Interpreting a 2-D Error Map

This optional analysis creates a two-dimensional map of the positions of your errors on the display and enables you to view, pan, and zoom about your accumulated error data in an easy and convenient way. Detected errors are first analyzed to see if they comprise burst errors or non-burst errors. The definition of a burst depends on the current Global Burst Criteria setting. Once this is determined, a color pixel is illuminated on the screen (the color depends on the type). Which pixel to light up is calculated by thinking of your data stream as a collection of sequential data blocks. These data blocks, or segments, are received in a serial fashion and “stacked up” next to each other when building a 2-D Error Map. Then, a raster-scan type image is built by placing the segments next to each other, from left to right on the display.

The size of the data block to use when creating the 2-D Error Map is up to you. It is common to have the data block size be the same as that used during Correlation analysis. Like Correlation analysis, it is also possible to define the boundaries of data blocks using the external Marker input.

You can also define the starting point of your error map using the Synchronize Analysis with Marker feature. If this is selected, errors received by the analyzer before the first Marker signal will not be mapped. Once the first Marker is seen, error mapping begins.

When error mapping begins, errors in the first bits of the accumulation period are mapped into the lower left hand corner of the display. As more bits come into the analyzer, the error map moves its way up the first column until the block boundary is found. Once the block boundary is found, the error mapping moves over one column to the right and starts again at the bottom of the display.

You may also display the positions of Marker signals. This works fine even if markers are not used to define block boundaries, and can be very helpful when trying to correlate rare errors to an external event of some kind. If this external event can be translated to a TTL-level marker input, then you can include the position of these on your 2-D Error Map. In this case, for example, you might see that large bursts were always preceded by your external marker event.

Non-burst and burst errors are displayed in different colors. It is common to have random background non-burst error phenomena simultaneously with a burst error phenomenon. By coloring them differently you can analyze them separately. In fact, by using the Global Error Removal filter, you can qualify the types of errors you want to include in your 2-D Error Map and create maps that have only burst errors, or only non-burst errors. Again, the Global Burst Criteria are used to define a burst error.

The location of any Squelch events can also be viewed on the error map. Squelched intervals are areas of the data stream that cannot be analyzed because the error event rate preceding the squelched time was so high as to cause the BitAlyzer to stop collecting error data momentarily, while it processes what it has. Errors and bits from squelched intervals are not included in the analysis of error statistics.

Panning and zooming about your 2-D Error Map allows you to quickly view the results of an extended experiment or zoom in to see the details of how a specific burst error occurred. Panning and zooming are done, as in all the charts, by selecting the type of operation from the rotator knob pull-down list and then either dragging the display or turning the knob. Cursors can also be enabled on the 2-D Error Map. Cursors can be used to find the segment number and position offset size for any position in the map.

At zoomed-out scales where many, many blocks and data bits are represented on the display, the 2-D Error Map analysis combines neighboring errors to determine if a given pixel is turned on or not. This is done both in the x-axis and y-axis. In this way, for example, it is possible for any one pixel on the display to really represent thousands of underlying data bits at any one time. When this occurs, burst errors take priority when deciding what color to select for that element. If there is one or more burst errors inside the underlying data bits, that element will be colored as an element with burst errors present. An element colored as a non-burst error type has no burst errors in any of the underlying data bits it represents.

Here are some examples that help demonstrate how the 2-D Error Map tool and its panning and zooming features are helpful when trying to understand your errors.

Because the human eye is a good correlator, the BitAlyzer optionally includes the 2-D Error Map feature that displays a two-dimensional map showing the geometry of error locations in a digital channel. This geometry can be due to physical mechanisms such as a scanning recording head, a rotating medium such as a hard disk, or the packetizing and blocking of data in a channel.

In the 2-D Error Map, the vertical axis (Bit/Symbol Offset) shows the location of errors in each segment, while the horizontal axis (Segment) shows the progressive segment number. Another way to visualize this is as a scanning of the input data with the vertical dimension from the bottom to the top of the graph representing the segment size. Each subsequent segment is displayed in the next position along the horizontal axis. Errors that occur in the same place in the segment will be seen as a horizontal line along the graph.

There are many ways this visual correlation can give information that is very difficult to obtain in other ways.

In the following example, analysis was done on playback from a digital rotary head tape recorder. Each head pass recorded 34,848 bits so the data block size was set to that value.

At approximately bit position 24,000 across the block, there is an error that repeats and which is seen as a horizontal line on the graph. This line of errors is actually a scratch in the tape that increased the error rate in the immediate vicinity of the scratch. In other applications, a line or band of horizontal errors can indicate a pattern sensitivity that occurs in the same place in the block, every block.

A specific error pattern has been circled in this picture. The graph can be zoomed and panned to examine single bit errors. In the following picture, the graph has been zoomed to see the error pattern in much greater detail—in fact, the detail is at bit error level.

This picture shows the individual errors as well as the bursts that occurred at this position on the tape. Cursor A shows that it is at Block 372,754 along the horizontal axis. Cursor B shows that it is at bit 21,063 within Block 372,754. This error group was probably due to a cosmetic defect in the tape’s surface, because it occupied multiple head passes.

While this example showed the analysis of a tape recorder, there are many applications where the ability to visualize the error pattern provides information that might be otherwise impossible to obtain.