Neither did John. We ran through the numbers a few pages back and it's beyond inconsequential. John has not argued that point, other than listing some courses he took last century.
Ahem, gentlemen, I still would like an answer here. Not at this point concerned about how strong the effect is (I get your point SY) but if indeed the non-linear cap results in 3rd order HD. John?
jan
Jan, you can worry yourself about anything. You know, 'Doppler distortion' in a cone loudspeaker is most probably a higher degree of nonlinearity than anything we are talking about here.
Audio design is about 'tradeoffs'. It just so happens that a single jfet IS MORE LINEAR than a single transistor. It is also true that a single tube is MORE LINEAR YET!
However, what about noise?
Well, the best tube that I can think of would be the 7788 pentode.
I have a few of them here. It might take 5-20 7788's in parallel to get to a 10 ohm equivalent resistance. Kind of expensive. BUT LINEAR! Low input capacitance too!
You could use a bipolar transistor, but even IF the BETA was really, really linear, the exponential V/I conversion will be MORE NONLINEAR, than the SQUARE LAW operation of the jfet. This can both be graphed and measured. I have done both decades ago.
Even with the somewhat nonlinear characteristics of the bipolar transistor, think about RFI. Jfets are worlds better at resisting RFI, than a simple bipolar.
Now what about biasing? Bipolars need a voltage difference between the base and emitter to work at all. Jfets do not need this voltage difference and are on at a reasonable operating level. This makes autobiasing simple and elegant.
Have bipolar designs been made that are equal or even lower noise that the Vendetta circuit shown here? Of course, but NOT without tradeoffs, of ab added input capacitor, etc., etc. and a VERY NARROW optimum operating range. They can have low input capacitance, however, and that may be necessary in some situations. Not in the MC case.
Audio design is about 'tradeoffs'. It just so happens that a single jfet IS MORE LINEAR than a single transistor. It is also true that a single tube is MORE LINEAR YET!
However, what about noise?
Well, the best tube that I can think of would be the 7788 pentode.
I have a few of them here. It might take 5-20 7788's in parallel to get to a 10 ohm equivalent resistance. Kind of expensive. BUT LINEAR! Low input capacitance too!
You could use a bipolar transistor, but even IF the BETA was really, really linear, the exponential V/I conversion will be MORE NONLINEAR, than the SQUARE LAW operation of the jfet. This can both be graphed and measured. I have done both decades ago.
Even with the somewhat nonlinear characteristics of the bipolar transistor, think about RFI. Jfets are worlds better at resisting RFI, than a simple bipolar.
Now what about biasing? Bipolars need a voltage difference between the base and emitter to work at all. Jfets do not need this voltage difference and are on at a reasonable operating level. This makes autobiasing simple and elegant.
Have bipolar designs been made that are equal or even lower noise that the Vendetta circuit shown here? Of course, but NOT without tradeoffs, of ab added input capacitor, etc., etc. and a VERY NARROW optimum operating range. They can have low input capacitance, however, and that may be necessary in some situations. Not in the MC case.
Its the non-linear capacitance -- if you know why and how a ceramic cap dielectric causes thd when voltage drop is across it, then you know why jFET's C causes distortion across it. Walt Jung has addressed this issue... especially obnotious when input Z is high (47K).
-Thx RNMarsh
-Thx RNMarsh
Last edited:
I've only asked straightforward technical questions about the basics of your design choices. If that's "pestering," I plead guilty.
I see you're punting on Jan's questions as well.
In this case, source Z is on the order of 10R, or about 5000 times lower. So would you agree that it's not important for MC first stages?
Using varicaps to generate harmonics is common enough so I would think you are correct in theory but I suspect at the levels here the effect is tiny.
Try a couple of back to back varicaps and see what the real effect is.
No problem John, Richard gave me a hint.
I believe it should be because the attenuation between the series R and the (varying) capacitance (to (virtual) ground) varies through a signal cycle. That means that the 'gain' varies during the signal cycle. That's pretty much the definition of HD, isn't it?
And assuming it is symmetrical for pos and neg amplitudes, it must be odd order.
jan
Once again I must make an appeal for terminological clarity. Jan touches on it: variation, per se, is sufficient as a distortion mechanism. Now, AS WELL, with voltage-dependent capacitances in a typical circuit, the variation as a function of something (usually voltage) IS a nonlinear function. But for the student attempting to understand, to constantly use the term "nonlinear" when what we really mean is "constant" --- i.e. NOT variable with, in this case, voltage --- is misleading.
I remarked on a Facebook post that some historian of engineering ought to determine when this gratuitous use of "nonlinear" began, especially when the term linear started being applied incorrectly when what is meant is non-varying or constant. I conjectured that it might have arisen with transistor data sheets when translated, as it is ubiquitous in Japanese bipolar datasheets: "linear beta" is seen a lot, never what is really meant which is constant beta (as a function of collector current). One sometimes sees, elsewhere, "flat" beta, which is also correct.
Now the devotee of semantics will perhaps point out that y = constant is a linear equation, and she is correct: it fits the canonical y = mx + b where m = 0. But it's also true that the Moon revolves around the Sun.
In that linear equation example, there are an infinity of equations where m is nonzero, and they are all linear equations --- but only one class of them has m = 0. So when someone says a transistor has "very linear beta" I sometimes say So that means it could have a beta of 100 at 100uA, and a beta of 200 at 200uA (say)?
The problem is, I think, we believe we sound smarter when we throw "nonlinear" into our discourse. And since everything under the sun by way of physical systems IS nonlinear at some level of discrimination, we are usually sort-of correct. But for one attempting to understand the distinctions it is misleading, particularly when what is meant is constant.
Maybe a tutorial thread on linear and nonlinear systems is in order. Of course there is ample online material, if not necessarily audio-oriented.
I don't think that is the case; I believe it is only used for tuning purposes, NOT to generate DH.
jan
Very much so.
(JD, you are aware of the ageing effect of high dosages of medium wavelength ultraviolet on the epidermis?)
Last edited:
It's a fine balance between looking older and looking fit and healthy. I think I'm still ahead
jan
Thats my understanding too, and that is what varicaps do. RF isn't really my field, I've only worked with varicaps as tuning elements for receiver front ends. But yes, capacitance that varies with voltage.
They can be used to generate frequencies way beyond the reach of conventional oscillators, up to hundreds of Ghz I believe.
- Status
- Not open for further replies.