Eye diagram

The matlab plot doesn't look like a proper eye diagram to me - so I assume you're asking about how to set things up in Matlab?

Also, I didn't say in the other post that there is no such thing as a regular sampling frequency - I said that there isn't a sampling frequency unless you told it to sample. In order to compare with Matlab, I used this example netlist - which comes from my app note on Cadence Online Support (with a small modification to strobe the output):

 // prbs.scs

vclk (clk 0) vsource type=pulse period=80n val0=0 val1=1 rise=0.3n fall=0.3n delay=3n
v1 (high 0) vsource type=prbs period=40n val0=0 val1=1 rise=0.3n fall=0.3n delay=3n //seed=1942
del (withdelay high) randdelay sd=1n rise=0.3n
rl (withdelay jitter) resistor r=100
cl (jitter 0) capacitor c=50p

r1 (high 0) resistor r=1k

ahdl_include "randdelay.va"
tran tran stop=400u strobeperiod=0.5n

Notice I've added strobeperiod=0.5n to force there to be fixed rate outputs every half a nanosecond.

This uses  this VerilogA model to add random jitter (of sorts):

// VerilogA for randdelay

`include "constants.h"
`include "discipline.h"

module randdelay (op,ip);
output op;
input ip;
electrical op,ip;
parameter real sd=2.0n;
parameter real gainsd=0.1;
parameter real del=5.0n;
parameter real rise=0.1n;
parameter real thresh=0.5;
parameter integer seed = 23133;

integer vseed;
real randnum,gain,delayed;

analog begin

  @(cross(V(ip)-thresh)) begin
    randnum=$rdist_normal(vseed,0,1);
    randnum=randnum*sd+del;
    gain=$rdist_normal(vseed,0,1)*gainsd+1;
    if(randnum<0) randnum=0.0;
  end

  delayed= transition(V(ip),randnum,rise,rise);

  V(op) <+ gain*(delayed-thresh)+thresh;

end

endmodule

I run the simulation using spectre prbs.scs and then in Matlab I do:

jitter=cds_srr('prbs.raw','tran-tran','jitter')
eye=commscope.eyediagram('SamplesPerSymbol',160,'SymbolsPerTrace',1,'SamplingFrequency',1/0.5e-9, 'MaximumAmplitude',1.25,'MinimumAmplitude',-0.25)
eye.update(jitter.V)

There are 160 samples per symbol, because the period of my signal is 80ns, and I have a sample every 0.5n - so there are 160 samples per symbol. I want to show one Period in the trace, and the SamplingFrequency is the inverse of my strobeperiod. I've told it the Max and Min amplitude to scale it OK.

In ViVA I use the Eye Diagram assistant and tell it to use a period of 80n and I start from 160n (just to avoid a startup glitch, but that's not critical). The corresponding function is:

eyeDiagram(v("jitter" ?result "tran") 160n 400.0u 80n ?intensityPlot t)

(the intensityPlot bit is just to make it look prettier)

I've captured the two results side by side. As you can see, they look rather similar (The colours in the Matlab graph don't show up as brightly for the rarer points, but other than that you can see they match well).

Regards,

Andrew.

The matlab plot doesn't look like a proper eye diagram to me - so I assume you're asking about how to set things up in Matlab?

Also, I didn't say in the other post that there is no such thing as a regular sampling frequency - I said that there isn't a sampling frequency unless you told it to sample. In order to compare with Matlab, I used this example netlist - which comes from my app note on Cadence Online Support (with a small modification to strobe the output):

 // prbs.scs

vclk (clk 0) vsource type=pulse period=80n val0=0 val1=1 rise=0.3n fall=0.3n delay=3n
v1 (high 0) vsource type=prbs period=40n val0=0 val1=1 rise=0.3n fall=0.3n delay=3n //seed=1942
del (withdelay high) randdelay sd=1n rise=0.3n
rl (withdelay jitter) resistor r=100
cl (jitter 0) capacitor c=50p

r1 (high 0) resistor r=1k

ahdl_include "randdelay.va"
tran tran stop=400u strobeperiod=0.5n

Notice I've added strobeperiod=0.5n to force there to be fixed rate outputs every half a nanosecond.

This uses  this VerilogA model to add random jitter (of sorts):

// VerilogA for randdelay

`include "constants.h"
`include "discipline.h"

module randdelay (op,ip);
output op;
input ip;
electrical op,ip;
parameter real sd=2.0n;
parameter real gainsd=0.1;
parameter real del=5.0n;
parameter real rise=0.1n;
parameter real thresh=0.5;
parameter integer seed = 23133;

integer vseed;
real randnum,gain,delayed;

analog begin

  @(cross(V(ip)-thresh)) begin
    randnum=$rdist_normal(vseed,0,1);
    randnum=randnum*sd+del;
    gain=$rdist_normal(vseed,0,1)*gainsd+1;
    if(randnum<0) randnum=0.0;
  end

  delayed= transition(V(ip),randnum,rise,rise);

  V(op) <+ gain*(delayed-thresh)+thresh;

end

endmodule

I run the simulation using spectre prbs.scs and then in Matlab I do:

jitter=cds_srr('prbs.raw','tran-tran','jitter')
eye=commscope.eyediagram('SamplesPerSymbol',160,'SymbolsPerTrace',1,'SamplingFrequency',1/0.5e-9, 'MaximumAmplitude',1.25,'MinimumAmplitude',-0.25)
eye.update(jitter.V)

There are 160 samples per symbol, because the period of my signal is 80ns, and I have a sample every 0.5n - so there are 160 samples per symbol. I want to show one Period in the trace, and the SamplingFrequency is the inverse of my strobeperiod. I've told it the Max and Min amplitude to scale it OK.

In ViVA I use the Eye Diagram assistant and tell it to use a period of 80n and I start from 160n (just to avoid a startup glitch, but that's not critical). The corresponding function is:

eyeDiagram(v("jitter" ?result "tran") 160n 400.0u 80n ?intensityPlot t)

(the intensityPlot bit is just to make it look prettier)

I've captured the two results side by side. As you can see, they look rather similar (The colours in the Matlab graph don't show up as brightly for the rarer points, but other than that you can see they match well).

Regards,

Andrew.