Parts Numbering Scheme

When I managed a parts inventory for electronic components one of the major wish list items was for some sort of picure of the actual components. Many items would have a similar description and value and multiple returns would result from a search thus requiring some intelligence in order to determine which particular item was the one required for an application. A picture based inventory system would have solved most problems but may incur some additional work and equipment to implement.

One piece of commercial software might be a contender for such an inventory system

Here are the ways a couple of big companies have done it. I'm not suggesting that these are optimal for you, just giving some examples.

General Electric major appliance part numbers are (were?) of the form

W[A-Z][0-9]{1,2}X[0-9]{1,4}

The leading "W" is constant and I am told it stands for "white goods" - refrigerators, washing machines, etc. ("Brown goods" are TVs, stereos, etc... they had a leading "E".)

The second letter indicates what type of product the part is for. R= refrigerator, B=oven, E=dryer, H=washer, etc. X was used for common parts like screws, washers, and bolts that could be used in many different products.

The next two digits indicate a major subsystem or group of related parts on the machine. For example, in refrigeration, timers are 9, and compressors are 87.

The "X" was a fairly constant separator. Historically it was the only separator, but "J" and "M" have also been included in newer part numbers.

The last one to four digits are just a serial number to tell the parts apart - they don't encode any characteristics of the part. About the only thing you could tell from these numbers is that newer parts had bigger numbers.

People typically didn't pad the numbers (WR9X330) but the computers did (WR09X0330). The lookup was smart enough to accept either style and find the right part. The "canonical" form was all upper case, but as far as I remember, it was case insensitive... wr9x330 or wR9x330 would all be acceptable to the computer as the same part number.

Volkswagen automotive part numbers are described at

The parts list for an old Systron-Donner frequency counter I have shows parts with a leading letter and four or five digits. The leading letter mostly corresponds to the part type: (R)esistor, (C)apacitor (no distinction between ceramic, electrolytic, etc), (S)witch, X for sockets, (V)acuum tube, CR for diodes, etc. Integrated circuits and transistors seem to have all-numeric part numbers. The digits just seem to be assigned in order of use.

My response to a couple of other points that have been brought up:

Allowing upper case and lower case letters in your part numbers does double the number of possible values in each digit, but IMHO will drive you crazy. If nothing else, most existing part numbering schemes are case insensitive, and you may tend to confuse people if they assume your scheme is the same. I think it won't be obvious to a lot of people that an A0123 and an a0123 are two totally different items.

The two extremes of part numbering are probably an all-numeric part number that has no inherent meaning (73853), and an alphanumeric string that tries to describe everything about the part (RES 1K 0.5W 5% METAL FILM). Both of them will drive you crazy. The all-numeric ones are too easy to confuse with one another. The alphanumeric string only works if you rigidly enforce some standard when you're putting new parts into the system, which never happens.

One slight advantage to the all-number system is that these are easy to copy correctly over the telephone. An alphanumeric system is harder to communicate - was that a D123 or a P123 or a B123? - unless you can train people to use some kind of phonetic alphabet (Alpha Bravo Charlie, Able Baker Charlie, etc).

If you have reason to suspect the part numbers may get garbled, or if you're just paranoid, you can include a check digit.

It may be helpful for the humans to include dashes, dots, or spaces in the part number. A big reason why local phone numbers in the US are seven digits expressed as three digits, dash, four digits (555-1212) is that Bell System research showed that most people could remember at most seven digits, and they remembered the digits better if they were broken into a couple of groups. The computer can store the numbers without any of this stuff, but it should accept user input with or without the dashes, dots, etc.

When you're trying to decide how long to make the number, also consider how many numbers you will need total. Disk storage is cheap, and adding one more digit multiplies your total available part numbers by at least

If I were trying to do this from scratch for electronic components, I might do something like this:

The initial letter would identify, broadly, the component type: (R)esistor, (C)apacitor, magic (S)moke, etc.

The next two digits would indicate more attributes of the component. For resistors, this might include metal film, carbon film, wirewound, through-hole, SMT, wattage, etc. Maybe this would be just one digit.

The final four digits would just get assigned as needed.

I would create a database to go with this, with the part number as the primary key of the main table. The main table would also have additional columns of descriptions that apply to most parts, like tolerance, through-hole or SMT, and so on. It would also have a reasonably-sized string field for future use. It's not "clean" design to have to stick two or three new attributes into this string field later, but sometimes it beats having to add a column to the table. There might be other tables for less frequently used descriptions, also indexed by the part number.

To make the production people happy, one of the pieces of data would be the part marking for parts that are color coded or too small for a complete part number. In other words, the list going to production would say things like "R-42-5309, resistor 1K 5% 0.25 watt metal film, brown-black-red" or "T-12-8670, transistor 2N2222A surface-mount, 8T5".

A place I worked at also used the Parts and Vendors software that has been mentioned. I don't know anything about it other than that it exists and it was possible to run it on several networked computers accessing a common database.

Whatever database or other software you use, and whatever backup plan you have, I would also suggest occasionally dumping the part number database tables to CSV or similar delimited plain-text files. If you need to change software vendors for whatever reason, most databases can be persuaded to accept a CSV or delimited-text input. To get your data into the new database, you might have to change single quotes to double quotes or escape the commas or whatever, but this is really easy to do with the right tools, such as awk, sed, perl, and friends on a Unix system.

I hope this helps!

Matt Roberds

What does the stockroom look like? Do you have smt diodes in one bin and relay racks in the next?

John

We have spent a number of years developing and refining our inventory and production management system. We use sequential part numbers that have nothing to do with the type of component. Our database system includes lots of information about each part number that assist in searching and identification. Each part number can encompass several different preferred and alternate items. So, we have the top level part number, and then one or more items under each part number.

For each part number we identify:

Category (resistor, IC, PCB, assembly, chemical, etc.) Description Storage location (inventory, offsite storage, hazardous lockup, fridge, etc.) Is part warranted, and if so for how long? Serial number required? History (pricing, purchasing, receiving, picking)

For each item under each part number we identify:

Manufacturer Part number Marking Link to PDF datasheet Photograph(s)

We have other stuff in there too, but this gives a good overview.

Our search tool makes a descriptive part number unnecessary. We like the shorter part number, since there is less chance for error when entering a part number, and it is quicker to enter. Also, whenever a part number is entered, the system displays the part description so that an error can be easily detected.

================================

Greg Neff VP Engineering

Suggest you go find out about relational database design. Once you understand how that works, most of what you want is easy to implement. Look for keywords like 'Natural key' 'Surrogate key' '1 to many relationship'.

You might want to avoid easily-confused choices such a 0/O, 1/I, O/Q...

I worked for a Japanese company that started off with a system where their part number was whatever the part number of the vendor who sold it to them was. Of course you can predict what happened next; two vendors making two parts, both called 98243, followed by a new numbering system.

I asked them how the system could have possibly worked in Japan, and they told me they simply asked the vendors not to use the same part numbers...

In my system (1-2345-6789) all subassemblies start with 8, followed by the top assembly number, followed by a four digit part number that is invariant from model to model.

For example, the chassis hardware subassembly for our audio panel model 564 is

The chassis hardware subassembly for our navigation receiver model 523 is

Is that what you had in mind?

Jim

components

If we had a part number to shelf & bin lookup on a database, we could keep all the resistor 1% SMD reels on the same shelf, even though the part numbers are simple numbers. That means our storeroom makes best use of space. We could print out some sheets which let us do the lookup without a computer in the storeroom. We could even keep our existing parts numbers, warts and all and reorganise the parts on the shelf.

I'm just not sure that we are big enough to keep it all going. The stock guy is has got a lot of other work to do already and is a hands-on sort of guy. There are just 6 or 7 of workers. If I want a part, I wander into the storeroom and get it.

I think the choices are becoming clear :

- reassign all part numbers in a way which lets us store parts by number. This system is simple, but storage by part number can still come unstuck as time changes parts and storage.

- use a database to connect part numbers to shelf locations. This is the ultimate system but requires discipline.

Thanks, Adrian.

components

lists,

order is

often in

significance

reel

resistor

sequence,

characters

system

I worked for company that operated a system very similar to this. It was fine for production, I believe. For the engineering design department it was a farce. So much so, that the lazy ones amongst us could design quicker by generating a fresh set of parts for each project. Management praise was forthcoming. But because of over-ordering on each project, the stores gradually grew and grew. When I investigated I found over 200,000 GBP of unused stock, some of the parts being of high value due to their age and not easily usable any more. The manager was not impressed and told me to keep quiet as it added to the value of the company (although I thought that auditors would not fall for that one).

My own preference is for a number system that at least includes an initial digit that classifies the part, e.g. 1=RES, 2=CAP, 3=IC, 4=RECT, with sub-assemblies being stored with reference to the drawing number system (start another thread?)

Graham Holloway

(snip)

We have areas of different general packaging types (small reel, large rail, rail, tray, bulk, etc,), and within each area the parts are in our part number order. This works fairly well, although we sometimes have to deal with parts in different packages. It's not perfect, but it's good enough.

================================

Greg Neff VP Engineering

(snip)

It sounds like engineering didn't have any constructive input to the system. Clearly, an engineer wants nothing to do with meaningless part numbers. To alleviate this problem, we handle engineering part selection in a descriptive manner. Our OrCAD part libraries are a controlled part of our inventory system. If I need a 1K 5% 0402 resistor, I open the RES_SMD_0402 library, and select 'R 1K 5% 0402' from that library. Nothing hard about that. Each part in each OrCAD library must include a field that is our unique part number. If we don't have 'R 1K 5% 0402' in the library, then we generate a new part number in the system and then generate the part in the OrCAD library. Generic library parts are prohibited. When we generate the BOM, we include our part number field. Our inventory control system scans the BOM into the system. Once it's in the system, we can generate work orders for x number of that assembly, including inventory pick lists, buy lists, etc.

================================

Greg Neff VP Engineering

Your goal is always to make it easier. If automation makes it harder, then you have failed in the implementation. But on the other hand if your stock guy is the only one who knows where stuff is, and he leaves...

-- Regards,

Adrian Jansen adrianjansen at internode dot on dot net Design Engineer J & K Micro Systems Microcomputer solutions for industrial control Note reply address is invalid, convert address above to machine form.

Just use the Digikey number. If it can't be bought from digikey, you shouldn't be using it in your design.

Now, you might just have to move to the USA, but ...

PN2222A

Digikey p/n 497-2406-ND

What do we do in the UK, where Digikey have left out 50 or so pages from this year's catalogue?

Paul Burke

Just try to keep from emulating Mouser. Some of their part numbers are simple 6-digits from the days in the late 1960s when Jerry Mouser was selling offshore capacitors out of the back of his pickup in San Diego. Some of their part numbers are the manufacturer's part number prefixed by a

Jim

"No Discount" I think.

John

What do you do when more than one Digikey p/n maps to your part spec? Even in this case I see 12 Digikey P/ns from 7 Manufacturers.

Robert

Thank goodness he didn't choose to prefix every P/N with "ND" or something silly like that.

Best regards, Spehro Pefhany