To Filter or to not Filter: Part 3
An Op-Amp digression
In our previous part 2 article, we mentioned the “O-Word” already, so let's digress a little.
The Op-Amp is represented by the Triangle with the + and – sign inside in the triangle in the active filter above.
When you read some advertising copy in high-end audio you may believe that Op-Amps are the latest and best thing since sliced bread and one might even believe some of the fellas using them invented them. But Op-Amp's are really old.
The first functional design is normally attributed to Karl D. Swartzel Jr. of Bell Labs in 1941 and was used in a Radar assisted Artillery director during World War 2.
The word “Operational Amplifier” for the type of amplifier circuit it embodies is documented in 1947 and the first “integrated op-amp” using tubes was introduced in 1953
Historical timeline
1941: A vacuum tube op-amp. An op-amp, defined as a general-purpose, DC-coupled, high gain, inverting feedback
amplifier, is first found in
U.S. Patent 2,401,779 "Summing Amplifier" filed by Karl D. Swartzel Jr. of Bell Labs in 1941. This design used three
vacuum tubes to achieve a gain of 90 dB and operated on voltage rails of ±350 V. It had a single inverting input rather than differential inverting and non-inverting inputs, as are common in today's op-amps. Throughout
World War II, Swartzel's design proved its value by being liberally used in the M9
artillery director designed at Bell Labs. This artillery director worked with the SCR584
radar system to achieve extraordinary hit rates (near 90%) that would not have been possible otherwise.
[14]
All the way through to:
1972: Single sided supply op-amps being produced. A single sided supply op-amp is one where the input and output voltages can be as low as the negative power supply voltage instead of needing to be at least two volts above it. The result is that it can operate in many applications with the negative supply pin on the op-amp being connected to the signal ground, thus eliminating the need for a separate negative power supply.
The LM324 (released in 1972) was one such op-amp that came in a quad package (four separate op-amps in one package) and became an industry standard. In addition to packaging multiple op-amps in a single package, the 1970s also saw the birth of op-amps in hybrid packages. These op-amps were generally improved versions of existing monolithic op-amps. As the properties of monolithic op-amps improved, the more complex hybrid ICs were quickly relegated to systems that are required to have extremely long service lives or other specialty systems.
Recent trends. Recently supply voltages in analog circuits have decreased (as they have in digital logic) and low-voltage op-amps have been introduced reflecting this. Supplies of ±5 V and increasingly 3.3 V (sometimes as low as 1.8 V) are common. To maximize the signal range modern op-amps commonly have rail-to-rail output (the output signal can range from the lowest supply voltage to the highest) and sometimes rail-to-rail inputs.
(source wikipedia:
http://en.wikipedia.org/wiki/Operational_amplifier#Historical_timeline).
Over the years the Op-Amp has slowly become the default building block for Amplifiers, both discrete and integrated and for the last few decades Op-Amps have been the dominant choice when it comes to “general purpose” Audio amplifiers (and outside audio).
They are manufactured in a huge variety at all price levels and with a massive array of different specifications and optimisations for specific jobs. Some are even sold as being optimal for audio.
Unfortunately, this ubiquity has also led to a lack of understanding of what happens inside these Op-Amp's. There are circuit structures that are common to almost all Op-Amps, there are inherent limitations that are too.
These days we pick up the Datasheet for an Op-Amp and we read:
“The XXX Op-Amp is a JFET-input, ultralow distortion, low-noise operational amplifier fully specified for audio applications. Features include 5.1nV/√Hz noise and low THD+N (0.00005%).”
Surely it ticks all boxes?
Come on, zero point how many zeros THD?
Noise in Nanovolts?
It even says “J-Fet” there and “J-Fets” are in fashion this year we hear.
It's even a special audio grade part. So let's just use that one, okay?
Actually, at iFi WE DO USE THAT ONE.
But not because of these numbers. Or the J-Fets. Or the “Soundplus” moniker.
Sorry to disappoint but our reason is more prosaic. Correctly implemented it sounds as good as anything we have tried and much better than most.
There are some other numbers for that chip that are not headlined. They have more to do with what happens in the real world.
For example, the bandwidth is only 11MHz at no gain and gets substantially less wide as gain is increased. And while the distortion is low at 1kHz and under ideal conditions, there are many things that make it worse, not the least raising the frequency so at 20kHz we have wipe off one zero of that very low distortion figure and at several MHz three to four of the zeros.
It also is not so great with low impedance loads, without adding buffers distortion goes up, so strike another zero if we drive 600 ohm. Actually, the gain is very load dependent!
Now it is not such a low distortion device anymore, is it?
One thing we do not really want to do is to make this Op-Amp filter signals in the region of several MHz or have it driving headphones. We have to take that over by different means if we want to use this Op-Amp for it's undeniable qualities.
If we understand the limitations and possible problems we can design our circuits accordingly and avoid the pitfalls of the limitations and take best advantage of the exceptional audio performance.
If we are simply members of the “Op-Amp of the month” club, we may get all sorts of results, maybe good, maybe bad, maybe indifferent.
Having Op-Amps or not is not a reliable indicator of quality.
Even the best Op-Amp's in the world can be implemented so ham-fisted that the result is poor, using Op-Amps that seem rather old and pedestrian correctly can give surprisingly good results in the real world.