Hard Lessons Learned (watch out for land mines)

Steve Owens wrote:

 

nor do their libraries contain parts that they actually sell

I think one should design a circuit based on the most appropriate components (appropriate for the design) that one can source. Requirements include availability, cost, performance, the ability to second-source, and so on.

If you are designing a circuit based on whatever you can find in a CAD software library then this is not typically how the design process works.

To address Steve's point, I would argue that the hallmark of a useful library be that it contains parts that are readily available, at reasonable cost, with good perfomance, and have second sources?

Hi Mark,

There is no interface in EAGLE between a library and inventory/stock levels at distributors, so there can be no assurance that a part in a library is readily available, nor of good performance.

There have been cases where manufacturers have rapidly deprecated a product (e.g. CC3000 comes to mind, within the space of just a couple of years) and the replacements are not pin-compatible.

No-one likes functionality to be removed in later releases of software, so I think it is right that libraries do not have content deleted over time to match the state of affairs with manufacturer supply and demand.

I know some people believe EAGLE libraries to be like a menu in a restaurant, but it is not the same; disappointment if they saw cheesecake on the menu and it had run out is not comparable to

a CAD package's library.

Just as importantly, there is a trust issue - no-one would invest time and money building a prototype based on unverified parts libraries. A project I'm working on is getting ready for parts purchase

(perhaps 500 parts per board)  and I will look very silly in front of my colleagues if the parts are not available because I didn't check, or if they do not fit because I didn't verify the landing pattern.

Basically similar to 'don't outsource what is strategic' I guess.

The solution is to create or verify parts libraries. And the process is established and works - as an example in over 100 boards I've created, while mistakes have been made, I don't think I've ever had a

landing pattern issue on a prototype that could not be corrected without more than a couple of wire links on the first prototype. Basically, 'get it right first time' is the motto, just as relevant for

business and for hobbyists because none of us have time to waste re-ordering boards. As a quick visible example, one of my published projects in a blog post has 92 parts on it (and the schematic and

board were entered in EAGLE) - not a huge amount of parts, but not insignificant either - and the first board worked first time, no need for a second revision.

So, is it fair to say then that you agree that a useful library does contain parts that are readily available, at reasonoable cost, with good perfromance, and have second sources.  What I read in your response is that to achieve these objectives one has to carry out due diligence and not blindly rely on pre-packaged libraries; is that a fair intrepretation?

Hi Mark!

I'm saying one usually chooses the best parts for the task  (taking into account cost, performance, availability decisions), and then finding and verify a library part, or designing your own.

The libraries that were created in EAGLE, were possibly "useful" parts at one time or another. Should they be deprecated and removed now that they are "useless" in the eyes of some? I think no, for the reasons mentioned in the last post.

Besides, some parts like TBA820M are considered useful (and available) in some parts of the world, but are definitely not suitable for designs in other parts of the world.

What is useful to us may be useless to others and vice versa. But yes, one key point is that you wouldn't know it was useful until you'd verified all parts you need to use yourself.

Some parts libraries attempt to get around this issue by having users 'rate' the part.. so people get visibility of a statistical confidence of sorts.. much like an Amazon book rating or 'look inside'. It doesn't guarantee the book is useful, but provides some confidence.

As another real-world example: I have a great (to me) toroid library, and it was built using custom software that I have tweaked over the years.

I can design parts like this, and it even tells me exactly where each wire turn should sit:

But, wound toroids are somewhat esoteric, expensive and unobtainable unless you contract with a manufacturer to custom-wind them for you (or carefully hand-wind for half an hour).

That library is useless to most people, extremely useful to me.

What perhaps the original poster's real issue boils down to (I'm guessing), is that some people need a hand to realise what are "popular" or "jelly-bean" parts that will usually be in stock, have reasonable performance,

and are likely to still be available from suppliers when they actually make use of the library. If users know that, then they can have some confidence that a library contains the part, the library

part has been sufficiently debugged through popular use, and the distributor's inventory contains the part.

An approach sometimes seen is a 'jelly-bean' library, or perhaps some parts list marked as 'jelly-bean'; companies like Adafruit and Dangerous Prototypes have libraries that are kind-of like this, already available for EAGLE.

(Called generic parts libraries or similar I think). They probably are fine, but personally I would check each part myself, no matter the rating.

Am 13.10.2015 um 19:51 schrieb shabaz:

Hi Mark!

 

I'm saying one usually chooses the best parts for the task  (taking into

account cost, performance, availability decisions), and then finding and

verify a library part, or designing your own.

 

In the company in which I worked before my retirement, every

component has been managed by a unique ID (eg XYZ123456). This

designation was used in Eagle as attribute in our own Eagle

component libraries. Availability, price and suppliers were

exclusively managed in a similar SAP

(https://en.wikipedia.org/wiki/SAP_R/3) software. This is in my

opinion, the only real solution for a professional use of Eagle

libraries.

For my personal use I look to the web site of my favorite seller

and do it like Guillaume.

Freundliche Grüße / Kind regards

Friedrich

-

... use NNTP:
news.cadsoft.de and a

functional news reader like Thunderbird!

I'll chip in here although I don't use Eagle because the discussion of libraries is close to my heart.

I totally agree with the point made or implied earlier that the existence of a part definition in a CAD tool library is not  a good reason to use it.

But I'll go much further - although market competition forces CAD tool makers to build up these libraries they are mainly useless and encourage very bad engineering practice.

The only standard library I use is the IPC footprint library and even then I usually tune the footprints before use.

I don't use any other pre-made stuff because:

it may contain errors and checking is difficult

it doesn't reflect the process that will be used on my boards

the schematic symbols (for complex parts) are very unlikely to suit my design intent

I have schematic symbols for standard parts  that I have used on zillions of designs.

the library part will often use a numbering or naming scheme that doesn't work for me so needs to be edited anyway.

So I cringe every time I see a request on E14 or elsewhere for  a library for some part or other as if it was some special deal to make one's own.

MK

Exactly, I've seen the same thing with custom part codes.

The part stock, availability, suppliers etc is usually managed separately. And separate part codes for assemblies and 'transformed' parts, e.g. cable assemblies or (manually) matched components.

Another example, a satellite company will store batch information and never mix with later purchased parts, even if they are from the same manufacturer and supplier, to reduce risk.

So parts are "exclusive" to a particular build, spares and to any subsequent re-orders of complete satellites.

Exactly, I've seen the same thing with custom part codes.

The part stock, availability, suppliers etc is usually managed separately. And separate part codes for assemblies and 'transformed' parts, e.g. cable assemblies or (manually) matched components.

Another example, a satellite company will store batch information and never mix with later purchased parts, even if they are from the same manufacturer and supplier, to reduce risk.

So parts are "exclusive" to a particular build, spares and to any subsequent re-orders of complete satellites.

First off, I am impressed with the skill and attention to detail exhibited in your custom torroid library.  Nicely done!

Second, I appreciate the deep experience brought forward in this dicussion.  The thoughts expressed here are actually valueable in a real world context.

Hi Mark,

That toroid library was a pain ; ) it takes into account wire thickness, and ideal spacing to minimise capacitance between the turns.

And I had forgotten so much trigonometry..

I have a spreadsheet to calculate the number of turns based on desired inductance basically, and use that as input to the tool to auto-generate the part.