- Written by Chris Heinonen
- Published on 24 March 2014
The Arcam AVR750 Receiver On The Bench
The Arcam AVR750 testing uses an Audio Precision APx585 analyzer. For all tests except for the power amplifier section we use the multichannel RCA pre-outs. This allows us to test the preamp section of the AVR750 and then test the power amp section separately. Dr. David Rich has contributed most of the analysis here since he knows this far better than I do.
One interesting thing that we found is that the maximum output levels at preamp out must be set below unity gain. Using the analog input with a 2VRMS signal the maximum output level the AVR750 can drive without going into protection is 900 mVRMS. Two issues with the power amp prevent use with an external amplifier
- The gain of the power amp is higher than normal. It thus clips sooner then a standard power amp. The typical amp gain is 20. This Arcam amp has a gain of 38. With the higher gain 900mV RMS drives the Arcam to full power (150Watts average). For an amp with a gain of 20 the clipping level would be 1.7VRMS. The gain on the Arcam matches the gain on their external amplifiers.
- Some AVRs can operate when the amplifier is clipping. The Class G amplifier of the Arcam is more sensitive to be driven into sustained clipping since clipping forces the amplifier to stay on its high voltage rails for a much longer percentage of time than a sine wave or music below clipping.
With a digital signal at full scale (0dBFS) another issue is noted. With the level control adjusted to unity gain the signal is at 1.6VRMS. This is too large and the volume control must be reduced to bring the output back to the 900mVRMS maximum level before clipping. Why would Arcam design a product that would clip the amplifier at digital full scale? The answer is they assume that some attenuation will occur in the process of normalizing all the speaker in the room to the same SPL. They assume the floor standing front channels will require more attenuation then the rears owing to the likelihood the front channels are closer to the chair and probably more efficient. If the attenuation needed is -5dB the maximum output is restored to 900mV with the volume control at unity gain, which optimizes the SNR. For our tests the attenuation of all channels was 0 dB.
THD+N with a 1 kHz sine wave at 1.6VRMS measures in at 0.004%. The same value was found for THD alone. The amplifier did not shut down with levels greater than 0.9VRMS because the AP distortion test is performed quickly and the level above power amp clipping for a short period of time.
The filtered (20 Hz - 20 kHz) signal-to-noise ratio checks in at 103dB relative to 0dBFS at 1.6VRMS. SNR is lower here since the DAC contributes noise as well as the analog electronics. How this SNR would change if the volume was reduced so 0dBFS at the preamp was set to the clipping value of power amp (900mV) is unclear. It is dependent which of the two noise sources (DAC or volume control) dominates.
Noise from the power supply comes in at -120 dB or lower. With respect to the tones around 1kHz Nick Clarke, Engineering Director of ARCAM writes:
"These low level idle tones are a function of the jitter reduction. They are close in and at a very low level so much more acceptable than the standard jitter spectrum of HDMI".
The jitter reduction system used by Arcam is proprietary.
The frequency response is down 0.5dB down at 30kHz with the digital sampling rate at 192K samples per second. This rolloff occurs earlier than other products we have tested.
The THD+N Ratio vs Frequency chart shows a surprising rise at 2 kHz. Talking to Arcam this is due to the substitution of an inferior capacitor in this signal path by the Far East subcontractor. This is said to be corrected in current production units.
THD+N vs. Level shows distortion and noise dropping until 900mV at which point it starts to rise. The signal level starts at 5 uV, which is dominated by noise and would not be heard.
The DAC linearity chart shows that from -95 dBFS to 0 dBFS is flat but then the error starts to come into the graph and is -105dBFS is 3dB.
Using a -60dBFS signal at 1 kHz, we see a noise floor that is flat out to 24 kHz and then starts to rise. This is indicative of a Delta Sigma DAC, much like the noise produced by DSD.
The start of the rise is dependent on the DAC design. With some DACs the rise starts at a higher frequency and sometimes cannot be seen in this plot as shown in previous reviews.
Time domain plots at -60dB and -90dB are not shown. The AP did not produce the plots as it was unable to trigger on the waveforms for reasons that are not clear. We have not seen this problem with other products tested.
Looking at analog measurements now, the THD+N is 0.0013% at 900mV and the SNR is 110.8 dB. The SNR value is lower than some other products tested since we are limited to referencing it to 900mV RMS power amp clipping point. The SNR would be 117.7dB if referenced to 2VRMS as we do for other products. The Arcam preamp can source 2VRMS and would achieve this SNR were it not for the power amp shutdown
The 117.7dB value is 6dB better than the analog stages we have tested with an LSI AVR chips. This is discussed in detail at the link below:
Frequency response is flat from 20Hz to 30kHz and down 0.4dB at 80kHz.
Power supply noise is below -135dB except for a 60Hz spur at -125 dB.
Crosstalk is -90dB from 20Hz to 2kHz. It then moves from -90dB at 2kHz to -65dB at 50kHz at 20dB per decade. The flat floor below 2 kHz indicates coupling between the channels is no longer dominated by capacitive coupling.
The figure below is THD + N vs. level (VRMS) at 1kHz. Note this graph the signal starts at 1mV and the graph for the THD + N vs. level in the digital section above starts at starts at 5 uV. The digital graph is this expanded making it look different than the graph below.
The figure below is THD + N vs. level (VRMS) at 10kHz.
The THD + N vs. frequency at 900mV is shown below.
Note: The Arcam limits us to 900mV RMS full scale to prevent power amp clipping so you cannot compare the results to products for which we use a 2VRMS level. At 900mV the distortion will be lower.
The spectral of a 1kHz signal at 900mV is shown below. Note the 2nd harmonic is below -100dB (0.001%) and the 3rd harmonic is below -110dB with all higher order harmonics below -120dB (0.0001%). This spectra looks much better as a result of the use of the Cirrus volume control but again direct comparison is not possible since the other products were tested at 2VRMS.
The noise floor level of the graph, related to but not the SNR value from 20Hz – 20kHz, is lower than what is seen in reviews we have done with the AVR LSI.
The spectral of a 1kHz 19kHz – 20kHz IM is shown below. Each input is at 450mVRMS so the peak to peak amplitude of the composite waveform does not clip the amp. All in-band spectra are below -120db and the out of band 2nd harmonics are below -110dB. Again this spectra looks much better than most other AVRs tested as a result of the use of the Cirrus volume control but one most note that direct comparison is not possible since the other products were tested at 1VRMS for each input level.
Running two channels into 8 ohms the AVR750 manages 139 watts at 0.1% THD+N and 146 watts at 1.0% THD+N at 1kHz. At 4 ohms we see 207 watts and 221 watts. Both are better than the specs from Arcam, though those use 0.2% THD+N as a target.
Looking at THD+N vs Watts, we see a steep rise at 50 watts to 0.01% from 0.0015%. This is the rail switching from Class A to Class G. From there distortion is flat until we hit out maximum power output at 140 watts and it rapidly increases in distortion. Go to Page 5: Conclusions