- Written by Stephen Hornbrook
- Published on 07 July 2014
The Meridian Prime Headphone Amplifier / Prime Power Supply On The Bench
Measurements by David A. Rich, Ph. D.
I did not have a schematic because I did request it in time so I will have to infer what is going on based only on the measurements. The measurements below were made with an Audio Precision APx585 analyzer which has been generously supplied for our use by Audio Precision.
The first thing I noticed about this product was the power supply voltage was only 20VDC. That is half what even a cheap preamp would have coming off the full wave rectifier. After the regulators a typical unit would have a supply of at least +/-15V referenced to ground.
This is a big deal since the 20V supply limits the maximum voltage that the headphone amplifier can supply. Let us assume the regulated rails take us to +/-9.0V referenced to ground and lets assume the amplifier needs 1.0V of headroom with no load. That brings us to +/-8.0V p –p or 5.6VRMS at clipping.
The curve below shows THD vs. measured level at the headphone terminal for loads of 330ohms, 150 ohms, 33 ohms and 16 ohms. The volume control was set to maximum which is provides a gain of 3.3.
At clipping (0.1% THD) the unit supplies 4.5VRMS into 330 ohms. A little less than the back of the envelope calculation above. Other headphone amplifiers we have tested made it past 10VRMS.
At 33 ohms the unit is current limited 3.2VRMS at clipping or 310mW average. At 16 ohms clipping is at 2.1VRMS or 275mW average. 300mW average is a very good result with many headphone amplifiers we have tested producing about twice the power but a couple could supply only 100mW.
What is very impressive is how fast the distortion declines if the unit is not driven to clipping. The graph below is 2V RMS out into 16 ohms. Here the volume control was set for unity gain not the maximum gain used for the graph above. Even at 20kHz for the worst channel the THD is 0.003%.
Keeping the output at a constant 2VRMS but increasing the load to 330 Ohms reduces the distortion at the frequency extremes by a factor of 2.
Making a measurement at the preamp out under the same conditions (100k Ohm load) looks almost identical.
For comparison the Emotiva XSP-1 had about half the distortion at line out. It was not tested at headphone out.
The Meridian uses a passive volume control (metal on carbon) arguing this yields lower distortion. The previously reviewed Benchmark, Funk, and OPPO make the same argument.
The passive volume control can impose an SNR penalty since the amplifiers gain is always in the circuit. A digital controlled analog volume control will apply the gain only when needed.
The SNR relative to 2VRMS was 109dB into a 330 ohm load at headphone out with a brick wall filter at 20kHz. The SNR measured 110dB relative to 2VRMS at the preamp output. The Emotiva preamp was 8.5dB better than this. We did not characterize the headphone amplifiers performance of that product however.
The spectra of 1kHz sign wave at preamp out is shown below at a 2VRMS signal level. Even the 2nd harmonic is at -114dB relative to the test signal. Note the X axis of the graph is in units of dbV. 6dBV has a value of 2VRMS. The 3rd is at -116dB down. Most of the higher harmonics are lower than the noise floor.
Line noise was -121dB down at 60Hz and -126dB at 180Hz (expanded graph below 1kHz not shown). No other spurs were seen above the noise floor. The external power supply made no difference. The headphone output has the same spur levels.
The 19kHz – 20kHz IM spectra is equally impressive as the single tone distortion results. Both test tones are at 1VRMS. Everything in band is below -130dB. This preamp is free of any dynamic distortion process. Even the out of band 2nd harmonics are below -100dB.
USB performance to preamp output.
For a 24/192 sample, the input the digital reconstruction filter limits the frequency response to 40kHz. Only bats need worry about this.
Shown below is the THD from preamp out with the volume control set for 2VRMS at 0dBFS. It is good but does not provide the excellent performance of the preamp with an analog input. The distortion is at 0.005% in the midband but rises to an average 0.02% at 20Hz and 20kHz. At 30kHz the digital reconstruction filter come into play.
The 1kHz spectrum correlates with the 0.05% THD. The 2nd harmonic is -88dB down relative to 2VRMS (6dbV). The other harmonics are below -102dB and do not contribute much to the THD value. Compare this spectrum to the equivalent analog measurement above which has a 2nd harmonic at -114dB down ref 6dBV.
The 19kHz – 20kHz IM spectra is not as extraordinary as the analog preamp (graph shown above) but all inband spurs are -110dB down. The out of band harmonics are at -90dB. Both test tones are at 1VRMS (0dBV). The level needs to be reduced since the two synchronous tones will have the same peak amplitude of a single test tone at 0dbFS.
The rise in the noise floor of the spectra compared to the analog measurement is correlated with the SNR performance. The SNR is down 3dB to 107.5dB. This is 17.5 bits equivalent. This is about a 1.5 bit less than the best products we have tested with a cascade of a DAC and a preamp. The best DAC box we measured with no preamp (the more expensive Auralic Vega) had a noise level of -123dB. Analog preamps are the limitation at the present time. The Auralic Vega used the ESS ES9018 a DAC also used in the Oppo USB DAC / headphone amplifier. I was not supplied with any information on the DAC used in the Meridian.
On the positive side (no pun) the power supply components were below -125dB down from dbFS with the internal power supply. Any other spurs were covered by the noise.
The noise can be seen in the time domain spectrum of a 1kHz sine wave at -90dBFS with a 24bit depth.
The spectrum of this waveform is free of any spurs above the noise floor.
The effects of the noise can be seen in the deviation in level as the signal generator is reduced digital full scale. A 1dB change occurs at -104dBFS and a 3dB change is at-108dBFS. For a different prospective the Yamaha CX-A5000 11.1 channel Pre/Pro at a similar price to this unit with external power supply had a 3dB change at -112dBFS. The Yamaha used the ESS ES 9016 DAC.
For Auralic Vega DAC box (no preamp in the system) had a 3dB change occurring at -124dBFS.
To sum up the measured performance of the analog preamp is excellent with the exception of the SNR which was 4dB below the best. The headphone amplifier is not far behind within its limited voltage swing for higher impedance headphones. If you have a high impedance headphone, that is not efficient, you want to check the Meridian can produce the level you want to achieve. Low impedance headphones will present no problem for this unit.
The DAC adds more noise into the system. The only thing that could be identified between the Meridian and the best available technology in a blind test with closed back headphones would be the noise. I must note the volume level might have to be set to the point that a full scale digital signal would cause hearing damage.
The $1250 external power supply made no difference. It does not increase the 20V DC power rails which represent a significant design compromise in headphone amplifier.