Measurements of Amplifier Peak Output Voltages Under Dynamic Conditions and Into a Real Loudspeaker Load - Part II.
by Peter van Willenswaard

Click here for Part I

(An earlier version of this story was published in Stereophile, Sept. 2000. I now expanded the score from three to five amplifiers, redid all the measurements using the same vertical scale in order to make comparisons easier and added an analysis of the effect of feedback)

(In Part I we explored the behaviour of various amplifier/loudspeaker combinations with sinewaves, artificial and musical transients. Valve amplifiers, and the Single-Ended variety especially, were shown to vastly outperform their transistor cousin under musical conditions. The conclusion was:

A transistor amplifier performs similarly into either a resistive or a real loudspeaker load. Externally augmenting its traditionally very low output impedance (high 'damping factor') to a value of 3.5 Ohm as commonly found in non-feedback valve amps helps mimicking valve amp behaviour. Valve amps prove capable of peak output voltages belying their nominal power output figure by a factor 2 to 7! Apparently, amplifiers that have 'ideal' near-zero output impedance waist energy that the loudspeaker has in store and is ready to use for building up extra sound pressure, by short-circuiting this energy out.

The feedback issue was also looked into and added some more puzzles. But feedback or non-feedback, most valve amps laugh in the face of the poor solid-state amp, even with the external 3.5 Ohm applied.)

In Part II we will try to look a little deeper into the material, by examining the behaviour of the current under musical conditions, and compare that to the voltage shape. Also presented are the results of a different voltage/current bias for 300B SE.

Let us take a look again at the voltage output of the 300B SE on the tambourine stroke that occurs 5 seconds into the 1rst track of the 'Touch' CD.

 

Figure 1

Fig 1: 11 W 300B SE valve amp loaded with AN-E loudspeaker, signal from 'Touch' CD at 0:05 into first track, driven to maximum produced 37 Vp (negative). Scale is 10 V/div. 

There is heavy voltage clipping on the positive side, but voltages going negative seem to meet no limit whatsoever. So what happens in the current domain? A 0.1 Ohm resistor was inserted in the loudspeaker return and the voltage across it measured. This voltage is a direct translation of the current circulating in the amp-output/cable/loudspeaker loop, 100 mV representing 1 Ampere. The current complement of fig. 1 looks like this:

 

Figure 2

Fig 2: current measured with inserted 0.1 Ohm resistor, so 100 mV represents 1 A; scale is 100 mV/div. Same amp, speaker and stimulus as in fig. 1.

 

Here we find the opposite! With possibly an exception for the onset of the tambourine stroke, the current is clipping where the voltage looks without limit, and vice versa. Also, the current looks somewhat like an momentaneous integration of the voltage. Both observations suggest an inductive behaviour in either speaker or output transformer (or both). But remember from Part I that when loaded not with the speaker but with a pure 8 Ohm resistor the 300B SE reached only 19 Vp with this signal, not significantly more than the 15 or pushed very hard 17 Vp reached with sinewaves. So the contribution from the output transformer is probably minimal. But look at the 3 A peaks, even if they are not wholly be generated by the 300B, it still has to deal with them and it does! (11W sine into 8 Ohm means only 1.9 Ap.)

Note that the 3 A peaks coincide with an 18 V voltage level, suggesting 6 Ohms momentaneous impedance, in agreement with the voice coil DC resistance once the inductance effect has faded out. On the negative half we see 36 Vp and 1.8 Ap, ie. 20 Ohm momentaneous, but here the current keeps clipping and the inductance remains active. A loudspeaker can't be that asymmetrical. This must have to do with the Single-Ended and thus asymmetric amp.

How do the other amps fare? A few examples:

 

Fig 3: 4 W EL84/triode-mode PP valve amp loaded with AN-E loudspeaker, signal from 'Touch' CD at 0:05 into first track, driven to maximum produced 15 Vp. Scale is 10 V/div.

 

Fig 4: current measured with inserted 0.1 Ohm resistor, so 100 mV represents 1 A; scale is 100 mV/div. Same amp, speaker and stimulus as in fig. 3.

This looks like two positive halves of SE coupled together in mirror fashion - which it is, as this is a Push-Pull amp! Nothing new or unexpected here. Applying 13 dB of feedback around this amp (and adjusting the input level to compensate for the decreased sensitivity) hardly changed anything: only the small 'overshoots' in the current waveform became less pronounced.

Also no surprises with the Dynaco SCA-35:

 

Fig 5: Dynaco SCA-35 EL84 PP Pentode loaded with AN-E loudspeaker, signal from 'Touch' CD at 0:05 into first track, driven to maximum produced 27 Vp. Scale is 10 V/div.

 

Fig 6: current measured with inserted 0.1 Ohm resistor, so 100 mV represents 1 A; scale is 100 mV/div. Same amp, speaker and stimulus as in fig. 5.

 

Nor with the Saba:

 

Fig 7: Saba Telewatt ECL82 PP Pentode loaded with AN-E loudspeaker, signal from 'Touch' CD at 0:05 into first track, driven to maximum produced 17,5 Vp. Scale is 10 V/div.

 

Fig 8: current measured with inserted 0.1 Ohm resistor, so 100 mV represents 1 A; scale is 100 mV/div. Same amp, speaker and stimulus as in fig. 7.

 

But now take a look at the 25 W transistor amp:

 

Fig 9: 25 W transistor amp loaded with AN-E loudspeaker, signal from 'Touch' CD at 0:05 into first track, driven to maximum produced 22 Vp. Scale is 10 V/div.

 

Fig 10: current measured with inserted 0.1 Ohm resistor, so 100 mV represents 1 A; scale is 100 mV/div. Same amp, speaker and stimulus as in fig. 9.

 

The inductance effect in the current waveform is almost gone, the current it trying hard to look as much like the voltage as it can! Is this because the amp's 0.1 Ohm output impedance totally controlling what the speaker is allowed to do? No, for when the amps' output impedance is artificially increased with an external resistor (outside the feedback loop) to 3.5 Ohm, the value of the 300B SE or that of the EL84 PP, the character of the current waveform does not change:

 

Fig 11: 25 W transistor amp with 3.5 Ohms in series with output, system loaded with AN-E loudspeaker, signal from 'Touch' CD at 0:05 into first track, driven to maximum produced 22 Vp. Scale is 10 V/div.

 

Fig 12: current measured with inserted 0.1 Ohm resistor, so 100 mV represents 1 A; scale is 100 mV/div. Same amp, speaker and stimulus as in fig. 11.

 

Look back at the feedback valve amp, the Dynaco, the Saba, the HF-309 with feedback applied. They all feature an output impedance of about 0.5 Ohm, not as low as the transistor's 0.1 Ohm but close and a lot lower than the artificially created 3.5 Ohm. So, although we found in Part I that an externally increased output impedance helps a transistor amp to mimic some of the voltage behaviour found in valve amps, the current waveform clearly tell us whether it's valve or transistors. Conclusion: what we see here, is the output transformer at work, feedback or no feedback.

Now all of the above has been measured with one stimulus (tambourine) only, so it would be interesting to see what happens at lower frequencies. The 'Touch' CD opens with a hit on a big drum; the tambourine sounds like 'toc', the drum is 'booom'. As before, I adjusted the levels to on the edge apparent subjective distortion. Here is what the 300B SE makes of it:

 

Fig 13: 11 W 300B SE valve amp loaded with AN-E loudspeaker, 'booom' signal from 'Touch' CD at 0:01 into first track, driven to maximum produced 26 Vp (negative). Scale is 10 V/div.

 

Fig 14: current measured with inserted 0.1 Ohm resistor, so 100 mV represents 1 A; scale is 100 mV/div. Same amp, speaker and stimulus as in fig. 13.

 

The voltage peaks are now closer to the levels reached with sinewaves, but even where it clips (positive side) the 300B reaches a few Volts more with this low frequency dynamic signal. The current peaks are impressive: how about 4 Amps from a 300B?

I took the Saba Telewatt as an example of the Push-Pull camp:

 

Fig 15: Saba Telewatt ECL82 PP Pentode loaded with AN-E loudspeaker, 'booom' signal from 'Touch' CD at 0:01 into first track, driven to maximum produced 15 Vp. Scale is 10 V/div.

 

Fig 16: current measured with inserted 0.1 Ohm resistor, so 100 mV represents 1 A; scale is 100 mV/div. Same amp, speaker and stimulus as in fig. 15.

 

Only 2 V less than with the tambourine, but remember that the Saba was the least spectacular of all valve amps. But 3 Ap (fig. 16) from a tiny ECL82 PP is nice!

 

Bias considerations for the 300B SE

Given the voltage clipping of the 300B one the positive half and the current clipping in the negative half of the signal, one might wonder if different bias conditions for the 300B valve would enhance its dynamic performance.

The setup of the 300B SE amp used is rather straightforward. The input valve is (half) a 6SN7, RC coupled to (half) an E182CC feeding a 1:1 interstage transformer (ITX) in its anode circuit. The E182CC runs at 200 V and 15 mA. The secondary of the ITX connects the grid of the 300B to an adjustable negative supply (about -80 V) and the 17.7:1 output transformer (OPTX) connects the 300B anode to a +420V high-voltage. The 300B filament is bridged with two 47 Ohm resistors tied to ground via a 1 Ohm resistor to allow monitoring of the standing current through the 300B on a 100 mV meter.

The OPTX is phase-inverting: when the anode goes down, the hot speaker binding post goes up. When tested with a sinewave (amp loaded with 8 Ohm) the anode first clips positive, at +600 V a voltage excursion of 200 V above the +400 V idling value; pushing harder just produces more flattening of the sinewave top. Meaning the valve is into cut-off, the anode current has become zero. To the negative side, however, a voltage excursion of 300 V is easy, the anode eventually clipping at +80 V.

So it was tempting to lower the high-voltage supply and increase the anode current reserve. I went down to +290 V and upped the anode current from 75 to 100 mA. Clipping on sinewaves now became more symmetrical (promising!):

 

Fig 17: 300B SE valve amp, V+ 290V, Ia 100 mA; loaded with 8 Ohms, sinewave just into clipping, 12.5 Vp.

 

But Vpp had sunk a little to 25, implicating a power output of 9.8 W against 11 W originally. Well, lower voltage higher current means that probably a lower anode load impedance would be needed for optimal power transfer, meaning a different OPTX. A 16 Ohm tap loaded with an 8 Ohm resistor would have done the trick, but there was no such tap on my OPTX. Loading the 8 Ohm tap with a 4 Ohm resistor would also cause a lower primary impedance. I measured that, it generated 19-20 Vpp max. A 16 Ohm tap then would have delivered 1.41x this voltage into 8 Ohm, meaning some 27.5 Vpp, a 10% increase in voltage and hence about 12.5 W of power output. So, yes, another OPTX would have been better in this case, preferably one with several primary and secondary taps in order to find the optimal match.

Nevertheless, let's take a look at dynamic performance:

 

Fig 18: 300B SE valve amp, V+ 290V, Ia 100 mA; loaded with AN-E loudspeaker, signal from 'Touch' CD at 0:05 into first track, driven to maximum produced 34 Vp (negative).

 

Fig 19: current measured with inserted 0.1 Ohm resistor, so 100 mV represents 1 A; scale is 100 mV/div. Same amp, speaker and stimulus as in fig. 18.

 

Vpp has gone down from 57.8 to 48.8 V (compare to fig. 4 in Part I), mainly due to an almost 6 V sacrifice in the positive half of the screen. Even the 10% correction (see comment following fig. 17) would make up for only half of the decrease. How unexpected! Does this mean we should have gone the other way: more high-voltage, less anode current, higher impedance OPTX? Difficult to say, as raising the supply voltage on an existing amplifier isn't easily done.

The currents in fig. 19 are lower too, in accordance with fig. 18, but current CLIPPING has remained at -1.9 A. Strange.

All in all, it doesn't look like an improvement, alas. A multi-tap OPTX would be needed to explore this further.

Let's take another angle. The positive voltage clipping under dynamic conditions, at the output, loaded with a speaker, corresponds with the negative clipping limit of the anode at an absolute +80 V. Compare figs. 17 and 18 (400V/75mA or 290V/100mA is of no importance here). And remember the OPTX is inverting. Now what happens in the negative output half? The anode speeding up to +600 V runs out of current upon reaching that point, but the OPTX's inductance doesn't accept that sudden current change and reacts with a voltage surge: hence the large negative output peaks (by the way, the primary and secondary waveforms were identical apart from being out of phase; just to be sure nothing silly is happening inside the OPTX I checked this with an oscilloscope). This seems logical, but why doesn't this happen when loaded with an 8 Ohm resistor? Not even when driving with short sine bursts, leaving the 300B no time to drift away from its set bias (tried in Part I)? A second big question arises however from the current plot. If in the negative half the voltage overshoots, why is the current clipping there, and not following? Because, if the overshoot is produced by energy reflected by the speaker, then why don't we observe it in the transistor+3.5 Ohm case?

If this is a unique speaker+OPTX thing, than what mechanism is at work there, how can we get our fingers behind this?

Questions in abundance, but no answers as yet.

All material copyright Audio Note (UK) Ltd., unless otherwise stated. Audio Note (UK) Ltd. reserve the right to change, modify, add to, or remove any of the content of this site at any time without prior notice. Further, images on this site may differ from current models or specifications.