Accuphase DG-68 Digital Voicing Equalizer Measurements

Sidebar 3: Measurements

The Accuphase DG-68 is a complicated product, comprising an analog–digital converter (ADC), digital signal processing (DSP), and a digital–analog converter (DAC). It also has digital inputs and outputs that bypass the ADC and DAC stages. JVS used it as a complete analog–analog processor, so, in addition to examining the individual performance of the ADC, DSP, and DAC stages, I looked at its behavior as set by JVS.

I used my Audio Precision SYS2722 system (see the January 2008 "As We See It") to perform the measurements, repeating some tests with the magazine's more-recent APx555 system. While Accuphase specifies the coaxial S/PDIF input as accepting data sampled at rates up 192kHz, it wouldn't lock at rates greater than 96kHz to the coaxial digital output of either Audio Precision system. The optical S/PDIF input accepted sample rates up to 96kHz, as specified.

Looking first at the DG-68's performance as a D/A processor, the maximum output level at 1kHz with its Automatic Level Control switched on was 1.7V from both the balanced and single-ended outputs. With its rear-panel switch set to pin 2 of the output XLR's positive, the Accuphase preserved absolute polarity (ie, was noninverting) from both sets of analog outputs. The output impedance was a low 47 ohms from both outputs at 20kHz and 1kHz, rising to a still-low 83 ohms (balanced) and 66 ohms (unbalanced) at 20Hz.

The Accuphase's DAC offers a single reconstruction filter with a classic minimum-phase response (fig.1), all ringing following the single high sample. This filter's ultrasonic rolloff with 44.1kHz data (fig.2, magenta and red traces) reaches full stop-band attenuation just above half the sample rate (the vertical green line). The aliased image at 25kHz of a full-scale tone at 19.1kHz (cyan, blue) is suppressed by 100dB, and the harmonics associated with the 19.1kHz tone are all very low in level. Fig.3, taken with digital data sampled at 44.1kHz and 96kHz, shows that the D/A response rolls off sharply just below half the sample rate.


Fig.1 Accuphase DG-68, DAC mode, impulse response (one sample at 0dBFS, 44.1kHz sampling, 5ms time window).


Fig.2 Accuphase DG-68, DAC mode, wideband spectrum of white noise at –4dBFS (left channel red, right magenta) and 19.1kHz tone at 0dBFS (left blue, right cyan) into 100k ohms with data sampled at 44.1kHz (20dB/vertical div.).


Fig.3 Accuphase DG-68, DAC mode, frequency response at –12dBFS into 100k ohms with data sampled at: 44.1kHz (left channel cyan, right magenta) and 96kHz (left blue, right red) (0.5dB/vertical div.).

Channel separation was superb, at>126dB in both directions below 1kHz, decreasing to a still-superb 112dB at 20kHz. The DG-68's analog noise floor was very low and free from power supply–related spuriae (fig.4). An increase in bit depth from 16 to 24, with dithered data representing a 1kHz tone at –90dBFS, dropped the Accuphase's noise floor by 21dB (fig.5), which implies a resolution between 19 and 20 bits. When I played undithered data representing a tone at exactly –90.31dBFS, the waveform was symmetrical, with the three DC voltage levels described by the data free from noise (fig.6). A very slight DC offset of +25µV was present in the right channel's output. With undithered 24-bit data (fig.7), the Accuphase's very low analog noise floor allowed it to output a clean sinewave, even at this very low signal level.


Fig.4 Accuphase DG-68, DAC mode, spectrum with noise and spuriae of dithered 1kHz tone at 0dBFS with 24-bit TosLink data (left channel blue, right red) (10dB/vertical div.).


Fig.5 Accuphase DG-68, DAC mode, spectrum with noise and spuriae of dithered 1kHz tone at –90dBFS with: 16-bit data (left channel cyan, right magenta), 24-bit data (left blue, right red) (20dB/vertical div.).


Fig.6 Accuphase DG-68, DAC mode, waveform of undithered 1kHz sinewave at –90.31dBFS, 16-bit data (left channel blue, right red).


Fig.7 Accuphase DG-68, DAC mode, waveform of undithered 1kHz sinewave at –90.31dBFS, 24-bit data (left channel blue, right red).

As a DAC, the Accuphase produced very low levels of harmonic distortion with full-scale digital data; the second harmonic was higher in the left channel (fig.8, blue trace) than the right (red trace). However, at just –117dB (0.00014%), this is inconsequential. This spectrum was taken with dithered 24-bit data into a high 100k ohm load. When I reduced the load impedance to the current-hungry 600 ohms, the distortion spectrum hardly changed. Intermodulation distortion with an equal mix of 19 and 20kHz tones, each lying at –6dBFS, was similarly very low (fig.9), with the difference tone at 1kHz lying at –134dB (0.00005%). The aliased images of the primary tones can be seen in this graph, but they lie at –99dB and –107dB. Again, substituting 600 ohms for 100k ohms hardly changed the level of the intermodulation products.


Fig.8 Accuphase DG-68, DAC mode, spectrum of 50Hz sinewave, DC–1kHz, at 0dBFS into 100k ohms (left channel blue, right red; linear frequency scale).


Fig.9 Accuphase DG-68, DAC mode, HF intermodulation spectrum, DC–30kHz, 19+20kHz at 0dBFS into 100k ohms, 44.1kHz data (left channel blue, right red; linear frequency scale).

When I examined the Accuphase's rejection of word-clock jitter via its S/ PDIF inputs with 16-bit J-Test data, it did very well (fig.10). All the odd-order harmonics of the undithered low-frequency LSB-level squarewave lie at the correct levels, indicated by the sloping green line, and while power supply–related sidebands can be seen at 60Hz and 120Hz on either side of the high-level tone at one-quarter the sample rate, these can only be seen because of the very low noise floor offered by the DG-68's analog output. Repeating the test with 24-bit data gave a clean spectrum (fig.11).


Fig.10 Accuphase DG-68, DAC mode, high-resolution jitter spectrum of analog output signal, 11.025kHz at –6dBFS, sampled at 44.1kHz with LSB toggled at 229Hz: 16-bit TosLink data (left channel blue, right red). Center frequency of trace, 11.025kHz; frequency range, ±3.5kHz.


Fig.11 Accuphase DG-68, DAC mode, high-resolution jitter spectrum of analog output signal, 11.025kHz at –6dBFS, sampled at 44.1kHz with LSB toggled at 229Hz: 24-bit TosLink data (left channel blue, right red). Center frequency of trace, 11.025kHz; frequency range, ±3.5kHz.

Turning to the DG-68's A/D coverter, I examined its performance in the digital domain by connecting its coaxial digital output to the SYS2722's digital input. The sample rate can be set to 44.1kHz, 88.2kHz, 176.4kHz, or 352.8kHz. I mainly performed the ADC measurements at 88.2kHz. The unbalanced analog input impedance was 19.5k ohms at 20Hz and 1kHz, dropping to 10k ohms at 20kHz. The balanced input impedance was twice the unbalanced, as expected.

Accuphase's specifications say that the analog input clips at 1.78V with the gain set to "0dB." I found that the analog input clips (when the THD+noise reaches 1%) at 1.85V (fig.12). When you clip the analog input, the word "Peak" lights up in red in the display and you need to reduce the input sensitivity with the control panel.


Fig.12 Accuphase DG-68, ADC mode, THD+noise (%) vs input level in volts.

At levels lower than 1.85V, the DG-68 digitizes its analog inputs with very low distortion. Fig.13, for example, shows a spectral analysis of the Accuphase's digital output with an analog signal 3dB below clipping. While some distortion harmonics can be seen, these are vanishingly low in level. The second harmonic is the highest, but at –129dB, left, and –140dB, right, it is insignificant. Note the circa 15dB of headroom visible in this graph. This is to allow the DG-68's equalization to apply up to 12dB gain without the level exceeding 0dBFS.


Fig.13 Accuphase DG-68, ADC mode, digital-domain spectrum of 1kHz sinewave, DC–10kHz, at –3dB ref clipping voltage (left channel blue, right red; linear frequency scale). .

I then examined the DG-68's behavior from analog input to analog output with the DSP equalization bypassed. With the Automatic Level Control (ALC) active, a 1kHz analog input signal at 1V resulted in an analog output level of 949mV, ie, an insertion loss of just 0.23dB. I switched ALC off, which set the output level to "0.0dB," and adjusted the level with the input gain adjust buttons on the remote control: "–6dB" gave an analog output level of 1.341V, ie, a gain of 3dB compared with ALC on; "0dB" resulted in an analog output signal of 2.674V; "5dB" resulted in 4.756V (exactly +5dB); but higher gain settings clipped the output.

Fig.14 shows the frequency response at all four ADC sample rates, with and without the subsonic filter active. The response at 352.8kHz (blue and red traces) is identical to that at 176.4kHz (green and gray traces), with a rolloff reaching –3dB at 55kHz. (The specifications say the –3dB frequency is 50kHz.) The subsonic filter offers a sharp rolloff below 40Hz, reaching –3dB at 20Hz. Some slight passband ripple can be seen, this presumably due to the ADC's antialiasing filter.


Fig.14 Accuphase DG-68, analog input–analog output mode with and without subsonic filter, frequency response at 1V into 100k ohms with data sampled at: 44.1kHz (left channel blue, right cyan). 88.2kHz left cyan, right magenta), 176.4kHz (left green, right gray), and 384kHz (left blue, right red) (1dB/vertical div.)

I examined the DG-68's impulse response from its analog input to analog output by using the shaped impulse sampled at 384kHz I used for my 2019 ADC article. I played this file with the Pure Music 3.0 app on my MacBook Pro, sending the impulse data to a Mytek HiFi Brooklyn DAC via USB. As the signal's sample rate was 384kHz and I was testing the Accuphase with its sample rate set to 88.2Hz, the ringing of the Brooklyn's own reconstruction filter at 192kHz would be two octaves above the DG-68 ADC's Nyquist frequency and would not affect the result. Fig.15 shows the result with the equalization bypassed. Other than the doubled sample rate, the impulse response is the same minimum-phase type seen in fig.1. However, it is now inverted: The balanced input XLR jacks are permanently wired with pin 3 positive, the opposite to how the output XLR jacks had been set. The unbalanced analog inputs preserved absolute polarity.


Fig.15 Accuphase DG-68, analog input–analog output mode, impulse response (one sample at 0dBFS, 88.2kHz sampling, 5ms time window).

Finally, I examined the effect of the DG-68's DSP, with the settings that had been created by JVS, by switching on "VC/EQ." Fig.16 shows the effect on the impulse response. As expected, the equalization's digital filters have modified the impulse response. (A digital filter operates by multiplying or "convolving" the input data with its own impulse response.) The result in the frequency domain of JVS's preferred settings (without smoothing) is shown in fig.17; it reveals the DG-68's versatility—the "Voicing" IIR filter has 67 1/6-octave bands, the "Equalizer" IIR filter has 80 bands, all offering ±12dB adjustments. (The DG-68's spectrum analyzer screen echoes traditional analog analyzers in that pink noise appears flat.) Significantly, switching in the equalization did not reduce resolution or increase distortion.


Fig.16 Accuphase DG-68, analog input–analog output mode with VC/EQ active, impulse response (one sample at 0dBFS, 88.2kHz sampling, 5ms time window).


Fig.17 Accuphase DG-68, analog input–analog output mode with VC/EQ active, 88.2kHz sampling, frequency response at 1V into 100k ohms (left channel blue, right red) (2dB/vertical div.)

The Accuphase DG-68's measured performance indicates that no compromises have been made to achieve its versatility.—John Atkinson

Accuphase Laboratory Inc.
US distributor: Axiss Distribution Inc.
17800 South Main St., Ste 109
Gardena CA, 90248
(310) 329-0187

Anton's picture

I would think your room has to be OK for it to be able to produce a decent room curve to begin with.

Your idea of continuing to tweak your room while you have the toy is genius.

It would be cool to have a "mother module" to use to set the room, and then a much less expensive "servant module" or chip that could talk to your DAC that you could then plug in to keep 1-3 curves while allowing the mother ship to continue her sonic journey.

Archimago's picture

Nice to see audiophilia using (at least not afraid of) EQ and finding benefit in "room correction" techniques.

So, basically, from a technical perspective is this performing 2-channel 35-band parametric EQ (settings up to 50kHz)?

Is there any time-domain correction being done when fed with room information (doesn't look like it?)?

teched58's picture

$24,000 seems a̶ ̶l̶o̶t̶ ̶o̶f̶ ̶m̶o̶n̶e̶y̶ very affordable for a 35-band parametric equalizer!

Anton's picture

Wow, what a great addition that would be.

RichT's picture

Something very unexpected here - the flat response curve is incredibly flat, I would estimate -3dB +1dB. This is a flatter response than most professional studios achieve, even the best. It’s totally flat 30Hz to 90Hz, which is absolutely remarkable. This Is very special room.

Kal Rubinson's picture

And not everyone wants flat. ;-)

MatthewT's picture

Perfect is boring.

Kal Rubinson's picture

And, under some conditions, undesirable.

MatthewT's picture

Is the sum of its flaws, and I love them all.

latinaudio's picture

are a necessity. What is certain is that the sound perceiver must be taken into account: the brain. And since no two human beings are alike, it will ALWAYS be necessary to adjust the bass, treble, midrange to each person... in order to have a pleasant experience. Tone controls ARE an unavoidable necessity, and hifi manufacturers builders would be arrogants if they don't accept it. So remember: it´s not only the room, are the persons, specially the owner of the equipment :)

Ortofan's picture

... published in the November 1973 issue of Audio magazine.

It includes a description of the then new Altec 729 Acousta-Voicette one-third octave equalizer and the Altec/HP 8050 real-time analyzer.

Presented is a rationale as to why the corrected frequency response should not be flat over the entire audible range.
The proposed optimum frequency response is flat up to 2kHz, followed by a roll-off above than point at the rate of 3dB/octave.

Kal Rubinson's picture

Good stuff at the time but acoustical science and audio componentry have moved on.........................

Ortofan's picture

... the latest developments in acoustical science (and audio componentry) suggest is ideal?

Kal Rubinson's picture

Toole's book is a good place to start.

Ortofan's picture

... those under the Harman umbrella, agree with Dr. Toole's findings and implement them in their products?

Kal Rubinson's picture

I suspect that all the current AVRs and prepros incorporate them in their EQ options.

zuman's picture

I've been listening seriously and buying good gear for over 40 years, and I firmly believe that the effects of some "tweaks" CAN be measured while we haven't yet discovered how to measure other audible tweak results. However, I also believe that many tweaks are placebos. Is the DG-68 sufficiently sensitive to identify/quantify/adapt to some of the supposed "night-and-day" effects of some of these tweaks?

Anton's picture

Of course, no matter the outcome no minds would be changed.

We are audiophiles, after all. ;-D

pbarach's picture

I wonder how this unit's results would compare to some of the other (much less expensive) room correction systems, e.g., Audyssey, ARC, Trinnov.

Kal Rubinson's picture

Hard to compare correction systems which reside in different hardware.

georgehifi's picture

Love to have this thing, but I would have liked to have heard the comparison of what Rossini Ring/Dac was like from the Accuphase's digital outputs compared to it's own ESS dac (never been a lover of the ESS converters). (was it done???)

Cheers George

AudioBang's picture

Back in the late 90s I bought a Sigtech for $7,300 which, in addition to room EQ, performed time alignment, cancellation of late arriving ceiling, floor and wall reflections out to 50mS, and had the ability to compensate for spectral build up [from room corners, etc].
I owned a pair of Dunlavy Vs at the time, later upgrading to the SC VI.
Since the Dunlavy's are all time-aligned to begin with, as expected, the impulse response did not visually change after a correction filter was applied. I too found it easy to get caught up in jumping from correction filter to correction filter and being distracted from the music. The Dunlavy recommended listening position was against the long wall so the speakers could be placed with a wide listening angle - up to 120 degrees! [per the manual] and obtain the smoothest bass response by eliminating room nulls. I liked mine at about 100 degrees as there was a level of psychoacoustic wonder at hearing a phantom center image with the speakers 12 feet apart while still presenting a reasonably proportioned soundstage. I later learned from the SigTech measurements, that with the listening position against the rear wall, the room buildup in the bass was enormous! A significant contribution in my system that the Sigtech made I felt, was ameliorating the room build up against the rear wall as well as smoothing a 12dB delta from 30hz to 90hz - I suspect from the interactions between the bass drivers positioned at the floor and the top of the cabinet. The low frequency measurements of my speakers from Dunlavy's very large anechoic chamber showed a similar 12dB peak to null although at slightly different frequencies from my room. Also, I recall, not all of that 12dB difference could be flattened as you wouldn't want to add unprecedented gain at high listening levels at the null which could blow the drivers. In a nutshell, in my room, taming the bass [overhang] was a significant factor in arriving at more transparency and I felt that manually treating the room [absorption in my case] for first sidewall reflections worked better than relying on the Sigtech's ability to cancel them out - even though these reflections were eliminated on the filtered impulse response. The measurement capability of these reflections was instrumental in confirming the before and after effects of midrange smearing that was eliminated after applying absorption from first reflections. I was not successful attempting to fix bass problems with corner traps.
The 2khz -3dB/octave rolloff was the recommended default filter. The flat filter sounded dry and sterile and was not pleasurable to listen to although I look back and wonder if part of the reason was that the recommended rolloff starting at 2khz may have masked the sub-par 90s digital in my system [Wadia 270 Transport/27ix DAC]. For what it's worth...

John Atkinson's picture
AudioBang wrote:
Back in the late 90s I bought a Sigtech for $7,300 which, in addition to room EQ, performed time alignment, cancellation of late arriving ceiling, floor and wall reflections out to 50mS, and the ability to compensate for spectral build up [from room corners, etc].

Stereophile's December 1996 review of the SigTech system will be posted to the website the week starting August 9.

John Atkinson
Technical Editor, Stereophile

tonykaz's picture

I recall setting up our Listening Rooms at Esoteric Audio in the 1980s.

We ended up with a Semi-Anechoic type result that needed the Largest Mono Amps pushing thick MH-750 Music Hose Speaker Cabling.

Do our brains re-calibrate room accoustics for us ( to some extent ) ?

I suspect that Room Problems are exasperated by Loudspeakers with too-powerful Bass circuits and no balancing electronic circuits..

B&O make that gigantic Beolab 90 that features built-in Room Correction.

Full Range Loudspeaker Manufacturers should include some sort of Bass management system as an included component. ( Genelec )

Leaving Accoustic Engineering in the hands of purchasing Consumers seems irresponsible.

Accuphase Gear is Georgous

Tony in Venice Florida

AudioBang's picture

I concur that our brains recalibrate room acoustics for us - to some extent....
Interestingly, per MF and other Stereophile reviews on the ISO Acoustics speaker isolators, I tried three sets on my Dunlavy SC VIs [650 lbs each] and they completely altered the bass. On the one hand, I agree with one of the reviewer's comments that they were "different" and on the other, the bass overhang that I am used to [assuming] from room modes, was reduced an order of magnitude as if "where did it all go?" But the transparency, depth and expanded soundstaging these footers offered was unexpected and unprecedented. I'd take these results 10X over any EQ solution. Because the rebuilt 4-way crossovers reside outside the speaker, while I was contemplating trialing the footers, I had disconnected the bass completely out of the picture to test whether the vibrational bass coupling to the floor made any difference to the midrange/treble. I didn't hear any difference. But after installing the footers everything transformed at the magnitude that MF stated in his review. I can only assume that even with just midrange and tweeter operating, something vibrationally is happening to/within the 650lb cabinet causing blurring to occur. My brain can not connect how that happens. BTW Stereophile, thanks for your value-add here. This was one of the most profound improvements I've experienced.

tonykaz's picture

Can we hear from the Loudspeaker Manufacturer on these devices.? Why don't they include these with every speaker?

I probably consider them to be like Engine Mounts for a vibratingly harsh Car, Truck, Boat Engines.

NoiseVibrationHarshness is a science, we should be able to explain these phenomenons and bring about consistant effects/affects.

In the 1980s I sold large quantities of Audio related do-dads, Audiophiles love to tinker around with this stuff.

Tony in Venice Florida

Kal Rubinson's picture

Can we hear from the Loudspeaker Manufacturer on these devices.? Why don't they include these with every speaker?

Good questions and there are a couple of speakers with them on their way to market The general answer is that it will increase the price of the speaker and manufacturers have to compete on price in the real world.

It evokes the memory of an amp manufacturer who told me that I could not realize the full potential of his product without a particular after-market power cord. I was surprised at that and asked why he would not just include it. He replied that it was a matter of price. I bought one anyway (who would not want his new amp to sound its best) but I found it was absolutely not audibly different from the stock cord. He should not have opened his mouth.

Ortofan's picture

... is including shock-mount feet.
Speaker prices range from $1,850 to $4,250 each.

Kal Rubinson's picture

PSB, too.

tonykaz's picture

Good Better Best has been a Sales Standard in all sorts of Product types..

I'd expect better from the Full Range Transducer System manufacturers, just as we've come to expect ( and rely on ) Stereophile Reviewing and Editorial Standards.


Audiophiles are typically DIY kinds of folks.

I'm a Fan of Meridian type of Music Systems approach from Manufacturers. I can enjoy a high performance designed full System from one Company.

Still Tweaking one's music systems can be a nice little distraction from the World problems crushing our shoulders.

Big Fat Stiff Power Cords are visually impressive, you could say "those power cords alone cost Thou$$$$$$and$$$ of Dollars!! ( even toss a handful of hundred dollar bills on the floor behind the speakers ( to show off even more )

Audio stuff can still be fun! I think

Tony in Venice Florida

tonykaz's picture

I just heard that the NY Auto Show is cancelled.!!!

We're not out of this China Plague yet

Tony in Venice Florida where masks are still optional

Kal Rubinson's picture

But what about the NY Audio Show?

tonykaz's picture

It seems that we are entering a new phase of China Pandemic hysteria with this Fresh Strain that can be understood but isn't..

So, we might usefully predict that Big Shows will set the pace for all the little Shows like Audio and/or things like Tool Shows.

CDC is predicting the New Strain will take the rest of this year to run it's course ( in USA that has less than 100% Vaccinations depth ).

Global Vaccination percentages have barely scratched the surface so we are a looooooonnnnnnnngggggg way from Herd Immunity!

Will it predictably take the Standard 3 to 4 years for this blight to run it's course ???

Will the politicians and media continue to foster Health condition Panic among the innocent polis ?

I have the feeling that Streaming and DACs like the dcs Bartok will have overwhelmed HighEnd Audio by the time this Virus is tamed.

It's time to sell off my Koetsu & Vinyl collections

Tony in Venice Florida

AudioBang's picture

I think you would be a very cool neighbor to have Tony :)
I enjoy your industry stories and commentaries from your experiences.