Voltage source vs. current source amplifiers

Thank you Al, as always. I finally got it.

Hi Kalali,

As I had mentioned a theoretically ideal voltage source has an output impedance of zero. A real world amplifier cannot have an output impedance of zero, but what allows any amplifier to act as a good approximation of an ideal voltage source is an output impedance that is close enough to zero to be an essentially negligible fraction of the impedance of the speaker.

As you realize nearly all solid state amplifiers do not have output transformers or output autoformers, and consequently do not have multiple output taps. And their rated maximum power capability into 4 ohms will, in almost all high quality designs, be considerably more than their maximum rated power capability into 8 ohms. But while that fact is an **indication** that the amp will act as a voltage source (when operated within the limits of its maximum voltage, current, power, and thermal capabilities), it is not the **reason** the amp will act as a voltage source. The reason is the amp’s very low effective output impedance.

And the McIntosh solid state amplifiers which have output autoformers have effective output impedances that are low enough to be essentially in solid state territory, as opposed to tube amp territory. While as you realize they do not provide the increase in maximum rated power capability for low impedances that is typical of solid state amps. And as I indicated earlier the reason they don’t is that their maximum power ratings are based on the assumption that an 8 ohm load will be connected to the 8 ohm tap, a 4 ohm load will be connected to the 4 ohm tap, and a 2 ohm load will be connected to the 2 ohm tap (with the taps presumably designed such that the output stage of the amp that is "ahead" of the autoformer will see a load impedance that is the same or very similar in each of those three cases, that apparently being the design philosophy McIntosh has chosen for these amplifiers). But what matters with respect to the voltage source question is how the amp will behave when a load impedance that varies as a function of frequency is connected to **one** of those taps. And that behavior depends on the relation between load impedance and amplifier output impedance, as long as the amp is operated within its capabilities.

Also, as I indicated the higher the amp’s damping factor is the lower its effective output impedance will be. The specs you cited for your amp indicate a somewhat lower damping factor for the 8 and 4 ohm taps than I’ve seen for most solid state McIntosh amps that have output autoformers (which as I indicated are usually spec’d at ">40"). But even in the case of your amp’s 8 ohm specified damping factor of only 16, that corresponds to an output impedance of 8/16 = 0.5 ohms. Which is lower than the output impedance of the great majority of high quality tube amps, and is only slightly higher than the output impedance of many solid state amps, and is a small fraction of the impedance of an 8 ohm speaker.

Ergo, while your amp does not approximate a pure voltage source quite as closely as most other solid state amps, or quite as closely as some other McIntosh amps which incorporate autoformers, it still comes pretty close. And as I said, no amplifier can act as a theoretically perfect voltage source. All a real world amplifier can do is to closely approximate one.

Regards,
-- Al

In broad sense, it appears that the amplifiers that increase their power as the load impedance drops - ideally double as the load halves, are voltage sources and the ones that maintain the same power, or close to it, regardless of the load impedance are current sources. Going by the specs, it looks like almost all the tube amplifiers fall in the current source category. I’ve also read that most ESL design speakers such as Martin Logan generally act as a capacitor so a voltage source amplifier would be a better choice. In contrast, magnetic/ribbon design speakers such as Magnepans are resistive so a current source amplifier would be a better option. I guess dynamic speakers fall somewhere in between. Looking in the market, it looks like McIntoch is one of the (very) few companies that is making a current source non-tube amplifier. I’m wondering why a lot of folks find McIntosh very nicely paired with ESLs like Martin Logans. Similarly, Magnepans should pair very well with tube amps.

Most of this is false although tubes often do work well with Maggies but not for the reasons expressed.

First, a voltage source is an amplifier that can make constant voltage regardless of load. No amp can actually do this, but they can come pretty close. This is the phenomena where the power is doubled as the load (speaker) impedance is halved.

A current source amp is an amplifier whose output impedance is significantly higher than that of the load and so makes constant current regardless of load. This is not a tube or solid state amp in particular; to achieve this, the amp must employ current feedback, which is a bit different from voltage (traditional) feedback.

Actually no dedicated commercial amp has been made (at least to my knowledge) employing this principle (although Fisher and a few others had a variable feedback control on some of their amps in the 1950s that allowed for constant current operation).

There are however a large number of amplifiers that behave as 'power sources' (wherein they attempt to make constant power regardless of load). Tube amps without feedback (and tube amps where voltage and current feedback are equally applied) fall into this category.

If a tube amplifier employs a large amount of feedback (about 15db or more) it may well be acting as a voltage source.

So the topic is a bit more complicated than the OP suggests!

Further, regarding the Martin-Logan loudspeakers, these are ESLs and so their impedance curve is based on a capacitor. Its important to note that the efficiency of the loudspeaker is **not** a function of its impedance curve; for this reason a voltage source is inappropriate as it will make far too much power into the higher frequencies (which have a very low impedance) and not nearly enough into the lower frequencies (which have a high impedance). A power source, one that is capable of driving the load, will produce a flatter frequency response. Please note the caveat.

Some amps that are considered voltage sources with most speakers will do alright on MLs simply because they fail at being a voltage source on the load that the ML presents (often 0.5ohms at 20KHz, very difficult for many amps to drive, especially if a speaker cable is involved) so you can sort of get away with it with some amps. As a result, audition in mandatory!

Magnetic planars also have an impedance curve that is not based on a driver in a box. But they often have fairly resistive and flat impedance curves which if that is so, all that is required is an amp that can make the power.

Box speakers for the most part are designed to work with voltage sources (the idea of an amp being a voltage source being championed by MacIntosh and EV in the late 1950s; in this case the impedance curve is also a chart of its efficiency), but some box speakers are designed to work with power sources (horns, single-driver full range, some acoustic suspension and some bass reflex designs). Power source technology (which I call the Power Paradigm) is the prior art and is still around because many designers in high end audio recognize that the application of loop negative feedback can have deleterious effects (brightness and harshness) and so eschew the technique altogether.

To my knowledge the only speakers for a current source amp are open baffle dynamic driver loudspeakers. Nelson Pass has experimented with this technique and you can find information about that on his websites.

For more information about the voltage and power paradigms see
http://www.atma-sphere.com/Resources/Paradigms_in_Amplifier_Design.php

Regarding the McIntosh solid state amps that have identical maximum power ratings into 8, 4, and 2 ohms, paradoxically they are best considered to be voltage source amps. They are a special case among solid state amps due to their use of autoformers at their outputs. Their max power ratings reflect the assumption that an 8 ohm load will be connected to the 8 ohm tap of the autoformer, and a 4 ohm load will be connected to the 4 ohm tap, and a 2 ohm load will be connected to the 2 ohm tap. However, for a given tap the current and power delivered will increase essentially in proportion to a decrease in load impedance, with voltage being held essentially constant, as long as the current drawn and the power delivered do not exceed the maximum capability of that tap into that impedance. Although the amp’s **maximum** current and **maximum** power capability into a 4 ohm load that is connected to the 8 ohm tap will presumably be less than if that 4 ohm load were connected to the 4 ohm tap. Likewise for a 2 ohm load connected to the 4 ohm tap vs. the 2 ohm tap.

All of that can be seen based on the amp’s specified damping factor, which is ">40" for most or all of the solid state McIntosh designs which utilize output autoformers. A damping factor of 40 corresponds to an effective output impedance of approximately 8/40 = 0.2 ohms for the 8 ohm tap, and 4/40 = 0.1 ohms for the 4 ohm tap, and 2/40 = 0.05 ohms for the 2 ohm tap. Those numbers are low enough to essentially result in voltage source behavior. (A theoretically ideal voltage source has an output impedance of zero). While most tube amps have damping factors that are in single digits, and not much more than that in most other cases, corresponding to effective output impedances that are measured in ohms, not fractions of ohms.

Regarding Martin-Logan speakers, Ralph (Atmasphere) has stated in past threads that many of them have been designed with solid state amplification (having low effective output impedance) specifically in mind. Most other electrostatics, certainly including (among many others) the classic Quad ESL-57 that was designed before solid state amplification existed, are considered by most (although not all) audiophiles to be best suited to tube amplification, that will come closer to providing at least a rough approximation of constant power delivery into the speaker’s varying load impedance than to providing constant voltage. See his paper on the subject here.

Regards,
-- Al

Ralph Karsten (atmasphere) has discussed this subject in depth here. Do a search of the topic, and look for his contributions.

Only certain ML models have powered woofers - my Aerius doesn’t for example. But I agree kijanski has it right. The amp will need to double the power when the load is halved in order to maintain a constant voltage. In contrast, the tube amps almost always either maintain the power or in some cases even increase the power when the load is reduced. What prompted me to read up on this was my current set up where I have my McIntosh MC2200 - 200 wpc @ 8,4,2 ohms, driving my Aerius MLs and I have a Vincent - 150 wpc @ 8 and 300 wpc @ 4 ohms, driving my Def. Tech. BP10s. The "science" seems to suggest that I have to swap amps/speakers but the MLs sound better with the MC2200. I’m sure there are other factors in play but it was worth the experiment. Mainly wondering if others have found a correlation between the amp design and the type of speakers, etc.

Ideal voltage source provides voltage that is independent of the load.
Power delivered will be inversely proportional to load impedance.

Ideal current source provides current that is independent of the load.
Power delivered will be proportional to load impedance.

I was a technician, not an engineer, but I do know that a tube is a voltage amplifier and a transistor is a current amplifier. The Martin Logans have powered woofers so tubes can handle the relatively easy load of the mid/treble panel. The Magnepans don’t have them so need a lot more current for the low frequencies.