Bad idea but why? Perfect resistance circuit

[FriedMule]

Since nobody has used this idea do I assume it's because it's a bad idea, but I do not understand why.

Mr. EEV do in one of his video tell about a trace being a precise resistance per square, no matter if it's a square millimeter, a square inch or a square mile.
Therefore, did I think, if you got a high quality board made where the manufacture knows the exact resistance, and did make them in some very precise length, then should you be able to make traces that are very precise in their resistance.
In that way would you know that trace number 5 was 1.0000001 KOhm or something like that.

[The Soulman]

1) Although the length and width of traces can be machined fairly precise its not that precise, also the thickness of the copper
may vary a few percent here and there.

2) Copper is a poor material for making precision resistors because its resistance changes changes a lot with temperature.

[Rerouter]

Temperature coefficient of copper is about 3930 parts per million per degree C
The trace will get stetched as the board changes temperature about 17ppm per degree.
Any flex in the board will also impart a larger ppm error (strain gauge effect)
A given plated pcb blank has tolerances on its plating thickness, purity and plating thickness consistancy
Etching is a very uncontrolled process for perfect traces, zoom in far enough and its bumpy and uneven in all kinds of fun ways,

In short, It would only be "accurate" at a very specific temperature, with no external force on the PCB, and you would not get a say in what value that is, just roughly where it sits,

If you want a precise value, easiest way is to buy some constantan wire, use that for most of your resistance value, and your within 40 parts per million, or manganin wire for within 20 parts per million, if you where serious, you could roughly use calculated lengths of the 2 to trim down the ppm to near nothing, then trim the length to get the ohms where you want it,

In some old equiptment this was wound around a bakelight like form card, and in 1 refurbed unit literally around cardboard, to get a very precise custom resistance value. leaving the only issue being measuring the thing,

[FriedMule]

Okay thank you, i am in doubt you are both right, but when eev sometimes open up a high precision unit do you find resistors and capacitors, only made by copper traces on the pcb. How do they dare to rely upon that?

[Rerouter]

Capacitance and inductance is something different, that is RF, resistance doesnt play much of a role in that, and that more comes down to geometry tolerances, which are pretty good on average (spacing between layers, and spacing between traces on same layer)

What throws them astray is actually the weave pattern in the PCB material make hard to predict errors, and how much RF energy the pcb material absorbs (really fancy gear uses special material pcb's)

depending on requirements, that RF gear can use CNC metal shapes to accomplish similar effects, but PCBs can get you far enough for a number of applications (its always a tradeoff somewhere),

http://www.earf.co.uk/ae88_pcb.JPG

RF I like to refer to electron sheparding, when the frequency is high enough, the electrons get energetic enough to be pushed out a little into the air surrounding the trace and away from the center of the copper, all the RF magic on the PCB's is then shaping the EM feilds to funnel the electrons with the right energies where you want them, usually filtering out the ones you don't want.

[Zero999]

Okay thank you, i am in doubt you are both right, but when eev sometimes open up a high precision unit do you find resistors and capacitors, only made by copper traces on the pcb. How do they dare to rely upon that?

It won't make much difference to the capacitance/inductance.

The temperature coefficient can be useful, for example in applications such as a current sense resistor, so the current limit has a negative temperature coefficient.

[TimNJ]

Are you talking about laser trimmed resistor networks, like from Caddock, Fluke, etc.? If so, the element is not copper. It's a much more stable carbon/metal film.

https://xdevs.com/doc/xDevs.com/FX/networks_res_1.jpg

[ratatax]

I use pcb traces as resistors in a commercial product 
But it's only practical for low values (a few ohms) and when you don't need any precision. In my circuit I needed to increase the ESR of a capacitor ("damping") to reduce a voltage spike when the device is powered. Does the job !

[Dave]

Are you talking about laser trimmed resistor networks, like from Caddock, Fluke, etc.?

Probably talking about microstrip circuits from video teardowns of spectrum analyzers and such.

[perieanuo]

the ONLY pro usage of pcb trace for 'resistance' is as a fuse, it's really a low-cost silly idea, view the prices for smd's those days
pcb smd shunt for me is NOT for production, pcb traces can be delay lines or fuses...
someone from france said he put this in production, he's right, when I was doing r&d in france my boss was capable of cutting 2-3 smd's (really 2 cents less for the prod cost) and later, we had to spend precious time with some 5$ smart module returned to the company as warranty caused by this missing smd
well, at your home, you can do even heaters with your pcb traces if you want...
i'm caustic, but that's the naked truth, it' good engineering to cut cost, but there are limits
regards, pierre

[ratatax]

I think it's totally fine the way i used it. I needed about 1 ohm to damp a capacitor, and I don't care if the precision is horrible (even -50/+50%). It's not suitable to many applications also because it does waste a lot of pcb real estate. Adding a BOM line just for that resistor is kinda wasteful, even if the cost isn't that big