DS250DF410: turn off retimer function

Hi Brian,

Yes, the MUX functionality on this device is facilitated by analog signal chain and is not dependent on CDR.  Please see section 7.42 Raw mode(CDR Bypassed) in the programming guide.  I shared the mysecure folder with you.  Can you add some details on why the customer is interested in disabling the retimer function?  Is the data rate not supported by the retimer?

One item to note is that this device has package level polarity inversion.  This is not an issue while operating in straight mode since there is an inversion on RX and TX, resulting in a net non-inverting signal chain.  However, when using the MUX, this leads to cases where there is a polarity inversion to the signal.  While operating in retimer mode, this is correctable through the use of TX FIR filter.  However, in CDR bypass case, this polarity inversion cannot be corrected at a device level.  This may not be a system issue if system receiver has the ability to handle a polarity inversion.

Thanks,
Drew

Hi Drew

thx for your quickly reply

our customer issue is they can't link up the rail to intel CPU.

in other project. DS100MB(only MUX function)  can work well but in this case CPU can't read signal.

our customer hope they can save SFI function and no need for KR.

below are code and SCH .

##  Timux Setting (bypass CDR)

i2cutil -mw 0x18 0xfc 0x01   #select CH0

i2cutil -mw 0x18 0xff 0x03    #write for all channels, read for the specific channel defined in 0xfc

i2cutil -mw 0x18 0x00 0x04  #Reset all channel registers to power-on default values

i2cutil -mw 0x18 0x31 0x00  #adapt mode 0: no adaption

i2cutil -mw 0x18 0x1e 0x09  #puts device into RAW mode

i2cutil -mw 0x18 0x2d 0x38  #Enable EQ boost override

i2cutil -mw 0x18 0x03 0x00  #Set EQ boost value as 0

i2cutil -mw 0x18 0x8e 0x01  #vga_sel_gain=1

i2cutil -mw 0x18 0x13 0xb0  #set EQ gain to 1

Best

regards

brian

Hi Team,

We need to onsite support next week.

Could you guide us how to link up the rail to intel CPU and  save SFI function and no need for KR.

Many Thanks,

JImmy

Hi Brian and Jimmy,

DS250DF410 supports 10 GbE in retimer mode.  Can you clarify why the customer wants to disable retimer mode?  Typically I would expect better performance in retimer mode than in CDR bypass.  Did the customer configure the device for 10 GbE?

Additionally, if the customer is able to use retimer mode, I would recommend implementing a polarity flip when "cross" is selected in the MUX (i.e RX0 --> TX1).

Another question: What is the insertion loss of the channels in the block diagram you shared?  If you don't have measurements/simulation data, estimates are fine.

Thanks,
Drew

Hi Drew,

Thanks for your great support.

1. DS250DF410 supports 10 GbE in retimer mode. May I know why you want to disable retimer mode?  Typically I would expect better performance in retimer mode than in CDR bypass.

= = > In the similar topology, The Intel CPU  can detect Broadcom MAC, that is:

Broadcom MAC  <= =>TI DS100MB203(Mux/Buffer) <= => Intel CPU can link up(Learn other project)

Broadcom MAC  <= =>TI DS2502DF410(Re-timer) <= => Intel CPU can’t link up

TI DS100MB203 can but TI DS2502DF410 can’t, that’s why we want to disable re-timer mode for debugging purpose

Did you configure the device for 10 GbE?

= = >We set the DS2502DF410 working at 10.3125G

1. Another question: What is the insertion loss of the channels in the block diagram you shared?  If you don't have measurements/simulation data, estimates are fine.

= = >We preliminary exclude the SI issue because we can see the eye open in the intel CPU RX side, as the following picture

         We think that it should be good for link up

          The purpose of the on-site support is to find re-timer setting to solve the issue, and if proving TI DS2502DF410 can’t be use in our application

they want to know the exact root cause and need  your suggestion for another solution.

Many Thanks,

Jimmy