9
\$\begingroup\$

I've never seen a BOM that explicitly declares manufacturer part numbers for the components (except the obvious ones, like a microcontroller) in any kind of circuit schematics. According to my knowledge, "if part numbers exactly matches, then datasheet of the parts should exactly match and practical performances might match to datasheet or should perform better."

Lately I'm informed that this is not the case and we should provide a manufacturer part number for every part in the circuit, including resistors, capacitors and such.

I do understand that 555-X would have differences from 555, but shouldn't the exact same component codes be compatible between brands?

If not, how would I declare the minimum requirements for the components? How would a client know which parameters were important during the 555 variant selection? How would that client know why I choose for example TI-555 over the rest of the variants, so how would he or she decide if ST-555 would satisfy the circuit's requirements or not?

\$\endgroup\$
7
  • 6
    \$\begingroup\$ The people who buy the parts might not know which parts are interchangeable and which ones aren't. What kind of person/organization is your client? A manufacturer wants exact part numbers so they can't be liable for buying the wrong parts, but another engineer should be fine with "1k resistor" \$\endgroup\$
    – user253751
    Dec 29, 2020 at 18:56
  • 3
    \$\begingroup\$ Yes it's tedious, but it's critical. That is, unless you wanna be explaining the rash of field failures to your boss someday.... \$\endgroup\$
    – Kyle B
    Dec 29, 2020 at 19:14
  • 1
    \$\begingroup\$ @user253751 Regarding to those invaluable answers, it now wouldn't matter what kind of person/organization my client is, because very same issues might be experienced in my own company, even by myself. If, for example, I forget what parameters I had considered while selecting - eg. - that MOSFET, I probably will struggle to select a counterpart. Part selection process should also be very well documented. \$\endgroup\$
    – ceremcem
    Dec 29, 2020 at 22:32
  • 1
    \$\begingroup\$ @Mast then you can put "1k 2% resistor". Point is, another engineer usually doesn't need the precise part number \$\endgroup\$
    – user253751
    Dec 30, 2020 at 14:06
  • 1
    \$\begingroup\$ @user253751 Agreed, but you don't make the BOM for the next engineer. You make it for whoever is doing the assembly and buy-in. \$\endgroup\$
    – Mast
    Dec 30, 2020 at 14:19

7 Answers 7

16
\$\begingroup\$

It obviously depends from the component and what it is used for in a circuit. Take for example a simple 1k resistor in a certain package.

A manufacturing facility would have component engineers that deal with alternative parts, and since they cannot know what the resistor does in a circuit, they also have no clue what kind of resistor they should select. So if you don't specify at least one specific resistor part or multiple different specific parts, they don't take the risk and put just something there, they will ask you for instructions suggest alternatives that you must choose from.

Sometimes, it could be that any 1k resistor with 10% tolerance for a LED is fine, but if it needs to be a precision 1% resistor for some sensitive thing, it makes a difference. Also for high speed AC circuits the resistor material makes a difference, a wirewound resistor won't work while a film resistor would. For low noise a carbon resistor may be out of the question. For high pulse current applications a standard resistor is not acceptable. There are just so many different kinds of 1k resistors, even if you did state the tolerance, package, wattage, and many other things.

And no, the part numbering does not mean that chips with identical part from different manufacturers behave identically under all circumstances. There might be a reason why you chose a specific manufacturer. Or if there is not, then indicate it that it should not matter. But then you should also be sure that any manufacturer is fine, perhaps by looking at datasheets and/or testing the component in the circuit before accepting the use in design.

Sometimes even a same chip but from different batch or with different silicon revision fails to work as intended in a circuit.

\$\endgroup\$
2
  • \$\begingroup\$ ..and 1K resisters have power ratings too. \$\endgroup\$
    – MaxW
    Dec 30, 2020 at 17:37
  • 2
    \$\begingroup\$ @MaxW Yes, power rating was mentioned. \$\endgroup\$
    – Justme
    Dec 30, 2020 at 17:39
11
\$\begingroup\$

Sometimes it matters, sometimes it doesn't. If it matters, and your customer finds out before you do through failing product, then it matters a lot.

Most parts are generic, some parts are not.

The most egregious instance of different performance semiconductors sharing the same marking I've come across recently is the TLV431. The TI part only goes up to 6 V, whereas the OnSemi and the DiodesInc parts go to 16 V.

The other biggy is high value, high K ceramic capacitors, with their horrendous voltage coefficient of capacitance. There's no point researching and standardising on a particular manufacturer and package size of capacitor, if your buying department then buy the cheapest rubbish in the same nominal size.

What my company did was to specify a list of acceptable manufacturers for each part. Let's say it took less testing to get a new resistor onto that list than a new semiconductor.

\$\endgroup\$
5
\$\begingroup\$

Sometimes it matters greatly, sometimes it matters not at all.

I've been on the receiving end of "it matters greatly" a couple of times.

  1. I was a technician in a Motorola factory that built the CQM6000. The command board (microprocessor and audio signal processing) could be placed in a test mode by attaching a pull-up to a specific accessory connector pin at power on. The text jigs all used a pull-up to the 12V power supply. This worked fine for thousands of radios - until one day it didn't. That digital signal went to the input of a 74XX something or other latch operating at 5V. The sudden change from "test mode works" to "WTF?" was that somebody, somewhere, bought that latch from a different supplier. The new ones didn't like that 12V pull-up on a 5V input. The old ones shouldn't have worked, either, but they were more tolerant. I was the lucky fellow who got to track down the cause of the test-mode failures. I knew how the test-mode worked because I had a test rig that made use of it. All the radios with test mode failures worked on my rig, but not in other ones. It turned out that mine used a pull up to 5V like it should have where other test rigs used a pull up to 12V. The pull up to 12V was used because the input pin and the 12V pin were right next to each other in the accessory connector. The test rig plug had a 1206 resistor soldered to the pins to activate the test mode. Once I figured out what was going on, all the test rigs were modified to use a proper pull up to 5V. Problem solved -but production was slowed for several hours while we looked for the cause and got it fixed.
  2. A small company I worked for had a custom controller board made for a repeater built of a couple of MC Micro (M110) mobile radios. The existing controller for the repeater used 9V digital logic, and the stuff we needed to connect to it used 5V digital logic. The board that the company had made by an independent contractor used some 4000 series CMOS ICs in the interface between the two systems to bridge the 9V/5V gap. The prototype worked, so we had the whole series manufactured. The finished boards didn't work. The ICs in the protoype and the ICs in the final product were from different manufacturers. The ones in the prototype had different voltage levels for high/low than the ones in the final product. We had to buy a bunch of ICs (from the correct manufacturer) and replace them in all of the finished boards because we didn't have time to have the controller redesigned.

Another case for "it matters greatly" is when selecting coupling capacitors for radio frequency circuits. A 100pF capacitor isn't just a 100pF capacitor - if the frequency is high enough. At high frequencies, you have to look at the inductance and self resonance of the capacitor. Two optically identical capacitors with the same value can be totally different at high frequencies.

A case for "it doesn't matter" are the typical pull-up resistors in digital circuits. Pretty much any 10k (or whatever) resistor will work. You don't need high precision or low inductance or any thing else special. Just (approximately) the correct resistance and the circuit is "happy."


You need to have some idea of what's critical and what's not.

Take the examples above.

The problem in the factory wasn't something that should have been anticipated. The service manuals all said "pull up to 5V for service mode." The guy who wired the test rigs saw the 12V handy (and 5V more difficult to arrange) and wired it the easy way. It worked, so nobody complained.

The repeater problem should have been designed to be more robust, though, so that it wouldn't matter which IC was used. It depended on the "typical" voltage levels for high and low rather than being designed with the maximum variations in mind (which the datasheets also give.) 5V was right at the edge of the low end of "high" for the ICs used, so the design should have considered it to be a critical point and either specified a specific IC - or better, taken a different route for converting the logic levels.

To get back to your question, take the example of the 555. If you are using it in a typical circuit, then it probably doesn't matter whether it's an LM555 from TI, a UA555 from Fairchild, or whatever. If you are going for very long timer periods or low power operation then you might need to specify a TLC555 (a CMOS variant of the 555 designed for low power consumption.)

\$\endgroup\$
5
\$\begingroup\$

In a safety critical world such as flight control computers, a single manufacturers part maps to a single internal part number. It is important to note that the qualification process includes every specific part used.

I have seen X7R capacitors from multiple reputable vendors that exhibited very different capacitance under the same DC bias, for example.

Much the same sort of issue exists with (advertised as) form fit and function replacement integrated circuits. The output bypass requirement for various flavours of 1117 low dropout parts are different from each manufacturer particularly in terms of output ESR; what works for one manufacturer part might cause a different vendor part to oscillate (and probably destroy the part).

For those reasons (and others including operation over temperature) a single part from a specific vendor is used. When that cannot be used (availability or obsolescence comes to mind) a lot of analysis and testing might be required.

Not all types of equipment require this, of course. The key is to understand what can be used.

\$\endgroup\$
2
  • \$\begingroup\$ Unfortunately, many specification for parts are written too loosely to allow correct behavior to be guaranteed based upon specifications alone. I've seen quite a few parts, for example, where the required minimum hold time was less than the guaranteed minimum propagation time. Robust circuit design could have been much easier if clocked devices specified two ranges of threshold, such that setup time would be specified relative to when clock reached the low-end of the sampling range, hold time relative to when it reached the high end of that range, and minimum and maximum propagation times... \$\endgroup\$
    – supercat
    Dec 31, 2020 at 20:09
  • \$\begingroup\$ ...would be specified based upon when it reached the low and high ends, respectively, of the output-switching range. That would make it possible to guarantee correct interoperation between fast and slow parts if the maximum sampling threshold for the downstream parts was less than then minimum output-switching threshold of the upstream parts, and the clock slew rate was controlled if necessary to ensure that the downstream device would sample old data before the upstream device could possibly change it. \$\endgroup\$
    – supercat
    Dec 31, 2020 at 20:13
5
\$\begingroup\$

It depends. When making a hobby project to be prototyped in China (for example Seeedstudio or Makerfabs) a description like "0805 LED green" is enough. Your PCB will get any LED flavor-of-the-day and will be fine.

Professional manufacturing is a totally different world, especially in oil&gas and automotive sectors. All companies I have dealt with had lists of "approved suppliers". Component described as "0805 1k 1%" is not enough. It must have the full part number, so a 1k from ROHM may be okay, but 1k from Kemet may not. Even if they are functionally 100% identical. Moreover a "1k from Kemet" may not be okay, and a "1k from Kemet, AEC-Q100-qualified" (automotive standard) has to be used. Oil&gas requires extensive paperwork (when was this component made, who handled this component, how did it travel through the distribution chain etc.) so the final component you buy may be a "1k from Kemet with proper paperwork" from the very same assembly line as all others, just with 10x the price.

Using the full manufacturer and part number also helps when the design needs some particular properties of the component that are not immediately obvious. For example - a capacitor can be basically described in terms of: package (size), capacitance, tolerance, max working voltage. Your circuit may also depend on ESR (important for SMPS), inductance, self-resonance frequency, ESD susceptibility, temperature coefficient, capacitance-vs-voltage behavior, lifetime-vs-temperature, RoHS. Now... if you list 10 parameters next to your "generic" capacitor in the BOM the assembly house will probably ask you to just provide an exact part number to avoid guessing and risking a wrong component.

Of course using generic components in your design is good practice from the business point of view. If the design is flexible then different suppliers can be used depending on circumstances. Panasonic has a long lead time on a particular capacitor?! Just buy a similar one from AVX etc. This is ultimately good for the end price and delivery time.

\$\endgroup\$
1
  • \$\begingroup\$ Medical is pretty much the same as oil&gas - each component is traceable back and forth. And I understand aerospace also works like that. Automotive doesn't, as far as I experienced. \$\endgroup\$
    – MSalters
    Dec 31, 2020 at 11:40
4
\$\begingroup\$

It’s a tricky one. Sometimes you’ll specify a single source (e.g. Texas Instruments) and sometimes you’ll specify a component value (e.g. 10k with 5% tolerance), but sometimes you’ll need to provide a list of acceptable alternatives. It’s up to you to specify.

\$\endgroup\$
0
2
\$\begingroup\$

I've never seen a BOM that explicitly declares manufacturer part numbers for the components (except the obvious ones, like MCU) in any kind of circuit schematics.

I always use full part numbers of parts (including where to buy them) in my BoMs except for generic components like resistors and capacitors. For resistors and capacitors I use a coding system (in effect a part numbering system) that defines the resistor or capacitor such as this: -

enter image description here

I even put relevant parts of that table onto every schematic I produce.

If you want something built right with the minimum possibility of error or ambiguity, add the information into your BoM. Don't leave anything to chance.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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