Differential Pairs matching
rajat_jain_785 , 05-30-2021, 10:09 AM
Hello,
I saw a video on Differential Pairs by Rick Hartley.
He mentions that differential pair length matching is not as critical as it has been considered by many pcb designers. He says it takes 50mm difference of length for a 10Gbps differential pair to fail. This fact doesnt make sense to me.
He also says that he has almost always never length matched a differential pair and it has always worked for him. Then what is the reason we spend so much time and energy trying to length match critical differential pairs to within say 5mil or 10mil.
Pls say what you think about this.
Regards
RAJAT
I saw a video on Differential Pairs by Rick Hartley.
He mentions that differential pair length matching is not as critical as it has been considered by many pcb designers. He says it takes 50mm difference of length for a 10Gbps differential pair to fail. This fact doesnt make sense to me.
He also says that he has almost always never length matched a differential pair and it has always worked for him. Then what is the reason we spend so much time and energy trying to length match critical differential pairs to within say 5mil or 10mil.
Pls say what you think about this.
Regards
RAJAT
Paul van Avesaath , 05-30-2021, 02:27 PM
Hi Rajat,
He mentions the 50mm @ 100Mhz, but @ 10GHz it stated +/-1.52mm that is a long way from 50mm
still that is a far cry from the 5 mil that you are trying so hard to reach..
take care that when there is a large different there will be common mode currents!! that will create EMI and will screw a lot with your system. so even if the diffpair will function properly.. he does not say it will function properly in a system when there are many diff pairs..
so the answer is to keep the EMI (common mode voltage) as low as possible.
also it brings down the ISI (inter symbol interference)
also please be aware of what he told about the GND via's near a layer switch.. this is very VERY important.
Anyway just always try and do your best to route them as neat as possible.. " if it looks good it will work good"
if you are unsure of what he is saying then please try and keep lines neatly coupled (far or close whatever) do not go blindly towards what he is saying, just because he is at the Altium summit. he is not wrong, but there is more comming into play when designing a full system. in theory all is arguments are correct but again in specific ideal situations that he describes.
Paul.
He mentions the 50mm @ 100Mhz, but @ 10GHz it stated +/-1.52mm that is a long way from 50mm
still that is a far cry from the 5 mil that you are trying so hard to reach..
take care that when there is a large different there will be common mode currents!! that will create EMI and will screw a lot with your system. so even if the diffpair will function properly.. he does not say it will function properly in a system when there are many diff pairs..
so the answer is to keep the EMI (common mode voltage) as low as possible.
also it brings down the ISI (inter symbol interference)
also please be aware of what he told about the GND via's near a layer switch.. this is very VERY important.
Anyway just always try and do your best to route them as neat as possible.. " if it looks good it will work good"
if you are unsure of what he is saying then please try and keep lines neatly coupled (far or close whatever) do not go blindly towards what he is saying, just because he is at the Altium summit. he is not wrong, but there is more comming into play when designing a full system. in theory all is arguments are correct but again in specific ideal situations that he describes.
Paul.
Use our interactive Discord forum to reply or ask new questions.