Suffixes, eeproms, ftdi chips , digikey and misery...

[free_electron]

Right. Long story. I just spent a whole day trying to figure out some really oddball problems.
Heres the setup : i am working on a design using an ftdi 2232h chip. I got about 100 boards made.
I am using th 2232h in dual fifo mode so i want to program the eeprom to load the correct driver and use. A custom string so my software can find the correct unit.
I start flashing the first boards and everything goes fine. After about 20 boards the programmer software can't find boards anymore...

The annoying thing is everytime you plug in a new board windows wantss to check the updates, load new drivers and assign new comports. So i thought.. Hmm maybe i ran out of comports...
So i tried cleaning the registry and uninstalling. No avail. Boards were not recognized by ftprog.

So i grabbed usbspy to snoop the bus. The board enumerated fine, windows saw it as a dual usb serial port. So the board works and is enumerated. Yet the programmer can't find the board ?!? Wtf.

Now i know there are revisions of that chip so i doublechecked i had the latest version. Since we got these from digikey that pretty much ruled out fake devices too.

By this time half the day was gone and frustration was high. 80 boards not programmable.. .!?,!,

I got on the phone with ftdi. They were baffled as well . They snet me beta versions of the new driver to see if that would solve anything. After scrubbing the drivers off , telling windows not to download automatically and some more timewasting ... Zilch. Exactly the same crap. 80 non working boards.

Bloodpressure was raising boiling point by now. Especially since i needed to ship these tonight.
I need to know if it was the ftdi or the boards. So i decided to do a test. I desoldered an already programmed eeprom and put it on a non programmable board to rule out problems between ftdi an eeprom.

Lo and behold. The device enumerates correctly , shows up with my custom strings and i can reprogram it. Ftprog finds it without a hitch...

That's odd.. These boards were all made on pick and place with parts from the same order.
I look at the markings.. Two different markings... Wtf ?

I got some atmel 93lc46 laying around in a bag, they are different footprint but i hotwire them in. So8 vs sot23-6 . Detects and programs fine. So its not the low voltage that poses a problem.. Darn it. What is going on. Look on the datsheet ( i had a national semi datsheet laying around , nope all is well )

So what the hell is the mystery part on my board.. I grab the datasheet from my order code. Son of a gun ! The nonprogrammable parts are microchip 93lc46a where the programmable ones are microchip 93c46b

Now Study the microchip datasheet a bit more. It turns out that the suffix for a microchip 93..46 had a special meaning...
An A suffix means the device uses 8 bit transfer frames. A b suffix means 16 bit frames. A c suffix means the frame size is programmable using an extenal pin...

Guess what.

-The ftdi ONLY works with 16 bit transfers.(clearly stated in their datsheet. I knew that. )
- i ordered 93c46b and digikey shipped me 93lc46a ....
- only floody bucking microchip has this 8 /16 bitmode controlled by the suffix. The original 93c46 from national nor the atmels nor any other vendor has that suffix crap. They listen to the org pin. But not those nutcases from Arizona...

Aaaaaaargh. A whole day wasted and my life probsbly shortened with a whole week.

I called digikey and they are replacing the chips overnight. They shipped the wrong ones. I ordered the right ones. ( i used the same partnumber from the ftdi bom which has the 93c46b listed. )

We had ordered these in two batches. I had 20 correct ones...

Point is
-microchip are bass ackwards with their damn crap. No other manifacturer has this weirdness.
-Digikey does mess up sometimes
- part suffixes do matter
- always use datsheets of the exact part and manufacturer you are installing.

Grmbl.... Stomp stomp stomp -slam-

[AlfBaz]

But not those nutcases from Arizona...

Priceless

[Amarbir[Lynx-India]]

Well,
    and by gods grace you have 20 ok once .If you did not have those Damn  .You did a job well .If you were in india i would have a beer with you and increased your life by 2 weeks 

[marshallh]

Haha, I've been there. BTW the reason all FT2232H's have to reload drivers is because they have unique serial numbers generated for each... Download USBDeview... you will see hundreds of FT2232's in there
Recommend directly programming the 93LC46 in system via pogo pins, use an arduino if you must...

Except my issue was with a serial flash rom... I had spec'd a M25P40 512kbyte spi flash rom on the boards.
The assembly house came to me around the time they got to building it (~2 weeks after sending BOM) and said that part was out of stock on Digikey. Well sure enough, they were out, and lead time was 6 weeks.

I had many customers waiting.. I said here, use this slightly newer revised SPI flash rom. Same company. Same specs. I combed over the datasheet and couldn't find out what could possibly be different.

Bahahaha, I was in for a surprise. The "small" run of 100 assembled boards shipped from China and NONE of them worked. Everything was perfect.

I swapped the M25VP40 for an old spare I had. That did it.

Meanwhile, poor marsh swaps out 100 spi flash roms on every board.

Lessons:
1. NEVER spec replacements YOU HAVE NOT TESTED
2. If possible find these suitable replacements BEFORE you have a bunch produced.
3. Run four-corners/stress test on the replacement part combos too!!

[David_AVD]

With the FR232RL you can tell Windows to ignore the different serial number and not enumerate as a different COM port each time.  Plugging into a different USB port (for the first time) still causes the driver search.

[Neilm]

At least you can re-work the boards. Just before Christmas I had a load of prototype board delivered and built for some pre-production units. Fully loaded with all the components is probably £100 a board. Got them - they all worked fine and went to install them in the new test instruments.

Then found that a slot had been routed to small - the boards wouldn't fit. As there were components both sides, I couldn't get them routed out. £4k down the tube, the PCB manufacture say "Oops, we will replace the PCB". 

Still, if that is the worst thing that happens prior to going into full production I'm happy.

[AndyC_772]

Those little serial EEPROMs are a law unto themselves, I've come across many different varieties exactly as you describe - some are 8 bit, some 16 bit, and some use the ORG pin. I never regard them as interchangeable without checking the exact part number first.

Oddly enough, it's not the first time I've heard of Digi-key supplying the wrong part. I still remember the afternoon I wasted trying to figure out why a new power supply design wasn't working, only to finally spot that the NPN transistor I thought I'd ordered was actually its PNP complement in a mis-labelled bag 

[JoeyP]

Oddly enough, it's not the first time I've heard of Digi-key supplying the wrong part. I still remember the afternoon I wasted trying to figure out why a new power supply design wasn't working, only to finally spot that the NPN transistor I thought I'd ordered was actually its PNP complement in a mis-labelled bag 

I had nearly the same thing happen with Digi-key. I ordered some solid state relays, and they shipped me relays with a part number off by one digit, which happened to be exactly pin-compatable with what I ordered, except opposite logic sense for the input.

[notsob]

At least the last stuff up on supplied parts I received was obvious, in stead of 1% resistors, I got a bag of 10A HRC fuses, even tho the label had the correct part # for the resistors and was marked as resistor, crap a tiny resistor versus a fuse thats 10mm x 38mm, thats quality control.

[RCMR]

Farnell/Element 14 have also had their snafus.

I ordered a 3-leg precision voltage reference and they sent a 2N2222 transistor (but charged me for the more expensive voltage reference).

I sent pictures of the errant device, they acknowledged the problem but never did put the order right.

It's only a small thing -- but it's the putting right that counts so I don't buy from Element14 any longer and no longer recommend them to others.

[mazurov]

In my experience, Digikey sends wrong parts all the time. Also, they are often the most expensive shop. They are on the bottom of my list of suppliers, I buy from them only if no one else has the part.

[AndyC_772]

It's only a small thing -- but it's the putting right that counts so I don't buy from Element14 any longer and no longer recommend them to others.

That's a shame, I use Farnell all the time and am generally very impressed with them. They're my primary source for components, followed by RS, Mouser and Digi-key (in no particular order).

[poorchava]

A colleague from work had a funny incident with Farnell. He had ordered some high temperature sensor in TO-5 package, and he was quite amazed when he received like 4kg package with huge power resistor inside . Luckily Farnell acknowledged their mistake and  fixed the situation

[amyk]

At least the last stuff up on supplied parts I received was obvious, in stead of 1% resistors, I got a bag of 10A HRC fuses, even tho the label had the correct part # for the resistors and was marked as resistor, crap a tiny resistor versus a fuse thats 10mm x 38mm, thats quality control.

HRC fuses are not cheap... use the mistake in your favour, sell them.

[MartinX]

At work I ordered pins for Molex micro fit from Farnell and got NiCd batteries instead, big D sized cells, 50 of them. Economically that was one hell of a trade but I did the right thing and returned them, considering the huge stock of different things they have and the volume of orders they must handle every day it is not often they mess up, the same goes for Digikey. It would be better if they messed up in an obvious way like with my connector pins instead of delivering almost the correct part, is there no way to ensure mistakes are sufficiently large so they are obvious in a logistics system? Some sort of error amplifier....

[SeanB]

No way, the people picking will always make mistakes, either in data entry or picking. Most hopefully get caught by a checker, but that is not always so.

[jeroen74]

In the 15+ years I have dealt with Farnell, I found mistakes are very rare. I once got 10 instead of 1 of those DIL IC test clip probe thingies at now €28 a pop. Expensive mistake for them.

More recently I had to deal with bad packaging, like receiving TQFP144 FPGAs with bent pins or OLED displays inside a slightly too small box so the fragile foil tab got damaged.

Other then that, they're a superb company. Same for RS.

[marshallh]

I have only once ever got wrong parts from Digikey, set of 10 diodes instead of flash ROMs. I emailed them and they said they already had replacements in the mail no charge.

Now on the other hand, Sparkfun has zero clue how to package parts, I ordered 50 fpgas because they were clearing out their stock (blinky arduino types can't deal with fpgas, i suppose) and ALL of them came in loose pink plastic, fragile QFP208 packages with bent pins. The assembly house would just laugh if I tried to have those placed and soldered. I think maybe I will make a stacked cluster hashing board or something.

Hard to see but the corners of all the pin rows are shoved in. Will take a lot of time with some tweezers to straighten all those out. Pity the poor bastard who actually removed these from the perfectly good reel/tape and shoved them in the packaging machine..

[jeroen74]

Amazing

[mazurov]

I've been receiving repackaged parts from Mouser, too. Not in plastic bags, but in some low-grade cardboard boxes. Even if a part is originally packaged on a 360-pieces tray and I order 500, I get all 500 in their boxes, 50 pieces in a box, 1-2 out of a box with a bend pin.

I like Avnet - when I buy from them I get a package  directly from the manufacturer.

[jeroen74]

Fairly recently RS introduced the option to choose the packaging type.

Recently I ordered fifty SOIC-20 ICs and forgot to select the production packaging option. So I got fifty grey metallised bags each containing one IC in cut-off tape (original packaging though)  Had I selected the option I would have gotten just one big tape 

In the old days Farnell used pink zip lock bags, but they replaced that with soft heat-sealed stuff that I hate.

[David_AVD]

Pity the poor bastard who actually removed these from the perfectly good reel/tape and shoved them in the packaging machine..

They may have been packaged in trays originally?

[AndyC_772]

Pity the poor bastard who actually removed these from the perfectly good reel/tape and shoved them in the packaging machine..

Pity Sack.

[amyk]

Now on the other hand, Sparkfun has zero clue how to package parts, I ordered 50 fpgas because they were clearing out their stock (blinky arduino types can't deal with fpgas, i suppose) and ALL of them came in loose pink plastic, fragile QFP208 packages with bent pins. The assembly house would just laugh if I tried to have those placed and soldered. I think maybe I will make a stacked cluster hashing board or something.

Did you at least get a partial refund for that? I think bent pins from improper packaging would count as damage.

[JoeyP]

It just happened again. I ordered 10 zener diodes from Digikey as part of a large order, and received instead 5 resistors with a label showing the correct part number for the zener.

I have to say, that I have very rarely received incorrect parts from Digikey. This is only the second instance that I can remember out of I don't know how many hundreds, maybe more than a thousand orders over the years. In general they are my first choice for parts. But judging by comments in this thread, I'll need to start checking their shipments more closely.

The rep at Digikey told me they've had a large influx of new workers since December 2012, and have also increased the number of parts in stock resulting in closer spacing of parts bins. He said it's been keeping his department busy correcting orders.