Thanks
The heatsink is connected to the earth wire from the wall.
Oh, i just fixed it by twist the signal wires with the ground
wires. Can anyone explain this to me why is this happen??
And btw, the F5 can drive headphone seriously nice,
headphoners - should try it!
Poked around to find some J176 P-Channel JFET's from Jameco -- I took a couple FFT's with a J74BL and the J176's -- both biased to the same Class-A bias level:
The noise out of both amplifiers is about the same at some 10's of microVolts.
An externally hosted image should be here but it was not working when we last tested it.
The noise out of both amplifiers is about the same at some 10's of microVolts.
Thanks
The heatsink is connected to the earth wire from the wall.
Oh, i just fixed it by twist the signal wires with the ground
wires. Can anyone explain this to me why is this happen??
And btw, the F5 can drive headphone seriously nice,
headphoners - should try it!
Regarding headphone use...can you explain how this is wired up please? Surely it's a bit overkill for headphones
Regarding headphone use...can you explain how this is wired up please? Surely it's a bit overkill for headphones
With a headphone jack, one ground to headphone jack, then red out from each channel to the other two connections.
KH1=AndrewT, since you followed up this quoted post with a stinging PM 2 minutes apart, I don't owe you a thousand words.
Since you're a smart fellow, solving other members' issues with your theories, why don't you tell us what happens given a single Source follower 2sk170 jfet say Idss=7mA in a circuit with a stable Vds=9V; Id=6mA; say Vgs is sitting at -0.02V (reference post 8699 attachment) -aside from temperature change contribution on Vgs/Vds/Id delta, what other factors contribute and by how much? At Idss or just below it what are the transconductance values for both n and p-channels?
KH2=EUVL, I don't owe you either. Btw how did you bias a p-channel with negative Vgs? For us mere mortals please put a NOTE that the slopes were overlayed for comparison. Also, your graphs and findings I find very useful if I decide to bias at lower than 70% of Idss the 2sk170/2sj74 jfets.
KH1/KH2 Tag team, huh?
I actually find his idea of using the tangent line slope and intercept brilliant in solving for transconductance. Useful for calculus challenged me.
The gm of the J74BL is around 0.040S, while the J176BL is around 0.007S so to get the gain of 10, and the right bias R1, R2 probably have to be changed.
The J74BL circuit has THD% @1W, 8R of around 0.002%, the J176 is around 0.06% -- this without a lot of experimentation.
I guess the feedback resistors may have to change as well.
Well, while we are on the subject of gain and complementary devices and such, I'ld like to bring up a point which has been touched on but, maybe not explained thoroughly. Maybe I don't even understand it properly
If all the device pairs in this amp were complimentary when matched, P-N JFETs and P-N MOSFETs, we would have 1 important condition met to help achieved minimum distortion, and DC offset. I beleive the P JFET however, has higher gm than the N. This mismatch will cause more gain for the negative portion of the signal than when the signal swings positive? Will it not? The P JFET drives the N MOSFET though, and it also has less gm than the P MOSFET causing somewhat of a correction of the higher gain created by the input pair? Since the amp is configured symetricly the opposite will be going on in the upper side of the circuit and with the positive portion of the signal. Ideally this would be canceling the non complementary effects created by the lack of ideal complements. Now I may have a few positives and negatives messed up here but isn't this "the devil is in the details" NP talked about?
Granted there is loop feedback to help correct all this but the idea is to create a simple circuit without non-linearites, gain compression, etc. etc. Isn't this what Euvl is attempting to better with his use of different source Rs and more linear output devices? Nelson has indicated, without thoroughly explaining, that building the F5 as described, you have a good chance of equalling what FirstWatt does. Changing things will require some tweaks to get you back to that "sweet spot".
Now there is more to this but,,, is anyone following me here? Is something like what I'm trying to describe going on and the reason for all this discussion?
If all the device pairs in this amp were complimentary when matched, P-N JFETs and P-N MOSFETs, we would have 1 important condition met to help achieved minimum distortion, and DC offset. I beleive the P JFET however, has higher gm than the N. This mismatch will cause more gain for the negative portion of the signal than when the signal swings positive? Will it not? The P JFET drives the N MOSFET though, and it also has less gm than the P MOSFET causing somewhat of a correction of the higher gain created by the input pair? Since the amp is configured symetricly the opposite will be going on in the upper side of the circuit and with the positive portion of the signal. Ideally this would be canceling the non complementary effects created by the lack of ideal complements. Now I may have a few positives and negatives messed up here but isn't this "the devil is in the details" NP talked about?
Granted there is loop feedback to help correct all this but the idea is to create a simple circuit without non-linearites, gain compression, etc. etc. Isn't this what Euvl is attempting to better with his use of different source Rs and more linear output devices? Nelson has indicated, without thoroughly explaining, that building the F5 as described, you have a good chance of equalling what FirstWatt does. Changing things will require some tweaks to get you back to that "sweet spot".
Now there is more to this but,,, is anyone following me here? Is something like what I'm trying to describe going on and the reason for all this discussion?
Last edited:
I am definitely no expert, but this will only happen when the amp goes into class B. Won't it?
I thought the top half and the bottom half of the circuit see the full single while in Class A.
The bottom half may see more gain than the top half or vice versa but that is where the two feedback loops come into play.
I suppose if all the parts were truly complimentary then you would get pure third hamonic with no second. Technically speaking this is preferable but it may not appeal to everyone when listening to music.
For me the performance of this amp is already good enough at low power levels.
A while back I mentioned you could play around with the jfet source resistors to get better performance or taylor the sound to your liking. Then Nelson said to try using a pot which I have indeed used which means you can tweak the amp with an adjustment of the pot for best performance or for best listening experience which may or may not be in the same spot.
Edit: I think linearity is more important than having purely complimentary parts. Since it is possible to have completely complimentary parts which aren't linear. Anyway that is just my feeling about this issue.
Last edited:
I never said it wasn't a good idea. I just thought it was bloody obvious.
Well if it helped any one then that is great, I thought it might help clarify what people were discussing.
let's say that one funny guy solve all these issues looking at curves/numbers/needle/screen of THD zeiger , and solve everything in few iterative hops , with additional source resistor on right side ?
trick question :
who's that funny guy ?
eeeeeeek!
wrong answer !
(another trick question - is his approach over engineering , or just good engineering ?)
trick question :
who's that funny guy ?
eeeeeeek!
wrong answer !
(another trick question - is his approach over engineering , or just good engineering ?)
In designing a high linearity amplifier, one tries to get good linearity and high bandwidth by circuit design first in open loop, so that the negative feedback is only there to cope with varying load impedance, and not working too hard to also correct for linearity error. At least this is how I would design my amplifiers.
FETs, whether Mosfets or Jfets, are approximately quadratic devices. That is to say their main linearity error is 2nd order. You can compensate for 2nd order error in two ways :
1. by using a balanced circuit,
2. by making sure the negative and positive halves of the push pull is fully complementary.
The original F5 is single ended push-pull, 2 stage design. So option 2 applies. The Fairchild devices are not complementary, in fact not nearly so. But they also have quite different Vgs. Nelson made used of that fact, and get the circuit to work in such a way that the combined characteristics of the two stages in the two halves are nearly equal. But this equality is also affected by the Vgs of the Mosfets and the Idss of the JFETs. Hence the need to use a distortion analyser to trim.
I use a different approach in my X'ed balanced circuit. I choose devices which are complementary in the first place, or I apply correction means (e.g. source resistor) to make them behave as such. When I know that my devices at each stage is complementary, they will self compensate all even harmonics when the two halves are fully symmetrical. You may say this is "super symmetry" also, not only left and right, but also top and bottom.
Of course such circuits will not compensate for odd harmonics. If you deliberate make the top and bottom half non-symmetric, you have a chance of trading 2nd harmonics for 3rd (increase 2nd to reduce 3rd). And then you use the left & right halves of the balanced circuit to self cancel the now increased 2nd, and end up with even lower THD. But this is not the entire story, because the high order harmonics are much more disturbing to the human ear than 2nd and 3rd. And they are much more difficult to fiddle with.
If you wish to follow the last option, you certainly need a very good distortion analyser. And you need to know what you are doing.
Patrick
FETs, whether Mosfets or Jfets, are approximately quadratic devices. That is to say their main linearity error is 2nd order. You can compensate for 2nd order error in two ways :
1. by using a balanced circuit,
2. by making sure the negative and positive halves of the push pull is fully complementary.
The original F5 is single ended push-pull, 2 stage design. So option 2 applies. The Fairchild devices are not complementary, in fact not nearly so. But they also have quite different Vgs. Nelson made used of that fact, and get the circuit to work in such a way that the combined characteristics of the two stages in the two halves are nearly equal. But this equality is also affected by the Vgs of the Mosfets and the Idss of the JFETs. Hence the need to use a distortion analyser to trim.
I use a different approach in my X'ed balanced circuit. I choose devices which are complementary in the first place, or I apply correction means (e.g. source resistor) to make them behave as such. When I know that my devices at each stage is complementary, they will self compensate all even harmonics when the two halves are fully symmetrical. You may say this is "super symmetry" also, not only left and right, but also top and bottom.
Of course such circuits will not compensate for odd harmonics. If you deliberate make the top and bottom half non-symmetric, you have a chance of trading 2nd harmonics for 3rd (increase 2nd to reduce 3rd). And then you use the left & right halves of the balanced circuit to self cancel the now increased 2nd, and end up with even lower THD. But this is not the entire story, because the high order harmonics are much more disturbing to the human ear than 2nd and 3rd. And they are much more difficult to fiddle with.
If you wish to follow the last option, you certainly need a very good distortion analyser. And you need to know what you are doing.
Patrick
I don't think so Thanh. As any class A comp pair, like this, has a device with rising current and lowering Vds, and a device with lowering current and rising Vds, there is a gain increase in one and a gain decrease in the other. Even if it's on a curve, the idea is to balance the gain compression and expansion so the curves are complementary and the rise of one cancels the decline of the other. Something like class B is not even close and requires more feedback to maintain any sort of linearity. We are tryinng to operate in the middle of a curve and not just above each one's off point were one is shuting down and the other looking to turn on.
Yes, they do. But we don't want to rely on feedback to correct everything. It would be cool if we didn't rely on feedback to correct anything
This might be correct. I was just concentrating on other details. It's not easy to explain this bis and English is my first language
Yes, this was my point also, and I believe the difference between ProFET and F5, and Euvl's approach, and all of this discussion of comp devices and tweaking gain, source Rs, etc etc.