Encoder Homing Frame (Speech Codec Test Signal)

Description

The encoder homing frame is a specialized test signal defined in 3GPP speech codec specifications, primarily used for verifying the correct operation of speech encoders and decoders in mobile communication systems. It is defined across several specifications including TS 26.090 (Adaptive Multi-Rate speech codec), TS 26.190 (Adaptive Multi-Rate Wideband speech codec), TS 46.020 (Enhanced Full Rate speech codec), and TS 46.060 (Enhanced Full Rate speech codec comfort noise aspects). The frame consists of 160 identical samples, each 13 bits long, with the least significant bit set to 'one' and all other bits set to 'zero'. When these samples are written to 16-bit words with left justification, each sample has a hexadecimal value of 0008.

The concept of a homing frame is rooted in the need for deterministic, reproducible testing of speech codecs. Speech codecs are complex digital signal processing algorithms that compress and decompress voice signals. To verify that an implementation of a codec conforms to the 3GPP specification, test sequences are used where the input signal and expected output are precisely defined. The encoder homing frame serves as a known input that, when processed by a correctly implemented encoder, produces a specific, predictable output bitstream. This allows test engineers to verify bit-exact compliance with the standard.

The homing frame gets its name from its ability to 'home' or reset the codec's internal state to a known condition. When the encoder processes this specific input pattern, the internal state variables of the codec (such as adaptive codebook contents, filter memories, and gain parameters) converge to a deterministic state. This is particularly important because many speech codecs are stateful — their output depends not only on the current input frame but also on previous frames. The homing frame provides a way to bring the codec to a known state before beginning a test sequence, ensuring reproducible results.

The test sequence SEQ05.INP described in GSM 06.07 defines the encoder homing frame in the context of the GSM Enhanced Full Rate (EFR) codec. This sequence specifies the exact input samples that constitute the homing frame and how they should be applied to the encoder. The specification also defines the expected behavior of the decoder: when at least two decoder homing frames are input consecutively, the speech decoder must produce the encoder homing frame as a response to the second and any further decoder homing frame. This creates a round-trip verification: encoder homing frame → encoded bitstream → decoder homing frame → decoded output should equal the original encoder homing frame.

The 13-bit sample format with value 0008 hex is significant because it represents a very small amplitude signal — the least significant bit of a 13-bit sample. This low-amplitude, DC-like signal is chosen because it exercises the codec's handling of near-silence and low-level signals, which are important operating conditions for speech codecs. The fact that all 160 samples are identical means the signal is essentially a DC offset, which produces predictable behavior in the codec's prediction filters and gain adaptation mechanisms.

In practical codec testing, the encoder homing frame is used as part of a broader test suite that includes various input signals designed to exercise different aspects of the codec. These test sequences are defined in the relevant specifications and are used by manufacturers to verify their implementations before submitting equipment for type approval. The homing frame is typically used at the beginning of a test sequence to establish a known starting state, and may also be used at the end to verify that the codec returns to a known state after processing complex signals.

Purpose & Motivation

The encoder homing frame exists to solve a fundamental problem in speech codec testing: how to verify that a codec implementation is bit-exact with the 3GPP specification when the codec is a stateful, complex signal processing algorithm. Speech codecs used in mobile communications must produce identical output regardless of the manufacturer, because the encoded bitstream must be decodable by any compliant decoder in the network or in another device. Without a standardized test signal that brings the codec to a known state, it would be impossible to create reproducible test sequences that all manufacturers could use to verify compliance.

Before the introduction of homing frames, codec testing relied on comparing implementations against reference software using arbitrary speech samples. However, because codecs are stateful, the output for any given frame depends on the history of previous frames. This made it difficult to isolate and diagnose discrepancies between implementations. The homing frame concept, borrowed from earlier speech coding standards and formalized in 3GPP specifications, provided a deterministic reset mechanism. By defining a specific input that drives the codec to a known state, it became possible to create test sequences where any deviation from the expected output could be unambiguously attributed to an implementation error.

The specific choice of 160 samples of value 0008 hex reflects the frame sizes and sample formats used in 3GPP speech codecs. The 160-sample frame corresponds to 20 milliseconds of speech at 8 kHz sampling rate, which is the standard frame length for narrowband codecs like AMR and EFR. The 13-bit sample format with the LSB set reflects the internal precision of these codecs, which use 13-bit linear PCM as their input format. The value 0008 hex (which is 8 in decimal, or the smallest non-zero positive value in 13-bit representation) was chosen because it produces predictable, stable behavior in the codec's adaptive algorithms without triggering any special cases or saturation conditions.

The requirement that the decoder produce the encoder homing frame in response to consecutive decoder homing frames serves as a powerful verification tool. It tests not only the decoder's correctness but also the fundamental symmetry of the codec design. If an implementation fails this test, it indicates a significant error in the codec's signal processing chain. This round-trip property has been maintained across all generations of 3GPP speech codecs, from GSM through LTE and into 5G, demonstrating the enduring value of this testing approach.