Test Results
The first test (Test 1) performed on the model was to synthesize a standard 741 op amp in terms of GBP and manipulate it for differing SR. The results should show very linear behavior up to fp, and a hard limit or sudden distortion rise as slew limiting is reached. Conditions were set up for a unity signal gain inverter, with a noise gain of 20 dB, using the test circuit of Fig. 39.
Figure 40 shows the results of Test 1 for a THD swept frequency test, at an output of 7V rms. Conditions A, B, and C are approximately 0.4, 1.8, and 18 VM S respectively. The different circuit conditions to yield these SR are noted. As should be noted, since GBP and the feedback conditions are identical for all three of these tests, the only variables are SR and Vth.
As can be noted from the A and B curves, these conditions produce a sudden distortion increase when the SS of the test signal equals the amplifier SR. The high SR of condition C prevents the limit from being reached, for any test condition.
Note that Vth increases, going from A to C, in the same proportion as SR. For a case of transient signal condition, the photos of Fig. 41 show how this same amplifier behaves for the three conditions set down in Fig. 40, but with a different method of measurement.
Figure 41A shows waveforms for the "A" test condition (SR = 0.4 V/ NS) for a signal condition of a 5-kHz, 20-V p-p square-wave input. The top trace shows the Vo waveform, which clearly resembles a 741 type response (31, 38), changing 20 V in over 40 NS. Inside the loop, the error voltage V1 is shown at the bottom. Here it is seen that V1 saturates negative, then positive, for the corresponding (+) and (-) slew intervals, respectively. It is clear from this photo that the slewing evident in Vo is a result of saturation in V1.
Figure 41B shows waveforms for the "B" test condition (SR = 1.8 V/ NS), with a 20-kHz, 20-V p-p square-wave input. At the top, the Vo waveform shows that slewing is present, as is evident by the linear (+) and (-) slopes. This is confirmed by the V1 waveform, which again indicates saturation of the 1st stage for these corresponding times. This is similar to Fig. 41A, but the difference is that for this higher SR condition, the slewing intervals are simply shorter (note scale factor differences do not be misled by same general wave shape). Figure 41C is very interesting, because it demonstrates that a sufficiently high SR and Vth can completely prevent saturation of the first stage and maintain operation within the small signal region entirely. Conditions of these photos are an SR of 18 V/ NS. However, the feedback conditions described above in conjunction with the 20-dB noise gain result in an amplifier closed-loop, small-signal bandwidth of 95 kHz. This in turn is equivalent to a single-pole, low-pass filter with a time constant of 1.7 NS. For a 20-V p-p output from this filter (the amplifier), the maximum signal slope is 12 V/uS. For Fig. 41C, the signal input is a 20-V p-p 50-kHz square wave, and it can be noted that there is no slewing evident in Vo. The waveform is exponential in shape with a risetime of about 4 pS consistent with the small signal relationships.
That slewing is not present is also confirmed by V1, which shows that the error voltage remains below the clipping level. Note that the highest amplitudes of V1 occur at the peak SS of Vo or at the transition points of the square wave.
This particular test confirms in another way the point made in Part I of this series, that slewing can be prevented by maintaining the amplifier small-signal bandwidth at a lower frequency than the power bandwidth. In 41C, fc is 95 kHz, but fp is 290 kHz, and no slewing is evident.
With this same model, experiments were also conducted to examine the sensitivity of the amplifier to open-loop bandwidth (fo). Test two conditions were commonly set up as described in Fig. 42, which resulted in an SR of 1.3 V/ uS and an fr of 540 kHz. For this test circuit with R4 present at 10K, fo becomes 16 kHz and the open-loop gain is 30 dB. With Rx open, the feedback is then 24 dB. With R4 open, the circuit becomes a classic op amp, with a very high open-loop gain and fo very low. Note, however, that GBP remains unchanged for either condition.
For condition A, where R4 is 10K, THD curve A indicates that slew limiting is reached at 18 kHz. V1 (A) is a plot of the rms error voltage versus frequency. Since it is essentially flat with frequency, it is testimonial to the wide open-loop bandwidth. Note that V1 increases to its clip level at 18 kHz, coincident with the slew rate limit point.
The B condition shows corresponding results with R4 removed, and the most obvious difference is the (apparent) increase in fp. Error voltage V1 (B) now increases 6 dB per octave with frequency, the inverse of the integrator's gain rolloff what is necessary to maintain a flat output versus frequency for the overall circuit.
The apparent increase in SR for condition B is not an increase for this condition, but rather reflects a less than potential maximum SR for the A condition. This is so because the 10K resistor loading the integrator absorbs a portion of the charging current available to C1 for slewing.
These points are also brought out in the square-wave photos of Fig. 43. This shows response of the circuit of Fig. 42 to a 5-kHz, 20-V p-p square wave for conditions A and B. For these test conditions, the transient performance is shown in Fig. 43. The slewing in Vo shown in 43A shows a quasi-linear ramp or a combination of ramp and exponential waveform caused by R4. Since R4 constrains the open-loop gain to a relatively low value, this is also reflected in the large error voltage shown in V1 (bottom). The voltage V1 is clipped for the slew intervals (as expected) but also shows a very large potential (10 V) for the steady = state waveform positions. This excessive error voltage reflects a relatively large gain error for this circuit.
Figure 43B shows the Vo and V1 response for the same input drive but with R4 removed or condition B. Note that in 43B the slewing intervals are shorter and linear, as would be expected due to the constant and larger C1 charging current available. The error voltage shown in V1 is much lower in the steady-state periods, reflecting the increased gain available in the integrator. The low gain error is also reflected (more subtly) in the greater amplitude in Vo, compared to Fig. 43A. This test indicates that, by both THD and transient response tests, there is no inherent advantage to a wide open loop, small-signal bandwidth. By contrast, there are definite disadvantages to the constraint such operation can place on amplifier characteristics, such as limited LF loop gain and also some sacrifice in SR. And, while it is not apparent from this particular experiment, loading an integrator stage in a conventional amplifier will usually degrade the open-loop distortion characteristics.
Fig. 43--Transient performance of synthesized opamp model with different
open-loop gains to a 20-V p-p, 5-kHz square wave. Top traces are outputs,
bottom traces error voltages.
A, R4 = 10k; B, R4 = open.
(Scales: 10 V/cm, 20 p S/cm.)
Fig. 44--Test circuit to examine Vth criteria.
Predicting A Non-Slew-Limited Response
We are now at a point where the information developed can be merged into a set of relationships useful in designing a non-slew-limited amplifier or an amplifier which is free of SID and TIM, by definition. This evolves in a fairly straightforward manner from the relations just discussed.
A non-slew-limited amplifier is simply one which cannot be made to slew for any signal input level below that which causes amplitude clipping. Input waveform shape is unrestricted and may include all waveshapes up to and including square waves. The square wave (as discussed in the sine square box of Part II) is the most rigorous test to which an amplifier can be subjected because of its very high SS (infinite, for an ideal square wave). Therefore, if an amplifier can be proven to be free of slewing distortion for a square-wave test for all signal amplitudes in its linear range, it is by definition non-slew limited and will be largely free from SID or TIM problems.
All amplifiers will have by design a small-signal bandwidth, fc. This bandwidth will either be determined by the feedback configuration or an input pre-filter. The amplifier will then band limit a square-wave input signal to a bandwidth of fc. For simplicity at this point, we will assume this to be a single pole rolloff. For such a filter response it can be shown [33, 67, 70] that the signal slope of the resulting band-limited-output square wave is
(53)
where Vpp is the peak-to-peak amplitude at the filter output, fc is the small-signal bandwidth, and SS is in V/uS. That this signal slope is much higher than a sine wave at fc (passed through the same filter) can be shown by the relation of the two slopes. A sine wave at fc will be down by 3 dB in amplitude, which can be expressed by modifying equation (1) by multiplying it by x/7/2, yielding
(54)
(55)
Since this is nearly three times the signal slope of a sine wave at the frequency fc, it is clearly a more rigorous test.
That it is the most rigorous test comes from the fact that the SS of the unfiltered square wave is infinite. It is clear then that an amplifier which passes a square-wave test without nonlinear distortion appearing in the output tends to be an optimum design. The question now arises, how can this be guaranteed? We already know that to guarantee freedom from slew limiting we must, as a minimum, guarantee that the amplifier SR is in excess of the output SS for all possible signal conditions. For the non-slew-limited amplifier, this will encompass the signal slopes of square waves up to the rated output. We can set up a criterion to provide this with only a few parameters. Initially, let us consider a conventional feedback amplifier which follows the relationships discussed for Vth, SR, and GBP. By general feedback theory, we can express the bandwidth of this amplifier as:
fc = GBP ß (56)
where fc is the small signal bandwidth, GBP is its gain bandwidth product, and ß the feedback factor. For this initial part of the discussion we will assume no other filtering, and the amplifier alone determines the bandwidth, as just outlined.
To guarantee no slew limiting, we desire that SR?. SS. To provide this, we can write an inequality, substituting the appropriate equivalents for SR and SS, as they pertain to this amplifier. SR is as described by (44), and SS by (53). The inequality is
(57)
With simplification, we can express this in terms of Vth as
Vth >= Vpp fc / GBP (58)
Equation (58) gives us an expression tor a minimum Vth, but we can further simplify it by substituting (56), which yields
Vth >= Vppß. (59)
The rather simple appearance of this expression may hide its rather profound implications. Since Vppß is in fact equal to the peak-to-peak input voltage, this relationship states that Vth should be in excess of the maximum pp input amplitude.
In other words, the input stage (alone) will not overload when driven with a full-scale input signal [47, 67]. That the criterion works can be illustrated with some data just presented. In test 1, condition C it was observed that the experimental amplifier did not slew limit when subjected to a full-amplitude square-wave input. For condition C, Vth was 3V and the SR was 18 V/ pS. If a minimum Vth is calculated from (59) for this amplifier, it is found to be 2V. Therefore condition C satisfies (59), since 3V>2V. On the other hand, if condition B is examined, it will be noted that Vth is only 0.33 V, and slew limiting did occur (Fig. 41B). Here the criterion was violated; i.e., 0.33V< 2V. Another example, more in the line of a real amplifier, was the variable-feedback amplifier from Part I, discussed in Figs. 3 and 4. If Figs. 4a, 4b and 4c are re-examined, it will be noted that slew limiting is evident in condition A and some in B. Condition C is a non-slew-limited case.
Since the gains in this case were 20, 40 and 60 dB, respectively, ß is correspondingly 0.1, 0.01, and 0.001. As the output level is 20 V p-p in all cases, it can be noted that conditions A and B violate the minimum Vth criterion, which says that Vpp should be less than the 301A's Vth of 0.104 V. In condition C, the criterion is satisfied, and no slew limiting is evident.
Fig. 45 Transient response of a 301A, operated inverting with unity-gain
compensation, Cc = 33pF, to 20-kHz square wave filtered at 100 kHz. Top traces
are error voltages, bottom traces outputs. A, slew-limited response, and
B, non-slew-limited response. (Scales: 5 pS/cm both; A, 0.5 V/cm top, 2 V/cm
bottom; B, 0.1 V/cm top, 0.5 V/cm bottom.)
Fig. 46--Transient response of a 301A, operated inverting and adjusted for
slew suppression, Rx = 1.2 k, Cc = 5 pF. Top trace is error voltage, bottom
trace is output. (Scales: 5 l S/cm for both; 0.05 V/cm top, 2 V/cm bottom.)
It may already have occurred to some readers that this criterion is a most restrictive one, as it dictates very low feedback factors to eliminate slew limiting in the case of low Vth amplifier stages. Inasmuch as all directly coupled, undegenerated bipolar-transistor differential-amplifier pairs have a Vth of 0.052, this can quite logically explain TIM and slew limiting possibilities in power amplifiers, where Vpp may be upwards of 70 V. It is interesting to plug typical power amplifier numbers into the relationship of (59) to see what results. A 100 W into-8-ohm amplifier with a gain of 20 (26 dB) has a Vpp of 80 V and a ß of 0.05, which results in a required Vth of 4 V
... clearly many times in excess of the 0.052 V resulting when an undegenerated bipolar differential pair is used in the input stage.
As a historical comment, the vacuum tube, still favored by many, has a Vth on the order of 3 V, for a typically used type such as the 12AU7. Viewed in this light, it is quite easy to see why a vacuum-tube design is much less susceptible to SID type problems; not only did they have less feedback in general, but they could also easily accommodate much larger inputs without first-stage clipping (47). Viewed in just the above light, it is rather easy to conclude that the transistor audio power amplifier cannot be made to work. If, for example, we were to manipulate ß to satisfy (59) for the 100 W amplifier, using a Vth of 0.05, ß becomes 0.000625 (or less), which corresponds to a gain of more than 60 dB! While this probably is a completely impractical signal gain, it is possible to use special compensation "tricks" such as input compensation [25, 31, 63], which provide a low ß, but at elevated frequencies (above the audio range). Of much greater interest are practical techniques which can be used to design an amp for no SID, without having heavy restrictions placed on the feedback loop. This can be done by separating the filtering and amplification functions, so that each can be optimized separately.
If an amplifier is preceded by a low-pass input filter with a cutoff frequency of fc, the filter-plus-amplifier combination can control the output signal slope with relative independence of the feedback factor. There are still restraints upon the Vth (or Vpp) of the amplifier, however, they are lessened to a great degree.
For this discussion it is assumed that the amplifier operates linearly and its own natural cutoff frequency, as determined by (56), is sufficiently higher than that of the input filter so as to cause negligible interaction. For such a linearly operated system, the peak-to-peak output of the input filter can be scaled by the gain of the amplifier, and the SS resulting at the output is of the same form as (53), but the relevant Vpp is the rated output of the amplifier.
If we now write an inequality such that the amplifier SR is to be maintained greater than the output SS, it follows the initial development form to (58), which is repeated here
Vth >= Vpp fc / GBP (58)
Written thus, it can be seen that as fc is lowered and GBP raised, the Vth required can be lowered. Within certain constraints, this allows considerably more design freedom. Like the previous relationship, this is one best understood by examining some performance which illustrates it functioning.
For an amplifier where Vth and GBP are fixed (as the 301A example of Part 1, Figs. 3 and 4), the only relief from the slewing problem is to decrease feedback in accordance with (59) until the criterion for Vth is satisfied. However, when we have control over GBP, we can manipulate things effectively to minimize slewing problems as we can by changing Vth.
A test circuit which can be used to demonstrate the relationship of (58) is shown in Fig. 44. Here Al and the associated components form a 100-kHz single-pole filter, which drives the D.U.T., connected in an inverting circuit. This allows direct observation of the error voltage, thus this monitor shows directly when Vth is exceeded. The error voltage of the D.U.T. is buffered by A2, a high-speed FET amplifier, which furnishes a voltage gain of 10 to aid observation of low error voltages without loading the summing point. Rx is used to adjust the feedback of the D.U.T. test amplifier. A small (10 pF) feedback capacitor is used to minimize HF phase errors (which can obscure detection of slewing near threshold). To check the validity of (58), a hypothetical amplifier stage was set up to pass a 6-V p-p output signal, after being filtered by the 100-kHz input filter. (Such a stage, for example, could represent the last stage of a preamplifier, and the numbers quoted are reasonable design figures.) A 301A compensated for unity gain with a resulting SR of 1 V/ NS and GBP of 1.5 MHz was used, with Rx open. The results for this device are shown in Fig. 45.
Fig. 47-Transient response of TL070, operated inverting with unity-gain
compensation, Cc = 33pF, GBP = 1.5 MHz, Vth = 0.67 V, to 20-kHz square wave
filtered at 100 kHz. Top traces are error voltages, bottom traces outputs.
A, slew-limited response; B, nonslew limited response. (Scales: 5 p S/cm
both; out puts both at 5 V/cm; error voltage, A, 1 V/cm, B, 0.5 V/cm.)
The bottom trace of this photo, 45A, is the output, which as can be noted is severely slew limited for the 6-V p-p level.
The error voltage (top) is 1 V peak in level, well in excess of Vth, a confirmation that slewing is present in the output.
If (58) is an accurate predictor of slew suppression, it should be possible to adjust this stage to a point where slewing is not present.
If (58) is rewritten in terms of Vpp, as-< Vth GBP (60) fc we should be able to calculate a Vpp below which this is true, for this circuit. Equation (60), with the substitution of the appropriate conditions, indicate that slewing should disappear below 1.5 V p-p, the level where Vth is 0.1 V. A photo for these conditions (displayed similarly) is shown in 456. As the output level and Vth indicate, slewing is just barely discernible in the output waveform (bottom). For levels below 1.5 V it will be absent; above 1.5 V it will appear with increasing degree, with increasing amplitude.
Equation 60 can also be used to adjust GBP to a point where higher output levels are possible without slewing.
With the same 301A compensated with 5pF, its GBP became 10 MHz, which should allow the 6 V p-p output to be realized. For stability, Rx must become 1.2 K for this test.
The results, shown in Fig. 46, indicate that a 6-V output is realized without slewing. As can be noted, the error voltage is under 0.1 V (top) for this condition, indicating that operation is conservatively below the slew limit level. Equation 60 actually predicts a 10 V p-p output before slew limiting is reached.
Another demonstration of how the relationships of (58) and (60) operate is possible by using an amplifier with a radically different Vth to see if it predictably follows a similar pattern. This was done for a TL070 device, which for a similar compensation capacitance of 33 pF also has a 1.5 MHz GBP. However, due to its higher Vth of 0.67 V, the SR for this device and condition is 6.7 V/ NS. As should be noted, these conditions produce a test amplifier with 6.7 times the Vth and SR over the 301A. Figure 47A shows the output/error voltages for the TL070 compensated as noted for a 20-V p-p output. Slewing is evident in the output (bottom) and indicated by the 2-V peak error voltage (top) which is in excess of Vth. Equation (60) predicts that slew limiting should disappear below a 10-V p-p output, which is shown in 47B. Note that the error voltage is just over 0.6-V peak, and slewing is just barely noticeable in the output (bottom). If this amplifier is adjusted for a higher GBP, as was done in the 301A case in Fig. 46, it shows a similar improvement.
For this 10-MHz GBP condition, the output predicted by (60) would be 67 volts p-p or in excess of the supplies. The results at a 20-V level are shown in Fig. 48, and there is no slewing detectable at all.
It should be noted that these two examples do indeed demonstrate similar adherence to the relationship described.
If the results are compared for conditions where the error voltage is at the Vth level, for example Figs. 45B and 47B, it can be noted that although the two output levels produced are different (due to different SR and Vth), the error voltages are of a similar percentage of the output or about 6.7 percent.
This demonstrates that it is, indeed, possible to satisfy a common criteria (SR>SS) by different means, with similar errors by the different routes taken.
Another way of stating this is to rephrase an earlier statement, that Vth in itself is not a single totally important parameter-it is important to this subject to the extent it affects SR and input overload. The relationships set down in (58) and (60) are somewhat deceptive in this regard, as they do not contain an SR term. However, it should be remembered that these two relationships are fundamentally based on an SR criterion and, as such, contain terms which are useful towards manipulating or maximizing SR. In a very broad perspective, it should also be understood that it is incomplete to imply that input dynamic range, Vth, or other similar conceptual terms describe the entire situation in terms of a no-slew-limit guarantee, for they do not. As the experiments just described have demonstrated, even a low Vth amplifier can be effectively used. If its operating conditions are set up to provide an SR>SS, the obvious stewing distortion can be suppressed.
There is a great deal more which can be said about specific amplifier operating conditions and methods of suppressing SID by guaranteeing SR>SS. Unfortunately, however, the scope of all of these factors might be a complete article or series in itself. Therefore, we will limit comments on these points to the highlights at this time.
What the relationships just discussed show is that when the output of an amplifier stage is, by design, purposely confined to signal slopes less than the SR of that amplifier, the amp will not slew limit. Further, if SR is maintained greater than SS for all output levels up to (or above) the clipping level, the amplifier will not slew limit for any input level below clipping.
While this was demonstrated with a model consisting of a separate input filter followed by the amplifier under test, it also holds true when the filter is integral to the amplifier, i.e. the amplifier is an active LP filter. An amplifier can, in fact, be designed in this manner for slew suppression, as described by Leach [10]. However, the conversion of an amplifier to an integrator at high frequencies will usually result in more compensation being necessary for stability, hence there can often be little net improvement for this approach.
Fig. 48 Transient response of a TL070, operated inverting and adjusted for
slew suppression, Cc = 5 pF, Rx = 1.2 k.
Top trace is error voltage at 0.1 V/cm, bottom trace is output voltage at 5 V/cm.
Fig. 49 Transient response of a non slew-limited amplifier design, loaded
with 8 ohms, to a 10-kHz square wave.
Top trace is input at 2 V/cm, bottom trace is output at 20 V/cm.
In practice, effective control and design freedom are also realized when the slope limiting filter is placed before the amplifier. This allows reduced compensation and a high SR in the amplifier, with complete control of maximum signal slope by means as simple as a single RC input section.
An example of a power amplifier design based on these principles is described in reference [71], and it is worth noting that a commercial design [72] following these principles has received some good marks from audiophiles and subjective reviewers. To illustrate the point that this amplifier is indeed a non-slew-limited design, a full-level output (80 W) square wave from it is shown in Fig. 49, along with the input square wave. It is clear that the response is small-signal bandwidth limited only, and the 6 µS risetime does not, in fact, vary as a function of level.
The design techniques and experimental data described above for reduction of SS by pre-filtering at the amplifier input have all been based upon single-pole, low-pass filters.
While this type of filter has been shown to be quite effective for control of SS, and thus prevention of slew limiting, more sophisticated filter techniques are even more effective in reducing SS. It has been shown [12, 66, 67, 70] that higher order filters are even more effective for reduction of SS, compared to a simple first-order type, for a given cutoff frequency. There are, of course, trade-offs to be made in comparing one to the other, considering the higher performance against the increased complexity. Also, the damping of the filter must be considered, as well as its frequency. However, the increased complexity of a second-order filter really depends on exactly how it is realized and may not in fact be prohibitive. For example, Leach has shown in [66] how the amplifier itself can be used as the active portion of the prefilter, without undue stability constraints, in what appears to be a practical and attractive topology. Further, in [70] it is shown that a second-order Bessel LP filter alignment will produce approximately 1/2 the SS of a single-pole filter for otherwise similar conditions. Unfortunately, time did not permit detailed experimentation with these techniques for this article, but they appear to have significant merit towards the reduction of SID effects.
Generally, the above discussion describes two alternate means which can be used to design a non-slew-limited amplifier and thus prevent SID and TIM. A logical question which may be raised is, do they yield equal results in auditioning? While we do not at this point have subjective response data to definitively answer this question, informal listening tests by one author (W.).) tend to favor circuit topologies which are designed from a standpoint of equation (59), using linearized input stages, such as FET or degenerated bipolar devices. As time progresses, it is hoped that further listening tests will more clearly define the optimum choice between the two approaches.
Conclusions In this article we have attempted to cover a quite broad topic from a multiplicity of viewpoints, in both discussion and analysis. These different techniques of analysis all indicate a common pattern of distortion in feedback audio amplifiers, which is a function of the ratio of signal slope to amplifier slew rate, a dimensionless parameter we define as SR ratio. When the SR ratio is less than unity, this distortion is suppressed; when greater than unity, strong nonlinear distortion products appear, which are subjectively objectionable.
Control of this distortion, which we call SID, can be achieved by maintaining linear amplifier behavior, with an SR greater than the highest SS, or stated in terms of SR ratio, an SR ratio less than unity. Since SS is both frequency and amplitude dependent, it follows that greater SR in an audio amplifier is required for higher voltage output stages, where the SS is highest.
Control of this distortion can be exercised by appropriate selection of amplifier type, by specifying an SR sufficient to the application. In design, it can be achieved by providing a sufficiently conservative SR (on the order of 0.5 V to 1V/ NS per peak output volt) or by designing for a non-slew-limited response. A non-slew-limited amplifier has an inherent SR greater than its maximum possible output SS and will therefore never slew for any input signal, including square waves, or its SR ratio is guaranteed <1. It is characterized by frequency response which is small-signal-bandwidth limited, for any output below its clipping level. As such it has no major nonlinear distortion products due to slewing effects.
Such an amplifier is also said to be TIM free and may be described in this context as well.
It is recognized that there is considerable controversy on the relative importance of TIM and SID, their audibility, and some of the relevant design criteria. For this discussion, it is not our purpose to dwell excessively on the relative importance of SID, its audibility or other factors which are often subject to opinionated views. What we seek to do is describe means to quantify and control this distortion mechanism, and basically this is the only main point being addressed.
The existence of the distortion mechanism is, of course, not a subject of debate, and like other distortions in amplifiers, knowledge of its behavior patterns is valuable to either the circuit designer or the informed user of audio equipment.
We would, however, like to express caution with respect to certain alarmist commentary, for example those to the effect that low TIM or minimal SID is the magic elixir of quality audio. While this distortion source is quite important, so are many others. Once sufficient linearity and slew rate have been provided in a design, there may actually be little gained by boosting SR further (to far beyond that necessary). The optimum audio amplifier is best designed with all contributions to audible defects given proper perspective.
We appear to currently be immersed in a specifications race on the part of some manufacturers in this regard, which is not only unfortunate for the confusion it spreads (as to what is most important), but doubly so from the standpoint that if nonlinear techniques are being used to achieve high SR numbers, the user can actually pay a penalty in higher distortion! Another specifications race practice appears to be the quotation of amplifier maximum output SS for small signal condition as its specified SR. If an amplifier is operating linearly in non-slew-limited conditions, the output SS for a fixed signal will linearly follow the output level, and at no point will it reach the true amplifier SR, which is, in fact, a limit. It is therefore erroneous or misleading to quote a maximum SS as an SR in such a case, as the true SR limit is never reached. In our opinion, while such an amplifier has real merit, it might more clearly and suitably be described in such terms as "maximum linearly reproduced SS" or the qualifier added that it is a true non -slew -limited design, as described in the text. Using the terminology of SR implies that the amplifier can be made to slew; if, in fact, it cannot be made to slew this should be clearly stated, for it is a point which distinguishes the design.
(In Part II on page 44 in July under "Comparison of Test Methods," we made the statement that the square wave's fundamental amplitude was 12 dB larger than the sine wave's. The square wave itself is 12 dB larger in amplitude, as described in the sidebar.) We hope this discussion has served to bring together some of the various issues involved so as to create a new perspective for the reader. We recognize that some of the points made in this article have been made elsewhere and acknowledge the work of previous authors. We believe that the extensive bibliography will be helpful to the reader to appreciate this material, and to tie older data in with the new material presented within this article.
(Source: Audio magazine, Aug. 1979)
Also see: Phase, Time, Ears & Tape (Apr. 1979)
= = = =