Issue related to crosstalk

[oscargomezf]

Hi everyone,

I'm designing an easy PCB and I would like to know If I'm understanding correctly the concept of crosstalking.

There are a lot of tools to calculate the crosstalk, PCB toolkit from www.saturnpcb.com, http://www.skottanselektronik.com/, http://www.eeweb.com/toolbox/microstrip-crosstalk, ...etc

I prefer to use this: http://www.eeweb.com/toolbox/microstrip-crosstalk because it's very easy to access it and you have more variety in units (mm, um, ...etc) than the others.

Anyway, On my design, I don't have clocks. I only have signals that are the typical GPIOS, with very low frecuency.

I understand If I'm trying to calculate the crosstalk of a clock I'll need a Cross Talk Coeff around -20.0 dB (Obviously, the bigger, the better). But My signals are very low frequency so, What Cross Talk Coeff is good enough?

Best regards.

[mairomaster]

-20 dB seems nice and safe to me.

The recommended level of cross talk doesn't really depend on the frequency, it always screws up your signals. However, it is much easier to achieve this particular level at lower frequencies. Frequency is not even the right term in the case - cross talk depends on the rise time of the signal, separation between the tracks, distance of the tracks to the reference ground plane and some other things. Normally though, at lower frequencies you have lower rise times, since such are easier to achieve by the hardware and high rise times are not required with low frequency signals. Hence, normally at lower frequencies you have less cross talk because of the lower rise times (not always - check your driver rise time specifications if available).

I use the Saturn PCB calculator for many things, but particularly for the cross talk. It is not a really high-end one, but it is the most accurate free one I could find. It gives you a good idea of the situation if you set it up right.

[oscargomezf]

Thank you mariomaster. Yes when I made the Schematic and PCB Design Course with robertferanec I remeber that he said that around -20dB.

I was trying to do a lot of calculations with the Sturn PCB to try to achieve the target of -20 dB. The design depends on so much of the stack up specially of the height.

[robertferanec]

Signals like interrupts or reset can be sensitive for crosstalk. As mairomaster explained, the problem is rising / falling edge - so for example even if your system is not high speed and you have a high crosstalk between your GPIO and RESET, it can cause you problems (e.g. one of 100 changes on your GPIO may cause random reset of your board).

So, for example you could route signals from same bus closer to each other and crosstalk will not influence your system (even if there is a crosstalk, it will happen at the time which is not important e.g. which is not at the time when you read or write the signals from bus), but if you mix signals from different buses with different timings and you route them close to each other, that could cause random problems.

So, what I normally do, I keep bigger distance between signals with different timings.

[oscargomezf]

Thank you roberfarenec. For example, It this type of bus is an i2c, then Is it advisable to keep -20dB between sda and scl? And do these signals have to be the same length? I've route this signals with +/- 10% Do you think is good enough?

Best regards.

[mairomaster]

It is hard to go wrong with I2C, it is quite low frequency (normally 100kHz) and usually low rise/fall times. Because of that it is less prone to cross talk and the length matching is not critical. Just be sensible with it and you will be fine.

[robertferanec]

For I2C bus no length matching is required. We route the signals together because it's just more clear in the layout. We do not route the signals extremely close to each other, but as mairomaster mentioned, I2C is not so sensitive and you should be just fine with it.