optPulseMetric

Pulse response metric for optimization routines

Description

example

metric = optPulseMetric(P,N,dt,B) calculates the eye height, eye width, eye area, and channel operating margin (COM) metrics at the target bit error rate (BER) level B. The function calculates these metrics from the pulse response matrix P, using the number of samples per symbol N, and the uniform sampling interval dt.

If there is no eye at the target BER level B, then the optPulseMetric function increases the BER until a positive height is realized.

The resolution of the optPulseMetric function is limited by number of samples per symbol N.

The nonreturn to zero (NRZ) eye metrics are estimated at a BER level, where both the center and maximum eye heights are reported. If the eye is not open at the target BER, the BER is increased until a measurable eye contour is found. The actual BER used is reported by the value usedBER reported in the output metric.

Examples

collapse all

Define the system exploration parameters. Set the target bit error rate (BER) to 1e-9, symbol time to 100 ps, number of samples per symbol to 16, and channel loss to 4 dB.

targetBER = 1e-9;
SymbolTime = 100e-12;
N = 16;
dBLoss = 4;

Calculate the sample interval.

dt = SymbolTime/N;

Create the channel model at the specified loss.

channelModel = serdes.ChannelLoss('Loss',dBLoss,'dt',dt,...
'TargetFrequency',1/SymbolTime/2);

Define a CTLE System object™.

ctle = serdes.CTLE('WaveType','Impulse',...
'SampleInterval',dt,'SymbolTime',SymbolTime,'Mode',2);

Process the channel impulse response with the CTLE System object.

impulse = ctle(channelModel.impulse);

Convert the impulse response to a pulse response.

P = impulse2pulse(impulse,N,dt);

Calculate the fast optimization pulse metric.

metric = optPulseMetric(P,N,dt,targetBER);

Obtain the result of the optPulseMetric function throughout the whole eye.

In order to do that, first find the number of complete unit intervals in the pulse response.

nUI = floor(size(P,1)/N);
prMatrixSort = sort(abs(reshape(P(1:N*nUI),N,[])).',1,'descend');

Then convert the BER into the number of largest pulse response samples to get the usedBER contour. The mean height is the first row (largest value) of the sorted pulse response. The noise is the sum of sorted pulse response's next nBER sample intervals.

nBER = floor(min(abs(log(targetBER)/log(2)),nUI-1));
meanEyeHeight = prMatrixSort(1,:);
Noise =sum (prMatrixSort(2:nBER+1,:),1);

Calculate the statistical eye and the eye metrics.

[stateye,vh,th] = pulse2stateye(P,N,2);
th2 = th*SymbolTime*1e12;
[~,~,contours,~,EHmax,~,~,~,~,EW,~,~,~,~,eyeAreas,~,COM] = ...
serdes.utilities.calculatePAMnEye(2,targetBER, ...
th2(1),th2(end),vh(1),vh(end),stateye);

Plot the statistical eye and the results of the optPulseMetric function.

figure,
subplot(2,1,1)
hold all
imagesc(th2,vh,stateye)
plot(th2,contours,'m:')
title('Statistical Eye and Contour')
axis('xy');
si_eyecmap = serdes.utilities.SignalIntegrityColorMap;
colormap(si_eyecmap)
xlabel('[ps]')
ylabel('[V]')
ax = axis;

subplot(2,1,2)
th4 = [(0:N-1),NaN,(0:N-1)]*dt*1e12;
plot(th4,[meanEyeHeight,NaN,-meanEyeHeight]/2,'r--',...
th4,[meanEyeHeight-Noise,NaN,Noise-meanEyeHeight]/2,'m:')
legend('Mean eye level','BER eye contour')
title('Fast Optimization Pulse Metric')
ylabel('[V]')
xlabel('[ps]')
axis(ax) Compare the results of the statistical eye metrics and the fast optimization pulse metrics.

comparisonTable = table([metric.maxEyeHeight*1e3;metric.eyeWidth*1e12;...
metric.eyeArea*1e12;metric.centerCOM; metric.usedBER], ...
[EHmax*1e3; EW;eyeAreas; COM; targetBER],...
'VariableNames',{'optPulseMetric','StatisticalEye'},...
'RowNames',{'Max Eye Height (mV)','Eye Width (ps)','Eye Area [V*ps]',...
'Center COM','BER'});
disp(comparisonTable)
optPulseMetric    StatisticalEye
______________    ______________

Max Eye Height (mV)        379.36            367.69
Eye Width (ps)              93.75            79.347
Eye Area [V*ps]            17.902            17.847
Center COM                 11.597            10.853
BER                         1e-09             1e-09

Input Arguments

collapse all

Pulse response, specified as a column vector. The pulse response vector can be calculated from the channel impulse response vector.

Data Types: double

Number of samples per symbol, specified as a real positive scalar.

Data Types: double

Uniform sampling interval of the waveform, specified as a real positive scalar in seconds.

Data Types: double

Target bit error rate (BER), specified as a real positive scalar.

Data Types: double

Output Arguments

collapse all

Pulse response metrics, returned as a structure with these fields:

FieldDescription
maxEyeHeightMaximum height of the inner eye in the usedBER contour curve.
maxMeanEyeHeightMean eye level height at the maximum inner eye height position.
maxCOMChannel operating margin (COM) at the maximum inner eye height position.
eyeAreaArea inside the usedBER contour.
eyeWidthEye width of the usedBER contour.
centerEyeHeightEye height at the center of the usedBER contour.
centerMeanEyeHeightMean eye level height at the center of the eye.
centerCOMChannel operating margin (COM) at the center of the eye.
usedBERThe bit error rate (BER) level used in the metric calculation that gives the nonzero eye height.

Data Types: struct

collapse all

Channel Operating Margin

The channel operating margin (COM) is a ratio between the signal and the noise  and is given by the equation:

$\text{COM}=20{\mathrm{log}}_{10}\frac{\text{Signal}}{\text{Noise}}$

optPulseMetric estimates the signal amplitude from the pulse response cursor voltage (the mean eye level height) . The noise amplitude is estimated at a given BER by the sum of intersymbol interference (ISI) voltages.

 IEEE 802.3bj 93 A.

 Common Electrical I/O (CEI) Implementation Agreement OIF-CEI-04.0, section 16.3.10.2.

SerDes Toolbox Documentation Get trial now