h_a said:
This project was started a long time ago, and only the phono amp adventure (not finished yet!) detoured me for the last months.
The secret beyond the speed - for the last 30 years, I did not sleep more than 3-4 hours a day. It's biology, I guess, anyway I have to burn daily 25% extra time than the guy next door.
Very interesting project. 
Any MOSFET amplifier has built in unlinearity in the output stage, that translates to distortion under load.
The same kind of distortion that used to be called 'cross-over distortion' that we all should be familiar with.
Below see a clip from the 2SJ201 datasheet, that shows it's ID as function of Vgs. I have drawn a bit on top of the curve to show my point. If the line is straight, you have a linear (ohmic) region of operation, but if the line is bent, you have unlinearity. That under load becomes cross over distortion.
2SJ201 becomes fairly linear above 1A. (Only fairly). This means to get rid of the cross over distortion you have to use idle current of 1A at least.
Shown on the red line.
And even above 1A the line is not straight, but bends slightly. But below 1A the line bends heavily, which will cause large amounts of cross over distortion. You can reduce cross over distortion with more feedback, but as many many people have been forced to realize over the last 40 years, it's not going to help your sound.
You will see this if you measure your amp with a distortion analyser. The higher idle current the less distortion. But it doesnt get very good before you enter the Clas A region.
A bipolar transistor becomes linear at like 100mA, and above that has a almost ohmic emitter resistance of 70-100 mOhms. (Blue Line) This means a bipolar transistor is much easier to make cross over free, but on the other hand lacks the MOSFET's drive simplicity and lack of second break down (at high powers).
I remember in the 1980's Hitachi advertised that MOSFET amplifiers are much more linear than bipolar ones. But i never really understood that argument, as MOSFET's as source followers are very unlinear (as you see on the above datasheet cut). Soundwise the MOSFET amp will always have that sharp and metallic treble, that you can always distiguish on factory amps made with MOSFET's.
I suggest you should consider finding a clever way to get rid of that distortion before investing in this kind of linear MOSFET amplifier. (Any kind).
Any MOSFET amplifier has built in unlinearity in the output stage, that translates to distortion under load.
The same kind of distortion that used to be called 'cross-over distortion' that we all should be familiar with.
Below see a clip from the 2SJ201 datasheet, that shows it's ID as function of Vgs. I have drawn a bit on top of the curve to show my point. If the line is straight, you have a linear (ohmic) region of operation, but if the line is bent, you have unlinearity. That under load becomes cross over distortion.
2SJ201 becomes fairly linear above 1A. (Only fairly). This means to get rid of the cross over distortion you have to use idle current of 1A at least.
An externally hosted image should be here but it was not working when we last tested it.
Shown on the red line.
And even above 1A the line is not straight, but bends slightly. But below 1A the line bends heavily, which will cause large amounts of cross over distortion. You can reduce cross over distortion with more feedback, but as many many people have been forced to realize over the last 40 years, it's not going to help your sound.
You will see this if you measure your amp with a distortion analyser. The higher idle current the less distortion. But it doesnt get very good before you enter the Clas A region.
A bipolar transistor becomes linear at like 100mA, and above that has a almost ohmic emitter resistance of 70-100 mOhms. (Blue Line) This means a bipolar transistor is much easier to make cross over free, but on the other hand lacks the MOSFET's drive simplicity and lack of second break down (at high powers).
I remember in the 1980's Hitachi advertised that MOSFET amplifiers are much more linear than bipolar ones. But i never really understood that argument, as MOSFET's as source followers are very unlinear (as you see on the above datasheet cut). Soundwise the MOSFET amp will always have that sharp and metallic treble, that you can always distiguish on factory amps made with MOSFET's.
I suggest you should consider finding a clever way to get rid of that distortion before investing in this kind of linear MOSFET amplifier. (Any kind).
hannes: Douglas Self seems to agree with me: http://www.dself.dsl.pipex.com/ampins/dipa/dipa.htm
Look at fig 13. ;-)
Look at fig 13. ;-)
IRFP240 / 9240 are the same. You basicly have a changing resistor in series with your load. The resistor is for IRFP240/9240 from 0,2 to 1 Ohm value and changes with signal.
If you use af CFP coupling, this changing resistance is compensated by the short loop, but in case you use a source follower, no bias system will help you. Nigel did you use a distorsion analyzer, and test your amplifier under load?
If you use af CFP coupling, this changing resistance is compensated by the short loop, but in case you use a source follower, no bias system will help you. Nigel did you use a distorsion analyzer, and test your amplifier under load?
Hi Lars,
I'm sorry I have difficulties following you. Figure 13 on Self's site is just a comparison of various output stage topologies, not THD-figures/theory presented there.
That bipolars and fets have a different transfer function (exponential for bipolar, square law for fet) is fundamental and well understood in theory.
Your understanding of crossover distortion is certainly different to the generally used one, that is generated distortion due to switching off of one transistor in a complementary circuit and I don't see how that is affected by the plot you've shown.
As I'm a lucky guy I enjoyed looking at figures from distortion analyzer several times - and will do again soon - but the FFT-plots, as nice as they are - have been reluctant so far to tell me the source of their distortion.
All of this is anyway a little academic, as in a real circuit you also have to drive the output stage somehow and provide voltage gain. And that is usually were the fun starts.
Have fun, Hannes
I'm sorry I have difficulties following you. Figure 13 on Self's site is just a comparison of various output stage topologies, not THD-figures/theory presented there.
That bipolars and fets have a different transfer function (exponential for bipolar, square law for fet) is fundamental and well understood in theory.
Your understanding of crossover distortion is certainly different to the generally used one, that is generated distortion due to switching off of one transistor in a complementary circuit and I don't see how that is affected by the plot you've shown.
As I'm a lucky guy I enjoyed looking at figures from distortion analyzer several times - and will do again soon - but the FFT-plots, as nice as they are - have been reluctant so far to tell me the source of their distortion.
All of this is anyway a little academic, as in a real circuit you also have to drive the output stage somehow and provide voltage gain. And that is usually were the fun starts.
Have fun, Hannes
OK Hannes, i think we have different approaches to the subject, so i will only comment that the transfer function of the output devices is the actual and only cause of cross over distortion.
There is nothing academic about cross over distortion, it sounds terrible This means you have to deal with it. Driving the output devices is the easy problem to solve, so you should not let your design be compromised by your wish to have a easy solution for output device driving. ;-)
There is nothing academic about cross over distortion, it sounds terrible
This means you have to deal with it. Driving the output devices is the easy problem to solve, so you should not let your design be compromised by your wish to have a easy solution for output device driving. ;-)Mr. Lumba
I don't know what your background or mission is, but obviously you have to back that claim. My knowledge comes from years of exprience doing professional design on this kind of applications, so i know what i'm talking about.
Claiming something is complicated, without discussing the matter in depth, is usually a strategy used to mislead customers, and protect your market. I don't know if that is the case here.
However i do know that it's not complicated, and that cross over distortion is indeed a result of unlinear transfer function in the outputstage. There is nothing complicated about that.
I don't know what your background or mission is, but obviously you have to back that claim. My knowledge comes from years of exprience doing professional design on this kind of applications, so i know what i'm talking about.
Claiming something is complicated, without discussing the matter in depth, is usually a strategy used to mislead customers, and protect your market. I don't know if that is the case here.
However i do know that it's not complicated, and that cross over distortion is indeed a result of unlinear transfer function in the outputstage. There is nothing complicated about that.
Lars,
the switching and transfer characteristics of bipolar transistors do not give the smooth transition and linear performance your presentation suggests. Although the impact of crossover distortion in low quiescent current class AB bipolar output stages has been optimistically described by some reputable authors, supposedly based on unreliable measurements, however theoretical explorations and listening impressions contradict that. MOSFETs do not suffer from the numerous bipolar diseases, have a more agreeable distortion attribute due to advantageous operational principles, unipolar nature, superior input voltage/output current relationship and much wider frequency range.
the switching and transfer characteristics of bipolar transistors do not give the smooth transition and linear performance your presentation suggests. Although the impact of crossover distortion in low quiescent current class AB bipolar output stages has been optimistically described by some reputable authors, supposedly based on unreliable measurements, however theoretical explorations and listening impressions contradict that. MOSFETs do not suffer from the numerous bipolar diseases, have a more agreeable distortion attribute due to advantageous operational principles, unipolar nature, superior input voltage/output current relationship and much wider frequency range.
Mr. Lumba
Thank You for your response, which shows your presumptions are based on listening tests of various exotic MOSFET's for audio use, rather than usual industrial FET's like IRFP240 or 2SJ201, which i was referring to.
That's OK.
I do however agree with you, that it's possible to get decent results with MOSFET's of this type, as they seem to have a more stable high source resistance. However the power handling seems to be somewhat lower.
Thank You for your response, which shows your presumptions are based on listening tests of various exotic MOSFET's for audio use, rather than usual industrial FET's like IRFP240 or 2SJ201, which i was referring to.
That's OK.
I do however agree with you, that it's possible to get decent results with MOSFET's of this type, as they seem to have a more stable high source resistance. However the power handling seems to be somewhat lower.
I'm sure these 'exotic' MOSFET's can be a lot of fun listening to. However i'm looking at fishbone amplifiers, based on semi linear MOSFET like 2SJ201 / 2SK1530 or industrial MOSFET IRFP240 / 9240. VSOP belongs to this class of amplifiers, with a small signal input stage, and a lot of parallelled output devices.
Lars,
the 2SJ201 / 2SK1530 is a well-matched complementary pair, certainly very suitable for audio, utilizing the more efficient (but less linear) vertical structure, significantly more linear than the IRF-family even at low currents. Crss and Vgsth are acceptably small (two important parameters for linearity).
the 2SJ201 / 2SK1530 is a well-matched complementary pair, certainly very suitable for audio, utilizing the more efficient (but less linear) vertical structure, significantly more linear than the IRF-family even at low currents. Crss and Vgsth are acceptably small (two important parameters for linearity).
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.