Son of Ampzilla 2000 Circuit and Measurement Review.
Now, I really did not want the thing, because I hear no difference in the sound quality of modern amps down to the rock-bottom stuff. Please note that I am not saying that all well-designed amps (flat frequency response, low output impedance, well-designed protection circuits, noise levels below room noise levels, and distortion below 0.5%) of the same power rating sound the same. It is just that I cannot hear differences in real rooms under normal operation. So how do we find out whether a difference exists?
The publisher somehow convinced HF to go one more time to the front lines and do some comparative listening tests. So HF got a Son of Ampzilla on his front porch.
What I really wanted from Spread Spectrum was the schematics of the Ampzilla 2000 and its Son. No UPS required here, just email. In opening the email I found a fascinating design. I will try to explain below what is special about this amplifier's design and why it is a breakthrough product. I will also examine measurements of the amplifier that show the clear progress promised by the schematics is delivered in the production product. I will then address the issue of the sound of the amplifier as inferred from the measurements.
Part 1: Circuit Design. Fully-balanced from input to output, the Ampzilla 2000 series holds a singular distinction among amplifiers. A balanced signal differs from a differential signal because the desired differential-mode signal
Vdiff = (Vin+) - (Vin-)
and the undesired common-mode signal Vcm = (Vin+) - (Vin-)
can both be extracted. The common-mode signal is undesired noise that was summed into the signal-carrying wires. It is not the desired signal because the two wires of the cable were driven out-of-phase. Consider an EKG machine where two electrodes measure the voltage difference across the heart. The EKG wires are exposed to all sorts of noise before they get to the recorder; something even as simple as line noise from lamps. At the EKG recorder the common-mode signal, which may be very large, is rejected and the differential signal is amplified.
Recall that a balanced connector has three terminals. The third approximately identifies the voltage around which the signal-carrying wires swing. In the case of the EKG system, the patient is set at ground potential with the third lead, a very high resistance connection. No current flows in ground in contrast with a standard two-wire connection between HiFi components. In a two wire connection ground is the return loop path and carries signal current.
The Ampzilla has a balanced input and the amplifier uses ground only as a reference. More importantly, the Ampzilla is fully-balanced at the output terminals that swing equal and opposite voltage. The two leads of the speakers are connected between the output terminals, and the speaker terminal is not connected to ground.
The transfer characteristics of a single-ended amplifier are
A([V.sub.in] - [VGND.sub.IN]) = [V.sub.out] - [VGND.sub.OUT]
Her, input and output grounds are shown separately. For the amplifier to work properly, the two grounds should be flat and constant, which may be tricky because the current that drives the speaker must return to ground. In an amplifier, a large metallic bus bar carries the ground current from the speaker to the center tap of the transformer; the metal attempts to keep the output ground to the center tap ground the same.
The transfer relationship of a fully-balanced amplifier is
A(Vin+ - Vin -) = (Vout+ - Vout-)
Ground is absent from this equation. Having no signal current in ground at the output is a big deal since it eliminates the need to establish power supply ground at the speaker's ground terminal.
In a differential system the output signals must swing around a fixed voltage. Consider a pair of differential signals that swing around 100V. The equation is invariant to this fact, but that would be impractical for a commercial amplifier since the signal would clip at the top supply rail. To have the differential amplifier signal to swing around to positive and negative supplies, a signal is introduced into the amplifier that can adjust the output of the amplifier to move around a fixed point. Typically, this fixed point is ground, and is established by the relationship:
[VCM.sub.out] = (Vout+) + (Vout-)
The fixed point is named VCM because this is the same equation used to calculate the common-mode signal. The common-mode voltage at the output is fixed to drive the cable. Additional common-mode noise may be imposed at the cable and be present at the other end of the cable. This causes the common-mode signal to be time-varying, but that should not occur at the output of the amplifier--the start of the cable.
The common-mode feedback circuit sets the common-mode output voltage. Think of it as an analog of a traditional DC servo circuit that keeps the single-ended output voltage at ground when no signal is present, thus eliminating any DC offsets.
The common mode servo in a balanced amp needs a ground reference, but no current flows in ground. One servo adjusts the two outputs to ground since the servo is trying to bring the common-mode signal to ground. Both outputs move together.
In a balanced amplifier, DC should not occur across the differential output signal (Vout+) - (Vout-) but it does. The origin of the differential DC offset voltage is the same as the offset voltage in single-ended amplifiers. Not all components match precisely, and the mismatch is reflected at the amplifier's output as a DC offset. To offset this effect, another DC servo is needed.
Of the two servos, one keeps the common-mode voltage at ground and the other pins the differential offset voltage to zero. DC servos are notoriously difficult circuits as they often break-out in oscillation. With the fully-balanced amplifier, complexity rules, since two interlocking servos are required that concurrently set the DC voltages at the output terminals.
With the grounds established, stabilizing the amplifier's gain and reducing its distortion are the next two goals. The process is similar to stabilizing a single-ended amplifier, but with a twist. Two wires are needed to route feedback from the differential outputs to the differential inputs. Recall that the input to the amplifier is balanced and referenced to the ground of the preamplifier driving the cable. There can be an offset between the common-mode input ground and the common-mode output ground. If the feedback signal is referenced to the common-mode output ground, the incoming differential signal from the input ground must be referenced to the output ground before feedback is applied. That requires the input circuitry of the amplifier to accept the common-mode voltage of the input signal but the input to the main amplifier must be set to the common mode voltage at the output since that voltage is being feedback by the feedback network.
How can we do this? Think of a dual center-tapped transformer. The primary of the transformer is connected with the balanced signal at its input. The center tap is connected to the common mode input voltage. Note we have no flow of current in the ground lead with respect to differential mode signals on the other two leads of the transformer. We do the same thing with the secondary of the transformer, but this time we connect it to the common mode voltage of the output to the center tap. By the action of the transformer we have re-referenced the common mode ground of the differential signal coming into the amp, but we have taken no current from the two ground references
In the Ampzilla 2000, the front-end of the differential pair of the amplifier is modified to eliminate the transformer. In so doing, the common-mode input signal becomes an auxiliary signal connected to the front-end, leaving the front-end to directly handle the differences between the input and output common-mode voltages using a highly innovative circuit topology.
To review: a balanced feedback loop stabilizes the AC parameter of the amplifier and two DC feedback loops set the common-mode output signal and remove the differential DC offset at the output. There are four loops from output to different internal nodes of the amplifier. If any of the feedback circuits are set incorrectly, then the amplifier will oscillate. If both outputs oscillate together, then common-mode feedback is the culprit. If both outputs oscillate in opposite to each other, the culprit is the balanced differential feedback (the AC path, the DC servo path, or both). James Bongiorno complicated matters by adding an extra input to the amplifier that feeds-forward the common-mode input voltage. Ugh.
The balanced amplifier has eliminated the nasty problem of returning high currents to a fixed ground. In addition has another highly desired property--the design allows us to take advantage of the fact that distortion that is even-order will have the same output polarity even when the input signal is anti-phase. Consider a second-order nonlinearity:
[V.sub.distortion] = [Vin.sup.2]
For a sine wave:
Vdistortion+ = [sin(x)] = 1/2cos(x) +1/2 where x = 2 [pi] [f.sub.t]
With the balanced signal - sin(x)
[V.sub.distortion] - = [-sin(x)] = ? cos(x) + ?
The distortion is now a common-mode signal and will be rejected. This works only for even-order harmonics. Single-ended amplifiers exhibit this property when they have push-pull output stages, though the whole amplifier is not dealt with in this elegant manner.
Not all high-end amplifiers follow the balanced amplifier paradigm. High-end manufacturers start with one of the Bongiorno single-ended topologies (e.g., Marantz 15 or SAE). The Doug Self text (Audio Power Amplifier Design Handbook by Douglas Self) offers additional detail to fine-tune a Marantz 15-based design. Tweak designers tinker with the parts, changing transistor types (often forced because older transistors have been discontinued), and move bias currents and voltage currents around as they design for a specific power output (really the amplifier's ability to swing voltage) and ability to drive low-impedance loads. The circuit is packaged in a fancy box along with large transformers, large primary filter caps, and other accoutrements to justify high prices. If the feedback loops begins to oscillate, then the designer can add a cap to part of the feedback loop or change a cap value at another part of the amplifier until it stabilizes. The same applies to a DC servo, when the designer brave enough to add one (otherwise, the DC offset is adjusted by a trim pot). Feedback systems belong to the field of control theory. Stabilizing feedback loops (such as power amplifiers or an airplane's control wings) is called compensation theory. For most designers of high-end amplifiers with limited knowledge of control theory, the process of compensating an amplifier is black magic.
The designer can attempt to borrow from Bongiorno when designing a fully-differential amplifier, as with the single-ended topologies, but the designer soon runs into the problem that these amps are too complex for cut and try.
As we said above, we have to two DC servos. One sets the common-mode voltage, and the other sets the differential DC offset to zero, which gives use to feedback loops plus the balanced feedback. To that we add the balanced feedback that controls the AC parameters of the amplifier. If any of the feedback circuits are improperly specified, then the amplifier will oscillate. Trying to stabilize the two DC loops and the balanced AC loop becomes impossible using the "putting this cap put here might help" approach. The rules of thumb might make one loop more stable but cause another loop to cut loose.
An engineer who understands control theory is needed so that the entire design can be modeled initially on paper. Using rules of compensation from control theory, the stability parameters of each feedback loop can be calculated and the proper compensation components can be added in the correct place in the circuit at values that have been calculated analytically and not guessed at. Subsequently, computer simulation verifies the calculations. The Matlab/Simulink control systems tool box takes hours of pencil and paper work and reduces it to a few minutes of computer design time.
But students can easily fall into the same trap that the cut and try designer uses on a prototype. The computer should not be used by the student or anybody else to determine a method to stabilize a system but only to confirm that the pencil and paper values have been calculated correctly. If the computer gives bad results it is back to the pencil and paper to find the error in the analysis. Actually, James Bongiorno skipped the computer until recently, and did the work, in its entirety, by hand.
Stabilizing practical audio power amplifiers can be especially tricky because of their undefined load impedance. You have to design to a set of worst-case loads trying to predict how crazy of an impedance a high-end speaker can present to the amp. Those crazy impedances are often found in speakers designed by tweaks who use the same cut and try design procedure. A skilled speaker design ensures that the input impedance of the speaker is amplifier-friendly.
More stability problems occur if the speaker lead is shorted together or to ground. Even if it happens in an instant this impulse can send the amplifier into uncontrolled oscillation that can turn output transistors into toast--and potentially your speakers as well. One favorite test to see if an amp is really stable and well-protected is to take a file across the output terminal. Sparks fly as the amp is opened then shorted then opened again. A good amp goes into protect mode or goes to sleep. An unstable amp can fry half of its components under this test. Another way to get an amp to go unstable big time is to cut the power to it unexpectedly. The type of thing that happens during a thunderstorm. A well-stabilized amp should emit a little blip and then go to sleep. Amps can also go unstable when the power returns.
Most balanced amplifiers have fallen into obsolescence, although the fruits of the engineers' labor has often been maintained for posterity in a variety of publications. Engineers at Sansui, led by Susumu Takahashi (one of the most prolific engineers in audio engineering, he spent his whole professional life at Sansui, which did not promote the work of individual engineers to the public. He thus remains unknown to the most audiophiles.) published many enlightened articles on the analytical aspects of the balanced amplifier; specifically, they clarified the distinction between a bridged and balanced amplifier.
For example, Takahashi pointed out that a bridged amplifier returns current into ground even if the ground terminal is not actually in use in bridged mode. A bridged amp thus has the advantage that one side of the bridged amplifier is pushing current into ground and the other is pulling current out of ground, so the total current returned to the power supply is reduced. This is good--but no current in ground is better. Also, a bridged amplifier has two single-ended feedback loops around each amplifier, not the balanced feedback of the genuine balanced amplifier. Sansui papers show how distortion from transistors in a true balanced feedback amplifier is reduced in comparison to a bridged amp. The Sansui balanced amplifiers remained with the company until its demise in the 1990s. Unfortunately, Takahashi never published again once Sansui collapsed. Perhaps he retired when Sansui went out of business. Other Japanese balanced amplifiers appeared, but had limited sticking power. Except for the ones designed by James Bongiorno, I am unaware of any others in production.
Please note that many amplifier manufacturers use the term "balanced amplifier" in their literature or on their websites. Having looked over much of the printed or electronic information presented by the manufacturers, I believe most are not offering a true fully balanced amplifier that meets the following four criteria. If all four criteria are not met, then all the benefits outlined will not be available.
1) A global differential feedback signal from the differential output of the amplifier to a differential input of the amplifier for the purpose of stabilizing the complete open loop amplifier's differential gain.
2) Common mode feedback injected into some part of the main amplifier such that the common mode feedback can drive the common mode output of the amplifier to ground.
3) A fully differential stage (differential in and out) incorporated at the front end of the amplifier. Both the differential feedback signal and differential input signal terminate in to this stage. This stage amplifies the differential mode error signal and its output is also a differential signal. The remaining part of the open loop amplifier can consist of two identical single-ended amplifiers or it may contain more fully balanced stages. The key takeaway point here is that a pair of single-ended, continuous-time, open-loop amplifiers without an interlocking differential stage cannot be configured to be a true balanced amplifier even if parts I and 2 are satisfied.
4) Ground is used only as a reference when the amplifier is used with balanced inputs. No signal current of any kind flows in ground.
The only way to determine whether an amplifier meets these criteria is to examine the schematic of the amplifier. Many highly skilled engineering teams, knowledgeable in the analytical design of power amplifiers, including compensation theory, could produce a stable fully balanced amplifier that meets the four criteria above with no reference to the design work of James Bongiorno. Much of the fundamental information the design team needs can be found in the Sansui papers published in the AES journals as well as other journal articles in the AES and IEEE press. The IEEE papers deal with design of balanced integrated systems for a wide variety of applications not related to the design of audio power amplifiers, but they do contain much important information that can be used to make fully balanced power amps. If a manufacturer besides Spread Spectrum is producing a fully balanced amplifier, I would welcome this information and pass it on to the readers.
Not content to solve the hard problems outlined above, Spread Spectrum Technologies multiplies the design difficulties in an attempt to improve the amplifier's performance. The Ampzilla 2000 uses two stages of differential voltage amplification, instead of a single stage, to enhance the common-mode rejection ratio. In turn, a common-mode voltage must be set at both the first- and second-differential amplifier outputs (add another feedback loop). Then, to make things even more exciting, the input's common-mode signal is actively coupled to the first differential pair. That is how the Ampzilla 2000 resolves the differences in the common mode ground voltage at the input and output of the amplifier. I had to have James go through the circuit for me. It was too complex to follow on my own.
As a final touch of engineering elegance, James uses current-mode feedback (also used in some other power amplifiers to improve the time domain response to a transient, although a bandlimited CD would never emit a signal that requires this type of performance. A DVD-A sampled at 176 kHz might be a different story). Current-mode feedback works its magic by offering another factor, the value of the feedback resistor, into the compensation equation. If you look at a data sheet of a current-mode operational amplifier, notice the optimal values of the feedback components; these are not expressed as ratios. More parameters such as these send the cut and try folks out of the room muttering, but those who really understand compensation welcome another degree of freedom in the design equations. The novel aspect of the Ampzilla 2000 is that the current-mode feedback is balanced!
Finally we note the voltage gain stages are powered from regulated power supplies. This plus the balanced topology makes the amp virtually immune to power supply noise. No need for expensive external power supply filter here.
One thorny issue yet to be addressed is the conversion of the single-ended input (what most likely comes from your preamplifier) into a balanced one. To accomplish this, current must be injected into the input ground to complete the circuit around the loop from the preamplifier to the power amplifier. A transformer can do the job and so can a can an added active circuit placed in front of the amplifier. But that circuit is not simple to design and can introduce distortion into the signal path. What is done in the Ampzilla 2000 is to make use of the common mode feed forward circuit in the first differential stage of the amplifier that we discussed above. Instead of receiving the common mode input voltage level, as it does for a balanced input, it receives a portion of the single-ended input signal. The input stage is thus put out of balance and the balance is restored by the AC feedback loops. In other words the amplifier itself does the single-ended to differential input conversion.
The Son of Ampzilla 2000 is a modified version of the monoblock Ampzilla 2000. Separate windings of the transformer are used for each output stage. The chassis is the same size, although more stuff needs to be shoehorned into the box. The front-end voltage amplification circuitry must be doubled: there are two input channels to deal with. The space for the PC board that contains these electronics is so limited that some of the voltage amplification circuitry had to be left off the in the Son. One of the circuits that was eliminated linearizes the second voltage gain stage in a topology proprietary to Spread Spectrum. In migrating from the Ampzilla 2000 to the Son, a cascode stage on the second gain stage also fell by the wayside. The cascode stage improves the voltage gain and high frequency performance of the second gain stage and is relatively common in high-end amplifiers.
Most significant, of the missing circuits is the second stage cascode, which has another important property. Without the cascode, the output of the second gain stage moves the full swing of the amplifier, but the input to the stage only moves a fraction of that (the ratio of input to the output is the gain). For clarification: the input to the second stage is the base of a transistor and the output is the collector of that transistor. In the transistor, the capacitance of a junction diode between the base and collector is highly nonlinear and as the output swings the capacitance at the base changes value. This varying capacitance is seen by the first gain stage and distortion can result. The cascode transistor isolates the second gain stage collector from seeing the large signal swings that now appear at the output of the amplifier. If the collector of the second gain stage transistor is held at a constant voltage by the added cascode transistor the input capacitance of the second gain stage is also held constant. Several noted designers aver that the cascode is important to reduce high frequency distortion and have written text books (the Self text mentioned) and journal articles that discuss this issue.
Another potential problem relates to how the second voltage gain stage is loaded by the output stage, which directly provides the current to the load while isolating the second voltage gain stage from the load. Both the Ampzilla 2000 and the Son do things differently than most amps. The output circuitry magnifies the load impedance of the speaker by four orders of magnitude and presents it to the second gain stage. In most designs, a resistor, or set of resistors, is added to the output stage circuitry to swamp out the multiplied load resistance so that the input impedance seen by the second voltage gain stage is constant and independent of the load. Spread Spectrum Technologies prefers that the second gain stage see the magnified load impedance, because they claim it improves the amplifier's stability. For example, if the load is capacitive, then the loop gain of the total amplifier will decrease, thereby preventing oscillation. Given the number of high-end speakers with very strange load characteristics, the Spread Spectrum innovation can be important. The disadvantage of this approach is that it assumes the load impedance amplification factor is constant as the output stage drives more current. This is not how it works in real life, because transistor current gain changes with collector bias current.
The effect can be minimized by increasing the quiescent bias current of the amplifier. In the Son of Ampzilla 2000, the quiescent current might be reduced relative to the big monoblock, because we now have two output stages for stereo. The amount of heat that can be dissipated in standby mode should be the same for the monoblock as the stereo unit, but we have the same size of box with same size of heat sinks. Howard, found no evidence the unit was running warm so we can assume that the Son is optimally biased to the same level as its big brother.
Part 2: Measurements. We do not have any facilities at T$S to measure an amplifier of this performance capability to determine whether the problems we discuss above cause real measurable effects. Only an Audio Precision would do the job for an amp this good. Even in its simple domain configuration for testing analog components, the price rises to five figures. Lucky for us this time, Spread Spectrum Technologies has the Audio Precision Graphs on their web site http:// www.ampzilla2000.com/songraphs.html
We are most interested in the first graph on the website, which shows distortion with changing power level. As can be seen, the curves taken at 20Hz is wonderful. The distortion is 106dB down. That is a little better than ideal 17-bit DA converter. Things degrade a little at 1kHz, and at 20kHz, distortion is only average for a power amp sitting, at 0.05% or-66db down. That a 40dB difference, or a factor of 1000. What could be happening? What happened to the expected performance breakthrough of the true balanced amp? I think it is the compromises made in taking the mono block to the stereo amp.
Let's head back to the 20Hz curve and examine something amazing. In Audio Precision curves we expect the noise dominate the distortion as the signal level decreases. That is because the Audio Precision measurement is noise + distortion When we are the noise-dominated region, the curve goes up as the signal level goes down. That is because the noise is constant but the ratio of signal to noise is going down as the signal level goes down. We can see this effect in the 20Hz curve below one watt. We cannot see it at all in the lkHz and 20kHz curve. The distortion masks the noise all the way down. That is not because the distortion is high but because the noise is so low. From the break point in the curve I can calculate the noise. At 100mw, which is .9V into 8 ohms, the noise is down by 0.0025%. We thus find the broadband noise level is 22.5uV. From the graph we see the amplifier clips at about 120 watts, which is 31 volts into an 8-ohm load. The signal-to-noise ratio to clipping is 122dB. Unbelievable! Nothing else on the planet comes close. This is equivalent to the noise of an idea 20-bit data converter. Here is a power amp that can be connected directly to the data converter, and I do mean directly. We have no need for coarse analog volume correction when the power amp has this noise level. It really is amazing.
Now, the best data converters have balanced outputs and they have current output. Current outputs are just what we want for an amplifier with current mode feedback. A match made in heaven. One analog gain stage does it all. Current to voltage conversion and current gain necessary to drive a complex 4-ohm load. All that is needed is a digital volume control (properly dithered) in front of the DAC and you could plug your SPDIF fiber cable from your CD player directly in (you will still have to wait for an interface to send SACD and DVD-A across the link). The future can be now but Spread Spectrum sees an analog future and has no plans to build the power amp I just described. Drats, Drats, Drats!
Back to the data on the Ampzilla 2000 website. A noise spectrum curve is given We can find nothing about the intrinsic noise of the unit from this curve because the bin width of the FFT is not given (do not worry if this sentence makes no sense, it is included for the tech types only as a warning about interpreting noise floors from FFT data. If the number of bins is increased, the noise floor on the graph goes down, but it is filled over more bins. The broadband noise comes out to the same value.)
What we do see is a vanishingly low 60Hz feed through to the output and no harmonics of the line. That is what a balanced amplifier with great common mode rejection can do. We are witness to what happens when we have no ground currents. Amazing.
The frequency plot shown on the website shows changes in response with load. This is a rare plot to come from a manufacturer. Even more interesting is the response into a simulated speaker load. This is a test Stereophile runs but this resistively loaded curve still gives us lots of info.
With the curve down 0.5dB at 20kHz, we can calculate the equivalent output resistance of the amplifier at that frequency. It is 200milliohms. This can be translated to the spec called "damping factor," but let's leave it in this more understandable form of resistance. If you want damping factor then use the formula 8/rout. In this case we get 40. Referring to the graph again we see that no attenuation occurs at 1kHz, but that is where manufacturers love to quote damping factor. It is meaningless. The 20kHz number is what counts. Also of significance is that this number is achieved even though the Ampzilla 2000 has open loop circuitry at the speaker terminal output to suppress oscillation when the amplifier drives a difficult load. Tweak amps drop this network. Maybe some of the designers of amplifiers think things sound better with oscillations. Often the high frequency oscillations are hard to find in tweak amps--rest assured they can change the sound, but definitely not in a way you want it to go.
We do not have a PowerCube test system at T$S, but I would expect very good results from the Ampzilla 2000, given that it has no voltage or current limiter. The power fuse just blows. I am a little worried about the time the output devices see short circuit currents before the fuse goes. It is possible for the base junction to change slightly although I am told by the manufacturer that the time is well within the safe operating area quoted by the manufacturer.
Summing it up, we have a fully balanced power amplifier. 20Hz distortion figures and rejection of line hum show the advantage of the topology. Unfortunately, this particular amplifier is not the ultimate statement of the Spread Spectrum Technologies expertise. That ultimate expression of the design is a more costly monoblock that has more sophisticated circuits for both gain stages of the amplifier. One result of the design compromises made in the Son of Ampzilla 2000 is a distortion level at 20kHz that is not comparable with this amplifier's chief competitors from Bryston and Accuphase.
On the other hand this amp has noise levels so small that these other amps sound like wind machines in comparison. This is not a result of the balanced amplifier but represents a whole other set of technological optimizations in this amp.
Part 3: Does it sound better? There is no question that you can set up an experiment to hear the noise-level difference although you may not hear it, under normal conditions, in a room with a normal background noise and speakers of normal efficiency.
Does the noise modulate with the music signal thorough a nonlinearity in the amplifier? Can you hear that? I do not think so, but it may be possible to investigate it with a special technique called PC ABX. (A future article will discuss PC ABX in detail.) In PC ABX (www.pcabx.com) the output of an amplifier is captured on a computer. We can then manipulate the signal to see if we can unmask distortion. For example, we can play the signal we recorded back through the amplifier and record it again and keep the process going. If we can hear the difference of the manipulated sound file in comparison to the original sound file off the CD using ABX double blind test results we know we have found something. Maybe it is just the noise of the amplifier which goes up every time we send the signal around the amplifier. Maybe it is the frequency response error which also gets multiplied. And maybe it is something, such as noise modulation or the mechanism by which a fully balanced amplifier cancels distortion, that most of us cannot hear without the PC ABX magnification but some golden ear can. I would not place any bets on it however.
Back to the real world. There is also no question that you can set up an experiment to try to introduce ground loops into the system, and then you can see the hum melt away when you use this amp. Even more, there is no question that you can here the effects of the low output impedance at 10-20 kHz. But again you may need to set up a special evaluation procedure with something like the double-blind, PC ABX test so that the effects can be magnified. Perhaps the lack of any I-V current limiter can also be detected under special load conditions when the amp is driven to clipping, but again I think normal mortals would need a special test setup that magnifies the effect to hear it.
I may also be able to set up a test to make the 20-kHz distortion in this amplifier audible using a special test setup (PC ABX combined with a multitone signal that produces distortion products at frequencies that we can hear it). More obviously, I do not have to do anything special to show that this amp cannot produce the same sound pressure levels in a given room with a given speaker that a 500-watt-per-channel amp can.
Part 4: The Highly Hedged Recommendation. The question comes down to something like that of owning a Ferrari v. an entry-level Lexus The technology in the Ferrari is remarkable but you cannot experience it in normal driving conditions. It goes 200 mph in long straight stretches and the Ampzilla 2000 family has a signal-to-noise ratio of a million to one. The Ferrari brakes bring the car to a stop from 200 mph in 200ft and the Ampzilla 2000 family has a 20Hz distortion figure that only the best test equipment can measure. Still, it is nice to know the ultimate car is sitting in the garage and the Ampzilla 2000 is on your equipment rack.
On the other hand, the lower-priced, stereo-chassis Son of Ampzilla 2000 unit under review here is like the Son of Ferrari. I think that if anyone who is that serious about amps and who has the necessary bankroll would want to spend the money for any designer's best effort. I wonder if $2250 saving for a lower power amplifier with slightly simplified circuitry is the way to go if you are shopping for the amplifier equivalent of a Ferrari. Get a pair of Spread Spectrum's Ampzilla 2000 monoblocks at $5500 if you want the lowest noise and most advanced circuit design ever attempted in a power amp.
If I had the cash to spend and wanted to make a statement, I would go with the power amp with the most advanced circuit topology on the planet. I would go for a pair of the monoblock Ampzilla 2000s, but that is not how I spend my money; I am very happy with the $250 Onkyo receiver I am testing--and my Subaru.
|Printer friendly Cite/link Email Feedback|
|Article Type:||Product/Service Evaluation|
|Date:||Jul 1, 2005|
|Previous Article:||Son of Ampzilla 2000 stereo power amplifier.|
|Next Article:||Toshiba SD-4960 Universal Player.|