To date, many articles have appeared on the subject, from highly detailed technical analyses of the fine details of the HDMI protocol to the subjective perceptions of audiophiles who decry HDMI for sounding dull, harsh and lacking in detail. The truth is that HDMI is not very popular among audiophiles and the general conclusion is that jitter is the culprit.

Dedicated HDMI Audio Output

Many audio enthusiasts wonder how bad HDMI jitter is, and if there is a way to get rid of it. Well, it seems to be bad enough, so audio equipment manufacturers have begun implementing dedicated HDMI output ports specifically for audio, claiming that it is either jitter-free or with significantly improved jitter figures. OPPO even published a knowledge base article describing the jitter reduction circuitry in their top performer UDP-205. Let’s take a brief look at it.

Link: https://www.oppodigital.com/KnowledgeBase.aspx?KBID=129&ProdID=UDP-205

or PDF: Understanding the HDMI audio jitter reduction circuit in the OPPO UDP-205

The article explains the difficulty of HDMI audio transmission in technical terms – how the audio cues are generated from the video cues in a complicated way. Sometimes this is a task that is nearly impossible with certain video resolutions and audio sampling rates. The introductory section is very useful for understanding the basics of HDMI audio jitter problems.

A solution is then proposed – a fixed video clock capable of generating audio clocks, empty video data and a high-quality oscillator. The improvement is then presented using two methods.

The first is an eye diagram measurement of the HDMI video clock, presumably the TMDS link clock. It is claimed that the eye jitter drops from about 53ps to 50ps, without further details.

The second method uses the J-test audio signal played throughout the chain from the OPPO player to a “low-end mass market AV processor” via HDMI. The results of the second method are presented as two FFT plots of the resulting analog signal. The improvement is interpreted as a reduction of two sidebands by 15dB.

Discussion

Let’s take a closer look at the results. While the technical solution of a dedicated HDMI port for audio, by basically creating defined conditions for HDMI audio transfer within the HDMI protocol boundaries, is certainly clever and must have taken a lot of engineering effort to implement, the results and the way they are presented seem a bit problematic to say the least.

The first thing to note is that jitter as a number by itself without context doesn’t say much technically and virtually nothing about how an audio system will sound. To quantify the impact of jitter on digital-to-analog audio conversion (that’s what it’s all about), much more information needs to be available – such as the frequency bandwidth of the phase jitter or the actual spectral content, the type of jitter, etc. Moreover, the ultimate effect of jitter on perceived sonic performance depends on many other variables, such as the modulation type (PCM or DSD), the sensitivity of specific DAC chips to different types of jitter and many others.

Method #1

There is a nontrivial relationship between the TMDS clock and the regenerated audio clocks on the receiver side in terms of jitter. Presenting the advantage of the dedicated HDMI audio output using an eye diagram of the TMDS clock has very little value. Both oscilloscope plots presented look almost identical and the indicated improvement in jitter rate from 53ps to 50ps is really very small. Without further details on the test equipment used, it can be assumed that the difference is within measurement error.

Method #2

As for the second method, the famous J-test signal was specifically designed as a worst-case scenario of audio data change to quantify the sensitivity of S/PDIF & AES/EBU digital receivers and then individual combinations of receivers and DAC chips (the results are analyzed in the analog domain) to data-induced jitter. The clock recovery circuits (PLL) in digital receivers are usually the most sensitive subsystem when it comes to clock stability. The HDMI transmission protocol is very different from S/PDIF and AES/EBU, and the sense of using J-test signal in OPPO’s test scenario is questionable. The J-test signal could by its nature have some influence on the dynamic behavior of the internal circuits in the HDMI receiver and its clock recovery, but interpretation of the results should be very cautious.

The “before and after” FFT plots presented generally show a very noisy analog signal. It is very difficult to consider the lower sidebands as significantly reducing “audio jitter artifacts” in the presence of an enormously high noise floor that completely masks the spectral components of the original J-test signal (16-bit J-test assumed). If the poor appearance of the FFT plots is mainly caused by jitter, then the vast majority of the jitter energy is in the raised noise floor and not in discrete spectral sidebands originating in the structure of the J-test signal.

Conclusion

The testing methods and results provided by OPPO to demonstrate the improvements achieved with the dedicated HDMI audio output are inconclusive. The differences between the standard and the new enhanced HDMI audio transmission methods appear to be negligible, and no real advantage is to be expected in terms of audio performance.

Jitter Origin

However, there is great value in OPPO’s article, and that is the mere fact that HDMI audio cannot be improved in terms of jitter by a cleverly designed circuit in the HDMI source. HDMI is a complex protocol, and its primary purpose has always been video transmission, not audio. This is why the sound quality is so far behind the outdated S/PDIF and AES/EBU protocols designed specifically for audio.

The problem of HDMI audio jitter must be addressed on the receiver side, where the audio clocks are derived from the high-frequency TMDS clock with its own timing characteristics. This is where most of the jitter in the audio clock comes from as a result of using high-bandwidth PLLs with very low jitter rejection capabilities.

Instead of trying to clean up the jittery HDMI audio clock, the ^VanityPRO completely ignores the regenerated HDMI audio clock to begin with. Only the incoming data rate determines the nominal sampling rate. Then our two local high-quality audio oscillators and a smart control algorithm that monitors audio data buffer usage migrate the audio data from the jittery HDMI domain to the new audio clock domain with highly stable clocks.

Jitter measurement

The previous sections discussed that the test methodology OPPO engineers used to demonstrate the qualities of their jitter reduction circuit was not optimal. So, is there another way to measure HDMI audio jitter or its effect on analog audio directly? Well, there seem to be at least two or three options. The first is just one step beyond what OPPO did with their J-test method.

As mentioned earlier, the J-test signal was designed to reveal the jitter sensitivity of digital audio receivers combined with DAC chips. Had the OPPO engineers converted the received digital audio to S/PDIF or AES/EBU and used a “lower-end mass market DAC” feeding an audio analyzer, the J-test method would be much more valid. Most HDMI receiver chips have a direct 1-wire digital audio output (S/PDIF), so this step would be effortless. In this way, the entire HDMI connection to its transmitter and receiver would turn into a black box of signal transmission.

Only the accumulated effects of such a black box could be analyzed with the J-test signal via S/PDIF or AES/EBU using a traditional DAC device with “not-so-good” jitter rejection to highlight the differences in input signal quality. The only other thing to keep in mind is that the J-test results represent the combined effect of the signal quality and jitter rejection performance of the DAC device.

The second option is to measure the jitter of the S/PDIF or AES/EBU carrier wave using specialized test equipment. Most dedicated audio analyzers on the market have the ability to measure carrier jitter built in, and the results are displayed visually as a carrier eye diagram or peak-to-peak jitter amplitude.

The third option is to use an oscilloscope to analyze the jitter of the digital carrier or DA conversion (master) clock. If one has access to the internal circuitry of an HDMI receiver, it should be easy to measure the jitter of the HDMI interface directly from the regenerated audio clock provided by the HDMI receiver chip.

With the audio signal in S/PDIF or AES/EBU format, any other established method for measuring, analyzing and quantifying jitter from standard digital audio interfaces can be used. Despite the conversion of the HDMI audio signal to S/PDIF or AES/EBU format in the HDMI receiver, the signal still carries all the timing characteristics of the HDMI interface. Therefore, subsequent jitter analysis of the S/PDIF or AES/EBU signal can provide a very good estimate of the HDMI interface/audio jitter.

Vanity ^PRO Jitter performance

To demonstrate the effectiveness of the jitter attenuation/clock recovery of the ^VanityPRO, several jitter measurements were performed using different methods. The following equipment was used.

– HDMI Signal Source: OPPO UDP-203
– HDMI Audio Extractors: ^VanityPRO and Ligawo
– USB Audio Interface: Weiss INT204
– Stand-alone DAC: SMSL Sanskrit6
– TIE Jitter Analyzer: Keysight EZJIT
– Audio Analyzer: Prism Sound M1
– J-test

The first method used is with the HDMI audio converted to S/PDIF format and the J-test signal. The UDP-203 was connected to one of the HDMI audio extractors simultaneously via a 2m HDMI cable. The J-test file was played in FLAC format at 48kHz. The digital audio output of the audio extractor was connected to the SMSL DAC. The analog output of the DAC was then fed to the single-ended analog inputs of the audio analyzer.

Since Prism Sound pioneered the J-test method, the audio analyzer has a script to automatically generate the test signal as well as the results in the convenient form of the “J-Test Sidebands + Noise” figure. This takes into account not only the sidebands caused by data-induced jitter, but also broadband noise. In addition to the HDMI audio extractors, the UDP-203’s own digital audio output and a high-quality USB audio interface Weiss INT204 with J-test signal were also tested for comparison.

Ligawo

Vanity ^PRO

OPPO UDP-203

Weiss INT204

Discussion

We note again that the results discussed here are the combined performance of all elements throughout the signal chain. The J-test Sidebands+Noise figures attributed to the individual signal sources are also affected by the DAC used. Reviews of the DAC and its performance can be found online for comparison purposes.

The Ligawo Audio extractor reproduces raw HDMI audio performance without any advanced jitter attenuation. As expected, the noise floor is very high with a comb-like shape around the 12kHz audio tone. Given the other results and the raw performance of the DAC, it is safe to assume that most of the noise is due to HDMI jitter.

The FFT plots in the OPPO article do not provide information on the total analog signal level, so a direct comparison of noise floor levels is not possible, despite the fact that the unit conversion between dBu and dBV is trivial. The J-test sideband+Noise figures are -69.1dB and -69.2dB for the left and right channels, respectively.

The FFT plots for the ^VanityPRO show the noise floor at the same level as that of the DAC itself, with only a slight edge around the main 12kHz tone. The data-induced sidebands are the only sign of performance degradation due to jitter, proving that the ^VanityPRO is an exceptionally clean digital audio source. The J-test sideband+noise figures are -90.7dB and -90.4dB for the left and right channels, which is more than 21dB better than the Ligawo. If a DAC with better data-jitter rejection were used, the difference in performance would be even greater.

The OPPO UDP-203’s proprietary digital output was expected to provide a very clean signal, similar to the ^VanityPRO. Surprisingly, it also appears to be burdened by a fair amount of jitter, leading to an increased noise floor and non-harmonic outliers. The J-test sideband+Noise figures are -84.7dB and -84.1dB for the left and right channels, respectively, which is significantly worse than the ^VanityPRO.

The last device tested was Weiss Engineering’s INT204. It is a high-quality USB audio interface with fixed audio oscillators, which should produce a very stable output signal. Performance was slightly better than the OPPO UDP-203 with J-test sideband+noise figures of -86.9dB and -86.0dB for the left and right channels, respectively. The noise floor was close to the DAC’s own sound, but a fair amount of low-level non-harmonic tracks were present. These traces can be caused by noise leakage through the USB bus from an electromagnetically very noisy environment of a PC or by the USB interface itself.

Summary

Digital carrier analysis

The next method used was the digital carrier jitter analysis of the S/PDIF signal performed directly by the Prism Sound analyzer. The output of this method is the “Sampling Jitter” value in nano seconds peak to peak (ns p-p). The advantage of this method is that it is also possible to demodulate the sampling jitter and show its spectrum in the audio band. The same collection of devices was used and the S/PDIF signal was connected directly to the Prism Sound analyzer. The measurements were made with a J-test signal to create a worst-case data transfer scenario.

Ligawo

Vanity ^PRO

OPPO UDP-203

Weiss INT204

Discussion

The Ligawo HDMI extractor showed 28.3ns p-p sampling jitter and jitter spectrum with a high level of uniform noise and a discrete peak at 20kHz.

The ^VanityPRO had 1.7ns p-p sampling jitter and jitter spectrum with a much lower noise floor with some low-frequency discrete components. These components were present at the same level even when no audio signal was transmitted. Their presence in measurements of other devices may indicate a test method limitation or a system-related cause.

The OPPO UDP-203 showed 6.0ns p-p sampling jitter and a slightly increased noise floor with a small bump around 12kHz, possibly due to the J-test signal.

Finally, the Weiss INT204 had 8.0ns p-p sampling jitter with a demodulated jitter noise floor similar to that of the Vanity ^PRO , but with a higher level of low-frequency components.

Once again, the Vanity ^PRO showed the lowest amplitude of sampling jitter in the digital carrier analysis performed by the Prism Sound audio analyzer.

Summary

Master-Clock TIE Analysis

The last method used to determine the jitter levels between a standard HDMI receiver and the Vanity ^PRO is direct audio master-clock analysis with a specialized instrument from Keysight. In this case, the audio master clock provided by the HDMI receiver chip is analyzed for TIE (Time Interval Error) and compared to the TIE of the Vanity’s audio master clock ^PRO . This comparison objectively quantifies the HDMI audio jitter and audio clock quality improvement provided by the Vanity ^PRO achieved.

TIE is an accumulated edge timing error of the real clock signal relative to an ideal clock representation of the same frequency. The instantaneous measurements are accumulated and displayed as a histogram with statistically determined parameters. The definition of TIE and standard measurement methods are online to be found.

HDMI Receiver

Vanity ^PRO

Discussion

The values to look for in TIE measurements are standard deviation and peak-to-peak (or min and max). The standard deviation represents an RMS value. One thing to note is that both histograms show a standard Gaussian distribution. The HDMI receiver gives the TIE standard deviation (RMS) 471ps while the Vanity ^PRO has only 2.9ps. This is a ratio of approximately 162x or 44dB in favor of the Vanity ^PRO .

Another way to visualize the TIE measurement is with oscilloscope display persistence and delayed horizontal line. Edge instability from the trigger to the observation point accumulates over any number of clock periods and causes a color graded thickening of the displayed waveform. This visualization method is used in the Vanity ^PRO online presentation of the jitter attenuation functionality.

HDMI Receiver versus Vanity ^PRO

Conclusion

The purpose of this paper was to address the topic of HDMI Audio jitter in the context of quantifying jitter levels using various methods and demonstrating the jitter reduction capabilities of the Vanity ^PRO HDMI audio extractor. The results of direct and indirect methods of jitter measurement presented in this article prove that the Vanity ^PRO provides unmatched performance in eliminating jitter from the HDMI interface. This offers the audio enthusiast a unique opportunity to use any compatible HDMI audio source as a top-class digital multi-channel audio transport with bit-accurate digital audio transmission and virtually zero jitter for the most accurate sound reproduction.

This is a loose copy of the original English Audiopraise article,“HDMI Audio Jitter: Is there a way out?“