Graham's Blog Archive
A Different Approach to Record EQ: Part 4
You’ll have heard of a thing called frequency response? But do you understand loop-gain?
You will have seen a frequency response plot of something you bought like a cartridge, amp or speakers. There’s a “y” axis on the left which is amplitude or sensitivity or gain, expressed in decibels (dB) or voltage or watts etc. And along the bottom is the “x” axis which represents frequency – low to high – and it’s usually in “log” format to make it look sensible.
As you know what one looks like I’m not going to show you one, but if you don’t, just Google “frequency response plot” and look at the images result and you’ll see plenty of different ones.
The ones I’m going to show you below all have one thing in common and that’s the open-loop frequency response of a typical amplifier.
Open-loop means without any negative feedback, global or otherwise. It’s a plot of the absolute maximum gain it will do, and for op-amps (that’s a chip form amplifier suitable for preamp signals) that maximum gain is around 110dB (about 300,000). At such high gain there isn’t much of a flat frequency response, and often the flat portion only goes as far as 10-200Hz – it will be somewhere in that region.
All the diagrams below show two curves. The outer one is called the open-loop response “envelope” and represents the maximum gain (or amplification) it can do versus frequency. The inner one is the closed loop gain representation of whatever application the amp is put to.
After some point in the region of say 10-200Hz, the open-loop gain falls off, and mostly this is at a rate of 20dB/decade (6dB/octave). This is intentional and is brought down in this manner by what’s called a compensation capacitor (a “unity-gain stable” amp has this compensation capacitor built-in). If it didn’t the flat response would go a bit further, but then its higher frequency gain would fall at a much steeper rate due to parasitic capacitance, and the amp would be unstable and oscillate.
To use the amp for good audio, some negative feedback is used to reduce the gain to the wanted amount. As that gain can only exist inside the “envelope”, or “gain bandwidth product” as it is properly known, then it is depicted by a curve within that “envelope”. That curve is known as “closed-loop” and it depicts the closed loop frequency response.
You will note from the first diagram that the frequency response of the “flat” amplifier is “wider”, which is a property of negative feedback. In fact, the flat frequency response could be made wide enough to intersect the “envelope” and utilise the full gain bandwidth of the amp. However, in most cases the feedback will be compensated by a further capacitor which helps with high frequency distortion and improves stability margins.
The difference between the outer “envelope” and inner closed loop curve is called the “loop-gain”. Loop gain is the “amount” of negative feedback applied. I’ve indicated the magnitude of loop-gain by the arrowed lines. In a flat gain stage such as is illustrated above, you’ll see the loop-gain varies with frequency – more at low frequencies and less at high frequencies.
Negative feedback has a number of properties. The ones we’ll discuss here are input impedance, output impedance and distortion. Negative feedback increases input impedance, lowers output impedance, and lowers distortion.
The flat gain stage therefore has falling input impedance, increasing output impedance and increasing relative distortion the further up in frequency you go. It can be made less consequential but these characteristics remain.
The reason for illustrating these varying loop-gain properties is because Stanley P Lipshitz et al. made such a big fuss (amongst other things) about having constant input impedance for phono preamp inputs. If the loop-gain varies with frequency – as is shown for the flat response amplifier, then the input impedance is going to change with frequency. The type of phono preamp using flat gain stages are “passive EQ” types.
As I said, this problem can be made less consequential, but to the purist, it’s an anxiety that remains.
This second diagram shows the response of an RIAA active preamplifier as the inner curve. An active RIAA amplifier equalises its frequency response in its feedback network. It can be seen that the loop-gain is a bit more constant – the magnitude of loop-gain difference is smaller – so input impedance, output impedance and distortion are a bit more constant.
But, for an “integrator” amp it can be seen that the loop gain can be made much more constant! All other things being equal, and bear in mind that nothing is perfect, but if the closed-loop gain can be made to “track” the open-loop gain, then it’s loop-gain is going to be constant. And if loop-gain is constant, then input impedance, output impedance and distortion is more constant too.
I personally think this would be an answer to Lipshitz’s concerns, however, as he seemed to be a perfectionist I’d better ensure that my “integrator” amp input stage tracks the open-loop gain as close as I can get it.
Skipping back to part 3, I said . . .
“I therefore set the EQ to “flatten” the cartridge EQ and provide the record EQ at 50Hz” Rather than setting the “integrator” turnover to track the amp, I already decided on 50Hz, so I need an amp to track the “integrator”.
Fortunately there are a number of FET input op-amps that fit the bill. Looking at the “envelope” response in all the above diagrams you will see where the curve intersects the frequency axis. You need to know that and the op-amp’s open loop gain to find its turnover frequency – and that information is found in data sheets. Taking an example which has a unity gain bandwidth (gain bandwidth product) of 4MHz and an “Av” (gain) of 100,000 (100dB), the turnover frequency is given simply by: 4,000,000/100,000 = 40Hz.
OK, not exactly 50Hz but very close and a lot more closer than the other design options available.
So that’s the input stage: an integrator feedback op-amp circuit that’s about as simple as you can get, but has virtually constant loop-gain, and by that virtue should address the problems of varying input impedance, and address the concerns of Lipshitz et al., and output impedance should also be constant, not forgetting a constant distortion figure.
What it outputs is the off-record signal without the magnetic cartridge CV response – constant amplitude – only needing some very simple EQ to do the frequency corrections for the mid frequency record response. The diagram below shows a simple embodiment of the “Different Approach to Record EQ”. A very similar design to this was first demonstrated at the Aylesbury “Roadshow” event in 2014.
Part 5 digs deeper and examines a similar idea.
Other posts on this topic: