Why the rule of thumb: 1Mega ohm max when doing PCBs?

[eq1]

My curiosity was peaked when looking at this application circuit from linear:
http://cds.linear.com/image/Fig2_-_Wireless_Current_Sense_Circuit_Floats_with_Sense_Resistor.png

They put two 2M resistors in series instead of one 4M resistor. I've never really used anything bigger than 470K so I was wondering why someone would do that. Especially as the change is just a factor of 2 and not an order of magnitude. That just seemed unusual to me.

At first I assumed it was the parallel resistance of the FR4. I found online they guarantee 10^9 ohm/cm even when damp and hot. If I assume the R is 1mm, to keep the math easy, then that is 0.4% of 4Meg and of course 0.2% of 2Meg so in both cases one is not getting to 0.1% overall tolerance.

Maybe it's a way of controlling shot noise (i.e. the probability of two large simultaneous events is much smaller) or it's something related to the manufacturing process?

As part of searching for an answer I found the rule of thumb: Don't use anything larger than 1Meg. But there was not explanation as to why. Maybe this rule is the reason?

 

[gleem]

They put two 2M resistors in series instead of one 4M resistor. I've never really used anything bigger than 470K so I was wondering why someone would do that. Especially as the change is just a factor of 2 and not an order of magnitude. That just seemed unusual to me

Because a 4MΩ resistor of the preferred tolerance is not commercially available?

 

[berkeman]

My curiosity was peaked when looking at this application circuit from linear:
http://cds.linear.com/image/Fig2_-_Wireless_Current_Sense_Circuit_Floats_with_Sense_Resistor.png

Is that from an Application Note from Linear Technology? If so, they probably explain why they did that.

For high-voltage divider resistors, it's common to use a number of resistors in series, to add up to the desired standoff voltage.

 

[eq1]

Because a 4MΩ resistor of the preferred tolerance is not commercially available?

I never would've guessed it but I think this is the answer. There are a huge number of options over tolerance, temp co, and vendor on digikey for 2Meg SMT resistors but I only found one vendor that would sell better than 5% tolerance 4Meg SMT resistors. Thanks for the suggestion.

 

[eq1]

Is that from an Application Note from Linear Technology? If so, they probably explain why they did that.

It's actually from an LT insider blog post.
http://www.linear.com/solutions/7698

I didn't see anything as to why they selected the two 2Meg Rs over one 4Meg R. But the explanation from gleem seems to fit.

For high-voltage divider resistors, it's common to use a number of resistors in series, to add up to the desired standoff voltage.

I was thinking that at first since the application doesn't state what is putting current through the sense resistor but the whole circuit is powered from a 3.3V battery and is therefore floating relative to that source so I don't think that's the likely reason. Also, the non-inverting input of the opamp would be unprotected.

 

[UsableThought]

I didn't see anything as to why they selected the two 2Meg Rs over one 4Meg R. But the explanation from gleem seems to fit.

Apologies if you already did this - but did you look at the data sheets for the chips involved, esp. the "typical applications" sections?

I say this because I just remembered that whenever I've asked for help with my own feeble design efforts, especially when a chip of some sort has been involved, various expert builders have told me the same thing many times in many ways; usually however including an expletive: "Look at the ____ data sheet!"

So for example in this case, if you skim through the sheet for the LTC2063, there are quite a few references to resistors and also to voltage dividers, explaining particular needs for precision, heat dissipation, etc.; all of which might give you further insight: http://cds.linear.com/docs/en/datasheet/2063f.pdf

 

[Rive]

At first I assumed it was the parallel resistance of the FR4. I found online they guarantee 10^9 ohm/cm ...

The question is already answered I think, but this part...

Unfortunately the resistance of the FR4 is just does not apply for commercial production. At least, we could never measure anything even close once the PCB were populated (and washed).

 

[UsableThought]

Apologies if you already did this

Actually I'm looking at what I wrote and thinking it's naïve compared to the original post & followups - so never mind.

 

[Svein]

I suspect one of the reasons is to match source resistance as close as possible. The bias network source impedance is 2M (2 + 2 in parallel with 2 + 2) and the feedback resistor is 2M. Getting hold of 5 2M resistors with sufficient tolerance is much easier than getting different value resistors that match and track.

 

[Manolis]

My first post here, so first an introduction: I'm a 65 year old electronics engineer with practical (office & factory) experience. I've designed circuits & PCBs & equipment housings for optimum cooling. I've run a reliability testing lab. I've been a Technical Manager & QA Manager and I've run my own business. I started electronics at age ten. Now I'm "retired" and mess with electronics as a hobby. I note that I should also put this in the "new member" area.

In answer to the question, there's a couple of reasons:-

1. High value resistors are subject to leakage currents across the outer surface, made worse by handling and by poor cleaning processes. In other words, dirt & grease on the surface lowers the effective resistance and the shorter the resistor body, the worse the problem. Obviously, this problem also affects the guaranteed tolerance value.

2. High value resistors are often used in high-voltage circuits. The shorter the body, the higher the risk of arcing.

In both cases, connecting two resistors in series may cut the problems in half.

Note: with SMDs the problem is made worse by the very small size and by the fact that the final value (for accurate tolerance) is achieved by laser trimming - effectively digging a trench into the edge of the conducting layer. Also, the resistor body might be bare, porous ceramic material, which absorbs contaminants and can't be cleaned effectively. The problem is also made worse by PCB design. Where a component is subject to high voltage and/or leakage currents, the copper pads should be kept as far apart as possible and (obviously) tracks must not be run between them. In some cases, it might even be necessary to punch a hole or slot** to make sure that there's no conductive surface (= potential leakage path) directly between the connections.

** You might have noticed such a hole and assumed it was for cooling purposes. Ha ha. Think again.

Note: a common problem is for the designer to ignore the resistor's max. VOLTAGE rating. Either that or some pratt in Purchasing Department decides that he can order a cheaper version with a "similar" specification (except for the voltage). If no max. voltage is specified then it's safer to assume 50 volts than 500 volts! Otherwise the result could be expensive smoke.
(I have seen wrongly specified resistors burn up in a PSU. The manufacturer's answer was to replace the 1 Watt resistors with larger 2 Watt resistors. This move reduced, but did not solve, the problem. My solution was to specify 0.75 Watt resistors rated at 350 volts.)

 

[Svein]

2. High value resistors are often used in high-voltage circuits. The shorter the body, the higher the risk of arcing.

I had a similar reply in mind, but when I saw the circuit, I decided that no high voltage was involved.

Note: a common problem is for the designer to ignore the resistor's max. VOLTAGE rating. Either that or some pratt in Purchasing Department decides that he can order a cheaper version with a "similar" specification (except for the voltage).

Oh, yes. Fortunately, In my case, they were not allowed to change any component without asking the designer.