Created on 2012-11-08 23:33:00
Bit-accurate measurements are digital domain measurements that use test signals in which the value of every audio sample is known. In analysis, each sample that does not have the correct value is counted as an error. Measurement results may be total errors, error rate, and so on. A digital device or system that can pass such a waveform with no errors is said to be exhibiting bit-accurate performance.
APx500 generates and analyzes four signals used for bit-accurate measurements.
Bit Test (also called Bittest Random) is a pseudo-random waveform with values uniformly distributed between plus and minus full scale. An algorithms in the generator creates a stream of determinate sample values. A matching algorithm in the analyzer compares the value of every bit and reports errors.
Bit test exercises a wide range of levels and frequencies and is the most thorough of the bit-accurate waveforms.
Walking bits patterns
The walking bit waveforms are most useful when viewed in a data bits display, such as Signal Path Setup > Verify Connections > Error Rate. The walking pattern makes it easy to observe “stuck” bits in digital hardware. The walking bits waveforms will also show no errors when passed through a bit-accurate system.
In the Walking Zeros waveform, all bits are set to 1 except one bit, which is set to 0. This single low bit is continuously incremented from the least significant bit (LSB) to the most significant bit (MSB). When it reaches the MSB it wraps around to the LSB of the selected word length.
In the Walking Ones waveform, all bits are set to 0 except one bit, which is set to 1. Like Walking Zeros, this single high bit is continuously incremented from LSB to MSB and wrapped back to LSB.
The time required for one complete cycle of the walking pattern depends upon the output sample rate and the generator Samples/Step value. For example, with Samples/Step set to 48000 and sample rate set to 48 kHz, each step will last for 1 second.
The Constant Value waveform is a continuous stream of data samples at the same fixed value (digital dc). Constant Value mode aids in the investigation of data-dependent errors in digital systems.
Enter the value for the digital data here, in either hexadecimal (base 16) or D units. Specifying the value in hex is often easier, especially if the output bit depth is configured for less than 24 bits. For instance, if the output is configured for 16 bits, the lowest positive constant value is 000100 hex; this is easier to specify than 30.52 µD.
Note: Constant Value signals of all ones (1s) or all zeros (0s) are treated as invalid Constant Value signals, and will return errors. This implementation serves to flag digital connection errors, which can incorrectly produce a stream of constant ones or zeros. When using Constant Value signals, set the signal to any value except ones or zeros.
Notes on hex values in APx500 Constant Value
Truncation of hex numbers at lower bit depths
The output bit depth selected in Signal Path Setup defines the range of valid hex values for digital levels. For example, a bit depth of 24 requires 6 hex digits (1 hex digit = 4 binary digits). If the bit depth is reduced to 16, only 4 hex digits are required, so the 2 least significant hex digits are not used. If, while bit depth is set to 16, you enter 6 digit hex value (123456), APx will return the truncated value 123400.
Conversion of hex values to two’s complement
As is almost universal in digital audio systems, APx500 uses two’s complement notation for digital audio sample values. Two’s complement simplifies digital processing by expressing bipolar signals as positive-only numbers. In APx, hex values are always shown in their two’s complement form.
For example, entering ‘-123456’ with the output configuration in 24-bit mode results in the display ‘EDCBAA hex’, which is the two’s complement representation of –123456 hex.
Units available for Constant Value are