It has been several weeks since I first evaluated
the highly acclaimed NuForce Reference 8 amplifier. The results from my measurements and listening tests showed that much of its appeal could largely be attributed to marketing rather than the actual performance it delivered. I didnít shun away from publishing my findings which resulted in quite a bit of controversy and discussion surrounding these amplifiers. NuForce themselves joined this discussion by posting their feedback in a number of public forums. They didnít question my results, but commented that their current model, the Reference 9, was leaps and bounds above the now obsolete Reference 8 I tested. When asked to supply me with a Reference 9 so I could see whether I could substantiate their claims they kindly denied this request, as my website did not fit the profile to be eligible for receiving evaluation samples.
Nevertheless I managed to secure two Reference 9 amplifiers which had been updated to the latest revision, 9.02, just recently. Because I still had the Reference 8s in my possession I decided to simply compare them directly in order to evaluate whether the Reference 9 is indeed leaps and bounds above the Reference 8. As with the initial evaluation I used an Audio Precision measurement system to measure the performance of the amplifiers. But naturally I started off with a number of listening sessions comparing the Reference 8 to the Reference 9 and a number of other amplifiers.
The beefier power supply as found in the NuForce Reference 9.02
The listening sessions showed that with the Reference 9 NuForce has indeed improved upon the Reference 8; it clearly is a better amplifier, with better control, better imaging but above all a better tonal balance and clarity. When compared to other amplifiers in about the same price range it however fails to impress, it isnít a showstopper, but more or less a mediocre performer. One of the drawbacks is the somewhat small and narrow stereo image, something which I first reported on with the Reference 8, and which is still present in the Reference 9. Overall the Reference 9 was quite a bit better behaved than the Reference 8 nonetheless.
Much of the RFI issues as reported on in the Reference 8 article have now been resolved, but thereís quite some RFI being generated still that again affected my tube preamplifier, although to a lesser extent. What remains are the noises that are being generated when switching the amplifiers on and off; youíd expect an amplifier with such a price tag to be silent during power-up and shut-down, unfortunately it isnít. Obviously I was curious to find what changes have been made to these amplifiers so I again took the covers off and peeked inside at the amplifier internals.
The common-mode inductor and capacitor across the speaker terminals
Whatís immediately obvious is the fact that a much beefier power supply is now powering these amplifiers, which translates into a higher power rating. The amplifier printed circuit boards themselves however look to have undergone a few component changes, but overall look identical to those in the Reference 8. Upon closer inspection the TL082 dual-opamp thatís part of the modulator in the Reference 8 is now replaced by the much more capable NE5532 and the dead-time settings for the MOSFET H-bridge have been modified. It is common knowledge that with class-D amplifiers distortion performance is better with shorter dead times, and NuForce have now too followed on this insight. Furthermore a common-mode coil has been added in series with the speaker terminals in an attempt to cut down on RF emissions.
For reasons unknown however, NuForce has also opted to mount a 470nF capacitor across the speaker terminals. A typical common-mode filter consists of a common-mode coil with a capacitor mounted from each terminal to ground. A common mode coil with the capacitor mounted differentially is the same as a common-mode coil with the extra capacitor mounted on the input side, which doesnít help much to reduce RFI because the problem with the NuForce amplifier is common mode. In order to create a low-pass filter with the common-mode coil it needs to have a capacitor from each speaker terminal to the chassis. I donít have a spectrum analyser to quantify this, but the overall impression is that the RFI level of the Reference 9 canít be more than 10dB lower than on the Reference 8. Looking at the switching waveform it still shows an enormous amount of ringing. Simply getting rid of this ringing would improve RFI much more than a half-baked common-mode filter. Why NuForce still hasnít figured out how to deal with this issue is unclear, because the solution is common knowledge.