Hegel Music Systems H95 Integrated Amplifier-DAC Measurements

Link: reviewed by Doug Schneider on SoundStage! Hi-Fi on August 1, 2020

General Information

All measurements taken using an Audio Precision APx555 B Series analyzer.

The H95 was conditioned for 30 minutes at 10W (8 ohms) before any measurements were taken. All measurements were taken with both channels driven, using a 120V/20A dedicated circuit, unless otherwise stated.

The H95 offers two sets of line-level unbalanced (RCA) inputs, one set of unbalanced variable line outputs, and six digital inputs (three optical, one coaxial, one USB, one Ethernet). For the purposes of these measurements, the following inputs were evaluated: digital coaxial, analog (RCA) line-level.

We attempted to optimize the volume position to achieve the best signal-to-noise ratio (SNR) and THD+N measurements at the speaker outputs (8 ohms) for both analog and digital inputs. For all analog input primary measurements (unless otherwise specified), the best measurements were achieved with the volume was set to 99. At a volume setting of 99, 240mVrms was required at the input for 10W (8 ohms) at the output. For the digital input, we found a volume setting of 59 yielded 10W (8 ohms) at the output for a 0dBFS input. We found essentially no differences between the optical and coaxial S/PDIF inputs, as well as the USB input, in terms of THD+N.

Based on the accuracy of the left/right volume channel matching (see table below), the H95 volume control is likely digitally controlled in the analog domain. The volume range is 0 to 99, in increments of 1-3dB from 0 to 9, and of 1dB from 9 to 99. Beyond volume level 58, every second volume increment has no effect. When the volume is set to 99 (maximum), at the variable RCA outputs, there is a gain of -0.57dB (almost unity) for the analog inputs. For the digital inputs, 2.35Vrms was measured at the variable RCA outputs (volume at maximum) for a 0dBFS input, with a 16-bit and 24-bit input bit depth dynamic range (AES17 method, A-weighted) of 96dB and 110dB, respectively.

Volume-control accuracy (measured at speaker outputs): left-right channel tracking

Published specifications vs. our primary measurements

The table below summarizes the measurements published by Hegel for the H95 compared directly against our own. The published specifications are sourced from Hegel’s website, either directly or from the manual available for download, or a combination thereof. With the exception of frequency response, where the Audio Precision bandwidth is set at its maximum (DC to 1MHz), assume, unless otherwise stated, 10W into 8ohms and a measurement input bandwidth of 10Hz to 90kHz, and the worst-case measured result between the left and right channel.

Our primary measurements revealed the following using the analog line-level input, and the coaxial digital input (unless specified, assume a 1kHz input sinewave at 240mVrms or 0dBFS, 10W output, 8-ohm loading, 10Hz to 90kHz bandwidth):

For the continuous dynamic power test, the H95 was able to sustain 96W (1% THD) into 4 ohms using an 80Hz tone for 500ms, alternating with a signal at -10dB of the peak (9.6W) for 5 seconds, for 5 continuous minutes without inducing a fault or the initiation of a protective circuit. This test is meant to simulate sporadic dynamic bass peaks in music and movies. During the test, the top of the H95 was quite warm to the touch, but not hot enough to induce pain.

Our primary headphone measurements revealed the following using the analog line-level input (unless specified, assume volume at 99, 1kHz sinewave, 2Vrms output, 300-ohm loading, 10Hz to 90kHz bandwidth):

Frequency response (8-ohm loading, line-level input)

In our measured frequency response plots above, the H95 is just under -1dB at 5Hz (better than -0.2dB at 20Hz), and just under -2dB at 100kHz (better than -0.2dB at 20kHz). While Hegel does not include “-dB” points in their specifications, this measurement clearly demonstrates that the H95 is a wide-bandwidth (in audio terms) product, corroborating the manufacturer’s claim of a frequency response of 5Hz to 100kHz. In the chart above and most of the charts below, only a single trace may be visible. This is because the left channel (blue or purple trace) is performing identically to the right channel (red or green trace), and so they perfectly overlap, indicating that the two channels are ideally matched.

Phase response (8-ohm loading, line-level input)

Shown above is the phase response of the H95 from 20Hz to 20kHz, using the analog input measured across the speaker outputs at 10W into 8 ohms. The H95 does not invert polarity. The plot is clean, with only +20 degrees of phase shift at 20kHz, and less than +10 degrees at 20Hz.

Frequency response vs. input type (8-ohm loading, left channel only)

The chart above shows the H95’s frequency response as a function of input type. The blue trace is the same analog input data from the previous chart. The red trace is for a 16-bit/44.1kHz dithered digital input signal from 5Hz to 22kHz, the purple trace is for a 24/96 dithered digital input signal from 5Hz to 48kHz, and finally green is 24/192 from 5Hz to 96kHz. The behavior at low frequencies is the same across input types—better than -0.2dB at 20Hz. The 16/44.1 digital input exhibits a sharp brick-wall-type attenuation near the limit of its theoretical frequency range (22.05kHz), with the “knee” at roughly 18kHz (-1.5dB at 20kHz). The 24/96 digital input also exhibits a sharp brick-wall-type attenuation near the limit of its frequency range (48kHz), with the “knee” at roughly 35kHz. The 24/192 digital input frequency response is nearly identical to the 24/96 plot, despite the extended theoretical range up to 96kHz, which indicates that the H95 may be downsampling 24/192 data to 24/96.

Digital linearity (16/44.1 and 24/96 data)

The chart above shows the results of a linearity test for the coaxial digital input (the optical input performed identically) for both 16/44.1 (blue/red) and 24/96 (purple/green) input data, measured at the line level output. The digital input is swept with a dithered 1kHz input signal from -120dBFS to 0dBFS, and the output is analyzed by the APx555. The ideal response would be a straight flat line at 0dB. Both digital input types performed similarly, approaching the ideal 0dB relative level at -100 dBFS, then yielding perfect results from -96dBFS to 0dBFS. At -120dBFS, both channels at 16/44.1 overshot the ideal output signal amplitude by about 4dB, while the left/right channels at 24/96 undershot by 3 and 1dB respectively.

Impulse response (24/44.1 data)

The chart above shows the impulse response for a looped 24/44.1 test file that moves from digital silence to full 0dBFS (all “1”s) for one sample period, then back to digital silence, measured at the line-level outputs of the H95. We can see that the H95 utilizes a typical sinc function reconstruction filter.

J-Test (coaxial input)

The chart above shows the results of the J-test test for the coaxial digital input measured at the line-level outputs of the H95. The J-Test test was developed by Julian Dunn the 1990s. It is a test signal—specifically, a -3dBFS undithered 12kHz squarewave sampled (in this case) at 48kHz (24 bits). Since even the first odd harmonic (i.e., 36kHz) of the 12kHz squarewave is removed by the bandwidth limitation of the sampling rate, we are left with a 12kHz sinewave. In addition, an undithered 250Hz squarewave at -144dBFS is mixed with the signal. This test file causes the 22 least significant bits to constantly toggle, which produces strong jitter spectral components at the 250Hz rate and its odd harmonics. The test file shows how susceptible the DAC and delivery interface are to jitter, which would manifest as peaks above the noise floor at 500Hz intervals (e.g., 250Hz, 750Hz, 1250Hz, etc). Note that the alternating peaks are in the test file itself, but at levels of -144dBFS and below. The test file can also be used in conjunction with artificially injected sinewave jitter by the Audio Precision, to show how well the DAC rejects jitter.

We see a few peaks in the audioband just below -110dbFS, the rest are below -130dBFS. This is a good J-Test result, and an indication that the H95 DAC has good jitter immunity. When sinewave jitter was injected at 2kHz, which would manifest as sidebands at 10kHz and 14kHz without any jitter rejection, no further peaks, or worsening of existing peaks, were observed up to a significant 600ns of jitter level, beyond which the H95 DAC lost sync with the signal.

J-Test (optical input)

The chart above shows the results of the J-Test test for the optical digital input measured at the line-level outputs of the H95. We see very small peaks in the audioband at -120to 130dBFS. This result and the results of the jitter-rejection test are essentially identical to the coaxial input.

Wideband FFT spectrum of white noise and 19.1kHz sinewave tone (coaxial input)

The chart above shows a fast Fourier transform (FFT) of the H95’s line-level outputs with white noise at -4dBFS (blue/red), plus a 19.1kHz sinewave at 0dBFS, fed to the coaxial digital input, sampled at 16/44.1. The sharp roll-off above 20kHz in the white-noise spectrum shows the implementation of a brick-wall-type reconstruction filter. The aliased image at 24kHz and the reciprocal alias at 15kHz are down below -110 dBrA, or 0.0003%. The second, third, and fourth distortion harmonics (38.2, 57.3, 76.4kHz) of the 19.1kHz tone are much higher in amplitude, lying between -70 and -80dBrA, or 0.003% and 0.001%, respectively.

RMS level vs. frequency vs. load impedance (1W, left channel only)

The plots above show RMS level (relative to 0dBrA, which is 1W into 8ohms or 2.83Vrms) as a function of frequency, for the analog input swept from 5Hz to 100kHz. The blue plot is into an 8-ohm load, the purple is into a 4-ohm load, the pink is an actual speaker (Focal Chora 806, measurements can found here), and the cyan is no load connected. The chart below . . .

. . . is the same but zoomed in to highlight differences. Zoomed in, within the flat portion of the plot, at 1kHz, the deviation from a 4-ohm load to no load is about 0.03dB. At the frequency extremes (20Hz and 20kHz), the spread is larger, between 0.06 to 0.08dB. This is an indication of a high damping factor, or low output impedance The maximum variation in RMS level when a real speaker was used as a load is also very small, deviating by just under 0.04dB within the flat portion of the curve (100Hz to 5kHz), with the lowest RMS level, which would correspond to the lowest impedance point for the load, exhibited at 100Hz, and the RMS highest level, which would correspond to the highest impedance point for the load, at around 4kHz.

THD ratio (unweighted) vs. frequency vs. output power

The chart above shows THD at the H95’s output as a function of frequency (20Hz to 20kHz) for a sinewave stimulus at the analog input. The blue and red plots are for the left and right channels at 1W output into 8 ohms, purple/green at 10W, and pink/orange at the full rated power of 60W. The power was varied using the H95’s volume control. At 1W and 10W, the THD values are quite close, with the 10W figures slightly outperforming the 1W figures, especially at the lowest frequencies (0.01% vs 0.02% for the left channel). The general trend of the lowest THD figures appearing between about 300Hz to 1kHz (around 0.005%), with higher figures at both low and high frequencies (0.025% at 20kHz at worst) was consistent for both the 1W and 10W data. The 60W data showed much higher THD values, with a general trend of lower figures at 20Hz (0.3%), then rising with frequency and peaking at just below 2% near 20kHz.

THD ratio (unweighted) vs. output power at 1kHz into 4 and 8 ohms

The chart above shows THD ratios measured at the output as a function of output power, for an 8-ohm load (blue/red for left/right) and a 4-ohm load (purple/green for left/right). The 4-ohm data shows consistently higher THD values compared to the 8-ohm data (about a 5 dB difference). THD ratio values are fairly steady from 10mW to the “knee” for both 8-ohm and 4-ohm data. The 8-ohm values hover between 0.01% and 0.005%, trending towards lower values with increased output power, while the 4-ohm values are largely between 0.02% and 0.01%. The “knee” in the 8-ohm data occurs around 40W, hitting the 1% THD mark at 69W, just above the rated output of 60W. For the 4-ohm data, the “knee” occurs just above 60W, hitting the 1% THD at 97W.

THD+N ratio (unweighted) vs. output power at 1kHz into 4 and 8 ohms

The chart above shows THD+N ratios measured at the output as a function of output power, for an 8-ohm load (blue/red for left/right channels) and a 4-ohm load (purple/green for left/right). The 4-ohm data shows consistently higher THD values compared to the 8-ohm data (about a 4-5 dB difference). THD+N ratio values start at 0.1/0.15% (8/4 ohms) at 10mW, then down to 0.006/0.008% (8/4 ohms) at 30/40W (8/4ohms).

THD ratio (unweighted) vs. frequency at 8, 4, and 2 ohms (left channel only)

The chart above shows THD ratios measured at the output of the H95 as a function of load (8/4/2 ohms), for a constant input voltage that yielded 5W at the output into 8 ohms (and roughly 10W into 4 ohms, and 20W into 2 ohms). The 8-ohm load is the blue trace, the 4-ohm load the purple trace, and the 2-ohm load the pink trace. We find increasing levels of THD from 8 to 4 to 2 ohms, with about a 5dB difference between 8 and 4 ohms above 1kHz (and very close tracking below 1kHz), and a 40-15dB difference between 4 and 2 ohms from 20Hz to 20kHz, with the largest differences in THD exhibited at lower frequencies, where at 2 ohms and roughly 20W, the H95 exhibits 2% THD at 20Hz.

THD ratio (unweighted) vs. frequency at 8, 4, and 2 ohms (left channel only)

The chart above shows THD ratios measured at the output of the H95 as a function of load (8/4/2 ohms), for a constant input voltage that yielded 5W at the output into 8 ohms (and roughly 10W into 4 ohms, and 20W into 2 ohms). The 8-ohm load is the blue trace, the 4-ohm load the purple trace, and the 2-ohm load the pink trace. We find increasing levels of THD from 8 to 4 to 2 ohms, with about a 5dB difference between 8 and 4 ohms above 1kHz (and very close tracking below 1kHz), and a 40-15dB difference between 4 and 2 ohms from 20Hz to 20kHz, with the largest differences in THD exhibited at lower frequencies, where at 2 ohms and roughly 20W, the H95 exhibits 2% THD at 20Hz.

THD+N (unweighted) vs. output level at 1kHz for line outputs (analog and digital)

The chart above shows THD+N versus output voltage measured at the variable line-level outputs of the H95, with the volume set to 99 (-0.5dB of gain at the line outputs). The blue/red traces (left/right channels) are for the analog input, swept from 0.2mVrms to 2.4Vrms, while the purple/green traces (left/right) are for a dithered 24/96 digital signal fed to the coaxial input swept from -90dBFS to 0dBFS. We find that the analog input outperformed the DAC section by about 8dB up until about 650mVrms at the output. Above this point, THD+N values jump up from 0.002% to 0.06% (left channel) for the analog input, and from 0.005% to the same 0.06% (left channel) for the DAC input. The right channel faired better at about 0.03% at 2Vrms out. This behaviour of more than a 20dB increase in THD+N between 0.6 and 2Vrms is curious; however, in most home-use cases, analog levels from the variable line outputs of the H95 driving most power amplifiers would not exceed 0.6Vrms.

FFT spectrum – 1kHz (digital input, 16/44.1 data at 0dBFS)

Shown above are FFTs for a 1kHz 0dBFS dithered 16/44 input sinewave stimulus at the coaxial digital input, measured at the output across an 8-ohm load at 10W. We see that the second through fifth harmonics, at 2,3,4, and 5kHz, are between -80dBrA, or 0.01%, and -100dBrA, or 0.001%. The highest peak is for the odd third harmonic, just below -80dBrA. The other harmonics within the audio band measured lower, below -100dBrA, down to about -125dBRa, or 0.00006% at 20kHz. Below 1kHz, we see noise artifacts, with the 60Hz peak reaching -85dBrA (left), or 0.006%, and -120dBrA (right), or 0.0001%. The second noise harmonic (120Hz) is at -105dBrA (left), or 0.0006%, and -95dBrA (right), or 0.002%. The higher harmonics of these two peaks can also be seen at lower levels.

FFT spectrum – 1kHz (digital input, 24/96 data at 0dBFS)

Shown above are FFTs for a 1kHz 0dBFS dithered 24/96 input sinewave stimulus at the coaxial digital input, measured at the output across an 8-ohm load at 10W. We see effectively the same result as with the 16/44.1 FFT above.

FFT spectrum – 1kHz (digital input, 16/44.1 data at -90dBFS)

Shown above is an FFT for a 1kHz -90dBFS dithered 16/44.1 input sinewave stimulus at the coaxial digital input, measured at the output across an 8-ohm load at 10W. We see the peaks from noise artifacts on the left side of the main signal dominating (with respect to amplitude) the peaks due to the harmonics of the signal on the right side. The 60Hz peak due to power-supply noise is nearing -80dBrA, or 0.01%, above the signal peak at -90dBrA.

FFT spectrum – 1kHz (digital input, 24/96 data at -90dBFS)

Shown above is an FFT for a 1kHz -90dBFS dithered 24/96 input sinewave stimulus at the coaxial digital input, measured at the output across an 8-ohm load at 10W. We see effectively the same result as with the 16/44.1 FFT above.

FFT spectrum – 50Hz (line-level input)

Shown above is an FFT for a 50Hz input sinewave stimulus at the analog input, measured at the output across an 8-ohm load at 10W. The X axis is zoomed in from 40Hz to 1kHz, so that peaks from noise artifacts can be directly compared against peaks from the harmonics of the signal. The most predominant peak is from the third signal harmonic at 150Hz, reaching -85dBrA, or 0.006%. The second largest peak is from power-supply noise at 120Hz, reaching almost -90dBrA, or 0.003%, followed by the second signal harmonic at 100Hz on the left channel, reaching -95dBrA, or 0.002%.

Intermodulation distortion FFT (18kHz + 19kHz summed stimulus, line-level input)

Above is an FFT of the intermodulation distortion (IMD) products for an 18kHz + 19kHz summed sinewave stimulus tone at the analog input. The input RMS values are set at -6.02dBrA so that, if summed for a mean frequency of 18.5kHz, would yield 10W (0dBrA) into 8 ohms at the output. Here we find that the second-order modulation product (i.e., the difference signal of 1kHz) is just above -100dBrA, or 0.001%, while the third-order modulation products, at 17kHz and 20kHz, are are both just above -90dBrA, or 0.003%.

Intermodulation distortion FFT (18kHz + 19kHz summed stimulus, coaxial digital input, 16/44.1)

Above is an FFT of the intermodulation distortion (IMD) products for an 18kHz + 19kHz summed sinewave stimulus tone at the digital input, where the top plot is for a dithered 16/44.1 digital input. The input digital values are set at -6.02dBFS so that, if summed for a mean frequency of 18.5kHz, would yield 0dBFS at the input and 10W (0dBrA) into 8 ohms at the output. We find that the second-order modulation product (i.e., the difference signal of 1kHz) is at -100dBrA, or 0.001%, so slightly better than the analog input, while the third-order modulation products, at 17kHz and 20kHz, are both at -85dBrA, or 0.008%, so slightly worse than the analog input.

Intermodulation distortion FFT (18kHz + 19kHz summed stimulus, coaxial digital input, 24/96)

Above is an FFT of the intermodulation (IMD) products for an 18kHz + 19kHz summed sinewave stimulus tone at the digital input, where the top plot is for a dithered 24/96 digital input. There is very little difference between the 16/44.1 and 24/96 IMD FFTs, other than a lower noise floor on the 24-bit spectrum below 22kHz.

Square-wave response (10kHz)

Above is the 10kHz squarewave response of the H95 using the analog input, at roughly 10W into 8 ohms. Due to limitations inherent to the Audio Precision APx555 B Series analyzer, this graph should not be used to infer or extrapolate the H95’s slew-rate performance. Rather, it should be seen as a qualitative representation of the H95’s extended bandwidth. An ideal squarewave can be represented as the sum of a sinewave and an infinite series of its odd-order harmonics (e.g., 10kHz + 30kHz + 50kHz + 70kHz . . .). A limited bandwidth will show only the sum of the lower-order harmonics, which may result in noticeable undershoot and/or overshoot. The H95’s reproduction of the 10kHz squarewave is reasonably clean, and although the edges are slightly rounded, they are nonetheless devoid of undershoot and overshoot.

Damping factor vs. frequency (20Hz to 20kHz)

The final chart above is the damping factor of the H95 as a function of frequency. Both channels show a general trend of a higher damping factor in the midrange frequencies, and lower damping factors at both the lowest and highest frequencies. The left channel generally outperformed the right channel, with a peak value around 520 between 300Hz and 2kHz, while the right channel peaked around 400 between 200 and 300Hz. At 20Hz, the damping factor for the left channel measured 220, while for the right channel it was around 300. At 20kHz, the damping factor for the left channel measured 340, for the right channel it was around 190.

Diego Estan
Electronics Measurement Specialist