EC is negative feedback. I have to quote Cherry, which found there are exactly three ways to lower distortions
a) Bias, so that active devices are in a more linear region (that is, dynamic variations around the bias points are smaller, this is the Class A principle).
b) Distortion cancellation (use circuit symmetry wherever practically possible, to cancel out the harmonics).
c) Negative feedback (includes local feedback).
The big difference is how the EC negative feedback loop gain is created; EC (at least in the Hawksford implementation) uses a local positive feedback loop to create loop gain.
Yes, but.
It can only cancel even harmonics.
Tough luck, at best you are left with all the odd harmonics.
Bridge mode is a way to do that.
There is an element of semantics here. Using complementary distortions could be considered by some to be a kind of symmetry.
Sorry, may be I played with words, that was not my intention.
I know for sure ( from simulation, measures and maths ) that a bridge implementation when the two legs are exactly the same cancells all even harmonics and leaves the odd ones.
I am quite sure a single ended implementation does the same when the two legs of the totem pole are exact complementaries.
I know for sure ( from simulation, measures and maths ) that a bridge implementation when the two legs are exactly the same cancells all even harmonics and leaves the odd ones.
I am quite sure a single ended implementation does the same when the two legs of the totem pole are exact complementaries.
The simplest of all an op-amp inverter, the two resistors are pure third order (thermal) in distortion but cancel for an exact gain of -1.
Can't see the wood for the trees?
The concept is a working one. It is a restatement of side-chain processing, which among other things, does reduce distortion. It was overall an oblique comment that those who don't fully investigate silly notions, are the losers. hellokiity123 has been subjected to ridicule for the -290dB specification. -180dB is lauded as an achievement. Both are improbable. Trotting out the phrase 'industry standard' seems to validate everything.
The concept is a working one. It is a restatement of side-chain processing, which among other things, does reduce distortion. It was overall an oblique comment that those who don't fully investigate silly notions, are the losers. hellokiity123 has been subjected to ridicule for the -290dB specification. -180dB is lauded as an achievement. Both are improbable. Trotting out the phrase 'industry standard' seems to validate everything.
Scott if you use two complementary differential input pair, does the second harmonic get canceled or is it transformed to third harmonic? I've heard it mentioned that this method seems to result in an increased third H, which suggests it is a transformation into third H.
I would argue for cancellation. An LTP has a transfer function proportional to tanh which has only odd terms while single ended it's log. Remember also an LTP only sees 1/2 the input on either side and not all input stages recombine both currents in a mirror (for example) so you have to choose how to normalize things.
Resistor thermal non-linearity is different, power (hence T) is a square law but the voltage across the resistor has a sign so the current will have a third order term.
Marcel does a good job of highlighting distortion mechanisms that come into play just around/south of that hard-but-acheivable -160 dB range. Pushing to that level of performance is very commendable and I'll leave it at that.
I was thinking earlier today.
Is there a reason I can't measure the mid-point between the input and the output of an inverting amplifier with the QA401 to get a distortion reading with a fully nulled fundamental? The input and the output of an inverting amplifier should be equal and opposite so the remaining signal at the mid-point is only the leftover harmonics.
Is there a reason I can't measure the mid-point between the input and the output of an inverting amplifier with the QA401 to get a distortion reading with a fully nulled fundamental? The input and the output of an inverting amplifier should be equal and opposite so the remaining signal at the mid-point is only the leftover harmonics.
The degree of nulling of the fundamental depends on the phase shift characteristics of the output and the chosen input frequency. The principle can be checked in LTspice with an opamp. Also if there is gain, then a pot has to be chosen to find the null. The null point will be extremely sensitive and 0.1% error in balance will not achieve a null.
An experiment in LTspice shows that more than 40dB of distortion reduction is possible at 20KHz. This initial attempt points to a method of achieving the reduction figures claimed. However, any method of distortion extrapolation leads to a rabbit hole for a specification.
An experiment in LTspice shows that more than 40dB of distortion reduction is possible at 20KHz. This initial attempt points to a method of achieving the reduction figures claimed. However, any method of distortion extrapolation leads to a rabbit hole for a specification.
Last edited:
I don't understand. Why is a pot required? An inverting amplifier strives to make the mid-point between the input and the output equal to zero volts. So anything other than 0 volts is distortion. I don't see how having gain matters here because the dBr can be set to the output voltage. Even with some phase shift the fundamental will be many order of magnitudes less than the raw signal. It should be a good method of measurement all-round no?
You can measure the error signal of an inverting amplifier without needing a pot, and it will reduce the fundamental a lot, but the problem is again the second-order effects, especially resistor non-linearity and Cherry's coupling issue. Those are still there, even if the error signal would be zero.
Maybe you could make a null with a separate voltage divider, one placed outside the enclosure of the amplifier at a larger distance from the distorted current loops than the real feedback network, and made with more unit resistors than used in the real feedback network.
Maybe you could make a null with a separate voltage divider, one placed outside the enclosure of the amplifier at a larger distance from the distorted current loops than the real feedback network, and made with more unit resistors than used in the real feedback network.
You just described the DM ...
Maybe get that going again. Put a signal in and start following it through the unit until you find a place where there is an error. Then we have an idea what's wrong with it.
Start with all power supply pins on each opamp.
Jan
Last edited:
- Status
- Not open for further replies.