DS100DF410: work with default settings

Hi Eduard,

1. The retimer channel in question needs to have CDR rate programmed depending on the data rate you are implementing. Refer to the datasheet for details. In addition the Rx adapt mode and Tx output settings may be optimized depending on the pre-channel and post-channel electrical characteristics
2. Correct in that, if Masteer Mode programming is unsuccessful then the retimer switches to Slave Mode operation with default settings

Cordially,

Rodrigo Natal

HSSC Applications Engineer

Hi Rodrigo!

We are testing a 10GB link between two our boards. We use utilities iperf3 and netperf under Linux. We can not get a speed higher than 1.7 Gbit / s. Processor utilization does not exceed 25%. SPF modules and cable have been tested on another system. They are capable of running at 10 Gbit / s.

We tried various modes of adaptation of CTLE and DFE. This did not have a noticeable effect.
We looked at the data of the eye monitor. The signal is good.

You wrote that "The retimer channel in question needs to have CDR rate programmed depending on the data rate you are implementing". In the datasheet there is only this:

 

But there is no description of how to use it. Can settings 2F[7:4] help us or the problem somewhere else?

 

Additional Information:

1. We use referencelesse mode (0x36[5:4] = [0:0]). Our boards do not have RefCLK 25 MHz. We set this bits:

2. Now we try to OFF (bypass) CDR:

Data transfer rate has not changed (1.7 Gbit/s). Does this mean that CDR does not work?

The 1.7 Gbps rate is not supported in retimed mode by the DS100DF410. You can however configure the retimer in CDR bypas mode and operate at that rate. To set retimer to CDR bypass mode you must set channel register 0x1E[7:5]=b000 and then also set 0x09[5]=1

Cordially,

Rodrigo Natal

HSSC Applications Engineer

XFI/SFI differential lines works on 10.3125 Gbps. We have a LOCK INDICATOR PINS = HIGH.

Speed measurement on Keystone processor get 1.7 Gbps.  We tests with CDR on, and test with CDR bypass - the result has not changed.

How do we find out: is the reason in the retimer or in the processor?

Should we set bits RATE0, RATE1, SUBRATE0, SUBRATE1 in the register 2F?

Hi Eduard. SO is the retimer channel acquiring CDR lock? What values are you observing on channel registers 0x01, 0x02, 0x03, 0x27, and 0x28?

Cordially,

Rodrigo Natal

HSSC Applications Engineer

Hi Rodrigo.

Values board 1:
0x01 = 0
0x02 = 5C
0x03 = 0
0x27/0x28 = 0x30/0x51

Values board 2:
0x01 = 0
0x02 = 5C
0x03 = 1
0x27/0x28 = 0x2C/0x66

Hi Eduard,

Is 1.7 Gbps the actual measured bitrate on the interface or the effective data throughput of the interface?

The DS100DF410 retimers will not impact the overall throughput.

The data collected in the registers indicates that the retimers are locking and there is a valid eye.

Regards,

Lee

Hi Lee.

1.7 Gbps is the actual measured bitrate on the interface. We use utilities iperf3 and netperf under Linux.
We are looking for a problem and want to be sure that the retimer is working correctly. If you think that the retimer is working correctly, then we will look for a problem in the processor.