Certainly, but such C is usually almost constant and thus not per se a distortion mechanism. Even if plenty of other things are, as you point out. There are ways of dealing with them as well, although it gets rather elaborate and will not please the minmalists. And how much, after a while, is really necessary?
if they are standard spice models just rip them to your Ltspice, save relearning the interface, use the better simulator
Is the circuit board stray capacitance a problem when the CCS has a dynamic/effective impedance of many megohms?
I forgot to say that I was wondering if in fact it would be a problem. I really did not know. My simulations show that it is not an issue at all.
You have to realize that I've hurt my elbow and it's hard for me to type because of the pain, so I may not type in complete sentences or questions. I apologize for the confusion and thank you for your detailed response.
OK, but I just copied the schematic from Jung's article. I don't know what would be the function of gate resistors since there's no current flowing there, or at least it is insignificant.
I've used this circuit in some of my designs and it works as advertised, lowering distortion a little bit.
Sorry to hear about your elbow.
There are glass-epoxy boards that when made improperly show some very bad dielectric properties, which Tektronix (iirc) called "hook". They were using little areas of trace for high-frequency compensation, and saw some horrendous waveform distortions. There was a magazine article circa late '70's (?) that described this and had scope photos etc. of peculiar transient responses.
Now, back when board houses were as good as they could be, it could be supposed that hook became a thing of the past. And there are dielectric materials that are more exotic (Arlon etc., some of the Rogers materials primarily developed for RF apps) that probably don't suffer from it, but that are more expensive and not all that easy to persuade board houses to use. But having seen some of the evils perpetrated in far-flung regions when corners are being cut, I have no idea if hook rears its ugly head or not in our time.
driven guard rings, planes can bootstrap away much of the pcb dielectric stray C - someone even patented the concept for audio amplifiers
but again to optimise the VAS, CCS to the degree PCB parasitics are limiting means you have to have spent lots of effort on the VAS load - the output stage - for its much larger device parasitics, reflected load current to not be the orders of magnitude larger limiting terms
but again to optimise the VAS, CCS to the degree PCB parasitics are limiting means you have to have spent lots of effort on the VAS load - the output stage - for its much larger device parasitics, reflected load current to not be the orders of magnitude larger limiting terms
if it were still active I wouldn't pay any attention to it as legal threat "obvious to a practitioner of the art" and all
in fact I didn't let that patent stop me: http://www.diyaudio.com/forums/software-tools/101810-spice-simulation-67.html#post1563301
in fact I didn't let that patent stop me: http://www.diyaudio.com/forums/software-tools/101810-spice-simulation-67.html#post1563301
Last edited:
Does the fact that the impedance of these CCS designs changes with frequency, and often becomes lower in the higher areas of the audio band mean that the response of the VAS will change also? Instead of going for the maximum impedance, shouldn't we be more concerned with a constant impedance across the audio band even if that means the absolute impedance is less?
Gain in a VAS is a function of the load, so the gain changes with changes in the load. So, I guess what I'm saying is the bandwidth of maximum gain for a VAS is a function of the bandwidth of the CSS load. I suppose if there is enough gain over a broad enough bandwidth for the applied feedback, then this isn't a problem. It would be a problem for very high gain circuits I think because the bandwidth would be reduced.
That is a major question. I suspect the linearity to be more important which is why I am trying to find any FETs that are fast enough to be constant over the audio band even though they are not as stiff as a standard 2N5551 and red LED. Remember the "vast" differences between types of CCSs are trivial compared to just a resistor, and there is still a large camp who believe the resistor sounds better. The only way to tell is to build them.
The overt response of the VAS or IPS would not change. It is the smaller differences that are not clear what we are measuring. I suspect that the FBP topology is so widely used in VAS where the BJT/led is widely used in the IPS is a hint. In my DH-120 project, the single FET IPS actually modeled better when I looked at multiple levels and multible frequencies to for the FFT, but only with a very fast unobtainable FET.
Linearity is a misnomer. Constancy is the term applicable, in this case constancy with frequency. If it were somehow key then a resistor would be best.
In fact what is frequently done is that the output node of the so-called VAS is loaded anyway with one or two resistors to common. This provides a limit to the low-frequency gain which some believe they can hear the effects of. It also makes the overall amplifier's response to load impedance more consistent, that is, such that the loop gain at low frequencies through midband stays roughly constant regardless of load.
Putzeys argues that the concern about increasing open-loop gain at low frequencies is not a big deal, and shows the settling behavior for a step signal for two cases in his "The F-Word" article in Linear Audio. But people disagree, and the notion that more loop gain entails more canine tail-chasing is a prevalent one.
In fact what is frequently done is that the output node of the so-called VAS is loaded anyway with one or two resistors to common. This provides a limit to the low-frequency gain which some believe they can hear the effects of. It also makes the overall amplifier's response to load impedance more consistent, that is, such that the loop gain at low frequencies through midband stays roughly constant regardless of load.
Putzeys argues that the concern about increasing open-loop gain at low frequencies is not a big deal, and shows the settling behavior for a step signal for two cases in his "The F-Word" article in Linear Audio. But people disagree, and the notion that more loop gain entails more canine tail-chasing is a prevalent one.
If you look at what this is, you will see it is exactly the single FET, self biased with a resistor. I have not imported their model. I hope they picked a very fast FET. If so, it could be a winner. Looking at the data sheet does not give me a lot of hope.
Syntax. I would consider something that is consistent over frequency to be linear. We do both mean the same thing. I have not found the exact references to Leach where he states " ccs measure good but sound bad". For all I know, it is totally out of context. Now, given the huge difference between just a resistor and the least stiff, but most "consistent" ( self biased FET) , compared to the most stiff ( JFET cascode), which sounds better? That folks is the question at hand. Build and listen. I would bet the answer may depend on the quality of the power supply which is something not included in the models.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.