yes. It is my experience that absolute polarity can be audible depending on signal waveform.
IIRC... decades ago, I read a study on it (not an electronics pub) ... ear is not exactly symmetrical in its sensitivity to compression vs rarefaction pressure.
I did simple earphone tests to see for myself, also.
THx-RNMarsh
There are millions reading this -- 99% do not comment... so don't assume ALL know this interfacing issue already.
In fact, many engineers seem not to pay attention to it based on comments of subjective listeners -- who sometimes report different audible results in Their system of an amp being reviewed. IMO this is an example of a micro-view of what is being heard.
Input current distortion? That can be solved with other amplifier toplogies. Topologies Bob and Doug aren't interested in exploring, apparently....
In my always simplifying mind, wouldn't input current distortion be part and partial to the current source being 'leaked' from?
Anyway, cancel it with the topology.
THx-RNMarsh
Last edited:
Only a person living under a rock would think that was just a simple question.
There are numerous examples. Here are just a couple, where my work preceded yours and there was no reference to my book.
In your chapter 13 on "output inclusive compensation" (TMC) and 2-pole compensation my book is not referenced at all, in spite of the extensive coverage of those two topics; yet you included an obscure reference to a DIY thread.
In your chapter 22 on thermal compensation you did not reference my book in spite of the extensive thermal simulation modeling in my preceding book. In that same chapter on ThermalTraks you once again could not bring yourself to reference my book, instead making a reference to an obscure DIYaudio thread.
In your brief chapter on Class D amplifiers, you made no reference to my book, which contained 5 chapters on Class D amplifiers that condensed a lot of scattered class-D material into readable form, in spite of 17 references to scattered material.
There are others...
I referenced your book 5 times in my book. As near as I can tell, that is 5 times more than you referenced mine in your sixth edition. My references to your book were fair and appropriate. You did not invent everything, including the Thompson architecture that you took the liberty of terming Blameless. Virtually all of the features of your Blameless amplifier were conceived by linear IC op amp designers 50 years ago. Virtually all of the amplifiers you show, build or measure are merely variations on the Blameless. I cover a great many more amplifier architectures, which I jokingly term "Selfless"
.Yes, and I doubt that that amplifier achieves 0.001% THD at 20kHz full-power.
At least you were polite about what??
If you look back a few posts you will see that I made my position clear - that CM input distortion in power amplifiers is not a big problem due to the relatively small size of the input as a result of typical power amp gain. I also pointed out there that the problem in op amps is usually in unity-gain or low-gain situations. Once again, CM distortion is not the issue I have been chastizing you about; it is that dumb sharing of the LTP and VAS current sources you seem to use in virtually all of your amplifier examples.
Bob
Tryphon, you see we do not need the lounge or the BT thread for this kind of behavior.
I'm a bit surprised that someone that has designed an Amp that has a price tag at 49.500EUR
I have no idea why this can not be discusses by me, I see this input current distortion as a interesting problem/subject while inverting does not solve all. Why would one mention the subject here and (as it looks like) then drop it without further explanation and elaborations.
D Self lets know it is a mechanism in the BJT's base (current) and B Cordell then makes some remarks about it that are not further qualified. Why can't I get a more knowledgeable discussion going about this phenomenon?
And why does my 'history' (previous work) forbid me doing this?
Thank you for the information! Is this exactly the same measurement procedure a vendor does to get amplifiers slew rate values for presentation in their catalogs ?
Many thx,
Toni
I do agree on what you say here, but there is just this point: would it not be interesting to get some input from DS and BC on the subject? To me the moment I got interested is when they started to talk about 'unknown' origins/mechanisms (creating this distortion, e.g. why/how is this base-current non linear and what can we do), BC saying that DS did not explain and DS saying (sort of) it's just there.
Again, to me, I find it interesting to know what this mechanism is, please [DS, BC] enlighten me
Last edited:
I was under the impression that we had established that there are a number of mechanisms at play:
1. Modulation of transistor beta due to early effect (transistor beta is a function of collector-to-base voltage)
2. Modulation of transistor collector-base capacitance (transistor collector-base capacitance is a function of collector-to-base voltage)
3. Beta is (or can be, depending on the transistor) a function of Ic, even with fixed collector-base voltage.
4. Tail current is supposed to be pure DC but can be modulated by the signal if the tail current source is not properly implemented.
All these mechanisms can result in the base current of the input stage transistors being distorted. When that current passes through an impedance, this will be converted into a voltage distortion. If the input transistors are exactly matched, and impedances seen at the two inputs exactly matched, these distortion products are common-mode and will cancel. Imperfect matching will result in some of the common-mode distortion being converted into differential distortion, which, as far as the amplifier is concerned, is actually signal that should be faithfully amplified.
Which of the four above-listed mechanisms is the dominant factor in a given amplifier design, depends on the specifics of that design.
Have I missed anything?
P.S. I find the talk of inverting mode + input buffer interesting. So far everyone seems to have ignored the fact that in that case, the input buffer will experience a common-mode signal. So, somewhere in your circuit you are going to have to use a technique other than inverting-mode operation to ameliorate problems caused by a common-mode signal.
The inverting configurations inherent distortion advantage is not a base current issue (only) - a fet input stages show the same effect. In any case, (at least in a power amp setting,) I don’t think the buffer is is essential - the feedback resistance values are typically substantially larger than the (variable)output impedance of the preamp. If you did decide to implement a buffer, (which I did) i just used two single ended ones. Made with matched jfets and trimmed resistors admittedly - hence my “if you afford the inconvenience “ caveat.
Last edited:
No you are not missing anything (for as far as I see).
> 1. Modulation of transistor beta due to early effect (transistor beta is a function of collector-to-base voltage)
Was taken away in one of the earlier simulations by bootstrapping
> 2. Modulation of transistor collector-base capacitance (transistor collector-base capacitance is a function of collector-to-base voltage)
Mostly solved by bootstrapping (if not all) and constant current mode.
> 3. Beta is (or can be, depending on the transistor) a function of Ic, even with fixed collector-base voltage.
Yes, there is always that 'left-over' factor even when bootstrapped and running at a constant current.
> 4. Tail current is supposed to be pure DC but can be modulated by the signal if the tail current source is not properly implemented.
There is always noise and the 'act of God' and the (possibly) 'imperfection' of nature.
All these will add up and generate some, but one(I) would expect that (by approximation) this should be divided by beta to get a figure that will be (by approximation) reflected in the base current.
But, the simulation with absolute bootstrapping, absolute mirroring, absolute amplification in the VAS and OS end a zero impedance signal source still shows a sizable input-current-distortion, witch I can not explain from the above, and for witch I like to see some input from the 'publicists' among us.
P.s.:
1st picture: {rx} is in the 1st step 1fempto Ohm the 2nd step (9k) demonstrates the effect of input-loading and then distortion canceling by the transistors being 'exactly' equal getting to a high THD-figure (in both cases)
2nd picture: shows the input-current-distortion (near -66dB) showing that the only way to make an amplifier with better figures is to depend on cancelation of this distortion in the LTP (my opinion, and I would like to see some input on this).
Attachments
Last edited:
Surely your buffers are subject to early effect and collector-base capacitance modulation (for BJT-based buffer), or drain-gate capacitance modulation (for FET-based buffer)? Therefore tracking cascodes are now required in your buffers, no?
Indeed, Scott ;-)Tryphon, you see we do not need the lounge or the BT thread for this kind of behavior.
And it is a pity, don't you think ?
This discussion of the causes of distortion in input stages and the means of remedying them promised to be interesting. Here the subject turns into a rooster's fight.
I believe you may have some insights on the subject. If I remember well, you are the author of a *very low* distortion OPA, the AD797 right ?
And, apart LTP, you were one of the first to design a CFA ( AD524 ?). So, may-be some idea on how to improve it there too ?
I'm afraid not.
If i remember well, when i was tying to optimize an LTP with BC150/160 for max bandwidth, i ended with a feedback bridge of 1K/47 ohms.
The base parasitic capacitance has a double effect on bandwidth and on distortion at HF, because the phase turns and its non linearity.
> The inverting configurations inherent distortion advantage is not a base current issue (only) - a fet input stages show the same effect.
Where?
> In any case, (at least in a power amp setting,) I don’t think the buffer is is essential - the feedback resistance values are typically substantially larger than the (variable)output impedance of the preamp.
There are multiple problems here that I try to avoid, high impedances and a potentiometer connected directly to the pre-amp output or/and the power-amp input.
> If you did decide to implement a buffer, (which I did) i just used two single ended ones. Made with matched jfets and trimmed resistors admittedly - hence my “if you afford the inconvenience “ caveat.
There you go
Attachments
P.s. Interesting as it may be, the previous post, talking about JFET LTP's is a distraction from the subject, being, base-input-current-distortion. While it (the input-current-distortion) may seem interesting only for the inputs (of the LTP) it is actually interesting for any BJT (also when in-the-loop) in the circuit and it may not be possible to always fight this distortion by cascoding, bootstrapping, constant-current-loading or the application of JFET's.
Hi Harry,
All that you have said here is exactly right. But whether the input current distortion is big enough to matter depends on the numbers - and the particular design, as you point out. The key question to me is which of these possible sources of input current distortion is the dominant one in a typical design. It would be nice to narrow this down, because then we would have insight on how to reduce it, rather than chasing improvements to reduce a non-dominant source of the distortion.
It is also wise, both in simulation and measurement, to confine the view to the front-end only (IPS, VAS, pre-drivers), since the presence of the output stage is usually a dominant source of distortion and usually the output stage has little to do with input current distortion. The presence of the output stage driving a load can also create many red herrings in measurements, like those measurements related to input current distortion. In most amplifiers I simulate and build, I center-tap the emitter-emitter pre-driver bias resistor and use it to provide feedback and close the loop to simulate or measure the amplifier in the absence of the output stage. This generally results in very low distortion numbers that will unmask other distortions like input current distortion.
It is so easy to build a very good tail current source that it seems foolish to even begin to talk about input current distortion without first setting the stage with a good tail current source. For this purpose, a properly-implemented feedback current source with 2 transistors and 2 resistors is all you need.
Cheers,
Bob