TGM8 - my best amplifier, incredible bass, clear highs, no fatigue (inspired by Rod Elliot P3a)

Jump to Latest
#383
It was because I started simulating with transistor models that I had on hand, plus the BD140 is generally a good choice. But the KSA is said to be slightly better and I had some left over from a previous project so that's what I used. I didn't compare them in the actual circuit and I would expect both of them to perform well in the TGM8.

I don't have my notes to hand, but I don't remember it getting very warm. I did leave a little space to fix a small metal-tab homemade type of heatsink to it if needed but I never found it was needed. Mind you, I haven't pushed the amp to its limits - the amp is intended to have headroom to minimize clipping on peaks but normal listening will barely wake it up! I didn't intend it to run flat out, it's for music pleasure.
 
Last edited:
#387
What a good question - I've never asked that question yet I've also followed the mantra that low Cob is good for the VAS. Sitting here and thinking about it, I would hazard that it is a) based on listening tests, over the years and with many designs by many engineers, b) one effect of Cob is that it acts as a negative feedback loop from the collector to the base. This feedback reduces distortion. To work well you want to have a high quality capacitor in this location and we usually add additional external capacitance to bring up the total value in order to make the overall amplifier stable when using a global negative feedback loop. Dr. Self has written about this too, saying that enough capacitance here helps linearize the VAS stage. However, the intrinsic capacitance 'inside' the transistor is a poor 'non-linear' capacitor compared with what we can add externally, so minimizing the internal Cob so that the VAS performance is determined more by the external capacitor should be beneficial, c) perhaps there is a correlation between low capacitance and larger base widths for higher voltage operation that leads to less non-linearity from the Early effect.
 
Last edited:
#389
Bigun said:
What a good question - I've never asked that question yet I've also followed the mantra that low Cob is good for the VAS. Sitting here and thinking about it, I would hazard that it is a) based on listening tests, over the years and with many designs by many engineers, b) one effect of Cob is that it acts as a negative feedback loop from the collector to the base. This feedback reduces distortion. To work well you want to have a high quality capacitor in this location and we usually add additional external capacitance to bring up the total value in order to make the overall amplifier stable when using a global negative feedback loop. Dr. Self has written about this too, saying that enough capacitance here helps linearize the VAS stage. However, the intrinsic capacitance 'inside' the transistor is a poor 'non-linear' capacitor compared with what we can add externally, so minimizing the internal Cob so that the VAS performance is determined more by the external capacitor should be beneficial, c) perhaps there is a correlation between low capacitance and larger base widths for higher voltage operation that leads to less non-linearity from the Early effect.
Click to expand...

Very good reasoning.... with a low Cob bjt, we provide the correct amount of miller compensation depending on a good quality stable cap.
 
#391
I've never used JFETs before. Other than reading of subjective positive opinions on their 'sound' it has usually been easier to use a BJT in my designs. Given their wide parameter variability and generally low transconductance and power handling there is also a dearth of through-hole JFET parts available anymore.

For a Singleton input, combining the JFET with a BJT cascode will do a couple of things, because what it does is hold the voltage at the JFET drain at a lower voltage level and confine this voltage to a much smaller swing. This 'constant voltage' type of operation will reduce power dissipation in the JFET, allow use of higher voltage rails and reduce the input Miller capacitance so as to keep a wide bandwidth. It's said that the cascoding device will not add it's own signature to the sound, but it is going to have some impact because it is operating as a common base/gate amplifier so it is clearly in the signal path. The 'constant voltage' mode will also reduce distortion in the JFET (mostly 2nd harmonic), which may be good or bad depending on what you want.

A folded Cascode does essentially the same thing as a regular Cascode, except that the output signal has a different reference. All signal voltages in the amplifier are referenced to something - voltages are always between two points. The signal voltage could be 'reference to' or 'with respect to' ground (which the input and outputs usually are), or with respect to a power rail. In TGM8 the signal on the collector of the input device is with respect to the +ve power rail. A cascode doesn't change that, the output from the cascode would also be with respect to the +ve rail. A folded Cascode would flip the output to be with respect to whatever the collector (or drain) of the cascoding device is connected to - so it could be handy if one wanted to reference the signal to ground, or to the -ve rail. My favourite example is the Taylor Follower.

You have me interested in JFETs with all this chit chat.
 
Last edited:
#392
Last edited:
#394
As I understand it, the low transconductance of JFET's compared with BJT's makes it difficult to drive a conventional BJT VAS stage. I believe one of the reasons the Fetzilla team settled on a MOSFET VAS stage was as a consequence of that. With an LTP you could add a current mirror to improve the situation but this isn't possible with a single transistor input. But we don't want an LTP, right?

There are still JFETs available that have adequate Vgds for a moderate power amp without having to resort to cascodes.

What about an EF2 or compound pair VAS?
 
#395
I'm not familiar enough with the Fetzilla but I think the idea was to use all-FETs so a BJT VAS was out of the question. There are some considerations in favour of a FET VAS, I had considered it for TGM8. With a higher current they can be quite linear and lower transconductance might be good if you are aiming for low open loop gain - I believe Fetzilla was an attractive idea to Hugh when he saw the design because of the low open loop gain. Anyhow, it's a worthy topic to search on - the FET VAS.

So, on the contrary, my first inclination is to think the lower transconductance of the JFET means that you might prefer a BJT VAS in order to keep the open loop gain reasonably high. Of course, there are many ways to improve the open loop gain, if you want to, including the topology I've used in the TGM8 front-end. Somebody more familiar with these considerations will have to chime in !

No, we don't want an LTP 😀

What attracts me to the JFET is the manner in which it's characteristics become triode-like under certain operating conditions. Actually, more than that, it's possible to establish operating conditions in which the JFET is very linear - reminds me I should go back and re-read Nelson's article on Zen8 and Zen9 and the one on 'the sweet spot'. You might find this quite interesting too.
 
Last edited:
#398
Nice unit Bigun

Congratulations... it is a very nice amplifier... highly developed...improved......this is nice Bigun....very nice.

There are some guys that are really shameless...go buzzing around bothering people that works....envy is the name of these guys..... shameless because are kicked by everybody and do not feel ashamed to be around...very lonely and sad these folks that need to be bothering others to feel good.

I see you have one of them dear Bigun...in the reality the leader of the dark mob.

You know, as no moderator fix that abnormal situation we live in the forum, i have understood, that someone very powerful in our forum, maybe one guy from the management group, have created a fake subscription to be playing around with us... without being discovered..you know... a fake nickname to be using another fake personality...some kind of sadistic need to drive others to be mad....this is the only explanation i can have to understand why the man was not kicked out from the forum and still go buzzing around...no one goes against him...why?... because he may be very powerfull man to the management.

Why i understood this way...because one day.... three or four years ago....i was so bothered by a small gang that i understood that they had advantage on me...i was working as an idiot and they were having fun, for free, with me and other guys that had spent money and time to our community....i felt they were in advantage...and i was nearby to decide to create another subscription...another computer, another IP... another internet provider...a fake one..to start to bother these guys too...because this could produce much more fun than to be victim of them...... latter on i decide not to do...if they have free pass in the forum is because they are protected by the forum...maybe they are the forum.

Also..... as you see my dots..... they would discover me very fast....because these dots i use....no one use this....heheheheh....

I have talked with hundreds...everybody dislike that man behavior...no one approve....how he still survive around us?

Audio message sent to your inbox...home adress.

Carlos
 
Last edited:
#399
Bigun said:
A folded Cascode would flip the output to be with respect to whatever the collector (or drain) of the cascoding device is connected to - so it could be handy if one wanted to reference the signal to ground, or to the -ve rail. My favourite example is the Taylor Follower.

You have me interested in JFETs with all this chit chat.
Click to expand...

That is what I meant, referencing the signal to gnd, or in this case to -ve rail.
 
You must log in or register to reply here.
Top Bottom