Harshness, Clicks and Pops

Graham Slee wrote: Section 7.07 also derives an input stage slew rate formula, which is 0.3 x Ft, and using a common low noise opamp for phono stage use, the AD797, we find it's stable Ft is 30MHz. It's input slew rate must therefore be 0.3 x 30 = 9V/uS. It is advertised as 20V/uS, but that is its output slew rate. It is obvious therefore that it has slew rate enhancement, and that never sounds good, because to do it, the input stage must clip.

The only way to increase the slew-rate is at the input.

I can see how the CD-era "designer" could have thought - anything above 20kHz is inaudible - but the highs in analogue aren't the noise dumping-ground as in the red book. Screw around with analogue HF, and it will bite you!

God forbid, but if we ever ended up in a Soviet system, I'd hope the authorities make them design discrete power amplifiers before allowing them near op-amps.

The op-amp is a power amp in miniature. Do with a power-amp what I see done with op-amps, and you don't even get a chance of it living past switch-on! Dry sound? Are your power supply caps getting warm? What's next - Blackgates?

What you have to dig, brother is bandwidth with single-pole compensation, respect stability, and that 0.3Ft is all you're ever going to get in a low-noise world. The AD797, like the LM318, and all that follow in their footsteps, thought up cunning ways to take noise down and pump-up slew-rate.

You might be under the mistaken impression that these ICs only live under the hoods of phono stages and headphone amps - think again! They're used everywhere the opportunist manufacturer wants - I mean, look, man, it's so low noise!

Screw low noise! Stop crippling music!

Taking 0.3Ft, it begs a multiplier to give a larger result. How about if we added "m" to make it 0.3mFt? Where can we find "m"?

"m" can only exist if the linear portion of input latitude is increased. The joined emitters limit us to 42mV. What if we could make that 420mV? "m" would be 10.

If we take what can be achieved by single-pole compensation, say, a bandwidth of 4MHz, the 0.3Ft slew-rate is 0.3 x 4 = 1.2V/uS.

Multiplied by an "m" of 10, we obtain 12V/uS. And if multiplied by 20, we get 24V/uS. Isn't that great?

It's done by placing a resistor in each emitter leg such that the gm of the stage gets reduced by the value of "m." It has downsides - one being that open-loop gain has been reduced (you won't get as much negative feedback). The other is that noise increases. The reason that noise increases is because resistors "contain" noise at the rate of 4nV per /Hz per 1,000 ohms.

However, it has another benefit, not requiring such a considerable value of compensation capacitance. Therefore, if the input signal strays even further than the x10 or x20 we gave it, the overload has a much gentler time.

The technique is called "emitter degeneration."

The big problem with active stages revolves around misunderstanding slew-rate and output current.

Most op-amps oscillate into little more than 100pF and need in-loop or out-of-loop output series resistance.

Often you'll see the op-amp driving the EQ capacitors directly. The divider resistor to ground from the feedback pin might be seen as the equivalent of this series resistance, but the phase is different at that end.

The craze for zero output resistance in op-amps for headphone amps might now look a bit stupid? Headphone cables might have up to 600pF of capacitance. Those people are "enjoying" high-frequency instability, in which case, op-amp "rolling" gives a different type of instability.

One type of op-amp which can withstand capacitive loads is the bootstrapped output stage type. This type of bootstrap is nothing like the "pull-up" bootstrap of an ac-coupled power amp. It senses load capacitance and reduces bandwidth to bring it back into stable operation. It is a kind of frequency compensation, but outside the gain stage.

However, that in itself cannot prevent oscillation. Still, to avoid using a "Miller" capacitor in the VAS stage, it uses emitter degeneration to reduce the excess phase contribution by the input stage.
This is the innards of one channel of an AD826 op-amp, and instead of the input emitters being directly joined, they have series resistors. The series resistors reduce the transconductance (gm) and increase input latitude.

The slew-rate is a staggering 350V/uS. By 0.3Ft, with a bandwidth of 50MHz, it should be only 15V/uS but is 23 times higher due to the emitter degeneration resistors' size.

It tends to suggest that the input can take 1 volt before overload. That is the voltage often found in video, and that's what the op-amp is designed for. However, it is an op-amp, so that it will work down to DC, and as such, it works for audio frequencies.

What size of "C2" can it drive? With 50mA output current and by I/SR, we obtain 143pF. That is a shame, but if we can add sufficient output series resistance, we could get it to work. Placing nearly 1k in series with the high-frequency pole capacitor, it "flys" if we can accept that the RIAA curve isn't necessary at, say, 50kHz.

It was used in the original Era Gold, which might give you a clue as to how it sounds. Although the S/N was acceptable to the ear, it wasn't acceptable on paper and turned quite a few prospective customers away. I'm sure I'd still be using it to this day if it didn't have a popcorn-noise problem. Each device was tested by AD for its purpose as a video amp, and at such frequencies, that noise is not apparent.

We had 50% fails-on-test, and to continue making the Era Gold and Jazz Club, I had to redesign it in a way that didn't change the sound. To do that, you have to understand op-amps.

What you didn't see, is what's known as a "proprietary moment."

To the outsider audio can be a bit like that. Let me rephrase that - to an outsider with a serious desire to learn, audio can be a bit like that.

Everything can look oh so snobbish, but out of sight, something rather vulgar might be going on. A good example is the preamp of a dynamic mic. Oh yes you know, it's phantom powered, such a fine method of placing a DC voltage on the balanced inputs - just look at the network - look at the high tech...

Tucked away in the mic body is a single transistor preamp (or used to be). What? Trust that quality of signal to such a horrible-horrible single transistor. Don't they know such-and-such superstar will be using the mic?

By now, possibly a handful of old timers who visit my posts (hi friends!), will be getting the gist. Others might be thinking of sending me their nephew's college textbooks on how to design basic audio circuitry - it has happened - such people are beyond salvation but are easily offended by the Yorkshire pit language reply!

"How can it be that Yorkie's can learn things - surely all Northerners are dumb?"

Well, dear bigot, that might be why it took me 57 years to learn what I know.