A
Antti Lukats
Guest
Hi
I have a few questions (I belive the answers could be of general interest)
1) MGT POWERDOWN seems to be only possible way to tristate MGT TX, but there
is seems to be some latency when exiting powerdown mode, unfortunatly this
behaviour seems to be fully undocumented ? Any pointers numbers?
2) When AC coupled RX inputs are floating (or even short circuited rxn-rxp
with SMB loopback cable) the MGT receiver keeps still receiving noise, this
noise seems to have little different characteristics depending what MGT (in
the same FPGA) is used. In all cases the noise seems to have some pattern
with 4 data bits wide repeat cycle. This is probably normal, but it looked
weird, any other observations or some explanation?
3) AC decoupling capacitors - any thumb formula what nominal to use
depending on the the transmittion rate? Value of 0.01uF seems to be
recommended, but it causes the RX DC balance to take quite long time to
settle. When MGT TX is in TXINHIBIT mode the caps will fully charge, when
then TX is restored (to 50:50 DC balance data stream) the charge on the AC
decoupling caps makes the receiver lock to delay. During the time the AC
decouplinc caps discharge the RX is receiving either constant 0 or 1. And
that for a quite long time >= 100s of symbols.
4) There have been comments on this newsgroup that RocketIO can be used as
high speed serial to parallel converted when all smart features are fully
disabled - has anyone actually implemented this succesfully? Are there any
special consideration/constraint that apply? As of my understanding there is
some "life" inside the MGT (when all things are bypassed) that prevents this
type of useage (see commet [2] about the noise).
Thanks in advance,
Antti
I have a few questions (I belive the answers could be of general interest)
1) MGT POWERDOWN seems to be only possible way to tristate MGT TX, but there
is seems to be some latency when exiting powerdown mode, unfortunatly this
behaviour seems to be fully undocumented ? Any pointers numbers?
2) When AC coupled RX inputs are floating (or even short circuited rxn-rxp
with SMB loopback cable) the MGT receiver keeps still receiving noise, this
noise seems to have little different characteristics depending what MGT (in
the same FPGA) is used. In all cases the noise seems to have some pattern
with 4 data bits wide repeat cycle. This is probably normal, but it looked
weird, any other observations or some explanation?
3) AC decoupling capacitors - any thumb formula what nominal to use
depending on the the transmittion rate? Value of 0.01uF seems to be
recommended, but it causes the RX DC balance to take quite long time to
settle. When MGT TX is in TXINHIBIT mode the caps will fully charge, when
then TX is restored (to 50:50 DC balance data stream) the charge on the AC
decoupling caps makes the receiver lock to delay. During the time the AC
decouplinc caps discharge the RX is receiving either constant 0 or 1. And
that for a quite long time >= 100s of symbols.
4) There have been comments on this newsgroup that RocketIO can be used as
high speed serial to parallel converted when all smart features are fully
disabled - has anyone actually implemented this succesfully? Are there any
special consideration/constraint that apply? As of my understanding there is
some "life" inside the MGT (when all things are bypassed) that prevents this
type of useage (see commet [2] about the noise).
Thanks in advance,
Antti