General setup:
MacBook Pro (Decibel bit-perfect) --> shielded USB --> TEAC UD-501 --> shielded 6' RCA --> E-MU 0404USB --> shielded USB --> Win8 laptop
The TEAC has quite a "hot" XLR output and unfortunately clips the E-MU so I was unable to get an accurate reading without using volume attenuation (rest assured it does look very good, dynamic range is probably about another 6dB better that what I got with the RCA 24-bit tests; beyond the resolution of the E-MU).
.bmp)
A look at a 24/44 1kHz square wave at 0dBFS using SHARP digital filter. RCA peak voltage at 2.8V. Nice wave morphology. Yellow is right channel, blue is left; channel balance looks good.
PCM 16/44:
The most common digital sampling rate is of course "good" old 16/44 Red Book format. These days, any DAC worth it's salt MUST perform close to ideal at 16/44...
Summary (RightMark 6.2.5):
As you can see, the UD-501 was measured with the 3 digital filters (DF's - OFF, SLOW, SHARP). Comparison was made with the Logitech Transporter (ethernet), Touch (ethernet), and Oppo BDP-105 Blu-Ray player (USB). Unless explicitly adjusted, vast majority of DAC's utilize some form of the SHARP filter by default at least for 16/44. Overall, you see that 16-bit audio is absolutely no problem for any of these devices.
Here's the frequency response of the devices:
The most obvious thing to see here is that the Touch rolls off on the low end by 1dB compared to the others, and the digital filter "OFF" setting of the UD-501 rolls off quite early starting around 5-6kHz... Let us focus for a second just on the TEAC:
Interesting; the SLOW and SHARP settings are pretty self explanatory (the Transporter has similar settings). The OFF setting results in significant roll off even earlier and by about 18kHz, the OFF setting is about -2.4dB, compared to SLOW setting at -1.7dB, and SHARP at -0.1dB.
Hmmm, where have we seen that kind of filter "OFF" curve before? Oh yeah, the old TDA1543 :-). Here ya go:
What does this mean? Yup, the TEAC can function in "NOS [NonOverSampling] mode" with the digital interpolation filter turned off! Behold, stair-stepped NOS waveforms out of a modern DAC with DSD capabilities:
Digital filter OFF:
Digital filter SHARP:
That's really quite a trick from the TEAC!
Noise Level:
All pretty equivalent with fantastic level of functioning. Note the 60Hz powerline hum visible.
Stereo Crosstalk:
Very close; worst being the SB Touch. Same 6' shielded RCA cable used in all tests.
PCM 24/96:
24/96 is the "sweet spot" for high-resolution PCM DAC's these days. Certainly in DAC's I have previously tested, once you go above 24/96, there's often deterioration in the dynamic range. Furthermore, I do not believe there is any scientific evidence to suggest human ears can experience sound beyond the resolution encompassed by 24/96 so it's important that a true "high resolution" DAC be able to demonstrate adequate performance at this level.
Here's the "big board" summary with a number of devices tested (you probably need to click on the image for more comfortable reading):
As you can see, for comparison I've thrown in results from the E-MU 0404USB itself, Logitech Touch, ASUS Essence One, Logitech Transporter, and Oppo BDP-105. Obviously every one of these devices is capable of >16-bit resolution showing improved dynamic range beyond the 16-bit test above. "Top tier" devices are the TEAC, Transporter, Oppo; each measuring beyond 110dB potential dynamic range using the RCA output - essentially at the limit of the E-MU's abilities (the Essence One & E-MU would also be on this list when using balanced XLR or TRS cables respectively).
Frequency Response:
Since the graph got too busy, I removed the ASUS and Oppo - they basically look like the Transporter in terms of frequency response. Again, the Touch drops a dB down near 20kHz.
Here's the graph with just the TEAC settings plus the old TDA1543 NOS:
Here's the graph with just the TEAC settings plus the old TDA1543 NOS:
Notice again the similarity of the TEAC's digital filter "OFF" setting and the TDA1543 NOS DAC in the high-frequency end.
THD:
Interesting increase in high frequency noise with the digital filter "OFF". Looks like unfiltered delta-sigma noise shaping coming through?
PCM 24/192:
Next, one more step up in sampling rate:
For interest, I threw in the Logitech Touch with EDO plugin --> coaxial --> AUNE X1 DAC. Notice how well this combination measures! The AUNE X1 is only a $200 DAC and the combination produces very respectable measurements (and sound very good IMO). In comparison, I am disappointed in the Essence One going from 24/96 to 24/192. The TEAC and Oppo really hang in there with essentially identical results compared to 24/96 - great to see!
Frequency Response:
As I mentioned in the MUSE TDA1543 measurements, one way to improve NOS DAC performance is to feed it with higher sampling rate data... In doing so, you get closer to the performance of oversampling interpolation filters. You see this here - the higher the sampling rate, the closer the digital filter "OFF" curve gets to the "SLOW" and "SHARP" settings (in fact, you see in the next section, they become identical).
There's that early roll off with the ASUS Essence One previously measured.
There's that early roll off with the ASUS Essence One previously measured.
Noise Level:
Essence One getting a bit noisy at high sample rate compared to the others (realize it still has >100dB dynamic range though). Again, we see quite a bit of high frequency noise with the digital filter "OFF".
PCM 24/384 (more than DXD [352.8 kHz]!):
This is a "pseudo-test" actually. The fact is that the E-MU 0404USB is incapable of digitizing at 384kHz so what I did was upsample the 24/192 test signal using SoX so see if running the TEAC at the higher sampling rate will cause a measurable loss in the analogue output dynamic range or worsen noise characteristics within the measurable capability of the E-MU.Summary:
First 3 columns were measurements done with the TEAC running at 24/384 with various digital filter settings. The last column is the "SHARP" filter measured at 192kHz. There may have been very subtle loss in dynamic range. Some or even all of this could be due to the upsampling conversion algorithm. In any case, the measurements look excellent and it seems indeed the TEAC is able to maintain low noise even at the extreme sampling rate of 24-bit & 384kHz!
Frequency Response:
Note how the NOS-like digital filter "OFF" setting is identical to the other settings now. Basically, sampling at 384kHz is like 8x oversampling of a 44kHz signal (2x oversampling of 192kHz).
Jitter:
As usual, let us look at some FFT's from the Dunn J-Test. For simplicity, I'll just show the spectra from the SHARP filter setting.
USB input (16-bit and 24-bit spectra):
Coaxial input using CM6631A USB to S/PDIF:
TosLink input using CM6631A USB to S/PDIF:
TosLink input again fed by CM6631A with *24/192 upsampling*:
The reason I didn't bother showing any results from hardware upsampling to 24/192 in the tables above was because the numbers and graphs looked essentially unchanged. However, there is one situation where upsampling makes sense... The same reason Benchmark chose to use ASRC (Asynchronous Sample Rate Conversion) for the DAC1 and DAC2 - jitter reduction. Although by no means high, the sidebands are more pronounced using coaxial and TosLink interfaces. The sideband peaks around the primary signal clearly were reduced with 24/192 upsampling using the TosLink input. As usual, whether anyone can actually hear this difference in properly controlled testing is another matter!
Summary of PCM Results:
TEAC has created a machine which objectively compares very well to some other excellently measuring devices like the Logitech Transporter and Oppo BDP-105. It's great to see that even operating at the extremely high DXD-level sampling rates, noise level remains low and dynamic rage appears preserved.
What I found surprising was the option to allow the digital filter to be turned "OFF"; I don't recall any reviewers spending much time on this (even the AudioStream review just glossed through this and didn't comment on the sound). This setting puts the DAC into a "NOS mode" where digital interpolation is suspended - this appears novel especially in a device with low-jitter asynchronous USB interface and a true 24-bit (err... ok, 32-bit as if that makes a difference) DAC... In general NOS DACs these days are still based on obsolete decades-old DAC chips like the Philips TDA154x (16-bit) or Analog Devices AD1865 (18-bit) which tend to perform poorly on measurements. Although personally I am not a big fan of the roll-off and aliasing distortion, some have commented on subjective improvement by taking out the digital oversampling filter, so I definitely consider it a positive that TEAC offers this option for anyone to try (in real time with instantaneous A-B'ing no less just by turning the knob)! I can certainly see this option useful to tone down some of the overly "bright" digititis-inducing recordings. Looking at my pop CD collection, an example where this was demonstrable was Jason Donovan's disco-inspired Too Many Broken Hearts from Ten Good Reasons (first pressing, 1989) where the OFF setting was more tolerable after 3 minutes :-). As a compromise, the SLOW filter may be reasonable.
As I mentioned at the beginning, PCM remains the cornerstone of digital audio. These TEAC UD-501 results suggest that nothing has been sacrificed in terms of performance in the PCM domain. Note that the ASUS Essence One is also based on the PCM1795 chip in dual-mono configuration but doesn't measure as well, highlighting the importance of the electronics around it like the analogue output stage, power supply and USB/coaxial/TosLink interface circuitry affecting the final output quality. One thing I wish the TEAC had from the ASUS is the beefier headphone amp though.
Bottom line: these results are consistent with the excellent subjective sound quality described in the previous UD-501 blog post. I would happily present some kind of award if it meant anything :-).
What I found surprising was the option to allow the digital filter to be turned "OFF"; I don't recall any reviewers spending much time on this (even the AudioStream review just glossed through this and didn't comment on the sound). This setting puts the DAC into a "NOS mode" where digital interpolation is suspended - this appears novel especially in a device with low-jitter asynchronous USB interface and a true 24-bit (err... ok, 32-bit as if that makes a difference) DAC... In general NOS DACs these days are still based on obsolete decades-old DAC chips like the Philips TDA154x (16-bit) or Analog Devices AD1865 (18-bit) which tend to perform poorly on measurements. Although personally I am not a big fan of the roll-off and aliasing distortion, some have commented on subjective improvement by taking out the digital oversampling filter, so I definitely consider it a positive that TEAC offers this option for anyone to try (in real time with instantaneous A-B'ing no less just by turning the knob)! I can certainly see this option useful to tone down some of the overly "bright" digititis-inducing recordings. Looking at my pop CD collection, an example where this was demonstrable was Jason Donovan's disco-inspired Too Many Broken Hearts from Ten Good Reasons (first pressing, 1989) where the OFF setting was more tolerable after 3 minutes :-). As a compromise, the SLOW filter may be reasonable.
As I mentioned at the beginning, PCM remains the cornerstone of digital audio. These TEAC UD-501 results suggest that nothing has been sacrificed in terms of performance in the PCM domain. Note that the ASUS Essence One is also based on the PCM1795 chip in dual-mono configuration but doesn't measure as well, highlighting the importance of the electronics around it like the analogue output stage, power supply and USB/coaxial/TosLink interface circuitry affecting the final output quality. One thing I wish the TEAC had from the ASUS is the beefier headphone amp though.
Bottom line: these results are consistent with the excellent subjective sound quality described in the previous UD-501 blog post. I would happily present some kind of award if it meant anything :-).
0 comments:
Post a Comment