing. The stock hasn't grown in a decade, but the board is completely useless in remedying the situation. The only way the company advances is when some useless VP retires and a good person gets the job. I still hold stock in the company, and wish to hell TI bought it instead of buying National. TI knows how to run a business. In fact, after looking at some SiGe part from National, I apologize for dissing them.

So what you get with Maxim is some good parts, depending on the designer. Much of the talent left the company a decade ago, but there is a portfolio of existing designs to use as a basis for new products. So good designs, a decent process, but just average production. That is how some parts end up not being delivered. Not only does the company have to produce the good stuff, they have to produce all the parts that don't sell well else they get cursed for orphaning products. [For the most part, the only orphaned products were due to VTC being a sole sourced vendor.]

LTC was founded by engineers. The design staff is the "home of the gurus". It clearly plays a part in how they think of their product. Their catalog is thinner, but they have good margins. I have a lot of respect for LTC.

In theory that is...

I have evaluated some Maxim parts in the past. For some parts the specs aren't complete or the device works a little bit different than expected.

But Maxim is not the only one. Last year I inherited a design with an Attiny from Atmel. The datasheet says it will work at 1.8V. At one point the software guys started to complain the eeprom programming is flaky. I contacted Atmel and after some e-mails it became clear 1.8V is the bare minimum. The lowest brown-out level is 1.8V (+/-30mV) which (usually) means the device should work at a few tenths of volt more. Bye bye Atmel.

??? Maxim uses semis (trucks)?? Are they filled with hot air or congresscritters?

That's the only way to deal with them.

I stand by my claim. The part works work at electrical limits because the tests are (gasp) guardbanded. That is, all parts should exceed the datasheet, not fail to meet datasheet limits. If you have parts that don't meet the datasheet limits, you don't bug applications. Rather you go to factory QA. Those are the people that are in charge of disturbing fecal matter.

Now there are specs that are guaranteed by design and not tested. In the dark ages, such specs would be for things like input capacitance on pins. That is a test that is difficult to do with contactors. Over the years, to eliminate cold testing, some tests at cold were removed from

Check..cold testing is extremely difficult without a semi-sealed enclosure and without ways to purge moisture. Using dry ice in a small make-shift box is a bit dicey..

Guardbanded?! Intrumentation Manufacturers don't even do that now.

Guarranteed specs?!

Which company are your working at?

My experience with IC's meeting specs has worked out well, BECAUSE I DO THE GUARDBANDING! I found it best to apply leeway in my designs to accept "out of spec" parts, et voila! No Problems. But count on the spec? ....no

Assuming that parts exist that are that much better than your design requirements. It's hard when you're pushing the envelope.

Cheers

Phil Hobbs

ed

ou

its

he

ff

True, but I started in the EXTREMELY conservative design environment of Aerospace designing flight control and autopilots. [Actually desinged the Hardware computer interface for Lockheed's L1011]

When I changed jobs and went to HP, I found HP's design philosophy so sloppy that I thought they were a bunch of 'garage' designers!

It wasn't until I entered the game industry and designed arcade electronic games [three months to develop and product lifetime maybe 1 year!] that I truly discovered the real meaning of 'garage design'!

Well I can show you an example of an amplifier from Maxim which only does half the amplification stated in the datasheet. How is that for testing?

LTC also seems to go out of there way to let people know that they appreciate the "little guys" who are only going to buy a few parts from them. I've worked at places where we might have been buying all of $250/year of parts from them (and this through distribution), but they would still send out an AE and marketing guy once or twice a year and buy us lunch at that -- I can't imagine they made any money overall, but clearly they figured that someday our quantities might increase and it'd have been all worth it.

That sort of long-term thinking is quite refreshing these days.

Or someone at your company would move on and take the experience with them.

Within reason. Nobody needs a supplier who goes t*ts-up when you have designed their parts into a product.

LTC is hands-down the semiconductor company with the best customer service there ever was. The downside, of course, is that they can't do all this super great service for free. Meaning their parts are rather expensive. When I do a design for an aircraft or a spacecraft, ok. For mass market products where the client produces thousandas per month, not so much. For example, there is a surprisingly small selection of switch-mode chips used in mass products. Sometimes it's donw without a dedicated switcher chip.