Last edited:
It has taken me a while to come around to this for some reason, but I'm dumping the ECX10x16 simulation models; they're totally wrong. The SK1058/J162 models from Cordell should be the closest we will get. SWF, I would suggest you do the same, because the ECX models give wildly inaccurate simulation.
Here are my files.
- keantoken
Here are my files.
- keantoken
What do you mean, inaccurate compared to their performance in reality, or, inaccurate as in they're hard to get stable in a simulation?
OK, my friends and I have been doing extensive listening tests on the SWF/AKSA version of this amp, the one built from the boards I designed - and have almost sold out.
It is a very unusual and distinctive amplifier, I would say. There are clear distinctions between this and bipolar amps of a similar design.
FetZilla sounds very natural. It is churchyard quiet, one of the most silent amps I have ever experienced, up there with my NAKSAs and probably just as quiet. It has wonderful bass, not too tight (loop gain is just over 50dB), and very musical. Midrange is extremely clear, sounds rather like a very good tube amp. Top end is good, but not as good as the midrange, which is marvellous on vocals and acoustic.
It does something quite special with horns, guitar and piano. I really like the sound quality. It is a very fast amp, and without an input filter is 1dB down at better than 300KHz with VAS running at 13mA. Sound field is very wide, speakers simply disappear. Image depth is average, I would not rank it with the best in this quality, at which tube amps excel. But it still manages to sound very real and dynamic, with great vitality and verve to the sound, none of that twee, slightly fragile sound you hear from SETs.
It appears quite reliable. I have had it running for two weeks continuously with 400mA of quiescent and it never gets overly hot, with idle current rock solid, more stable than most bipolar amps. It sounds way more than 50W, quite amazing.
I have not built Mikelm's variant, which has all DC coupling, choke supply, and is thus a different beast. However, IMHO the principal determinant of the sound quality is the singleton jfet input stage and pmos VAS, which are common to both amps.
Cheers,
Hugh
It is a very unusual and distinctive amplifier, I would say. There are clear distinctions between this and bipolar amps of a similar design.
FetZilla sounds very natural. It is churchyard quiet, one of the most silent amps I have ever experienced, up there with my NAKSAs and probably just as quiet. It has wonderful bass, not too tight (loop gain is just over 50dB), and very musical. Midrange is extremely clear, sounds rather like a very good tube amp. Top end is good, but not as good as the midrange, which is marvellous on vocals and acoustic.
It does something quite special with horns, guitar and piano. I really like the sound quality. It is a very fast amp, and without an input filter is 1dB down at better than 300KHz with VAS running at 13mA. Sound field is very wide, speakers simply disappear. Image depth is average, I would not rank it with the best in this quality, at which tube amps excel. But it still manages to sound very real and dynamic, with great vitality and verve to the sound, none of that twee, slightly fragile sound you hear from SETs.
It appears quite reliable. I have had it running for two weeks continuously with 400mA of quiescent and it never gets overly hot, with idle current rock solid, more stable than most bipolar amps. It sounds way more than 50W, quite amazing.
I have not built Mikelm's variant, which has all DC coupling, choke supply, and is thus a different beast. However, IMHO the principal determinant of the sound quality is the singleton jfet input stage and pmos VAS, which are common to both amps.
Cheers,
Hugh
Last edited:
Stability itself is a poor indicator of model accuracy. It is more convenient and more direct to look at the capacitance values in the model to see if they are accurate. Their transfer curve is inaccurate; it is too smooth, giving highly optimistic distortion performance.
Furthermore, the P-channel model requires a way too large gate stopper in order to balance with the N-channel model.
- keantoken
A word on the mosfet VAS degeneration.
If you do the simulation and examine the distortion artefacts of the FetZilla and start moving the degen resistor around you notice the artefacts - and the THD - move up and down in step. More degen = more distortion, less deg = less distortion.
I have not seen any application in audio where degeneration does not affect OLG, even where a CCS is in use. If this were not so, then current would never change in the VAS - it would be rock steady - and thus signal would not pass to the output devices.
I may be wrong, but my entire experience with this and other amps points to degen having profound effect on performance. Place a sim probe on the gate of the VAS and you will clearly see a range of voltage operating. At +20dBU output (13W into 8R, 28Vpp) I get a 86mVpp variation at the mosfet gate - which clearly then passes to the source and thus the degen resistor, affecting OLG. The corresponding current variation at the degen resistor itself is almost 3mA, from 11.6mA to 14.5mA. The FetZilla uses the hybrid CCS/bootstrap circuit, of course.
Cheers,
Hugh
If you do the simulation and examine the distortion artefacts of the FetZilla and start moving the degen resistor around you notice the artefacts - and the THD - move up and down in step. More degen = more distortion, less deg = less distortion.
I have not seen any application in audio where degeneration does not affect OLG, even where a CCS is in use. If this were not so, then current would never change in the VAS - it would be rock steady - and thus signal would not pass to the output devices.
I may be wrong, but my entire experience with this and other amps points to degen having profound effect on performance. Place a sim probe on the gate of the VAS and you will clearly see a range of voltage operating. At +20dBU output (13W into 8R, 28Vpp) I get a 86mVpp variation at the mosfet gate - which clearly then passes to the source and thus the degen resistor, affecting OLG. The corresponding current variation at the degen resistor itself is almost 3mA, from 11.6mA to 14.5mA. The FetZilla uses the hybrid CCS/bootstrap circuit, of course.
Cheers,
Hugh
Last edited:
Hugh, For Mikel's own circuit he may be right because he uses a Jfet across the G-S of the VAS rather than a 1.2k resistor. Thus theoretically Vgs modulation will be nearly immaterial. However I totally agree for the versions which just use a resistor.
PS I think I've made a few nonsense circuit comments recently. It's because I'm still learning!
- keantoken
PS I think I've made a few nonsense circuit comments recently. It's because I'm still learning!
- keantoken
Hi KT,
I cranked up the simulator, and compared current variation through the VAS and its degeneration resistor with the 1K2 resistor from VAS gate to +Vcc, and with a 3.5034mA CCS - a perfect one no less - from VAS gate to +Vcc.
In both cases, and for +20dB output (13W into 8R) the VAS current variation was identical, that is, 11.6mA to 14.5mA.
However, I then compared THD. With the specified 47R degeneration, and using Mike's CCS concept on the gate of the VAS, I saw 0.008815%, while with 0R degeneration, the THD increased very slightly to 0.008881%, with virtually no change in the H2 which stood at -84.5dB in both cases.
In both cases I still saw the current variation through the VAS at 11.6mA to 14.5mA.
So, Mike is right that with a CCS, degeneration or otherwise has little effect; however, there is still approx 40mVpp variation in the signal as applied to the gate of the VAS, clearly visible on the LTSpice graphs. Nevertheless, with a resistor between gate of VAS and +Vcc, degeneration does have an effect.
I suspect there are two effects working in opposite directions here; the very high loop gain of a bipolar circuit is more predictable, but the low loop gain of the fet circuits makes it more critical.
Hugh
I cranked up the simulator, and compared current variation through the VAS and its degeneration resistor with the 1K2 resistor from VAS gate to +Vcc, and with a 3.5034mA CCS - a perfect one no less - from VAS gate to +Vcc.
In both cases, and for +20dB output (13W into 8R) the VAS current variation was identical, that is, 11.6mA to 14.5mA.
However, I then compared THD. With the specified 47R degeneration, and using Mike's CCS concept on the gate of the VAS, I saw 0.008815%, while with 0R degeneration, the THD increased very slightly to 0.008881%, with virtually no change in the H2 which stood at -84.5dB in both cases.
In both cases I still saw the current variation through the VAS at 11.6mA to 14.5mA.
So, Mike is right that with a CCS, degeneration or otherwise has little effect; however, there is still approx 40mVpp variation in the signal as applied to the gate of the VAS, clearly visible on the LTSpice graphs. Nevertheless, with a resistor between gate of VAS and +Vcc, degeneration does have an effect.
I suspect there are two effects working in opposite directions here; the very high loop gain of a bipolar circuit is more predictable, but the low loop gain of the fet circuits makes it more critical.
Hugh
Last edited:
At 1KHz, there will not be much VAS current to speak of, but at high frequencies I think it will matter. This is because the MOSFETs require capacitive drive from the VAS, causing greater Vgs modulation, which imposed across the 1.2k resistor will cause a proportional current through it. Theoretically a current source here will eliminate this induced error current; but I think the effect would only be seen at very high frequencies.
Degeneration will not change the VAS current much, because the MOSFET drive requirements are not effected by it. Rather the direct effect of degeneration is to increase the voltage modulation across the G-S resistor, which loads the input stage.
So I think degeneration matters for phase and frequency response, not THD.
- keantoken
Degeneration will not change the VAS current much, because the MOSFET drive requirements are not effected by it. Rather the direct effect of degeneration is to increase the voltage modulation across the G-S resistor, which loads the input stage.
So I think degeneration matters for phase and frequency response, not THD.
- keantoken
thank you,
adding TMC network (or TPC - I am confused with this...):res 470 ohm to the output, 56p+120p caps in series instead of 33p vas
much improve high harmonics behavoiur
Last edited:
the hydraulic equivalent to the VAS is a water valve that is throttled to part open.
Increase the opening angle from the handle and more fluid flows. Decrease the opening angle and less fluid flows.
The VAS works identically. Apply current to the input Drain load and the resulting voltage changes the "opening" of the VAS Drain to Source Conductance. More current to the Drain load and more conductance through the VAS. Less current to the Drain load and less conductance through the VAS.
The VAS changes current to suit the input conditions and that change of current is what the output devices measure and respond to.
Last edited:
How much the current changes in the VAS completely depends upon the design you employ.
If you use a simple resistor as the load with 10mA steady state the current changes by about 10mA ( 5mA - 15mA ) for about a 20V peak to peak output
However with the CCS I use the current changes by only 0.5uA for the same output swing.
If you use a simple resistor as the load with 10mA steady state the current changes by about 10mA ( 5mA - 15mA ) for about a 20V peak to peak output
However with the CCS I use the current changes by only 0.5uA for the same output swing.
I would emphasize that different folks, different strokes.
I am guided by the sound quality; I'm more than happy to discuss the phenomenon in technical detail, BUT, we have to compare apples and apples.
Mike is right, his CCS in the VAS is very different in that his current barely changes. There will be a modulating voltage on the gate of VAS, higher at HF as KT states, but can I ask, what are we actually arguing about here?
If it's turning into a competition, I see little point, and will gracefully withdraw. Mike, you might like to start your own thread on your version because your amp is palpably different, particularly with its CCS between VAS gate and power rail, and will offer different sound quality as well. Two circuits in the one thread, both pursuing different development, is a bit confusing. Which amp is better is more a matter of subjective judgement.
Cheers
Hugh
I am guided by the sound quality; I'm more than happy to discuss the phenomenon in technical detail, BUT, we have to compare apples and apples.
Mike is right, his CCS in the VAS is very different in that his current barely changes. There will be a modulating voltage on the gate of VAS, higher at HF as KT states, but can I ask, what are we actually arguing about here?
If it's turning into a competition, I see little point, and will gracefully withdraw. Mike, you might like to start your own thread on your version because your amp is palpably different, particularly with its CCS between VAS gate and power rail, and will offer different sound quality as well. Two circuits in the one thread, both pursuing different development, is a bit confusing. Which amp is better is more a matter of subjective judgement.
Cheers
Hugh
Hi Hugh,
Actually, I was not really arguing - just presenting the two extreme possibilities for the sake of clarification
The funny thing I have just discovered is that whether I have CCS or resistor in IP stage and also whether I have CCS or simple resistor in VAS and 3 different combinations of these ( not including simple R in both which is too little OLG ) with vastly different OLG in the audio band, the ideal miller cap always seems to be about 33pF
Actually, I was not really arguing - just presenting the two extreme possibilities for the sake of clarification
The funny thing I have just discovered is that whether I have CCS or resistor in IP stage and also whether I have CCS or simple resistor in VAS and 3 different combinations of these ( not including simple R in both which is too little OLG ) with vastly different OLG in the audio band, the ideal miller cap always seems to be about 33pF
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.