Build or simulate
I simulated until I determined the best type of devices.Then I built.
I simulated until I determined the best type of devices.Then I built.
Hi onto, salam kenal. Kerja di Polytron Kudus ya? You sound knowledgeable about this EEEngine. Are you doing research on this topic?
I have to say that I am not familiar with all of the possible N-channel-only output stages, but I have never seen a quai-complementary N-channel output stage using source follower on top and CFP structure on bottom that I thought would be able to perform as well as a complementary output stage.
However, here is an example of an N-channel-only output stage that some might like, and would be symmetrical. Use a pair of N-channel devices in push-pull with an output transformer (as if they were tubes).
Cheers,
Bob
The quasi-complementary PP power buffer stage from post 94 (N-channel output stage using source follower on top and CFP structure on bottom) was developed from Mr. Nelson Pass. I have found it in the TAA Book - go to
http://www.diyaudio.com/forums/pass...omplementary-mosfet-citation-12-taa-book.html
At this time no P-Ch MOSFETs available - therefore that design. After release P-Channel MOSFETs from IRF several months later Mr. Pass updated this circuit and remove the CFP on bottom.
The topology you like is that one from follow URL's - so I think:
Zero Feedback Transformer Audio Power Amplifier
index
http://www.diyaudio.com/forums/solid-state/24744-push-pull-using-only-n-channel-mosfets.html
and what about the circlotron like follow URL's (my personal favorite of all push pull variantes)?
http://www.amplimos.it/images/CIRCLOTRON_50W.bmp
http://www.diyaudio.com/forums/solid-state/3891-circlotron-amp-using-n-channel-mosfets.html
http://www.davidsaudio.com/Nine.pdf
6moons audio reviews: Thorens TEM 3200
A URL collection from me of the most possible "N-channel-only output stages" you will find there:
http://www.diyaudio.com/forums/soli...e-ended-related-solid-state-output-stage.html
and there:
http://www.diyaudio.com/forums/soli...better-audio-non-complements-audio-power.html
For me the most interested topology for universal use to modify exist amps, if the orig. output power devices too expensive or no longer available seems to be follow:
http://www.diyaudio.com/forums/solid-state/154388-its-cheap-its-n-its-dirty-its-circlomos-4.html
unfortunately I haven't time to check out the behaviour until this day.
Attachments
Last edited:
There are 3 topologies that come to mind that make a completely symetrical amp using onlu N-type devices. They all have a phase splitter of some kind in common:
1) totem pole output with bootstrapped concertina type phase splitter (best made with a MOSFET - a BJT would show errors because of Ic not equal to Ie). The top transistor only appears to be working as a follower - the bootstrapping actually makes it common source operation.
2) totem pole output with P-type LTP as phase splitter (usually requires extra positive rail for best rail to rail performance, although bootstrapping can be used with some difficulty). Again, top device only appears as a source follower, since the drive is voltage derived from the LTP output current, terminated to the output, this is a form of bootstrappinf, so the top device again reverts to common source.
3) Circlotron. It requires differential drive with the same DC offset, and is actually possible to do it two ways - source follower based and common source based, but the latter is much more difficult to bias. The source follower version is actually the only topology that comes to mind that utilizes both MOSFETs as anything near source followers and does not require a transformer to merge the phases back together into a single output. The only real disadvantage here are the multiple floating power supplies, bias is actually very simple (don't confuse this with the Bongiorno patent that shows how to cleverly use the Vbe of a BJT output circlotron to compensate itself thermally).
1) totem pole output with bootstrapped concertina type phase splitter (best made with a MOSFET - a BJT would show errors because of Ic not equal to Ie). The top transistor only appears to be working as a follower - the bootstrapping actually makes it common source operation.
2) totem pole output with P-type LTP as phase splitter (usually requires extra positive rail for best rail to rail performance, although bootstrapping can be used with some difficulty). Again, top device only appears as a source follower, since the drive is voltage derived from the LTP output current, terminated to the output, this is a form of bootstrappinf, so the top device again reverts to common source.
3) Circlotron. It requires differential drive with the same DC offset, and is actually possible to do it two ways - source follower based and common source based, but the latter is much more difficult to bias. The source follower version is actually the only topology that comes to mind that utilizes both MOSFETs as anything near source followers and does not require a transformer to merge the phases back together into a single output. The only real disadvantage here are the multiple floating power supplies, bias is actually very simple (don't confuse this with the Bongiorno patent that shows how to cleverly use the Vbe of a BJT output circlotron to compensate itself thermally).
you mean this topologies:
1) totem pole output with bootstrapped concertina type phase splitter (best made with a MOSFET - a BJT would show errors because of Ic not equal to Ie). The top transistor only appears to be working as a follower - the bootstrapping actually makes it common source operation. ... is it really so? - read this:
The Tube CAD Journal: symmetrical solid-state output stages
http://www.amplimos.it/images/N-CH1.JPG
2) totem pole output with P-type LTP as phase splitter
http://peufeu.free.fr/audio/schemas/Kaneda_Mosfet.jpg
http://www.amplimos.it/images/2sk77 amp YAMAHA B-1.gif
Yes it is so, at least if you want to keep it DC coupled and reasonably uncomplicated.
From your examples, only the bottom amp in the John BRoskie link works it's outputs as source followers but it's AC coupled. The amplimos amp, and the kaneda amp are precisely the arhitecture I mentioned, as is the B-1, but in the case of the B1 it's using depletion mode VFETs (SITs) so the second stage uses N-ch MOSFETs and develops negative bias - which could be confusing you. Here we are talking about MOSFETs so in 99.99% of all cases enhancement mode devices are assumed, which lends itself to a P-ch (or PNP) LTP as the phase splitter in my example (2).
Regarding Broskie's amp, it is possible to do this DC coupled using current conveyors (also LTP's and even simple concertinas could be used to level shift the drive signals for the MOSFETs) but it gets complicated and not easy to manage re stability, as you can't disregard the extra stages.
The problem with all kinds of parallel symmetric (against complementary symmetric) topologies is, that somewhere in the circuit a kind of differential VAS and/or phase splitter means has to be used in such a way, that the two "phases" see different voltage swings and/or completely different output impedances.
Some incarnations of this parallel symmetric paradigm try to solve this by using differential VAS stages running at extreme bias voltage levels and heavy cascoding. The B-I e.g. uses a -200V (the -C voltage) supply to achieve this (and the negative bias of the V-FETs), but nevertheless there is still a bunch of problems:
- The VAS for the upper (push) transistor still sees (AC wise) the full voltage swing, while the corresponding path of the lower (pull) transistor sees only the gate voltage variation.
- Another drawback of these common parallel schemes is the very ill-informed use of non-inverting feedback, which btw. kills any symmetry from the input stage onwards (by means of unavoidable common mode distortions) ...
- Just another con is the inherent asymmetrical slew rate bahaviour (especially of the totem pole flavours of this paradigm) ...
The asymmetries introduced by these mechanisms can't be reduced very well, so why even think about such topologies at power levels of 200W (or even 22kW ;-), especially since these tend to love Class A operation ?
And finally, believe me, the Yamaha B-I produces lots of even order distortions, which it shouldn't do if the parallel symmetry idea could really work here (and even anywhere else) and it even doesn't sound better than (partially) complementary amplifiers of half the complexity and driver supply voltage and ten times better reliability ...
But then there is still the rather dogmatic question, if topological symmetry is an ultimate goal at all, especially since a bridge mode of any topology (and some matching) always leads into a naturally symmetrical behaviour ...
Some incarnations of this parallel symmetric paradigm try to solve this by using differential VAS stages running at extreme bias voltage levels and heavy cascoding. The B-I e.g. uses a -200V (the -C voltage) supply to achieve this (and the negative bias of the V-FETs), but nevertheless there is still a bunch of problems:
- The VAS for the upper (push) transistor still sees (AC wise) the full voltage swing, while the corresponding path of the lower (pull) transistor sees only the gate voltage variation.
- Another drawback of these common parallel schemes is the very ill-informed use of non-inverting feedback, which btw. kills any symmetry from the input stage onwards (by means of unavoidable common mode distortions) ...
- Just another con is the inherent asymmetrical slew rate bahaviour (especially of the totem pole flavours of this paradigm) ...
The asymmetries introduced by these mechanisms can't be reduced very well, so why even think about such topologies at power levels of 200W (or even 22kW ;-), especially since these tend to love Class A operation ?
And finally, believe me, the Yamaha B-I produces lots of even order distortions, which it shouldn't do if the parallel symmetry idea could really work here (and even anywhere else) and it even doesn't sound better than (partially) complementary amplifiers of half the complexity and driver supply voltage and ten times better reliability ...
But then there is still the rather dogmatic question, if topological symmetry is an ultimate goal at all, especially since a bridge mode of any topology (and some matching) always leads into a naturally symmetrical behaviour ...
Alas, different people have different views of what constitutes symmetry. The blind pusuit of any particular symmetry at the expense of common sense should be avoided. Just because both the push and pull output transistors are the same type in a quasi-complementary output stage does not make its operation the least bit symmetrical.
Similarly, those who pursue symmetry just because it looks cool on a schematic are just as likely to end up with a poor-sounding amplifier as anyone else.
No amplifier design that I know of is perfectly symmetrical in every sense. Pick your poison, but do so wisely.
Cheers,
Bob
BUZ90x spice models ?
Do you happen to have (LT)spice models of the BUZ90x's which actually work? I'm in great favor of those Magnatec/Exicon laterals (have used them several times with zero issues) but I have not found yet spice models that are realistic in any true sense of the word...
I agree, virtually no MOSFET spice models out there, either for laterals or verticals, are trustworthy. Most don't model sub-threshold conduction at all, expecting the MOSFET to behave as a pure square-law device.
Most models also do a poor job of modeling the nonlinear gate-drain capacitance in a way that is maybe satisfactory for switching applications but quite bad for linear applications. At least a proper VDMOS model from LTspice does a decent job of modeling the gate-drain capacitance.
Cheers,
Bob
Bob,
did you ever have a chance to evaluate/compare how close simulation with these better MOSFET models come in regards of (real world) group delay modulation (PIM) ?
I would expect, that a proper modelling of the non linear junction capacitances would also allow to have a rough idea, how an amplifier behaves in this direction (Hoping that the bipolar transistors models are as precise as well), at least for the audio range.
I agree; but the decisive factor for me is always the nature of the resulting asymmetry and thus the kind of the resulting distortions.
We know, that there are ear friendly sounding amps (low order distortions like tubes) and "unmusical" sounding amps (at least in the perception by a comparable test) - through high order distortions and compression effects by the power supply.
If the developer has choose a design to avoid the last, one has certainly a good sounding amp, even if the THD-Value is not particularly low.
If the developer has choose a design, that looks cool on a schematic (e. g. most used "true complementary" output stage) and THD-Values at 1 KHz just above (or even below) the detection limit, the probability of a really good sonic behaviour is low, especially by idle currents between 20mA and 100mA.
In this case it would be interesting to know, how sounds not OTL-amps like the topology from Mrs Susan Parker ("Zeus" or "Zeuz" Series) in Class AB between 20 - 100 mA quiescent current instead class-A - I have never heard. Who knows about such amplifier brands and models?
Last edited:
I think that, if complementary symmetry makes sense, then in output stages running at rather low bias currents, simply because one can provide a common bias voltage and temperature compensate bias currents quite easily. The popular error correction schemes based on feedback seem to be also easier to apply to complementary output pairs.
Try to do this with a pseudo symmetrical stage or the like and you have serious troubles (or a mess) in real life to stay around the correct bias currents, where gm variation is minimal (if gm variation can be held low with these concepts at all).
This doesn't mean that I want to promote low bias currents here (IMO no straight forward scheme is capable to produce a good sounding amplifier with bias currents in the mA range, except you use a power cascode in order to minimize thermal variations), neither do I suggest to use NFB based EC as a means to reduce distortion of a Class (A) B amp (because it e.g. doesn't reduce relative gm variation and its thermal dependencies at all, so it only conceals the inherent weaknesses of the approach by reducing or limiting the absolute error to a certain amount).
In one of my previous posts I wanted to say, that the visually tempting symmetry of the schematics doesn't necessarily lead to real electrical symmetry at all, so I would accept (partial) symmetry only if there exists a profound technical (!!) reason to use it and this can actually be a really feasible alternative.
nitroboy's question moved here:http://www.diyaudio.com/forums/solid-state/167520-best-replacement-transistors-amp.html
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.