Greg Smith
Today's question is simple enough: how important is the quality of your power amplifier to the sound of your audio system? The typical answers you'll get, however, are neither simple nor definitive. Some will tell you that all amplifiers sound pretty much (if not exactly) the same. That's a somewhat deluded viewpoint as far as I'm concerned. I'll state flat out that most of the amplifiers I try out, even similarly priced and rated ones, sound a bit different, and in some cases the differences are quite dramatic. Those on the more objective side of this argument say that it's probably because I'm not matching levels correctly, or point out other limitations in my testing methodology. What a bunch of nonsense. This part is simple--I have some well- regarded amplifiers here that, with certain speakers, just plain don't sound right no matter what volume they are played at. I can pick the problems out, and so can everybody else I ask to listen. And these people are placebo-free, mind you, because they aren't paying a bit of attention to the endless tweaking I'm doing when I'm not asking them to listen. While I'd like to live in the nice orderly world where amplifier sound can be measured easily and objectively, my ears won't let me get comfortable there.
This is especially true in the realm of budget equipment. While I might be inclined to believe it when I'm told that two $5000 solid-state amplifiers are indistinguishable, there is most certainly a vast difference in sound quality between components in the $500 range. You're not going to find an amplifier that costs that much and does everything right. This is fully in line with one of the basic principles behind my columns: you can't judge audio equipment as good or bad in a vacuum. You have to ask what it's good or bad for, and find the context it might be an appropriate choice for. The piece I'm looking at this month, the Rotel RB-956AX, is a perfect example. While it's a great performer is some respects, its limitations are obvious. It might be the perfect $500 six- channel amplifier for you; then again, it might be a horrible choice. Does that make it "good" or "bad"? That's quite the loaded question. Appropriately enough, loads will be our first topic.
Most people get introduced to the concept of resistance in early science class. Any item that conducts current provides a path for the electrons that make up that current to flow through. The measurement for how difficult that path is to travel gets called the resistance. Along with it, you probably got introduced to the idea of voltage and how you measure the total power involved as wattage.
What usually gets skipped over in the earlier introductions is the fact that these simple methods of thinking of things are only applicable in that form to direct current (DC). Audio applications involve alternating current (AC), and the simple resistance concept falls apart. See, when you're dealing with AC, the current moving along doesn't just have a simple amplitude, it has some frequencies involved as well. So while you can measure DC current easily, i.e. "that's 1A into 2 ohms which makes 2V and 2W", measurements of AC are far more complicated. The simplest AC signal, the sine wave, has an amplitude and frequency; i.e. "that's a 1V signal at 1000Hz". Converting these measurements into power or other relations involves some more concepts, like the idea of root mean square (RMS) power vs. peak to peak. This isn't totally relevant to the discussion here; you can grab any good electronics book for an introduction to the topic. The reason I bring it up is because I'd like to discuss the idea of impedance.
When you have a simple resistor, it has a simple resistance. You can read the label: "10 ohms". Now, that's a bit of a simplification; as the resistor heats, its value changes some, and there is some tolerance involved for the measurement, but under most circumstances the resistance is fixed close to the value given. When dealing with AC signals, we introduce some other components, like capacitors and inductors. These also resist the current, but not in such a simple manner. For example, if you take a signal at 0Hz (which is a DC signal), the capacitor will have a nearly infinite resistance, while the inductor will have almost no resistance. Use an AC signal like 1Khz and you have to do some math to figure out what the resistance is at that frequency for each type of component. This is the idea behind inductance: while similar to resistance, it varies based on what frequency you are trying to pass through. Like most of the things here, inductance is actually more complicated than that; you can actually split the measured inductance into purely resistive and reactive parts (the reactive part is that from inductive or capacitive types of loads), but for the discussion here we'll simplify by treating it as a single measurement.
Since audio circuits are made out of inductors and capacitors (along with even more difficult to model parts like transistors and tubes), they all have their own inductance profile that goes along with them. Speaker drivers are even more of a mess. The mechanical parts of the driver have certain physical properties. For, say, a normal woofer, you have a cone with a certain weight, a suspension with a certain amount of spring to it, and other such parts. When you apply an electrical force to the voice coil, the magnet makes each of these parts move around. How much they move depends on the frequency of the signal you applied. You can also measure the inductance of this whole system by applying test signals of various frequencies and seeing how much current is required to maintain a particular voltage.
This inductance curve works in addition to the driver's frequency response, which is what most people consider when looking at speaker measurements. Let's check out a typical woofer. I've spent a lot of time working with the Vifa M17SG-09-08, a traditional 6.5" model from a well respected company. The nominal impedance of this driver is listed as being 8 ohms; nominal in this case isn't quite the average value, but it's close to being the most commonly encountered one. Now, this driver has a frequency response going from approximately 34Hz to 5000Khz; typical usage is far less than that, but that's another story. Let's look at what the impedance curve does for this driver; here's a very much simplified version I just drew to approximate what Vifa's graph shows:
The frequency
scale is a bit erratically represented, and you can obviously
tell I'm no artist, but the point is to look at the way the curve
goes. The impedance is approximately flat at slightly less than
the nominal 8 ohms through most of the midrange (200hz-1Khz). In
the bass frequencies, there is a huge hump centered around the
driver's resonant frequency, and the impedance here peaks at
around 16 ohms. Above 1Khz, the impedance also climbs up, so that
by 20Khz it is over 16 ohms.
Why should you care about what driver impedances do? Lots of reasons. First off, you can see that the impedance of a individual driver can vary wildly throughout its frequency range. Usually the designer compensates for this (along with frequency response problems) by making the crossover more complicated. The end result is a flatter load and flatter response, but it actually makes for a more difficult load in some respects because all the additional capacitors and inductors that are involved in that flattening add their own impedance complications. By the time the crossover is done, you'll often find that instead of a big, broad hump, the bass impedance actually becomes the most difficult part of the curve, dropping lower than the midrange or treble ever do. The second reason you should be concerned about what the driver impedance does is an issue called output impedance.
All of the interfaces between your audio components have impedance matching concerns. The interaction of the amplifier output to the loudspeakers just happens to be the most complicated of them. Any amplifier you use will have its own output impedance, which appears in series with the loudspeaker load. The amplifier impedance at any frequency effectively gets added to the loudspeaker impedance.
Since output impedances tend to be small, it's easier to compare them by looking at their reciprocal, which is referred to as the damping factor. Let's consider some examples. Say we have an amplifier with an output impedance of 1/4 ohm, which translates to a damping factor of 4. At midrange frequencies, that amplifier driving our example Vifa driver will have a total impedance of 8.25 ohms. At the bass peak, we're looking at a total impedance of 16.25 ohms. If the speaker load were flat, every frequency would be equally effected by the output impedance, and it wouldn't be an issue. This isn't the case here. Working through the math, the midrange suffers a 0.13dB loss because of the output impedance, while the bass loses 0.06dB. That's only a 0.07dB difference, which you're certainly not going to find audible.
If you start having some more bizarre combinations, though, you can easily get to where there's an audible loss. If we were dealing with an amplifier output impedance of 1 ohm, and the speaker load ranged from 4 ohms to 16 ohms, there would be a 0.71dB difference, which you couldn't so easily dismiss as insignificant. Making all of this worse is that, just like the speaker impedance, the amplifier output impedance varies with frequency, so there is an even greater potential for mismatch when both loads are up for grabs.
In any case, the lower the output impedance, the less the interaction with the load. Something with a damping factor of 500 is essentially not interacting with the loudspeaker. Even a damping factor as low as 100 is difficult to even measure, much less hear. Typically, tube amplifiers have damping factors of 10 or less, which is in the territory where the interaction is likely to be audible. The subjective effect of all this is usually less controlled bass, and unpredictable interaction in areas where the crossover is complex (like the crossover point between drivers, which often occurs right in the critical portion of the midrange). Now, a high damping factor is not worth chasing at all costs; one of the more effective ways of raising it is to increase the amount of negative feedback in a circuit, which improves the measurements but can degrade the sound quality. It's a complicated balancing act to get all this right.
Output impedance interaction and feedback are just two of things that can make amplifier/speaker combinations sound different. It's impossible to quantify how much either factor is going to degrade the sound without having the speaker involved present; these quantities just don't matter without a reference load attached. The more difficult the speaker load, the more electrical interaction factors like these become important. This effects final products in various ways, even beyond the amplifier itself. Speakers designed to sound their best with vacuum tube amplification don't quite sound the same with solid-state amps, and vice- versa; the load interaction is just too different for the results to be the same. Cables get dragged into this discussion as well; after all, their frequency-dependent impedance adds to the amplifier's output impedance, which means that cable difference audibility depends not only on amplifier characteristics but speaker differences, too.
We'll get back into impedance as a factor later. Right now I want to introduce the star of this month's review.
$500. Six channel power amplifier, 40W/ch. Adjacent channels bridgeable to 100W/ch.
I continue to be impressed with Rotel's equipment in most areas. The budget audiophile would be hard pressed to get better equipment for their dollar than what Rotel offers. The RB-956AX offers excellent sound under many conditions, and is one of the most flexible units available to boot. You get about 250W of total power, configurable as anywhere from 3 to 6 channels offering between 40W and 100W of power each. The power output is grouped into three stereo pairs. Each of those pairs can be bridged into a mono unit. This is how you end up with so many different channel possibilities. If all three are run in stereo, that's 6 channels. Switch one to mono, and you've got 5 (one 100W mono and two 40W/ch stereo pairs). Two in mono and one in stereo makes 4 channels. Run all three bridged, and you've three 100W units.
This makes this unit very configurable for both home-theater and stereo installations. You could use all six channels in a 5.1 setup (5 channels and a subwoofer). Get a powered subwoofer, and you can bridge a channel to get extra power for your center channel. Forgetting about home theater, I've used the amp in 4 channel mode to drive a pair of loudspeakers and a pair of stereo subwoofers. And it's an affordable 100W/ch stereo unit (you don't have to use all the channels), with future expansion possibilities.
Considering that most high-end amplifier front panels consist of at most a power switch, this Rotel is fairly complicated. Along with the switch, there is a configuration area for each of the three channel pairs. Indicator lights show whether the channel is active in stereo or bridged mono mode. More importantly, there is a level adjustment for the three outputs. This can be highly useful for many reasons. Some preamplifiers (especially passive ones) don't have a high enough output level to drive the amplifier to full output; you can adjust the input level to correct for this. In bridged mode, you can use the level as a balance control for the mono channels. And if you're using surround, you can use these as a fine adjustment mechanism. Many surround decoders have channel balance controls where the steps are too large, or where the range isn't big enough. For example, one of the Harmon-Kardon units I use varies the channel balance by a minimum of 2dB. You can fine adjust with the analog level potentiometer on the unit. If you've got a huge sensitivity difference between your front and rear speakers, your decoder might not have enough of a correction range available; you could easily fix that with this unit, too.
I found a somewhat different use for these adjustments. One of my main speakers (the Klipsch Forte II) has an extremely high sensitivity at 99dB. With normal power amplifiers, this means that I barely turn up the preamplifier volume control to get to very high levels. The Proton D1200 unit I was using with them before I switched to this Rotel would get to deafening levels with the Rotel RSP-960AX preamp volume control at a mere 11 o'clock. With the new Rotel power amp, I turned the amp level adjustment down a bit. This meant I had to turn the preamp up higher to achieve the same volume, making my useful range of motion for the same range extended to about 3 o'clock on the knob. This not only made the volume knob more sensitive and therefore easier to set, it actually made the sound better. This is easily enough explained; let's postulate some numbers. Say the Proton amplifier I started with would output 50w/ch (which is ear-splitting into these speakers) with an input level of 1V. Using the Rotel amp instead, I could reset the level adjustment so that I instead needed an input of 2V to get 50w/ch. With all other things being equal, a 2V signal is less likely to be effected by background noise. If most the noise is constant no matter what the volume (as it often is), that's a 3dB increase in the S/N ratio. And a higher voltage is simply more likely to survive the transmission from preamp to amplifier without loss. This is why the power company has those huge transformers to increase the voltage across power cables--they lose less juice to the wire resistance that way. Experimenting with the interface between your preamp and power amp is a free tweak to try if you have amplifier level adjustments, and it is worth trying.
There are some things to be worried about concerning the level potentiometers. They are another volume control sitting in the signal path, and judging from their small plastic feel these aren't the best sound controls of their type available (although entirely typical at this price point). Also, it's been my experience that small, infrequently used volume potentiometers are the first thing to wear out and get noisy as equipment like this ages, which is something to consider. The rear panel is filled with inputs, binding posts, and bridging switches. You get your standard 5-way binding posts and RCA jacks here; not cheap ones, but not the real high-budget models either. Functional, but not fancy, which is exactly what I'm willing to pay for. Looking inside the case shows a very large power transformer and output transistors with sizable heat sinks for each channel, distributed all around the inside. The power supply capacitor section doesn't have the huge stiffening caps you usually find on heftier amplifiers (at 23 pounds, this is a middleweight chunk of metal), which may contribute to the limitations on the output capabilities we'll get into in a bit.
After you stop looking around, the real question is how the amplifier sounds. Despite my subjective-friendly comments about amplifiers sounding different, I will freely admit that this is a silly question. You really can't describe what an amplifier sounds like without qualifying it with what speakers you used. Do this with enough speakers and you can generalize a bit, but that still doesn't mean you can accurately predict in all cases how it will perform with speakers you haven't actually tried. I can tell you how the amplifier worked for me, but that doesn't mean it will work quite that way for you, too, if you are using different speakers.
With all the speakers I tried, the Rotel sounded very accurate, while leaning just a bit toward the laid-back side of totally neutral. This is typical of products from them, and it's almost always a blessing with budget products. The treble was well-extended without being too bright. In fact, it was so less bright than the budget solid-state amplifiers I normally use (a whole stable from inexpensive audiophile staples like NAD and Adcom) that I found myself pulling the grills off of some speakers that had always been too in- your-face before to do so with other amps. Grill cloth almost always tones down the treble a bit. Overall, from midrange to treble, this is as good of an amplifier as I've ever found in the $500 category. Note that I don't include the bass in that statement; there's a good reason.
I didn't bring up the whole impedance issue in the introduction for nothing. The RB-956 has a very low output impedance, yielding a damping factor of 180. This essentially means the interaction with the load is minimal. The real potential problem with this amplifier is how it handles low impedance loads, especially when bridged.
Solid-state amplifiers usually act as a voltage source. This means that they amplify the input signal and attempt to maintain an appropriate voltage to match it across the output terminals, regardless of what the speaker load does. So if the load drops from an impedance of 8 ohms to 4 ohms, the amplifier will (following Ohm's law) double the amount of current to keep the voltage the same. A halved impedance and doubled current means you get twice as much power out of the process. Accordingly, your typical big n' beefy solid- state amp will double the amount of power it can produce whenever you halve the speaker load. If it's 100W/ch into 8 ohms, it will do 200W/ch into 4 ohms and 400W/ch into 2 ohms. This Rotel amp is not such a beefy amp. There isn't even a specification given anywhere for the amplifier output into 4 ohms, and 2 ohms is not even mentioned as a possibility.
Let's look at what the specs do say. The amplifier is listed as being 40W/channel into 8 ohms. And it is only recommended for use with speaker impedance loads from 4-16 ohms when operated in stereo mode. In bridged mode, the specs say to use it with a minimum of 8 ohm speaker loads. Here's why. The way most bridging circuits work involves driving the speaker involved in parallel between the two channels. This essentially halves the impedance load that each channel sees. So an 8 ohm speaker presented to the amplifier in bridged mode presents a load like a pair of 4 ohm speakers. This is one of the reasons the bridged power you'll see rated on amplifiers is often more than twice the single channel power--since solid state amps increase their power into lower impedance loads, and bridging decreases the load, you get more power. This Rotel specs as being 100W/ch into an 8 ohm load when bridged. Not a big improvement over the 80W/ch you'd get if you simply doubled the power, which leads me to believe that the amplifier does not increase its power significantly into lower impedance loads.
This wouldn't be such a big deal if it weren't so audible. I started digging into these specs because something just wasn't right about the sound with some of the speakers I tried, and I wanted to know why. Here's what I noticed: if the impedance load on any channel dropped below about 4 ohms, the amplifier just didn't control the output properly. Typically, speakers are rated with their nominal impedance, along with a minimum impedance. Sometimes these ratings are given in a format like 8/4, where the 8 is the nominal and the 4 is the minimum. My Klipsch Forte II speakers are 8/4 ohm speakers. Running in stereo mode, the Rotel did a good job with them overall, although it was a touch light in the bass. When I bridged the amp, it lost all control over the bass, and the sound was not nearly as good. When I tried out my Magnapan MMGs, which are nominal 4 ohms speakers, I never got really satisfying results. In stereo mode, there just wasn't enough current to really move the planar- magnetic driver around, and bridged mode was totally out of the question. I got the best results from this amplifier with a pair of Boston Acoustics T1030 speakers; these are 8 ohm nominal speakers that don't dip far below that (I believe they are 8/6). There was nothing to dislike about the sound quality driving that easier load, and it sounded almost identical in bridged mode (perhaps a touch less bass, but made up for by the extra power).
For comparison sake, I tried all three of these speakers again with an Adcom GFA-555 (first series), a well-known reference. The Adcom has incredible bass-driving ability into loads even below 1 ohm, but I normally find it too aggressive in the treble to really like. With the Klipsch speakers, the Rotel was a much better choice because of the more relaxed treble, but the Adcom did coax more out of the bottom-end. The Adcom was by far a superior with the Magnapans, easily manhandling the woofer without any trouble, and the normal harshness was not noticeable with that combination. With the Boston speakers, the results were very similar between the two; the Rotel was a bit more laid back, and the Adcom had a bit more bass, but it was very close. Current production models from Adcom are less forward than my old unit, so don't read too much into this as a generic Rotel vs. Adcom comparison.
It was obvious that the low impedance abilities of this amplifier are less than spectacular. I wouldn't recommend ever using speakers that drop below 4 ohms in stereo mode, or below 6 ohms in bridged mode. Even with easier loads, I never felt there was the same level of bass impact with any of the speakers I tried compared with a more powerful amplifier like the Adcom. Then again, what do you really expect when comparing a $500 six-channel amplifier to a more expensive stereo only one?
The only other thing I was a bit disappointed with when using the Rotel RB-956AX was the turn-on sound. I got a louder thump out of this unit than any modern amplifier I've ever used. This might not be a big issue if your speakers aren't as sensitive, but for me the pop when the unit is powered on is loud enough that I get worried for a second whenever I hear it. I guess turn-on protection is something Rotel left out of the design at this price point, which isn't really a big issue for the typical speaker this unit would be connected to.
Other than the points mentioned, this amplifier was excellent to work with. I didn't notice any obvious incompatibilities with any of the preamplifiers I tried it with, which leads to me believe it's an easy load for them to drive. This is hardly a surprise, given that this amp is designed for the home theater market--surround preamps in general do not have particularly powerful preamp outputs. You can drive this unit to input overload with a mere 5V of input, which means you may want to be careful if you do have a high-output preamp section you're attaching it to.
Since I pointed out that this amp is rather limited in bridged mode, it's worth nothing that the stereo performance is a bit underrated in some cases. The power spec given in the manual is 40W/ch into all six channels. When I was just using two channels of the amplifier, it sounded quite a bit more powerful than that. I suspect the output transistors can swing more than 40W if the power supply isn't occupied with other channels to feed. Judging from the clean volume I could get out of it, I'd put the two-channel unbridged output of this amplifier at well over 40W/ch, possibly as high as a full 80W/ch. There didn't seem to be as much headroom over the spec in bridged mode; the bridging only seemed to get a small fraction more useful power when driving a single pair of speakers than I got in stereo mode. I almost always preferred the sound of this amplifier when no channels were bridged, and you don't lose nearly as much power that way as you'd think from looking at the numbers Rotel supplies (which just don't cover all the possible ways you could configure the amplifier).
Rotel's RB-956AX has flexibility as its calling card, but that card isn't accepted everywhere. When operated within its impedance limits, the sound is near the best I've ever heard from a $500 amplifier, especially with its smooth but detailed treble (which is often what separates the high-end from the lesser products in the market). Go outside those boundaries, though, and there are a number of better choices available.
Adcom used to run this ad in the early 90's promoting the combination of their separate preamp and power amp units as "the cure for the common receiver", giving far better performance for a modest marginal investment. It was a good ad, too; I bought a set. The combination of Rotel RSP-960AX surround preamp and Rotel RB-956AX power amplifier follows that same sort of philosophy; it's really quite superior to your typical surround receiver, and at $1100 for the combination it's not priced all that much higher. While I was hard on its impedance limitation while looking at it as a high-end amplifier, to be honest it does a better job at dealing with difficult loads than almost any receiver you'll find. Since most home theater speakers are built to be very easy loads precisely because the receivers are so wimpy, when used for that sort of application (the one it was designed for) this Rotel unit is nearly impossible to beat at this price. Since I personally have home theater as a secondary pursuit, it's just as important to me that the unit be a good choice for use in stereo mode with more difficult speakers, which is why I wrote the review the way I did. Considering its price, I think the RB-956AX did a good job at that test as well, although it certainly would not be my choice if stereo was my only plan for the future. If you're not sure where you're going with your audio system, or you already know surround is something you're interested in, the additional channels give you considerable expansion capability while giving admirable stereo performance under most circumstances.
I'd say it's a good unit; depending on what you're looking for, it might be a bad one for you. If this business were black and white enough that we could easily dismiss components as fitting in one category or the other, it would make buying much easier, but the arguments about equipment would certainly be a lot less interesting.
.....GS