Sony TA-ZH1ES: A DAC/Headphone Amp with a Sonic Signature?
View attachment 265148
Stop. やめる. Halt. Arrêt.
You probably shouldn’t read the rest of this post, it's triggering. There's a lot of speculation even though there are a good amount of measurements too.
If you're looking for a transparent DAC/headphone amp
or a great pure headphone amp, I recommend that you get something else
This post is really for those who are curious. You see, there exists a DAC/headphone amp which has commonly been described as not neutral but still sounds great
: The Sony TA-ZH1ES.
The Sony TA-ZH1ES is an almost 7-year-old product that is still in current production/current US availability in 2023. There was a production hiatus during COVID and it was assumed that Sony discontinued the product completely, and it has indeed officially been discontinued in Japan, but it's still readily available from US retailers after a 2 year hiatus of being out-of-stock. Nonetheless, it's pretty clear that the TA-ZH1ES is on borrowed time at this point.
If you want one, I would strongly recommend buying one while you still have the chance and buying the unit from a retailer with an in-home trial.
Impressions from the Internet
"The Sony’s honeyed rendering went too far
, mellowing out the sound balances of familiar recordings
(albeit in a very agreeable way)."
- Sound & Vision
, Steve Guttenberg (a.k.a. The Audiophiliac)
"...sound is quiet as death and joyously alive
"...headphone amps must be ultimately judged on the subjective quality of their transparency
—not just on the relative quantity
of transparency they effect."
Stereophile, Class A Product
The high frequencies are foggy and the reproduction of the high frequencies is ambiguous.
- 2/5 stars; user review posted at Kakaku.com
It's like the final form of SONY's house sound, so I don't think you should buy it if you don't like the SONY sound very much.
- 5/5 stars, user review posted at Kakaku.com
"I find the amp to be warm-musical-detailed ....not neutral
of the Sony TA-ZH1ES, participating in Head-Fi forums
"...gut für die Langzeithörtauglichkeit, bewirkt aber, dass das Ganze etwas wärmer, dunkler klingt als „Normalnull“."
...good for long-term listening, but causes the whole thing to sound a little warmer, darker than "transparent/neutral"
Un rendu sonore typé analogique sur une plateforme technologique purement numérique !"
An analog sound rendering on a purely digital technological platform!
Elements of Good Engineering Hygiene
TA-ZH1ES Balanced-2 output directly to E1DA Cosmos ADC (~8V output), no dummy load. Wireworld Chroma8 USB cable. Generic 3.5mm to XLR from Amazon.
1. Exceptional Channel Matching to 0.001 dB
One thing that subjectivists and objectivists can universally agree upon is that an "ideal" stereo product should have matched performance between the left and right channels. Let's look at the channel matching of the TA-ZH1ES as measured by the Grade A, E1DA Cosmos ADC.
The Sony TA-ZH1ES is level matched to 0.001 dB
(green versus orange). The TA-ZH1ES's channel matching was better than the channel matching of my "Grade A" E1DA Cosmos ADC
which is already an impressive 0.007 dB (purple vs green). This is well beyond the threshold of audibility. That's incredible and potentially useful since this is a headphone amp where the distance between the two transducers won't change the way it would between two speakers. Channel matching for home speakers isn't quite as critical since sitting in a different position in your chair can result in channel imbalances as well.
View attachment 265073
Here's a standard zoomed-out measurement showing the frequency response at 24/44.1 versus 24/192. It appropriately attenuates everything above 22 kHz and starts rolling off very early is already -0.1 dB at 6 kHz.
View attachment 265074
2. Clean 44.1 kHz roll-off*
The roll-off filter used by Sony isn't as fancy as the Chord DAVE so I won't call it exceptional but it if you compare it to something like the SMSL SU-10, you can see that it has the characteristics of a fast hybrid filter where the roll-off starts early and then drops more quickly above 18 kHz. Unlike the SMSL SU-10, however, there is no undulation/oscillation with the Sony in the audible range. This suggests that Sony's version of the Fast Hybrid digital filter implemented by their conversion to DSD is better than our SINAD leader. The audibility of the different DAC filters is not well established, but it's certainly nice to see a prettier curve from the standpoint of engineering.
View attachment 265075
3. Avoids AC mains noise*
The E1DA Cosmos has a noise floor of around -128 dB so it's going to mask mains noise to a greater degree than a APx555. Looking at this plot in dBr, in the region from 4 Hz to 400 Hz, you can see that any noise from the power supply is below the threshold of my test equipment except for something potentially at 120 Hz. I'm not sure a clean power supply should be considered exceptional in 2023 at the -128 dB threshold, but it's certainly not bad engineering to get to this level.
Some audio hobbyists have said that "groove noise" is a positive component of the vinyl LP listening experience. I don't believe mains noise falls into the same category. Even if you want to impart a sound signature that is not transparent, I don't see how adding mains nose is useful. But if you want mains noise, the Sony isn't for you...
View attachment 265077View attachment 265080
If you look at the Pass Labs HPA-1 which also purports to have a favorable "harmonic distortion" profile, you can see that there is a lot more mains noise in the PASS Labs product. Of note, the HPA-1 is not designed by Nelson Pass but Jam Somasundram for Pass Labs (previously Cary Audio). This seems to be Mark's (@eatapc
's) headphone amp of choice when talking about "sometimes there are things that don't measure well, but you like" in this YouTube video
I added the orange line to show the noise floor of the E1DA measurement of the TA-ZH1ES. The spikes above the orange line is how you can compare the Sony TA-ZH1ES to the Pass Labs HPA-1.
Elements of Sony Idiosyncrasies
4. Incorporates a flagship ADC and USB interface with no recording ability.
The unit has a line-input that is powered by an AKM AK5572EN. That was the flagship ADC from AKM at the time. The line input is always digitized, either to PCM or DSD. The unit has a USB interface. Sony being Sony, the line input stays inside the box and there is no way to bring this digital information to your PC. You can't the TA-ZH1ES to record anything. The line output runs through the whole signal chain and the highly processed FPGA/DAC so the line input fed a clean analog signal versus the USB input generates similar THD+N.
View attachment 265105
View attachment 265106
5. Overly complex amplification with decision to go with zero negative feedback
You may have noticed that there are two AD converters. That's because there are two amplifiers in the TA-ZH1ES (as is the case with the SA-Z1 which uses the second generation of this approach).
The musical input is converted to a 1-bit stream (PWM/DSD) via a FPGA. The S Master HX
"power DAC" amplifies the PWM or DSD signal before filtering it. Thus, the amplifier and DAC are a single stage, not separate stages. They are one and the same. This is no different from the S Master HX in the Walkman NW-WM1A
. What is unique is that, in parallel, the PWM or DSD signal is routed through another DAC, the output of which is amplified in the analog domain. This is used as a "feed-forward signal for error correction." This second stage is a conventional analog amplifier.
This has been described in English-language Sony marketing as "reducing noise and distortion" and given Sony's expertise in the noise reducing part of noise reduction headphones
and history/extensive experience with Class D amplification
, you might believe that to be true. I don't think this is accurate since if Sony wanted low distortion, it would make sense to have a negative feedback loop in the design.
Sony's Japanese engineering team has provided few technical details except for this slide I have screenshoted below and a screenshot from the Sony Canada website.
The slide explains that the analog component is used to provide a "squarer" square wave for the Class D amplification component. This is not feedback to the audio signal the way negative feedback loop works or a THX-style feed-forward loop works. They are making the pulses from the Class D sharper. The engineering presentatios DO NOT talk about decreasing noise or decreasing harmonic distortion and specifically talks about compensating the wave form error of a high amplitude pulse of the digital amplifier.
Sony's S-Master HX does its volume control by changing the height of the pulse. Sony says that when the amplitude is increased, the square wave fades slightly. Correcting this less perfect square of the Class D component is what is done with the analog amplifier. This suggests that the goal is to improve the square wave accuracy of the S-Master HX component not necessarily correction audio distortion and discrepancy between the audio source and amplifier output...
has explained superbly, the way you make square waves "squarer" is to increase the bandwidth and add more high-order harmonics
View attachment 265108
View attachment 265121
Note the phrase "...preserving the unique sound character of the non-feedback S-Master HX.
6. DSEE HX upsampling and DC Phase Linearizer
DSEE is Sony's upsampling algorithm is designed to add extra information to help extend frequency response beyond the cut-off of lossy compression or even standard 16/44. The standard version, available for free for anyone with a Windows PC is found in Sony Music Center
. It works with any high-resolution USB DAC and allows you to test the effect yourself to see if it's audible.
The version of DSEE HX in the TA-ZH1ES allows slight tweaking of the algorithm/aggressiveness through 5 different tunings: Standard, Male Vocal, Female Vocal, Percussion, and Strings.
What's different from random upsampling is that these tunings were developed by comparing Sony Music's actual studio DSD/DXD master files recorded in high-resolution and comparing them to the end-result 16/44.1 PCM or 16/48 AAC distribution format. The later version of DSEE HX AI automatically switches the algorithms based upon the content that is detected using maching learning. Some measurements of what DSEE HX does with even low-bitrate MP3s has been posted here,
but the audibility of this effect depends on specific content and whether you still have hearing to 16 kHz or not. IMD introduced downstream may drive the changes into the audible range.
View attachment 265122View attachment 265123View attachment 265124View attachment 265125
The DC Phase Linearizer is designed to tune the bass to match classic analog amplifiers by throwing the bass out of phase to 90 degrees as you go below 40 Hz.
7. Exceptional Build Quality
When this unit arrived, for a moment I thought that I had been shipped the wrong product. It's almost 10 lbs. The pictures on the internet make it seem like a desktop DAC but it's very large. Tactile feel of the buttons is superb and even the way your cables plug into the various jacks has a nice and firm feedback. Each of the outputs are independent and relay controlled. Even though the unit runs cool to the touch, the system has several thermal sensors monitoring the electronics, potentially ensuring product longevity. Having owned vintage McIntosh gear as well as modern McIntosh gear like the MC2102 and flagship reference electronics like the Marantz PM-10, I would say that the TA-ZH1ES is right up there with the best of the best. It's Made in Malaysia, but is as nice as the very best audio jewelry from USA, Japan, or Germany.
So, how does it measure? Like a non-feedback amplifier...
In short, poorly from the harmonic distortion standpoint.
View attachment 265128
View attachment 265126
View attachment 265142
View attachment 265143
View attachment 265144
View attachment 265145
Does this pass the objectivist's threshold of audibility?
There are two ways to assess this.
1) Null test between a 16/44.1 music source (the digital file) against a recording from the headphone output of the TA-ZH1ES
PK Metric is -67.3 dBr for the opening portion of Adele's "Rolling in the Deep".
That means that if you were Neo from the Matrix and could plug the original digital source directly into your brain and then could compare it to a recording that has gone through the entire TA-ZH1ES DAC and headphone AMP and then another ADC signal chain, you'd have a hard time determining the difference when using @pkane
's perceptually weighted comparison. Anything lower then -50 dB is unlikely to be heard and we're almost-70 dBr!
While the Sony TA-ZH1ES only has a 1 kHz SINAD about -86 dB, doing the same null test with the -112 dB SINAD UB9000 achieved a PK Metric of just -71.8 dBr. Comparing the Sony TA-ZH1ES recording against the Panasonic UB9000 shows a PK Metric of -66.3 dBr.
View attachment 265129View attachment 265130
2) Look at the measured performance and compare to the thresholds summarized nicely here.
- Dynamic range, linearity: 96 dB
- THD, IMD: -66 dBFS / 0.05%
- Noise: -85 dBFS / 0.005%
- SINAD: 85 dB
- Crosstalk: -60 dBFS
- Jitter: -110 dBFS, -100 dBFS around the main tone
- Frequency response: ±0.5 dB
- Channel balance: 1 dB
Using RMAA 6.2.5, the answer is yes, it should be transparent with A-weighted values. Notice that DSD Remastering decreases distortion at the expense of noise/dynamic range on these measurements. The crosstalk is potentially limited by the DC error of the E1DA Cosmos ADC.
View attachment 265136
Does DSEE HX pass the threshold of audibility?
Not according to the PK Metric.
Not according to the PK Metric for the one song that I tried, but you can try DSEE HX yourself with any DAC by grabbing Sony Music Center for Windows. Here, the PK Metric measuring the audio with and without DSEE HX is -101.2 dBr. The algorithm does generate the high frequency content as claimed.
View attachment 265140
View attachment 265141
Do I like it?
It sounds great. Clean and without fatigue. The problem is that I have a lot of inherent bias from reading what others have written. Is it more comfortable listening at extended levels because the amp is doing a good job or is it because my earpads/headphone band are now n+1 days of fitting to my head?
Is there a Sony Signature Sound?
There is an audiophile belief that zero negative feedback, or DSD can be more musical. No one says that it's transparent or truthful -- just that it's musical. This DAC/amp still gets a PK Metric that suggests that it's fully transparent through an entire signal chain with null testing. There are audiophile products like the Chord DAVE and Mola Mola Tambaqui which are acclaimed for their transparency. This Sony is acclaimed for its non-transparency. The Sony clearly measures worse than the DAVE or Tambaqui, so the measurements are in the correct direction such that it may be possible to hear the difference between this DAC/amp and something else.
In other words, if this measured perfectly, it would be easy to argue against a Sony Signature sound. This measures far enough from transparent to suggest that there may be a sound character to it. If you wanted to intentionally design a product with a sound signature that was generally good with all types of music, it would have to measure exactly like the TA-ZH1ES: low noise, clean power supply, lots of distortion high enough to meet the threshold of audibility but no so high to be intrusive all the time.
The channel matching is world-class suggesting careful thought that has gone into the design. The decision to go zero negative feedback is a decision that goes against decreasing distortion to the lowest possible levels.
Looking at the 19/20 kHz IMD of the Sony TA-ZH1ES, we can actually see that the distortion is similar to that of what Stereophile measured from the Western Electronic WE 91E (their Figure 11)
with the initial side bands at -60 dB. So, you have a product that measures poorly for a solid state product yet matches an expensive 300B tube amplifier...
View attachment 265227
When using real-world music, using null testing (which is more prone to falsely identifying differences as opposed to falsely saying two things are similar), the perceptually weighted PK Metric shows that the Sony reproduction of music is indistinguishable from a theoretical bit-perfect direct digital stream to the brain with a PK Metric of -67 dBr
. The build quality and ergonomics are excellent and it's a just a question if PK Metric audibly transparent is enough or if you want the satisfaction of having even lower distortion or noise. One objectively measured strength of the TA-ZH1ES is its exceptional channel matching and output flexibility, as long as you are within the power capabilities.
By the numbers, you can get lower noise and lower distortion, but by the PK Metric calculations, you will not be able to hear the differences. You do get flagship level channel matching with this product, a very nice remote, and lots of headphone jack options.
The Sony provides a non-feedback headphone amp design, DSD resampling options, and high-frequency reconstruction algorithms based upon looking at original high-resolution studio masters in DSD and DXD and comparing them to the resulting CD resolution and AAC lossy-encoded results.
The absence of negative feedback results in THD that is closer to a high-end $15,000 Western Electric 300B tube amplifier with zero negative feedback for tests like the CCIF 19/20 kHz IMD
. The TA-ZH1ES delivers what subjective reviewers have said: a solid state product with some sound characteristics of a tube amplifier.
Channel level matching is essentially perfect and better than the channel matching of the Grade A, E1DA Cosmos ADC.
It measures well enough to be transparent to a theoretically impossible perfect music reproduction chain using null testing with the PK Metric.
It measures poorly enough to have distortion levels as high as zero negative feedback tube products like the $15,000 Western Electric 91E.