This is a second review of the Denon AVR-X4700 AVR. It was kindly purchased by a member and drop shipped to me. It costs US $1,700 from Amazon including free shipping. I will explain the reason for the revision in a bit. For now, I will cut and paste some of the information from
the first review for completion.
The industrial design of the Denon AVR-X4700 is similar to many other AVRs:
We have the typical back connections:
In use, the unit was robust, never shutting down in 2-channel testing. However, the right side of the unit where I think the front left and right channels exist, got very hot. It was so hot that it was uncomfortable to touch the top of the case even though it is 1 inch or more removed from the heatsink. There is a fan under them but it did not come on. I worry about long term reliability of this unit so I suggest assisted cooling if you are going to use it anywhere near full power.
For testing, I turned off the eco mode and operated the unit with just two large front right and left speakers. For some of the tests, I selected the 11 channel mode which allowed me to reassign them to "external amp" which provides better performance in that configuration. Alas, there is no way to do the same for the center channel which is a miss. In this regard, there is no improvement over the last generation AVRs.
Reason for Second Review
The original review measurements showed significant degradation of AVR's performance when measured using HDMI input as opposed to Toslink input. This was surprising to me as I had not seen much discrepancy in the past reviews (sans jitter and such). This resulted in some harsh comments from me regarding the unit and vote of non-confidence.
Subsequent to the review, Denon Engineering contacted me and shared measurements with me which did not at all show the problem I was seeing. They ran the same measurements on the Denon AVR-X3600 (last generation model) that showed similar or same performance. Their work was first class and I could not find any fault with their methodology. I think I can speak for them that they could not see or explain why I would get different results.
The one significant difference in our setup was that they used an APx585 analyzer which has built-in HDMI output and I was using my desktop workstation as video output. So we chased many paths to try to identify why we were seeing such drastically different results. Many hours were spent in this analysis followed by late night conference calls to sort through this. Right when we thought maybe the explanation is too hard to find, we had a breakthrough. I realized that in my testing I was setting the AVR for 2-channel configuration with Front right and left speakers set to Large, and all other channels configured as "None." The output path form my PC however was 8 channels (determined by Intel GPU HDMI implementation in my Intel CPU). I had turned off all channels beside left and right assuming that would simulate simple 2 channel playback. Well, turned out this was the problem!
AVRs have channel mapping to deal with configuration differences between input audio streams and playback capability. You may for example use the AVR with just two speakers for living room sound while playing 7.1 channel Blu-ray content. AVR will then use its internal mapping to mix down the high input channel count to stereo. For reasons that are unknown at this point but is being investigated by Denon engineering, when feeding only two channels to AVR but in 8 channel configuration, and the AVR is configured as just Left and Right speakers, the noise level goes up substantially and there is potential for clipping on maximum level signal. The former was responsible for lower performance across many measurements. The latter was responsible for unusually high distortion in multitone test.
Fortunately the fix was simple. I simply turned on all the speakers for 7.1 configuration while still continuing to feed the AVR the same way I was before. With all channels configured, the stereo audio data was no longer changed since no mapping was required and performance shot way up. Since vast majority of you are using AVRs with multiple channels enabled anyway, the issue that caused the problem should not be something you see.
Note that the amplifier measurements were not impacted so we did not focus on them. I will just be cutting and pasting them from the original review to this thread.
These new measurements fully supersede the data in the original review. So please do not refer to the other set anymore.
AVR DAC Measurements
The heart of the audio subsystem in an AVR or processor is conversion of digital audio samples to analog. So we always start there by tapping the "pre-out" from the back of the unit instead of using speakers (to eliminate the effect of amplifiers). Let's start with HDMI input and leaving the unit as shipped (i.e. front amps connected):
[this is the old measurement -- I did not re-run it because it is limited by the amp being on as opposed to configuration issue above]
We see the typical high harmonic distortion due to internal amplifier stressing the power supply causing the DAC to underperform. Fortunately in this amplifier we can turn them off for the fronts as mentioned and this gets us an improvement: [new measurement]
View attachment 71648
The results are far better than in the original review and now actually outperforms the Coax input:
View attachment 71647
Using the more common HDMI result gives this ranking for AVR-X4700H:
View attachment 71649
And among AVRs/processors tested so far:
View attachment 71650
We can see the much improved dynamic range using HDMI:
View attachment 71651
On the right I am showing Denon Engineering results with kind permission from them. As you see, they are getting a few dB better result. This may be due to unit to unit variation or some other factor not yet investigated. I am not worried about it though as it is in the same ballpark.
We now have better than 18 bits of dynamic range which puts my mind at ease with respect to clearing the 16 bit hurdle of CD format.
The IMD performance is also much improved since the sloping down portion of this graph is dominated by noise:
View attachment 71652
Denon Engineering had better results here but due to me not specifying the tone configuration for this test, they used a different one as noted. Regardless, we are back to the type of performance I expect to see in good AVRs.
Noise floor in J-test is now the same for both inputs which is good:
View attachment 71654
Even though I use a lot of filtering, noise can still impact linearity test. Here are the much improved results:
View attachment 71655
This is extremely good, showing accuracy to almost 20 bits!
Now we get to the key test of multitone:
View attachment 71656
Gone are the high intermodulation spikes caused by overflow condition previously. And we once again land on better than CD distortion-free range which is good.
I noted in the original review that the DAC reconstruction filter attenuation was not as high as I wanted. Fortunately there is improvement here as well as noise was hiding the true response of the filter:
View attachment 71657
Likewise the original THD+N vs frequency test showed high level of distortion+noise+spurious tones which is now much improved:
View attachment 71658
In discussions with Denon Engineering, they made a good comment that this test pulls in filter artifacts due to its wide bandwidth yet it is labeled "THD+N." So I changed the title of the test to make it more clear that it is measuring everything that is not the one frequency under test. I also made another run at 192 kHz sampling so that the spurious images due to the filter get pushed out of band. This is reflected in the graph in green. I plan to run this test this way from now on.
With the AVR-X4700 (and all of 2020 series Denon AVRs) you can only turn off the amp for the front channels, or all the channels. If you want some other configuration, then you are dealing with amplifier clipping and lowering performance depending on how much you turn up the volume. To show this I measure at different output levels to identify the best performance:
View attachment 71659
As you see, the best performance is around 1.1 volts which produces over 100 dB SINAD (dashboard uses 2 volt output so gets lower SINAD). With the amp turned on, the highest you can go is 1.4 volts before clipping occurs and performance drops precipitously. So make sure to look up the specifications for your external amplifier to see what its "sensitivity" is that generates its maximum power. If it is below 1.4 volts, then you are good.
Streaming DAC Performance
I used the Heos App to stream our 1 kHz tone. Here is the dashboard:
View attachment 71834
This is the same performance as Coax input so the hardware pipeline must be the same.
I also tested using Airplay streaming via Roon player. That produced 92 dB SINAD due to truncation to 16 bit (limitation of Airplay).
EDIT: here are the output impedance measurements and performance into 600 ohm:
It is fairly high output impedance. So don't load it down below 12 k Ohm.
When I tried it with severely low 600 ohm load, it naturally dropped a lot of output level but distortion remained good:
-------------------------------------------------------------------------------------------------------------------------------------------------------------------
[The rest of these measurements are cut and paste from the original review]
Analog Pre-amp Measurements
I always test to see if analog input works better or as good as digital in AVR as that allows me to then compare the amplifier in them to output amps. Here is that performance:
We see that performance is a couple of dBs better than Toslink indicating the ultimate limit of the analog buffer/volume control/gain stage. Not a great news but does allow us to run the amplifier tests with analog input, telling us its best case performance.
AVR Amplifier Measurements
As usual, we start with our dashboard of 1 kHz tone, 5 watts output into 4 ohm load:
his is actually above average for an amplifier:
And among AVRs:
Notice that it is sitting very close to the 3600H indicating the same design.
Frequency response is identical in all CD modes which surprised me:
Maybe it never digitizes this input? Anyway, it is good to see such wide bandwidth and dead flat response in audible band.
Crosstalk is not that great but fortunately that is not a huge audible thing:
Dynamic range is nearly as good as 3600H:
We have a bit more power than 3600H:
And into 8 ohm:
Finally testing to see frequency dependency on power and distortion+noise:
Conclusions
It goes without saying that new measurements completely changes my outlook on AVR-X4700H. It now ranks in the upper tier of home theater AVRs. Not only that, the company behind it is excellent to work with. Despite my harsh original review, they came to the table ready to work through this with no ill will. I am especially appreciative of Rainer Finck of Denon/Marantz Europe who spent many hours with me on the phone and in email to brainstorm. He was also the conduit to Denon Engineering in Japan which worked hard to chase any theory we came up with. It was also a pleasure to work with Sound United personnel in US in final resolution and planning of this update.
This was one of the best experiences I have had working with a company on a review!
Anyway,
I am happy to now recommend the AVR-X4700H. I will take all the eggs you want to throw at me for not doing so the last go around.
-----------
As always, questions, comments, recommendations, etc. are welcome.
