Saturday, 26 October 2013

MEASUREMENTS: ONKYO TX-NR1009 as HDMI / SPDIF DAC... Are AV Receivers any good?

The rat's nest.

Running separate components like multichannel processors/preamps to monoblock amplifiers are generally considered the ideal, "cost no object" approach to home theater. In the real world, cost and space are considerations and AV receivers become the "Jack-of-all trades" central device that most of us have in the home theater setup. But like the proverbial "Jack", it's useful to also consider the second part of that saying... Just how bad  is he also "the master of none"?

In the last installment, we looked at passing an analogue audio signal through the Onkyo and found that noise can be an issue. Today, I want to demonstrate the quality if we were to just use this device as a DAC - a look at the digital portion. Some natural questions arise - how well did the designers shield noise from getting in (especially in light of the high analogue noise measured previously)? Is the jitter through the use of HDMI "bad" (compared to TosLink and coaxial S/PDIF)? How does it compare to other stereo DACs?

Based on the Onkyo specs sheet, the TX-NR1009 uses the TI PCM1690 6-channel + PCM1789 2-channel DAC chips. Both are rated as 113dB SNR. These DAC chips are often found in consumer AV receivers and are lower spec'ed than most stand-alone DACs like the TEAC UD-501's PCM1795 with >120dB SNR. Of course, you cannot just look at this specification and judge the quality of a DAC. Much depends on the analogue circuitry around that DAC so the measured results are more useful than just looking at the components individually.

Setup:

As usual, for the sake of full disclosure and opportunity to repeat/verify, here is the setup for these measurements:

Win8 AMD A10 "Trinity" HTPC --> HDMI/TosLink/coaxial cable --> ONKYO TN-NR1009 front stereo "pre-out" --> shielded RCA --> E-MU 0404USB --> shielded USB --> Win8 laptop
CM6631A device used for asynchronous USB --> coaxial / TosLink conversion duties.

HDMI driver: default AMD WASAPI. I used JRiver for playback.

Since I want to check the performance in a more "naturalistic" fashion, I made sure the TV was connected and on as well as my Blu-ray player (Panasonic BMP-TD220). Remember that in my previous post, plugging in the HDMI TV cable added significant noise to the analogue pass-through. All results were made with the Onkyo in "Pure Audio" mode to defeat any audio DSP/bass management.

HDMI cable: A decent looking 6' length capable of high speed HDMI (officially rated as HDMI 1.3 but fine with my HDMI 1.4 3D TV), brand named "ION" that I purchased for something like $20 about 2 years back at a local computer/supplies store.


TosLink and coaxial SPDIF cables I'll be using for comparison are the "Acoustic Research" branded 6' lengths I measured previously (see links for details).


Results:

As usual, I ran the output through my digital oscilloscope first to have a look:
Here is a 0dBFS 1kHz square wave sent through the HDMI and measured off the front stereo "pre-out" RCAs. Not bad - good square waveforms with excellent channel balance (sorry about the pixellation, usually screenshot looks better than that). With the receiver volume set to the "reference" of 85 (there is a little popup on the front screen when you hit 85 that correlates to the 0dB THX reference level) and in "Pure Audio" mode, the peak voltage is around 2.3V.

Here's a 24/44 impulse response:
Good linear phase impulse response, nothing fancy here. Absolute polarity also maintained by the Onkyo.

I. RightMark:
As usual, I used RightMark to look at the measured dynamic range, noise level, distortion levels; here's the summary for HDMI at the various bit depths and sample rates:

As you can see, I've also included for comparison the results at 24/96 for the Squeezebox Transporter and TEAC UD-501 (unbalanced outputs) - two of the best measuring DACs I've tested here with the same hardware/software.

Clearly the Onkyo is capable of hi-res with >16-bit dynamic range. With 24-bit data, it can do ~109dB dynamic range which equates to just over 18-bits! Not as good as the dedicated audio units like the Transporter or TEAC but pretty darn good for an AV receiver! This result is about equivalent to the AUNE X1 and ASUS Essence One using unbalanced RCA output - however, those DACs had better distortion numbers.

Some graphs to review from the 24/96 dataset:
Frequency Response
Noise Level

THD
The Onkyo rolls off a bit more in the high end, a little more noisy, and notably more harmonic distortion.

For fun, here's the spectrum off the Onkyo playing 24/192:
Yup, capable of 24/192 although the roll-off on the high end is obviously earlier than the TEAC UD-501 (Onkyo drops -3dB at 50kHz).

I was curious if the SPDIF (TosLink and coaxial) inputs measured just as well:

Yup. They all look pretty good. The graphs all look identical except for slightly more high end roll-off with the HDMI interface compared to the SPDIFs - not sure why.



With the dynamic range >16-bits, this test should be no problem for the Onkyo (HDMI input).

That looks very nice given that many very expensive DACs are not even capable of this degree of resolution! Again, this is an AV receiver! As I previously posted, even recently released DACs like the Wadia 121 Decoding Computer is incapable of this resolution.

III. J-Test for Jitter
As usual for my DAC tests, let's have a look at the Dunn J-Test spectra for both 16-bit and 24-bit signals. Here is the summary using the 3 digital inputs - HDMI, coaxial SPDIF, and TosLink SPDIF:



As you can see, the Onkyo is quite jittery in general whether HDMI or the other SPDIF interfaces. Although quite similar, I am somewhat surprised that the sidebands were more pronounced for the coaxial digital input! For comparison, here's the Transporter and TEAC:


Although the scale and dimensions are a little different, one can certainly appreciate just how jittery the Onkyo is compared to the others especially with the 24-bit signal. From this data, we see that the Onkyo itself has more jitter as a whole; specifically it's not any worse with the HDMI interface. We'll talk about jitter again in a little bit...

IV. Does sending a 5.1 channel signal degrade the measured performance?
I thought this would be interesting to check out. I left the RightMark test signal as the two front channels and added some AC/DC "Thunderstruck" into the center, rear, and LFE channels played back in JRiver as a multichannel FLAC through HDMI.


Beautiful ain't it?! The idea is to see if driving 6 channels (5.1) at the same time through the HDMI cable into the Onkyo's DAC will change the audio quality... For example, doing this might increase the noise floor, or perhaps worsen channel crosstalk since we've tripled the number of audio channels being processed.


Frequency Response

Noise Level
As you can see, there's very little difference whether 2 channels are playing or 6 channels. Great to see! Essentially no frequency response or crosstalk difference. However, there is a very small increase in noise level when playing multichannel... IMO audibly insignificant but measurable.

Summary:

Here you go folks! That's how a higher-medium end "modern" AV receiver measures as a stereo DAC. Of course, each model will be a bit different, but I suspect similar tiered receivers from Pioneer, Denon, Integra, Yamaha, H/K, Anthem, etc... should be comparable (won't know unless someone tests it out). Note that most magazines like Sound & Vision will measure receivers but usually in the context of power output and flatness of frequency response rather than on the accuracy of the digital-to-analogue conversion as I did here.

In some ways I am impressed and in other ways the results were as expected.

I was impressed by the low noise and very good dynamic range for example. To achieve almost 110dB in the audible spectrum is quite something especially considering the complexity of an AV receiver with all the potential electrical noise sources inside the box! The accuracy of the 16-bit -90.3dB waveform looks excellent; something which only the better stereo DACs or CD players would have been able to accurately reproduce a decade back. Likewise, the fact that the measurements remained excellent even with 6 channels being processed concurrently and only measuring about 1dB difference in the noise floor again demonstrates the engineering quality. Given the results I found previously with HDMI noise polluting the analogue input, I'm guessing that Onkyo put more attention in optimizing the digital side which makes sense since most people will be connecting digital inputs for multichannel sound.

As for the more "expected" results, let's talk about jitter...

A few years ago in 2009 this message came over the 'Net which I remember made quite an impression on me around how "bad" HDMI is as an audio interface (supposedly from Hi-Fi News & Record Review / Miller Research):
(I didn't notice it at the time, but that Denon AVR-3803A was a typo - the 3803 has no HDMI. It's actually the 3808.)

Later, a more comprehensive message appeared:

Hmmm, it looks like HDMI jitter can be cleaned out after all (eg. Arcam, Classe, Pioneer)! It's about the implementation, not necessarily the interface itself. If you read around these posts, one also finds that the jitter value and subjective sound quality do not necessarily correlate.

Let's think about the J-Test and what was found in measuring the Onkyo for a moment. The Dunn J-Test is a synthetic test of data jitter first published by the late Julian Dunn around 1994 which (in the 24-bit 48kHz version) superimposes an undithered LSB 250Hz square wave over a primary 12kHz -3.01dBFS sine wave which is of course an exact 1/4 of the sampling rate. This superimposition stimulates the effect of subtle timing inaccuracies (jitter) which can be demonstrated as accentuation of the sidebands measured in the spectral graphs.

Remember that this test is synthetic and stimulative. What you see measured is not something you're probably ever going to "hear" in real music! The noise floor is not going to be down to the last bit in 16-bit audio and essentially impossible with recorded 24-bit audio (unless it's purely computer synthesized music). Also, jitter is more pronounced in the higher frequencies (11kHz and 12kHz are used as the primary signals in the J-Test). Realize that the human hearing sensitivity is well on its way down by 5kHz (as can be seen by the Fletcher-Munson Curves). Furthermore, if we specifically look at the Onkyo's J-test spectrum, the most pronounced side bands are about -90dB below the primary signal. To make matter even less worrisome is that the tall sidebands are all +/-250Hz around the primary signal and the audibility would be masked even if one did have awesome auditory acuity at 11/12kHz and could hear a signal 90dB down! This is also why I feel adding up all those sideband peaks and calling it a number (whatever picosecond or nanosecond) is really not all that useful when it comes to audibility.

What I'm trying to say is this... Tests like the J-test can demonstrate that jitter is a real phenomenon. Engineers should pay attention to it when designing hi-fi equipment. A discerning audiophile should be aware of it and if able to, can measure it themselves and decide if the engineer did a good enough job. However, IMO, to say that jitter is somehow audible at these kinds of levels I think would be impossible. In fact, unless the jitter were ridiculously high (like Track 26 from Stereophile Test CD 2, where an insanely simulated 10ns sideband is inserted +/-4KHz around the 11kHz primary - again totally synthetic), the concept of jitter significantly deteriorating sound quality I believe is utter nonsense in the real world. That some companies would even consider using jitter as a reason for putative significant audible differences between passive "components" like cables is just not credible!

I had a listen to the Onkyo's output over the last few nights with some familiar music - Ella Fitzgerald "Sings The George & Ira Gershwin Song Book", Grateful Dead "American Beauty", and Keb' Mo' "Just Like You". Also had a listen to Sting's new album "The Last Ship". They all sounded nicely rendered as they should with a good DAC. Great details with my older Paradigm Reference Studio 80 v2's which will be my rear speakers in the new sound room. The wife and I both enjoyed Sting's "The Night The Pugilist Learned How To Dance" - cute.

So, even though the AV receiver might be the "Jack-of-all-trades", at least in this specific instance with the Onkyo TX-NR1009 as an HDMI DAC, he might not be a "master" at it, but I'd say he's a pretty decent tradesman :-).

Well, unless I dig up something else to report, I'm likely "going dark" for the next month as I head off overseas for some work and then the big move to the new home. I'll be sure to post some pictures in time... Enjoy the music everyone!

7 comments:

  1. Thanks for sharing! I love my Onkyo TX-NR809. I fancy I can hear a *slight* difference running direct via HDMI (i.e., using the Onkyo internal DAC, which is capable up to 24/192) and running analog in through a low-end external DAC (e.g., Dragonfly or Meridian Explorer), but no way I'd pass a DBT on it ;)

    As you say, pretty good performance from these AV receivers these days.

    ReplyDelete
    Replies
    1. Yes. The result is excellent. Your NR809 uses the same main DAC chip (PCM1690) with essentially the same feature set and topology so it should sound essentially identical.

      Enjoy!

      Delete
  2. Thanks again.
    Since you also ran these tests with the TV connected via HDMI I can't understand why the noise that was present in the analog pass-through tests isn't visible here. My assumption was that the signal path in the HDMI test would be near-identical with the exception of the addition of a DAC stage and I'm still not sure why this isn't the case. Can anyone educate me?
    I've long been skeptical about differences between competent power amps, but are there any objective tests of AVR power-amps vs. hi-fi integrated stereo amps or monoblocks?

    ReplyDelete
    Replies
    1. I believe the signal path must be different and better shielded from noise coming out of the DAC than the full analogue pathway. Perhaps a way to check this out would be to look at the noise level difference between the front RCA input vs. the rear.

      Since analogue isn't as important as digital input for me, I'm just glad the HDMI input seems clean.

      I guess one could compare the results from monoblocks vs. integrated amps in Stereophile to have a look at the difference. I'd love to get some amp measurements done.

      If anyone has tips about measuring amps as a hobbyist at home without buying an Audio Precision unit, please let me know :-).

      Delete
  3. Good to know that the Onkyo hold up as a good DAC & Pre Amp in "Pure Audio" mode ... but bad things happened when you turned on the video processing in your last post! Were you able to solve the ground loops issues? My NR901 is an older model but has a two prong power cord (hot, neutral) - perhaps to eliminate ground loop issues.

    Thanks for another informative write up. Since I have an older NR901 Onkyo I have since purchased a MF VDAC-II and Emotiva monoblocks (XPA-1s) . It was time for a change and my faithful AVR has been retired to pre-amp duties in pure audio mode of course :)

    Needless to say I am very happy with the new gear on good 16 and 24 bit material.

    All the best!

    ReplyDelete
    Replies
    1. Hi Leonid,
      Haven't had time to investigate the ground loop issue further; at least until I move in to the new house in about a month or so. Ultimately I think it won't matter since I'm going to route all 2-channel audio through the Emotiva XSP-1 preamp rather than the receiver. So long as the HDMI DAC piece is good, I'm happy!

      I actually have 2 Emotiva XPA-1L's waiting for me across the Canada-US border to pick up :-). I was waiting for the arrival of the XPA-1 Gen 2's (supposed 60W Class A bias) but at the end of the day, figured it wouldn't matter for my rather efficient 92dB speakers (the Paradigm Signature S8's) - might as well save some cash with the current sale price and power bills if I decide to run them in Class A bias mode for long :-).

      Delete