| Stefano |
| Hi to you all, is there anyone of you that has used that transformer as a phase splitter? I'm thinking about using N channel MOSFet only for the output stage of an amplifier in class A. Has anyone any imput? |
|
|
| SY |
| I've done this with my tube-fet hybrid driver. It works fine as long as you get the load impedance right (10K) or use an RC series network with a higher load resistance. If you do a search on "SYclotron", you can see the schematic of what I did. |
|
|
| Nelson Pass |
I think I would see if Jensen has something geared around a
lower impedance, as the Mosfet capacitance may exceed the
design of that particular transformer. |
|
|
| Stefano |
| Tank you for reply. I'm trying to reduce power dissipated, but i don't want to use complementary push-pull with hexfet since they are not so complementary. |
|
|
| Nelson Pass |
I realize that people might expect better performance from
identical plus and minus parts, but I think that you get those
benefits only under certain limited conditions. In general, I
don't find that inexact complements are much of a performance
problem, and are probably less problematic than inserting
a transformer in the circuit. |
|
|
| GRollins |
Nelson,
All right...I'll take the bait. If you can spare a minute, please elaborate on the case of the non-complementary complements (insulting complements, perhaps?). It seems intuitive that to get good results the Ns and the Ps should be as close as possible to precise mirror images of one another. It's not the first time that something "obvious" has turned out to be false, so I can't say that I'm surprised, but you've managed to make my curiosity itch.
Grey |
|
|
| Steve Eddy |
| quote: | Originally posted by Stefano
Hi to you all, is there anyone of you that has used that transformer as a phase splitter? I'm thinking about using N channel MOSFet only for the output stage of an amplifier in class A. Has anyone any imput? |
You could do that, though as Nelson pointed out, the high input capacitance of the MOSFETs could be problematic.
If you know what sort of capacitance you're dealing with, you can EMail Jensen and they can help you with a compensation network.
But you might also consider using an output transformer instead. The JT-11SS-DLCF would be a good candidate. It's less expensive than the 11P-1 and it's low impedance so you shouldn't have the problems with capacitance that you would with the 11P-1.
The thing you have to concern yourself with is if any of the sources are capacitively coupled. Because output transformers have such low primary inductance, in order to avoid low frequency resonant peaks you need to make sure that the coupling capacitance is sufficient. Instead of the usual 2uF or so that works ok in most situations, you'd need to bring it up to several hundred uF.
If your sources are direct coupled and you haven't much DC offset, you should be just fine.
se |
|
|
| Nelson Pass |
| quote: | Originally posted by GRollins
It seems intuitive that to get good results the Ns and the Ps should be as close as possible to precise mirror images of one another. |
It does, but consider that you want the nonlinearity of one
device whose current is increasing to match the nonlinearity
of its complement as the current decreases so that the two
distortions cancel. Perfect matching (both devices having the
same curve exactly) does not guarantee this.
Like a triode on a resistive load line, you get some first order
cancellation of distortion, but it's never perfect. |
|
|
| GRollins |
Aye, that's kinda what I meant by mirror image--one zigs whilst the other zags, net result being a straight line in the sense of plus one and minus one summing to zero. But if they aren't truly complementary you've got a pig and a bag, and they won't cancel one another. From what you're saying, I gather that even a little bit is a net improvement, regardless of precise mirror imaging.
Yes?
Grey |
|
|
| Nelson Pass |
Absolutely, and not at all trivial. I have seen reductions by
a factor of as much as 10:1 from such cancellations, carefully
chosen in differential pairs, but I haven't seen that much
improvement in complementary followers in output stages. |
|
|
| GRollins |
So why not in complementary outputs? What's different?
Grey |
|
|
| Nelson Pass |
The biggest improvements in cancellations occur in 2nd
harmonic, so if you can concentrate the characteristic
nonlinearity in 2nd harmonic, which is most characteristic
of single-ended operation, then that's where you get the
most bang for the buck. Your typical diff-pair is simply a
single ended common source/emitter device talking to another
such through the sources. As long as you don't overdrive the
inputs, you get a decent "sum of distortions = 0" effect.
As you increase the swing on a diff pair, they will start having
lots of 3rd harmonic. This is a good reason not to overwork
your input diff pair.
In complementary output devices, you usually see dominant 3rd
and higher harmonics (especially with matching) because the
voltage and current swing are large, and they don't
cancel much. |
|
|
| jam |
Mr.Pass
Is the third harmonic largely independant of bias level? |
|
|
| Nelson Pass |
It is very dependent in the diff pair case, less so for the
complementary followers (assuming we're talking about the
operating extremes, not the lower levels). |
|
|
|
