Digital surround receivers are by far the most complicated products we test.
Not only do they have two primary modes of operation — two-channel stereo
and multichannel surround sound — both using their digital inputs, but
today they may also be called on to handle multichannel high-resolution analog
signals from a DVD-Audio or Super Audio CD player. With all that going on, plus
amplification, radio tuning, and signal switching, receivers generate more lab
data of different types than any other products. Unfortunately, we don’t
have the space to print a full set of data in all reports.
Since most of the audio-performance tests we run on a receiver duplicate the identically named tests for a full-featured DVD player, they were already covered in our first “Behind the Numbers” article, on DVD players (July/ August 2001; on the Web at soundandvisionmag.com, in the Archives under “Buying Tips”). If you want the gory details on what excess noise and linearity error are and how theoretical considerations limit measured distortion and noise in the digital domain, see that article. The only difference when we do these tests on a receiver is that we look at speaker-level signals instead of line-level signals and use a reference output level of 1 watt into 8 ohms instead of 200 millivolts into 10 kilohms. For many of the tests the digital test signal is identical.
This equivalence between our DVD player and receiver audio tests is intentional to allow direct comparison of performance at two ends of a connection. You can tell, for example, whether the noise level through a receiver’s digital input is worse (higher) than that from the analog output of a CD player. If the player’s analog output is quieter, you’ll get better sound when playing CDs by feeding the receiver from that instead of the digital output.
Measuring Power
Many readers are most interested in a receiver’s maximum output power (“Output
at clipping”), which we test for both Dolby Digital and digital stereo operation.
That used to matter more than it does now, because many home theater receivers
have more power than they’ll ever be asked to deliver. However, following long
tradition in the audio field, we list these results first for each operating
mode.
But as with many of our tests, how we obtain power measurements differs in some important respects from industry standards and tradition. For example, we define the clipping point (where an amplifier starts to go into gross overload) as the output level where total harmonic distortion plus noise (THD+N) exceeds 0.3%. Manufacturers may use different distortion levels for their power specs, either higher or lower. Ours was chosen because 0.3% represents THD+N that’s 50 dB lower than a receiver’s output, so it would generally be inaudible given the high sound levels a receiver generates at its clipping point.
Sample
Lab Box
Watts
to dbW Table
Figure
1: Power Ratings
Figure
2: Bass Management
The dBW Difference
When it comes to the maximum output power figure itself, most measurements in
spec sheets and test reports, ours included, are given in watts, the standard
electrical unit of power. Unfortunately, watts are almost useless for this purpose
because they’re linear while our hearing is logarithmic. For example, a receiver
that delivers 15 watts at clipping is audibly more powerful (by 1.76 dB) than
one delivering only 10 watts, but the 0.2-dB difference in maximum output between
receivers delivering 105 and 100 watts is inaudible.
If the power figures are converted from watts to dBW, however, you can easily tell whether a difference in output at clipping is audibly significant, making meaningful power comparisons possible. Standing for “decibels referred to a 1-watt output,” a dBW figure indicates how much louder a receiver can play than an amplifier with a 1-watt drive level. For example, a receiver that measures 17.25 dBW (53 watts) at its clipping level can generate sounds up to 17.25 dB louder than a 1-watt amplifier.
Any power results that differ by more than 1.5 to 2 dBW are significant, regardless of the wattage figures involved. The table on the next page gives dBW equivalents for common wattage ratings, calculated by multiplying the common logarithm of the wattage by ten: dBW = 10 ¥ log (W). In our test reports, we round every wattage figure to the nearest 0.25 dBW since smaller gradations are sonically insignificant in product comparisons.
Stating power capabilities in dBW is useful in another way: you can simply add dBW power figures to a speaker’s sensitivity rating to get an approximate value for the maximum undistorted sound level that can be generated by a given receiver (or amplifier) with that speaker. With a speaker having typical sensitivity — say, 90 dB sound-pressure level (SPL) with a 1-watt input — a 17.25-dBW receiver will generate up to 107.25 dB SPL at 1 meter, which is very loud for a home theater or listening room (the SPL at the main listening positions may be somewhat lower since most people sit more than a meter away from their speakers). You can actually trade off receiver power for speaker sensitivity — as power in dBW goes down, speaker sensitivity should go up by at least an equal number of dB SPL. A very careful shopper can save money by buying a less powerful receiver and still end up with a system that can play very loud if the speaker sensitivities are high.
Power Testing
Here’s the test procedure we follow for our receiver (and amplifier) power measurements.
First, the receiver is connected to precision, heavy-duty load resistors that
function like ideal speaker loads. For most of our tests the load resistance
is 8 ohms, with one test (stereo output) also using 4-ohm loads. Theoretically,
the output wattage at clipping should double when 4-ohm loads are used instead
of 8 ohms, but usually it doesn’t quite get there.
In all cases, as an input test signal we use a 1-kHz pure tone (a sine wave) that increases in amplitude from a low starting level by around 0.5 dB per second. (Dialogue normalization, which keeps one Dolby Digital source from sounding a lot louder or softer than another, is set at a standard –27 dB, as it is on nearly all movie DVDs.) To make sure the receiver will overload, we crank up the volume control quite a bit — as much as 5 or 6 dB above the reference volume setting used to make noise measurements. As the input signal rises, we watch for the point at which THD+N shoots up in a graph of output level and distortion vs. time.
Since the AC line voltage can influence power measurements, and it tends to dip by several volts as the maximum output from a high-power amplifier is reached, we monitor the line voltage to the receiver during these tests to prevent the equivalent of a brownout. Using a variac (a large, variable, line-voltage transformer), we adjust the line voltage manually so that it stays within 1% of 120 volts (118.8 to 121.2 volts rms) when the receiver’s maximum output is reached.
In our stereo maximum output tests, we drive both channels (left/right front) simultaneously with the receiver switched to stereo operation. Some receivers internally rewire their power-supply circuits for multichannel operation, so testing only the front L/R channels while the receiver is switched for full 5.1-channel operation can give lower results than in true stereo operation. For our 1-channel Dolby Digital power test, we use the center (front) channel, as that is the one usually driven hardest in a multichannel soundtrack mix.
In our tests with five or six channels playing our reference-level (–20-dBFS) signal, we take special pains — using the receiver’s channel-balance controls — to make sure that the output from every channel is as close to 1 watt as possible. Imbalances here, such as when some channels are 3 dB below the others, can considerably relieve the receiver’s burden at full output. If any imbalances remain, we monitor the odd-man-out channel — the one that’s slightly higher than the rest — as it will be the first to overload. These all-channels tests are tough, and it’s not unusual to see large differences between our results and a receiver’s wattage specs (which are usually not made with all channels driven).
Over the years we’ve also seen some receivers whose one-channel and all-channels power tests gave dramatically different results, like 100 watts and “only” 40 watts, respectively. Here is another case where the dBW unit clarifies things. Our examples are equivalent to 20 and 16 dBW, respectively, a 4-dBW difference that would be audible and significant taken on its own. But add those dBW figures to a typical speaker sensitivity rating of 90 dB, and you get respective 1-meter SPLs of 110 dB with one channel driven and 106 dB per speaker with all channels driven.
Since even one speaker playing at 106 dB SPL would be quite loud, much less all five of them at once, it’s clear that our example’s startling 60-watt difference in power levels, or even the 4-dBW difference, is less significant in actual listening conditions — you might never turn the receiver’s volume control high enough to run out of juice! On the other hand, if the difference between the one-channel and all-channels tests is large (more than 5 dBW, say) and the overall levels in dBW are low with all channels driven (around 15 dBW, say, or lower), you may have trouble getting distortion-free reproduction of movie soundtracks at satisfyingly loud volumes. (Note that our procedures are optimized for receivers whose front and surround channels all have the same power rating, which is typical though not universal.)
Bass Management
One of the most important tasks a multichannel receiver is called on to perform
is bass management, which has two parts: 1) using high-pass filters to remove
deep-bass frequencies from the channels feeding any speakers that can’t reproduce
deep bass, and 2) mixing the extracted deep-bass information with the bass-only
low-frequency-effects, or LFE, channel — if there is one in the recording —
so the combination can be fed to the receiver’s subwoofer output. (The LFE channel
is the “.1” in a 5.1-channel system.)
To test a receiver’s bass management, we take frequency-response measurements of the main-channel and subwoofer outputs when the receiver is set for operation with all “small” main speakers, which on many recent models is called “normal.” The results, which show up in a graph like Figure 2, may help you choose a receiver whose bass-management system is the best match for your speakers. Or, coming at it from the other direction, you may be able to choose speakers that best match the requirements of your receiver.
The most important requirement a receiver puts on a speaker system (or vice versa) stems from the crossover between the main speakers and the subwoofer. In most receivers the bass crossover frequency is fixed at 80 or 100 Hz (the latter is shown in Figure 2). In order to get a smooth blend, the main speakers used with any receiver we test should have reasonably flat response down to the crossover frequency cited in our test report. The higher the crossover frequency (say, 100 Hz and above), the easier it will be to match the receiver with home theater systems that have small “satellite” speakers. If the crossover frequency is 80 Hz or lower, the main speakers in the system should be largish satellites or small tower models that have decent-size woofers.
A bad match here — specifically, very small satellite speakers and a low crossover frequency — can create an audible gap in the important midbass frequency range between where the main speakers roll off and the subwoofer kicks in. Receivers with variable crossover frequencies are considerably more versatile in this respect than fixed-frequency models.
The slopes of the high-pass and low-pass (subwoofer-output) filters are determined from the slanting portions of the response graph curves. Knowing whether the rolloff is steep or gradual is of less practical value than knowing the crossover frequency, but these numbers may prove useful in ultra-critical applications. In theory — and when it comes to playing low frequencies with real speakers in real rooms, theoretical considerations must be taken with mountains of salt — an ideal crossover network will have unequal high-pass and low-pass slopes. That’s because the main speakers will naturally roll off at the low end of their frequency range, usually at 12 dB per octave below their –3-dB point, and this rate must be added to the high-pass filter’s rolloff rate, which is often 12 dB per octave also. The result in this case (24 dB per octave) should equal the low-pass rolloff rate at the subwoofer output.
Besides the speaker responses at the crossover frequency, the relative phases of the main-channel and subwoofer outputs have to match for the blend to be as smooth as possible. All these conditions are fulfilled only when the subwoofer is down 6 dB at the crossover frequency and rolling off at 24 dB per octave, and the high-pass filters and main-channel speakers are both down 3 dB at the crossover point and rolling off at 12 dB per octave below that (this is the situation, minus a speaker-response curve, shown in Figure 2). These are also, by no accident, the basic requirements for the bass-management filtering system used in THX-certified products. Lately we’ve seen such behavior in non-THX-certified receivers too — a welcome development.
If a bass-management system has to be used at all — and it must be with most subwoofer/satellite setups — then it should operate the same way with all inputs, analog or digital, stereo or multichannel, in order to avoid changes in bass balance when you change the program source. We now perform some quick checks to see whether this is the case.
Unfortunately, the bass management in most (if not all) receivers currently on the market does not act at all on their multichannel analog inputs. As a result, depending on your speaker setup, a source heard through those inputs (say, a DVD-Audio disc) could end up sounding very different from one whose audio is connected digitally (say, a DVD-Video disc). In addition, in some receivers the bass-management system turns itself off when you switch from multichannel Dolby Digital or DTS playback to stereo playback from a digital source, as would happen if you played a CD in your DVD player right after watching a multichannel movie. Some models also provide no bass management for their analog stereo inputs unless one of the surround sound modes is activated, like Dolby Pro Logic or a DSP (digital signal processing) ambience mode. When a receiver has such quirks or deficiencies, we mention it either in the main text or in the tech notes accompanying our lab data.
In any case, even with a properly operating bass-management system in a receiver, you still need to properly set up the receiver and speaker system, preferably using a good test disc and a sound-level meter. Proper setup can compensate for many of the bass-management anomalies that might turn up in our lab tests.
The subwoofer output voltage and distortion measurements are both worst-case tests in which maximum-level low-frequency signals from each of the main channels as well as the LFE channel are added together and fed to the subwoofer output. The sub outputs of most receivers either grossly overload (meaning distortion is well above 1%) or are limited to some level short of what is needed to preserve full dynamic range. If an overload can be eliminated by adjusting the receiver’s subwoofer-output level-trim control, then we give the proper setting. (Not all receivers allow this — some subwoofer outputs always go into clipping with worst-case signals.)
If you set up a receiver we’ve tested with the subwoofer trim at the level we used or lower, the subwoofer output will not overload provided the main-channel level-trim controls are also all at their 0 or default settings or lower (which is quite often the case). In practice, worst-case signals rarely occur with program material, and the distortion generated by even an overloaded subwoofer output is usually filtered out by the subwoofer’s own circuits. In any case, most of the time it’ll be masked by the (at this point, very loud) program material from the main channels.
Multichannel-Input Tests
Since a receiver’s multichannel analog input is, for the time being, the
only path through which most of us are going to hear the multichannel music
from a DVD-Audio or Super Audio CD player, we’ve recently instituted a short
series of tests for it. They are designed mainly to see if the receiver degrades
the incoming high-resolution signals.
The most important number we run here is the A-weighted noise level — which, by a considerable margin, should have the lowest value of any A-weighted receiver noise measurement we publish, –90 dB or lower. With such performance, the background noise level in DVD-Audio or SACD playback will be thoroughly dominated by the noise of the player or the recorded program.
We also include an extended-range frequency-response measurement (up to 96 kHz) for those who think — mistakenly — that flat response far above 20 kHz is audibly important. Unless we find something really peculiar here (which would be pointed out in the notes), you can consider the multichannel-input response measurements irrelevant to a buying decision. However, if and when receiver manufacturers start applying bass-management processing to their multichannel analog inputs, you can be sure that we will add some response measurements of that as well.
Due to the introduction of these multichannel-input tests, and the space they’ll take up in our lab boxes, we will have to discontinue a few of the measurements we’ve run in the past, including channel balance and tone-control response. Not to worry, though: for a variety of reasons, these measurements are not particularly relevant today for making a buying decision, for using the receiver you choose, or for selecting source components and speakers that will perform well with it. The lab results we do publish for receivers represent our best attempt to put the audible performance of current receivers to the test.