Patrick,
do not forget that useful signal dIdout current is 4 x higher than dId in a single device, for complementary differential pairs, you can see it in the plots. E.g., useful output current of 2mA means only 0.5mA in a single device, biased for example at 10mA.
do not forget that useful signal dIdout current is 4 x higher than dId in a single device, for complementary differential pairs, you can see it in the plots. E.g., useful output current of 2mA means only 0.5mA in a single device, biased for example at 10mA.
john curl said:
It does not seem to depend much on Gm, in case we keep same signal Id. It highly depends on lambda. The question is, what is the lambda of real devices? Have you reliable information on this? Models seem to be strange here.
> do not forget that useful signal dIdout current is 4 x higher than dId in a single device
Yes, I am aware of that.
So let say I want to see 40V p-p at the speakers, and I have 1k resistors for R20, R21. I would have to swing +/- 20mA through each of R20, R21, which means +/- 10mA per JFET at the complementary diff pair. That is hard work for the JFETs.
If I increase R20, R21 to say 2.5k, I reduce the swing to +/-4mA, at a bias of say 10 to 12mA. So I still get some reasonable linearity. But then the bandwidth of my source follower at the output stage becomes tricky with an input impedance of 2.5k, unless I put a (e.g. JFET) buffer in between, which in turn adds to distortion.
And I am pumping something like 200mW into each JFET. So heatsinking required.
All can be done, just not as ideal as in the case the circuit is used as a line amp.
Patrick
Yes, I am aware of that.
So let say I want to see 40V p-p at the speakers, and I have 1k resistors for R20, R21. I would have to swing +/- 20mA through each of R20, R21, which means +/- 10mA per JFET at the complementary diff pair. That is hard work for the JFETs.
If I increase R20, R21 to say 2.5k, I reduce the swing to +/-4mA, at a bias of say 10 to 12mA. So I still get some reasonable linearity. But then the bandwidth of my source follower at the output stage becomes tricky with an input impedance of 2.5k, unless I put a (e.g. JFET) buffer in between, which in turn adds to distortion.
And I am pumping something like 200mW into each JFET. So heatsinking required.
All can be done, just not as ideal as in the case the circuit is used as a line amp.
Patrick
> Have you reliable information on this? Models seem to be strange here.
I have measured data on Id vs Vgs for all types of JFETs mentioned here. But then I think it will take you 10 minutes to measure them with a scope and a func gen.
How to fit them to the model, I have no idea.
Patrick
I have measured data on Id vs Vgs for all types of JFETs mentioned here. But then I think it will take you 10 minutes to measure them with a scope and a func gen.
How to fit them to the model, I have no idea.
Patrick
I speak about lambda, the channel-length modulation. This is the parameter that that violates square-law of JFET transfer function, in case it is non-zero.
> I speak about lambda, the channel-length modulation. This is the parameter that that violates square-law of JFET transfer function, in case it is non-zero.
That I did understand.
But if you have real measured data (of say 1000 points) of Id vs Vgs under defined conditions, and you have the equations from the spice model with Lamda, Vto etc., I would have guessed that it would not be particularly difficult to derive one unknown (Lamda), or even multiple unknowns (Vto, ....) from that ??
My maths is not particularly good. So I normally just do something like this by trial & error in an excel sheet.
Patrick
That I did understand.
But if you have real measured data (of say 1000 points) of Id vs Vgs under defined conditions, and you have the equations from the spice model with Lamda, Vto etc., I would have guessed that it would not be particularly difficult to derive one unknown (Lamda), or even multiple unknowns (Vto, ....) from that ??
My maths is not particularly good. So I normally just do something like this by trial & error in an excel sheet.
Patrick
It would be possible to compare measured transfer function with the math expression and to find matching parameters by iterration. It would be a lot of work and I do not see much sense in doing that. I think my role has been fulfilled, i.e. to show how circuit topology affects resulting distortion and to set assumptions for the circuit with very low distortion and almost no content of higher harmonics. Also, we know what will increase distortion and create higher harmonics. This is enough, IMHO, and anyone can take it as a hint, or not. Do not expect for me to analyze specific circuits, I have done enough work on this and spent considerable amount of time on this.
We are most grateful for your contribution.
I was merely pointing out a possible means of working out Lamda, since you raised doubts about the values you had used in your model about 6 posts before.
Patrick
I was merely pointing out a possible means of working out Lamda, since you raised doubts about the values you had used in your model about 6 posts before.
Patrick
john curl said:Good work, PMA, and welcome to Charles Hansen's and my world of audio design. You know, what if I do this, what happens?
Hi John,
You deserve some credit here as well. Welcome to the world of designers who see SPICE simulation as a valuable tool in audio amplifier design.
PMA has done an outstanding job here of demonstrating this, and I am very pleased to see this interaction and discussion. SPICE is only one tool in our tool-box, and its use does not guarantee a good-sounding amplifier. But its ability to allow us to explore the design space and gain insight into tricky questions like distortion, where many in the past have merely relied on their instincts and time-consuming experimentation, is very valuable.
Welcome aboard!
Bob
Bob, I have been using emulation since 1966. In 1971, I took classes at Berkeley from Don Pederson, claimed by the IEEE to be the father of Spice. I just learned from the beginning, its limitations, and never found it as important in making real circuits, although is great for 'what if' questions. I am still not using Spice for anything like a full circuit, and until I get a communication card running on the PC that Nelson Pass gave me (thanks again, Nelson) I don't think that I will be using Spice, except for what PMA has thoughtfully and usefully provided.
PMA, this is an early measurement of a 2N3819, and early but somewhat typical fet done by Dr's R.G. Meyer and Dr. Pederson, and found in their class notes from 1971.
Id-Idss = 6Vg + .5Vg (squared) + .17Vg (cubed)
Perhaps we can make something out of this.
It has been my experience that the same fet type can have different distortion, if it is made by a different process. We used to have problems using different manufacturers at times, because of this.
Id-Idss = 6Vg + .5Vg (squared) + .17Vg (cubed)
Perhaps we can make something out of this.
It has been my experience that the same fet type can have different distortion, if it is made by a different process. We used to have problems using different manufacturers at times, because of this.
PMA:
thanks for the considerable work in providing this info.
John:
thanks for pushing hard to keep everyone focused on-task
i've learned a lot and also picked up some great books along the way.
mlloyd1
thanks for the considerable work in providing this info.
John:
thanks for pushing hard to keep everyone focused on-task
i've learned a lot and also picked up some great books along the way.
mlloyd1
PMA said:
Richard, you should just do these "experiments" yourself. The demo version of Micro-Cap (that's what Pavel was using) is very easy to use and can build relatively large circuits.
Do a search and you will find models on DIY Audio.
But in the end, as John Curl will tell you (this is the Blowtorch thread, after all), the feedback of a complementary feedback pair will degrade the sonics. Perhaps not quite as much as a global feedback loop, but still a distinct sonic signature that is more about sound than music.
Do a search and you will find models on DIY Audio.
But in the end, as John Curl will tell you (this is the Blowtorch thread, after all), the feedback of a complementary feedback pair will degrade the sonics. Perhaps not quite as much as a global feedback loop, but still a distinct sonic signature that is more about sound than music.
PMA said:
Pavel, going back to your post #3842, I am still interested in the
input voltage on that sim because I wish to weigh its contribution
to the variation of Vds by Vds.
Good to see you here, Charles. 'Keep on Truckin!' I agree that we should keep things simple at first. Of course, there are many 'variations' that could be interesting. I didn't know that PMA was using Microcap 9, I have that too, but I thought he was using LT Spice, and I haven't got it up to speed, yet.
john curl said:
I could be wrong. I just made that assumption based on the screen shots.
Pavel, if you are still reading this thread, perhaps you would be so kind as to post the .CIR files on your website. In this way others could download them and have a starting point for further explorations, such as the kind of question that Nelson is asking. (It would also be great if you included the models you used.) Thank you.
I did it Charles, some time ago, I begin by this:
http://www.diyaudio.com/forums/showthread.php?postid=424902#post424902
I tested after the CFP (fet/bjt) in various configurations( included complementary differential) thd figures are very good :
http://www.diyaudio.com/forums/showthread.php?postid=425536#post425536
THD graphs refer to a complementary CFP but not differential.
Included an I/V stage:
http://www.diyaudio.com/forums/showthread.php?postid=610031#post610031[/URL]
I end up with fet only and you are right, I prefer, more music "less sound"
http://www.diyaudio.com/forums/showthread.php?postid=424902#post424902
I tested after the CFP (fet/bjt) in various configurations( included complementary differential) thd figures are very good :
http://www.diyaudio.com/forums/showthread.php?postid=425536#post425536
THD graphs refer to a complementary CFP but not differential.
Included an I/V stage:
http://www.diyaudio.com/forums/showthread.php?postid=610031#post610031[/URL]
I end up with fet only and you are right, I prefer, more music "less sound"
Hello all, I surely will be back, after a small break. I will try to get lambda from the equation that John has provided. Charles is right - yes, I am using the MC, though different version and grade than stated.
- Status
- Not open for further replies.