JFET input, MOSFET VAS, LATERAL output = Perfect!!

[Ian Finch]

Originally Posted by mikelm

....My parasitic is in region of 16meg hz and square wave ringing due to a difficult load ( 100nF ) is around 1 meg.
100 ohm gate stoppers cure the parasitic but exacerbate the square wave ringing so I'm hoping to steer between them with ferrites.....

I think you are on the right track with this approach, depending on the design of the leaded ( I assume a pre-wound slug or bead) ferrite.

Usually, above 10MHz, single pass beads are adequate for low level damping. I can't imagine core saturation being an issue at gate charging currents but others may care to comment on that.

As an alternative, I've also used 47R/100 pF Zobels on each device (gate to drain), and these work OK too - at least with mild cases of just a few volts P-P which is what I usually find with a 100MHz analogue 'scope. Mosfets seem to go much higher with oscillation than BJTs - perhaps as high as 100 MHz though the laterals I use seem to range 10 - 60 MHz, as you find.

This is pretty cheap and easy to try out so why not give it a bash whilst waiting? Perhaps too, say, 47R stoppers in combination may be less of a problem there with ringing. I haven't measured this myself and neither tried out dual die Mosfets which could be more problematic, so it could be interesting.

[GregH2]

I'm back from the dead and keen to start experimenting again...now to go read the last few months worth of progress

[mikelm]

Originally Posted by swordfishy

I'm back from the dead and keen to start experimenting again...now to go read the last few months worth of progress

Hi Greg,

Good timing from my point of view. After a longish break doing other things I just did a rebuild of my DC linked version - sounds even better than before

Actually it sounds very valve like.

I'm planning to post about my progress soon.

I wonder where you are with ur research ?

mike

[AKSA]

Greg,

I'm so glad you are back, I've been trying to phone you for weeks....

I have now sold 172 pcbs for the FetZilla, it has been very successful. IN fact, Group Buy #3 is now complete and mailout is this week.

Yesterday the amp was tested using Clio 5, the latest THD software, using a battery powered laptop. Very quiet. Brilliant results, too, very low THD, around 0.008%, all H2, at 1W output. Max output is only 40W, however, rather less than we expected, but still goes damn loud.

I will email you now, in fact.

Hugh

[GregH2]

Ah, great news and great to hear from you both. To be honest I haven't done a single audio related thing in months, but that hopefully changes this weekend. Mikelm, do post your final version.

Hugh, glad you're still here after so long. Great news about the group buys. I take it people are happy with the sound?

Amazing results with the testing too! I would never have thought 4 fets could measure so well while being so pleasing sonically. Why is this thing not a cult icon yet? haha.

If I had to criticise one thing about the amplifier, after months of listening, it would have to be a slight loss of competence and control during very complicated music with high intensity transients. That said, probably no more than any other mortal amplifier.

Look forward to progressing the art over the coming months. First job this weekend is to clear out the shed and move my audio "lab" from the guest room.

[kees52]

Transients are lost because of compression, we need to get free voltage swings, like in cascodes, then it will do oke, there are cascoded amps out there like kaneta, but as I heard here, it is a succes, little thing will happen and it keeps us busy, congratulations.

[AKSA]

Folks,

I have the Clio measurements on the FetZilla, kindly supplied by Graeme Huon and Paul Bysouth, highly competent EE friends of mine here in Melbourne, taken Monday 27th Feb 2012. They are taken at 1KHz, into an 8R resistive load, at outputs respectively of 1W (+10dB), 13.3W (+20dB), and 26.5W (+23dB):

Fundamental 10.00dB Fundamental 20dB Fundamental 23dB
H2 -93dB 5.012E-10 H2 -78dB 1.585E-08 H2 -78dB 1.58489E-08
H3 unmeasureable H3 -93dB 5.012E-10 H3 -88dB 1.58489E-09
H4 unmeasureable H4 -105dB 3.162E-11 H4 -93dB 5.01187E-10
H5 unmeasureable H5 -102dB 6.31E-11 H5 -103dB 5.01187E-11
H6 -96dB 2.512E-10 H6 -106dB 2.512E-11 H6 -110dB 1E-11
H7 unmeasureable H7 -110dB 1E-11 H7 -101dB 7.94328E-11

H8, H9 and H10 are all unmeasureable

THD by voltage 0.00274% (1W, +10dB); 0.01284% (13.3W, +20dB), and 0.0134% (26.5W, +23dB)
Measurements below about -80 dB can be assumed to be insignificant except in ultraclean designs

These figures are surprisingly good, but with output bias set to 370ma the first 5.5 watts of output are in Class A, so perhaps not so surprising after all.

I would add that the THD does not correlate well with the sound qualilty; rather, it is the profile of the distortion artefacts that affect the perceived sound quality for the human ear.

Not bad for a jfet, two mosfets, and two lateral output fets - only five active devices in total!

My thanks to Graeme and Paul for generously giving their time to measure the FetZilla!

Cheers,

Hugh

[Bigun]

the 'trick' with cascodes is to ensure some emitter degeneration of the cascode device. The 'memory distortion' guy called it a 'magic resistor'. I just like a little 'room' for the cascode to work within and some degeneration trades off a bit of ppm for a better harmonic structure. Maybe it's all bullcrap, but I used it in my TGM5 all the same and I like the sound

[AKSA]

Yes, you should drop approx 200mV across this resistor.

Quite by chance, I discovered this about 15 years ago working on a cascoded VAS, I was amazed, but could never explain it.

Hugh

[keantoken]

My suspicion is that a monotonic harmonic structure indicates an amp with generally well-behaved characteristics such as group delay (phase linearity) and consistent behavior over a range of operating points. IE, when the harmonics are monotonic, so is the group delay. Monotonic profile indicates that each device in an amp behaves in monotonic fashion; this also indicates that the current draw of the output stage is mirrored without major distortion all the way to the input stage; in effect, linear (monotonic) open-loop gain curve. This is the ideal characteristic for a high-performance amplifier; it ensures stable and consistent operation across a range of operating points and I think it should be a starting point for any amplifier design. This is opposed to super high-gain designs where the open-loop linearity may be very bad.

In so many words, monotonicity seems to be key. It ensures that the compensation and feedback in the amp work like intended and the amp does not fight with itself. If an amp is designed without open-loop linearity, you usually discover that it can be finicky, or it may require special external compensation to be stable with certain loads (zobels and cables).

This deviates from the conventional idea of looking at an amplifier as a voltage amplifier. Really, to make a stable and consistently well-behaved amplifier it needs to be seen not as a voltage amp or a current amp, but as an impedance amplifier, and both it's voltage and current response addressed. Current response, as well as voltage gain response, needs to be properly compensated in order for an amp to be consistently well-behaved and stable. Knowing this is key to making improved, reliable designs with our current technology, rather than trying to use faster and more linear semiconductors to support flawed designs.

These are my thoughts based on my studies on amplifier stability. Maybe I have tunnel vision, but who could blame me when I have a hammer and everything looks like a nail.