I don't understand what's so appealing with perfect bias stability.
I feel very comfotable with overbias (~160mA/device) at switch-on going to optimum at normal working conditions at moderate listening levels and finaly to underbiasing (60mA/dev or even lower) when the heatsink has a fever (~45deg.C).
I mean with proper design it should not make a big difference wheater bias current is slightly away from optimum point and negative thermal coefficient makes sense as thermal economics is concerned.
I use bjt triples( |<,|<,|<, ), complementary Vbe multiplier and pre-drivers out of main heatsink.
Adam
I feel very comfotable with overbias (~160mA/device) at switch-on going to optimum at normal working conditions at moderate listening levels and finaly to underbiasing (60mA/dev or even lower) when the heatsink has a fever (~45deg.C).
I mean with proper design it should not make a big difference wheater bias current is slightly away from optimum point and negative thermal coefficient makes sense as thermal economics is concerned.
I use bjt triples( |<,|<,|<, ), complementary Vbe multiplier and pre-drivers out of main heatsink.
Adam
Re: Re: Re: Re: A question on Lateral's mosfets
Hi Charles,
The zeners I used are In5245 15V zeners, with a capacitance of 25 pF. See the Vishay datasheet. There are two in series, back-to-back, so the net is more like 13 pF.
But, more importantly, they are NOT connected to ground. The pair is connected from gate to source, so their effect is bootstrapped, just like the 1000+ pF of gate capacitance that they are connected in parallel with.
They are insignificant.
Cheers,
Bob
Charles Hansen said:
Hi Charles,
The zeners I used are In5245 15V zeners, with a capacitance of 25 pF. See the Vishay datasheet. There are two in series, back-to-back, so the net is more like 13 pF.
But, more importantly, they are NOT connected to ground. The pair is connected from gate to source, so their effect is bootstrapped, just like the 1000+ pF of gate capacitance that they are connected in parallel with.
They are insignificant.
Cheers,
Bob
With all this discussion about MOSFET gate resistors, I thought I'd mention a Renesas app note that deals with the subject here. There's a section titled "Analysis of Oscillation in Source Follower Circuits". It looks like they are neglecting Cgd though, so the results may only be pertinent to laterals.
Re: Thermal Bias Instability
Why not put all 16 Thermal Trak diodes in series, take the voltage across all of them and divide it down to derive a sensor voltage. Or, have two diode strings, one for each sex output, divide, and apply each one to half of a complimentary Vbe.
This way, all outputs are monitored to derive an average. The diode string could be fed from a 25ma constant current source. Also, the voltage division could be varied for less than full diode increments.
Bob Cordell said:
Why not put all 16 Thermal Trak diodes in series, take the voltage across all of them and divide it down to derive a sensor voltage. Or, have two diode strings, one for each sex output, divide, and apply each one to half of a complimentary Vbe.
This way, all outputs are monitored to derive an average. The diode string could be fed from a 25ma constant current source. Also, the voltage division could be varied for less than full diode increments.
Re: Re: Re: Re: Re: A question on Lateral's mosfets
Yes, you are right. But only at idle.
When the parts are driven near clipping (and Vds heads towards zero), the Cgs of the vertical parts becomes much, much worse than the equivalent lateral parts. Look at Figure 12 of the app note Andy_C linked:
http://documentation.renesas.com/eng/products/transistor/rej27g0017_charalterstics.pdf
To see how this affects things (in a common-source circuit, I believe) when used for switching, compare Figure 11 (lateral) and Figure 14 (vertical). Look at those nice glitches in both the drive voltage and the output voltage caused by those non-linearities. Nasty, nasty, nasty.
Yes, this is also true, but not a huge effect. The best place to see this is Figure 1 of the Stereophile test report of our V-1 power amp:
http://stereophile.com/solidpoweramps/635/index5.html
Although not labeled on the graph, there are curves for both 8 ohm loads and 4 ohm loads. The high frequency response isn't as good into 4 ohms precisely because of the effect you mention. (The Cgs isn't fully "bootstrapped" out of the picture as the load impedance decreases.)
The way around the lower transconductance of the lateral parts is to simply use more of them. Depending on what vertical parts you are replacing, you will probably have to use 4 to 8 times as many lateral parts to get equivalent transconductance. The advantage of the lateral parts are:
a) Designed for audio, complements are easy to find.
b) Lower Cgd.
c) MUCH more linear Cgd -- lower distortion.
d) Easy to bias to zero tempco, no thermal compensation required.
e) Sounds better (try it and see!)
Disadvantages:
a) Higher cost.
Nelson Pass said:
Yes, you are right. But only at idle.
When the parts are driven near clipping (and Vds heads towards zero), the Cgs of the vertical parts becomes much, much worse than the equivalent lateral parts. Look at Figure 12 of the app note Andy_C linked:
http://documentation.renesas.com/eng/products/transistor/rej27g0017_charalterstics.pdf
To see how this affects things (in a common-source circuit, I believe) when used for switching, compare Figure 11 (lateral) and Figure 14 (vertical). Look at those nice glitches in both the drive voltage and the output voltage caused by those non-linearities. Nasty, nasty, nasty.
Nelson Pass said:
Yes, this is also true, but not a huge effect. The best place to see this is Figure 1 of the Stereophile test report of our V-1 power amp:
http://stereophile.com/solidpoweramps/635/index5.html
Although not labeled on the graph, there are curves for both 8 ohm loads and 4 ohm loads. The high frequency response isn't as good into 4 ohms precisely because of the effect you mention. (The Cgs isn't fully "bootstrapped" out of the picture as the load impedance decreases.)
Nelson Pass said:
The way around the lower transconductance of the lateral parts is to simply use more of them. Depending on what vertical parts you are replacing, you will probably have to use 4 to 8 times as many lateral parts to get equivalent transconductance. The advantage of the lateral parts are:
a) Designed for audio, complements are easy to find.
b) Lower Cgd.
c) MUCH more linear Cgd -- lower distortion.
d) Easy to bias to zero tempco, no thermal compensation required.
e) Sounds better (try it and see!)
Disadvantages:
a) Higher cost.
Re: Thermal Bias Instability
Again, I have not found this to be a problem in the real world. There are hundreds of thousands (if not millions!) of BJT amplifiers in the world that have paralleled output devices, but by-and-large they are pretty darned reliable. All of these theoretical considerations simply are not borne out in the real world.
Bob Cordell said:
Again, I have not found this to be a problem in the real world. There are hundreds of thousands (if not millions!) of BJT amplifiers in the world that have paralleled output devices, but by-and-large they are pretty darned reliable. All of these theoretical considerations simply are not borne out in the real world.
roender said:
You asked me what I thought. I told you, and now you want to argue with me.
My opinions are based on decades of building and listening experience. Since you don't want to listen to me, I would suggest building your circuit and listening to it. If you are happy with the way it sounds, then you are all set. Nothing else matters.
If you are unhappy with the way it sounds, then you have a big job. Everything, every single detail, both in circuit topology and parts quality, will affect the way the amplifier sounds. Then you have a job to figure out what is causing the problems you are experiencing.
Good luck.
Shall I presume that you have build it an listened to it? Then, I want to thank you for your advice.
I have build it and listened and I'm very happy with results, either sonically and stability, but your presumptions based on experience make me very curios to transform my output stage into T and listen to it and compare with original.
Again, thank you very much and I'm sorry for "incident"
I have build it and listened and I'm very happy with results, either sonically and stability, but your presumptions based on experience make me very curios to transform my output stage into T and listen to it and compare with original.
Again, thank you very much and I'm sorry for "incident"
roender said:
I haven't built that specific output stage, but I have listened to many amplifiers that use complementary feedback pairs. Even if that is the only feedback loop in the amplifier, I still hear a sonic signature due to the feedback. After listening to many amplifiers with varying designs and various ways of applying feedback, I came to the conclusion that the less feedback was employed in a design (and the shorter the loop), the more "natural" the sound quality became.
I simply took this observation to its natural conclusion -- eliminate all feedback. There have only been a handful of solid-state amplifiers made like this, and none were available at the time so I made my own. I have been using zero-feedback designs for 15 years now, and there has been nothing that makes me the slightest bit interested in using feedback in my designs.
For a nice example of a zero-feedback solid-state amplifier, please refer to this post:
http://www.diyaudio.com/forums/showthread.php?s=&postid=340375&highlight=#post340375
And if you remove some (or better yet, all) of the feedback from your amp, please let us know what you hear.
Oddly enough (since this is a DIY community) very few people have actually tried this. But the ones who have seem to also prefer the sound without feedback. Here is a link to the ONLY person I could find that actually tried it:
http://www.diyaudio.com/forums/showthread.php?postid=348739#post348739
Hi Charles,
I have tried varying the amount of feedback in some designs. My findings are that the sound quality improves to a point, then starts to become "compressed" or lifeless as feedback becomes too high. The amount of feedback depends on the design. The last one I played with turned out that around 9 dB was the best.
-Chris
I have tried varying the amount of feedback in some designs. My findings are that the sound quality improves to a point, then starts to become "compressed" or lifeless as feedback becomes too high. The amount of feedback depends on the design. The last one I played with turned out that around 9 dB was the best.
-Chris
Charles Hansen said:
Indeed, the experiment of removing the follower output stage
from the loop is very instructive and illuminating, however it has
been done many times, if only by a few.
At PL, the X series has had a history of releases of similar
topologies rendered in different sizes and biases, and for each
of them one of the last decisions has been to include the output
stage in the feedback loop or not.
This is an easier decision for me than for some - besides being
fearless, I use lots of outputs, high (or very high) bias, and big
(or very big) heat sinks. As a consequence, the measured specs
between including the output stage in the loop or not does not
make a huge difference, that is to say the amp's numbers are still
marketable.
For each piece, we evaluate the subjective performance with
and without feedback and make a decision. Some amplifiers were
released without output stage feedback, and others were.
It really came down to what our small listening group preferred,
and if the call was very close, we opted for no feedback.
The trend over the past year or so has been toward feedback,
at very modest levels.
This is, however, the entertainment industry, and as such is also
a fashion industry.
