ilimzn,
Can one get a good approximation of a device's input capacitance from a known source impedance and output frequency cutoff?
For example, NP states the following here:
Can one get a good approximation of a device's input capacitance from a known source impedance and output frequency cutoff?
For example, NP states the following here:
Can we surmise from this that the input capacitance is approximately 1300pF using a 600 Ohm source impedance drive and 200kHz cutoff?
I would love to hear it. Just use existing mosfet and be done with it.
Then I can sell my 2sk82 jfets at twice the cost.
At some point, a comparison between wrenchone's and diyVFET amps will be welcome. Maybe twice the cost is generous on your part. How large is your stash?
IMHO, diyF6 fits the Schade practice of triodizing the characteric behaviour of pentodes. The loop feedback of diyF6 emanates from a current sourcing high output impedance [~ I remenber 30 Ohms], and terminates in a summing junction. Thus, diyF6 is also game for comparison with diyVFET; fortunately noting that we have DIYers and professionals with autrhentic golden hearing to discriminate among them. For diyF6, one may consider including a 4 Ohm non-inductive power resistor in series with its output to ~ match that of diyVFET amp so as equalize damping of the common loudspeaker used in the test.
I guess that explains the remarkable lack of H2 in Nelson's sony sit amps... without nulling pots it's pretty amazing....
Certainly they're a pain to bias and the amps need regulated rails but perhaps this time, it's worth it..
A new thread in the Pass Labs forum that is full of therapeutic fun is at hand. I have no doubt that DIYers will appreciate your comparison, and revelations. Thank you.
If I recall correctly, the input transformer has a 1:2 step-up, so a 600 ohm source would be reflected as 2400 ohms, hence the capacitance would effectively be around 350pF. I have measured around 500 for a pair but did not take any precautions to minimize strays. It can get high at very low bias voltages but that's not what is normally used.
It is especially interesting since these were optimized from the original single chip design to work in a class D amp - where this would not be the main requirement. BTW the Yamaha B1 uses only N-ch parts, although selected, but it does have an AC balance pot.
Now, the amps themselves do not strictly need regulated supplies but then you need the bias supply to be modulated by the drain supply or you get standing current modulation with main supply ripple, which intermodulates with the actual signal. Because of this, it's simplest to regulate the supplies and you have solved that problem - and you can throw in a protection scheme for cheap too.
I don't believe NP is using a transformer in the SIT-1 and SIT-2.
I'm trying to get a feel for what Cis is for NP's PASS-SIT-1 FET, with no luck. No one has come forward with a datasheet.
How did you measure the input capacitance as 500pF?
nelson, i'm sure you will have tried these sony vfet complementaries in self-bias fashion - like they would have done with tubes (if they had P channel tubes that is ). Source resistors (maybe 8ohms on the P side and an ohm less for the N) bypassed with large electros.
Tradeoffs for sure vs what you show in the article but were the results not satisfying perhaps ?
Tradeoffs for sure vs what you show in the article but were the results not satisfying perhaps ?
Thanks Nelson.Operated Common-Source with 18 dB of gain, I get about 3 nF effective input
capacitance.
Based on 200kHz (-3dB) and 3nF input capacitance, you would need a source impedance of 265 Ohms. Does that seem correct?
Would you say the PASS-SIT-1 has about the same characteristic inter-electrode capacitances as say the IRFP140?
Sorry, I thought you were referring to a pair of Sony 2SK82/2SJ28.
In the latter case it's simple to measure using a capacitance bridge (HP4261). It is capacitor coupled to a suitably biassed pair, the coupling caps are sufficiently large so that the series connection results in a negligible error. Also, power supplies are locally decoupled to ground, where the other end of the brifge is connected. A small test signal is selected and the output is loaded with the expected load.
Then a DC offset is applied ('input signal') to check for capacitance variation respective to the input signal. This is incidentally why a small measurement signal is selected on the meter, not to produce an error due to capacitance nonlinearity.
Of course, the VFET pair is on a suitable heatsink as this is essentially a working follower, working at DC conditions into the rated load.
Just like Vgso and some other characteristics, the capacitance varies from VFET instance to instance within the same rank, but I measured ~500pF on average for the pair.
2SK70 has about 600-700pF on it's own and it's counterpart J20 almost double that. As i said, Sony's offerings are different because they were optimized for ultrafast switching, and you want the least possible capacitance there. NEC looked at electrical complementarity but unlike Yamaha decided not to produce a N-ch part with a higher Vdsmax to get there, so they ended up with unequal capacitances. They do however have signifficantly higher gm than the Dony parts - gm was what Sony sacrificed.
Yes, at low voltage (few V bias) you are looking at a huge varicap diode :{, plus Miller, also from a larger Cdg due to low Vds. Keep in ind I only measured common drain (follower).
Last edited:
Not wildly different. - of course that's also taking into account the 18 dB
gain through the capacitance from Drain to Gate, which is dominant.
For Common Drain, the apparent capacitance is much less, but I don't have
a figure off hand.
Guys,
sorry for running (half) mad here on the forum, but I heard about the Nelson Pass "VFET" amp project (burning hundreds or even thousands of V-FETs prior selection) only a few minutes before my first post in this thread ...
An yes it was 1. April in Germany, but to be honest, my posts were meant (almost) seriously .. at least at the time I wrote them.
And I expected to get all kinds of negative criticism of almost all people here. Just like stirring up a hornet's nest and that's exactly, what I wanted to provoke The massive reaction also showed to me, that some people were deeply hit in their unreflected believes.
Probably many of you interpreted things into the NP amplifier, which makes it superior in your eyes (and believing, what is written in a commercial Hifi press release) and I did the opposite (because I know things about SONY V-FETs others may not know). Who is right now ? Probably not even NP knows.
.... We don't have a schematics of it so far ... and that's good, that nobody tries to clone it, burning even more V-FETs for a horrible 2% of distortion afte matching a hundred ...
Zeljko:
What I said, doesn't very much contradict what you wrote.
The only thing I was definitly wrong about, is the topology (bridge mode, regulators) and part count of the NP amplifier (which I only heavily speculated based on the available photos).
For a bridged amp, the things certainly get a little more sane, but this also has it's own disadvantages (you design for a unloaded voltage swing of perhaps +-100V and probably can only finally use +- 60V with load). That disadvantage, which already exists for non bridged PP V-FET amps in a more mild form is btw. greatly reduced by running the V-FETs in cascode with bipolars (like SONY TA-N7B), as you also can read in official technical SONY literature. But this then wouldn't be an amp with a "pure" V-FET output stage anymore ...
Complementary+bridge symmetry certainly helps canceling second order nonlinear effects of junction caps, but doesn't help a lot for the remaining symmetrical modulation of the group delay and such an effect can often be heard easily, even if no widely accepted easy measurement routines (or a large promoting group of people) exists for that.
If probably some of the people here think, that a -3dB rolloff of a V-FET output stage at 20kHz without any additional FB is OK, this simply means that they may not consider odd order PIM in the range of >10us (not ns or ps) bad for sound. I at least do, as I would consider a 10us integrated Jitter in the audio range very bad in a DAC.
Declaring the 0.5A idle current in e.g. a TA-8650 as being Class B strictly follows the definition of Douglas Self, that Class B is where the minimal (static) crossover distortion is. Above the associated bias point you get gm rise (often called gm doubling), below you get gm droop close to zero.
This is IMO absolutely compatible to the commonly accepted definition, that in Class B the push and pull devices are exactly on for 50% of the time. This actually assumes gain devices with a constant gm, so no overlap of conduction. Unfortunately this is only of theoretical nature (for the text books), since such devices don't exist, so in a first order approximation the time of conduction of one side has to be weighted by its relative contribution to overall (added) gm. This leads to real world Class B operation having a significant idle current, whatever the actual characteristic is, hyperbolic, quadratic or logarithmic etc. or how you exactly define, what minimal crossover distortion exactly means, only the Class B associated bias current is different.
Probably the following is still not entirely accepted or understood:
All classic PP V-FET amps run roughly at the point, where minimal (third order) crossover distortions appears and this means Class B at least for those who define it like Douglas Self e.g. does.
What V-FETs really makes shine is the fact, that the bias in Class B is rather high and extremely constant over device temperature, a thing bipolar and MOSFET Class B (without tricks) can normally only dream of.
Any try to operate V-FETs in Class AB (What others may already call Class A) increases crossover distortion, instead of lowering it.
Nelson:
Could you please explain, at which impedance the V-FETs are driven in your design (if it is not a secret). In this case we have at least a rough idea, if they are driven in a rather sluggish way or with "iron fist" ...
What is the bias current per device ?
I don't have any idea, how many of the V-FETs you bought, actually went into this project and if you still have many ones left, which some day could find back their way into SONY amps as spares ?
Again:
Sorry for the very provoking style of my last posts, but this doesn't really change the core, of what I have to say.
And:
You all should not forget, that I'm probably the only person in the west (or at least in this forum), who listened to all the classic V-FET amps ever commercially produced (including the hyper rare HA-500F, except GX-1 tone cabinet amp), so I can probably best tell, what the "classic" V-FET sound is about and what not and how those compare to recent designs.
sorry for running (half) mad here on the forum, but I heard about the Nelson Pass "VFET" amp project (burning hundreds or even thousands of V-FETs prior selection) only a few minutes before my first post in this thread ...
An yes it was 1. April in Germany, but to be honest, my posts were meant (almost) seriously .. at least at the time I wrote them.
And I expected to get all kinds of negative criticism of almost all people here. Just like stirring up a hornet's nest and that's exactly, what I wanted to provoke
The massive reaction also showed to me, that some people were deeply hit in their unreflected believes.Probably many of you interpreted things into the NP amplifier, which makes it superior in your eyes (and believing, what is written in a commercial Hifi press release) and I did the opposite (because I know things about SONY V-FETs others may not know). Who is right now ? Probably not even NP knows.
.... We don't have a schematics of it so far ... and that's good, that nobody tries to clone it, burning even more V-FETs for a horrible 2% of distortion afte matching a hundred ...
Zeljko:
What I said, doesn't very much contradict what you wrote.
The only thing I was definitly wrong about, is the topology (bridge mode, regulators) and part count of the NP amplifier (which I only heavily speculated based on the available photos).
For a bridged amp, the things certainly get a little more sane, but this also has it's own disadvantages (you design for a unloaded voltage swing of perhaps +-100V and probably can only finally use +- 60V with load). That disadvantage, which already exists for non bridged PP V-FET amps in a more mild form is btw. greatly reduced by running the V-FETs in cascode with bipolars (like SONY TA-N7B), as you also can read in official technical SONY literature. But this then wouldn't be an amp with a "pure" V-FET output stage anymore ...
Complementary+bridge symmetry certainly helps canceling second order nonlinear effects of junction caps, but doesn't help a lot for the remaining symmetrical modulation of the group delay and such an effect can often be heard easily, even if no widely accepted easy measurement routines (or a large promoting group of people) exists for that.
If probably some of the people here think, that a -3dB rolloff of a V-FET output stage at 20kHz without any additional FB is OK, this simply means that they may not consider odd order PIM in the range of >10us (not ns or ps) bad for sound. I at least do, as I would consider a 10us integrated Jitter in the audio range very bad in a DAC.
Declaring the 0.5A idle current in e.g. a TA-8650 as being Class B strictly follows the definition of Douglas Self, that Class B is where the minimal (static) crossover distortion is. Above the associated bias point you get gm rise (often called gm doubling), below you get gm droop close to zero.
This is IMO absolutely compatible to the commonly accepted definition, that in Class B the push and pull devices are exactly on for 50% of the time. This actually assumes gain devices with a constant gm, so no overlap of conduction. Unfortunately this is only of theoretical nature (for the text books), since such devices don't exist, so in a first order approximation the time of conduction of one side has to be weighted by its relative contribution to overall (added) gm. This leads to real world Class B operation having a significant idle current, whatever the actual characteristic is, hyperbolic, quadratic or logarithmic etc. or how you exactly define, what minimal crossover distortion exactly means, only the Class B associated bias current is different.
Probably the following is still not entirely accepted or understood:
All classic PP V-FET amps run roughly at the point, where minimal (third order) crossover distortions appears and this means Class B at least for those who define it like Douglas Self e.g. does.
What V-FETs really makes shine is the fact, that the bias in Class B is rather high and extremely constant over device temperature, a thing bipolar and MOSFET Class B (without tricks) can normally only dream of.
Any try to operate V-FETs in Class AB (What others may already call Class A) increases crossover distortion, instead of lowering it.
Nelson:
Could you please explain, at which impedance the V-FETs are driven in your design (if it is not a secret). In this case we have at least a rough idea, if they are driven in a rather sluggish way or with "iron fist" ...
What is the bias current per device ?
I don't have any idea, how many of the V-FETs you bought, actually went into this project and if you still have many ones left, which some day could find back their way into SONY amps as spares ?
Again:
Sorry for the very provoking style of my last posts, but this doesn't really change the core, of what I have to say.
And:
You all should not forget, that I'm probably the only person in the west (or at least in this forum), who listened to all the classic V-FET amps ever commercially produced (including the hyper rare HA-500F, except GX-1 tone cabinet amp), so I can probably best tell, what the "classic" V-FET sound is about and what not and how those compare to recent designs.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.