Not too long ago, I built an SET amp, just to try tubes out. It certainly was different from transistor amps, and I learned to appreciate the sound, especially with acoustical music.
Recently, I tried using a mosfet in a circuit much like the SET, but of course with lower voltage/higher current, and a 70 volt PA type transformer on the output. My purpose was to see if triodes really were better, or if the transformer/circuit config was important. The result was pretty good, but it still sounded like a transistor amp, and not like a tube amp.
What I have noticed is that the triode gives a sound that has great depth and coherency, and easy, effortless dynamics. The transistor, even the mosfet, seems to compress dynamics and turn the sound into a 2 dimensional wall of sound that is in your face.... very detailed and clean, but a compressed image of the original.
Has anyone done any work to document this difference? Is my observation all wet, or in agreement with others? Just curious....
Kent
Recently, I tried using a mosfet in a circuit much like the SET, but of course with lower voltage/higher current, and a 70 volt PA type transformer on the output. My purpose was to see if triodes really were better, or if the transformer/circuit config was important. The result was pretty good, but it still sounded like a transistor amp, and not like a tube amp.
What I have noticed is that the triode gives a sound that has great depth and coherency, and easy, effortless dynamics. The transistor, even the mosfet, seems to compress dynamics and turn the sound into a 2 dimensional wall of sound that is in your face.... very detailed and clean, but a compressed image of the original.
Has anyone done any work to document this difference? Is my observation all wet, or in agreement with others? Just curious....
Kent
Hi, Kent-
I've heard similar differences between tubes and transistors, even down to the individual device, and I believe the particular effect you mention relates to the nonlinearities of the device's parasitic junction capacitances which all such devices have. In the case of silicon, these capacitances have a dielectric constant, or K, of 12, which is comparable to that of an electrolytic capacitor (which is not to say that these capacitances will have the same audible deficiencies as teeny tiny 'electrolytic' capacitors of similar value - just that they are likely to be somewhat of the same magnitude). And with tubes, the K=1.0000, which is ideal, and the world could do with more ideals, IMO.
I've heard similar differences between tubes and transistors, even down to the individual device, and I believe the particular effect you mention relates to the nonlinearities of the device's parasitic junction capacitances which all such devices have. In the case of silicon, these capacitances have a dielectric constant, or K, of 12, which is comparable to that of an electrolytic capacitor (which is not to say that these capacitances will have the same audible deficiencies as teeny tiny 'electrolytic' capacitors of similar value - just that they are likely to be somewhat of the same magnitude). And with tubes, the K=1.0000, which is ideal, and the world could do with more ideals, IMO.
thoriated said:
Um, okay. I'll forget you said this for a moment...
No
Also wrong, Al2O3 is circa K=5
And it still hasn't been proven that electrolytics sound "bad"...
So what of the mica supports? They're less than 1/32" thick, but still, surely their slightly higher dielectric counts for SOMETHING!?!??!?!?! ARGH!!!!
Of all the unbelievable things in your post, the least believable is that you didn't make the slightest indication of oh I don't know, IMPEDANCE for instance? You need to compare a ZNFB SEP (pentode) amp to a ZNFB SEM (MOSFET) amp. And with a better transformer too, any random 70V transformer is guaranteed crap.
Tim
Gently does it...
Guitar amplifier manufacturers persist in trying to say that MOSFETs are similar to valves. Trouble is, they're a good deal more similar to a pentode than a triode.
As for capacitances, yes, it's true that MOSFETs have capacitances between their electrodes that change dramatically with applied voltage (the semiconductor manufacturers are quite open about this in their data sheets), but careful circuit design should reduce their effects.
Guitar amplifier manufacturers persist in trying to say that MOSFETs are similar to valves. Trouble is, they're a good deal more similar to a pentode than a triode.
As for capacitances, yes, it's true that MOSFETs have capacitances between their electrodes that change dramatically with applied voltage (the semiconductor manufacturers are quite open about this in their data sheets), but careful circuit design should reduce their effects.
sch3matic -
Intrinsic silicon, as used in semiconductor junctions, does indeed have a dielectric constant of 12. You perhaps are confusing it with silicon dioxide, an insulator used to create gates & capacitors in IC fabrication, that has a K of around 4.
As far as electrolytic capacitor K, Nichicon says, at:
http://www.nichicon-us.com/english/lib/alminium.pdf
that the aluminum oxide formulations they use have K's ranging from 7 to 10. I was thinking '9' when I made my last post here. That is why I used 'comparable', perhaps slightly loosely, but still justifiably. IAC, I was giving sand junction parts the benefit of the doubt.
I'm not discussing whether someone's favorite electrolytic capacitor sounds 'bad', an only partially correlated issue, but the fact is, that, as a class, they are less linear and lossier than capacitor families with lower K dielectrics, and this is readily discernible audibly.
In fact, no serious & knowledgable audiophile denies that dielectrics significantly better than intrinsic silicon have audible deficiencies, and some better than intrinsic silicon have fallen out of favor among the more discerning, polyester being a case in point even though it is easily a better performing dielectric than used in electrolytic capacitors.
I indeed considered bringing up the fact in my original response that the transfer function of a standard MOSFET resembles that of a pentode, not a triode, but didn't want to obfuscate the issue at hand by doing so, or by being sidetracked by peripheral implementation issues.
I assume you are simply attempting a funny by bringing up any relative capacitive contributions of mica spacers or the glass envelope of a tube, because they are still dielectrically better than intrinsic silicon, yet contribute no more than a tenth of a pF or two, and usually substantially less, to any interelectrode capacitance.
So, I stand by everything I posted.
Intrinsic silicon, as used in semiconductor junctions, does indeed have a dielectric constant of 12. You perhaps are confusing it with silicon dioxide, an insulator used to create gates & capacitors in IC fabrication, that has a K of around 4.
As far as electrolytic capacitor K, Nichicon says, at:
http://www.nichicon-us.com/english/lib/alminium.pdf
that the aluminum oxide formulations they use have K's ranging from 7 to 10. I was thinking '9' when I made my last post here. That is why I used 'comparable', perhaps slightly loosely, but still justifiably. IAC, I was giving sand junction parts the benefit of the doubt.
I'm not discussing whether someone's favorite electrolytic capacitor sounds 'bad', an only partially correlated issue, but the fact is, that, as a class, they are less linear and lossier than capacitor families with lower K dielectrics, and this is readily discernible audibly.
In fact, no serious & knowledgable audiophile denies that dielectrics significantly better than intrinsic silicon have audible deficiencies, and some better than intrinsic silicon have fallen out of favor among the more discerning, polyester being a case in point even though it is easily a better performing dielectric than used in electrolytic capacitors.
I indeed considered bringing up the fact in my original response that the transfer function of a standard MOSFET resembles that of a pentode, not a triode, but didn't want to obfuscate the issue at hand by doing so, or by being sidetracked by peripheral implementation issues.
I assume you are simply attempting a funny by bringing up any relative capacitive contributions of mica spacers or the glass envelope of a tube, because they are still dielectrically better than intrinsic silicon, yet contribute no more than a tenth of a pF or two, and usually substantially less, to any interelectrode capacitance.
So, I stand by everything I posted.
Neither triode or mosfet that I used is ideal, and I knew that (my SET uses 6S4A's in parallel). The mosfet was an inexpensive IRF510 operated at 20 volts, half Amp (wanted to stay well within dissipation). Not much power, but enough to hear. Transformer was generic but in spite of that, it still sounded good (there just wasn't any bass). I wasn't trying to be picky in this test, just general impressions. I'm not even saying the mosfet sounded bad, as it was better than a lot of commercial equipment I've heard! I was just trying to compare two different types of devices with minimal difference to circuit topology, that's why I didn't try caps on output coupling, and I was concerned about impedance.
Hearing that mosfets are more like pentodes does tell me something though..... that's interesting. Sorry if I stirred up trouble here..... didn't mean to do that!
Thanks,
Kent
Hearing that mosfets are more like pentodes does tell me something though..... that's interesting. Sorry if I stirred up trouble here..... didn't mean to do that!
Thanks,
Kent
Apples to apples
Try conecting up the IRF510 with your original SET xfmer this way for a more even comparison (see attached diagram). Use the secondary, disconnect the primary and watch out for High Voltage on the primary. This connection provides some feedback on the Source connection to make it act like a triode. Also, try varying the idle current (via gate to source bias -- I didn't show biasing in the figure, use a fixed bias scheme with capacitor isolation for the gate input and a high value resistor (10K) from the bias supply to the gate, with bias supply relative to source pin. ) Most mosfets follow a varying power law transfer curve that changes from square law at low current to 3/2 power law at some intermediate current to nearly linear at high current. So by varying the operating point, should be able to get near tube sound at some point.
Try conecting up the IRF510 with your original SET xfmer this way for a more even comparison (see attached diagram). Use the secondary, disconnect the primary and watch out for High Voltage on the primary. This connection provides some feedback on the Source connection to make it act like a triode. Also, try varying the idle current (via gate to source bias -- I didn't show biasing in the figure, use a fixed bias scheme with capacitor isolation for the gate input and a high value resistor (10K) from the bias supply to the gate, with bias supply relative to source pin. ) Most mosfets follow a varying power law transfer curve that changes from square law at low current to 3/2 power law at some intermediate current to nearly linear at high current. So by varying the operating point, should be able to get near tube sound at some point.
Attachments
notes
You may need a large electrolytic cap in series with the speaker if you operate at high DC idle current due to the internal resistance of the secondary causing a voltage drop that will offset the speaker coil. Also, the SET transformer will have some maximum DC current it will tolerate on the secondary before it saturates magnetically. (If the xfmr spec for DC on the primary is known, can transform it upwards by the turns ratio to get secondary DC current rating.)
Don
You may need a large electrolytic cap in series with the speaker if you operate at high DC idle current due to the internal resistance of the secondary causing a voltage drop that will offset the speaker coil. Also, the SET transformer will have some maximum DC current it will tolerate on the secondary before it saturates magnetically. (If the xfmr spec for DC on the primary is known, can transform it upwards by the turns ratio to get secondary DC current rating.)
Don
i would try with some more voltage, say 30 - 50V, and 1 - 2 A. Take a look into Nelsons ZENs : http://www.passdiy.com/projects/zenlite1.htm
http://www.passdiy.com/amps.htm
http://www.passdiy.com/amps.htm
Re: Apples to apples
http://www.diyaudio.com/forums/showthread.php?s=&threadid=6338&highlight=
I built an amp like that. Has been used daily for almost 2 years now. I am still unable to hear any shortcomings in the sound.smoking-amp said:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=6338&highlight=
Um sorta
"Um no"
Well, obviously the mosfet by itself is much more like a pentode with high output impedance, giving a poor damping factor without Neg feedback. But lets look at how things work here.
For the triode, current emission depends on the electric field (to the 3/2 power) at the cathode, which for a triode is something like Vgk + Vpk/Mu. (The pentode has the plate voltage Vpk mostly shielded by the screen grid so as to have little effect.) Since the output AC signal on the plate is inverted, the internal feedback field from the plate acts as negative feedback, subtracting 1/Mu of the output Vpk from the Vgk field.
Similarly, for the mosfet, current emission depends on the electric field at the source due to the gate or Vgs (to some power between 2 and 1 depending on current level). The drain voltage is shielded by the channel pinchoff effect so has little effect like for the pentode.
In the "Apples to apples" circuit, the transformer is connected so as to subtract about 30% of the output voltage from the Vgs, acting as negative feedback to current emission in much the same way as the triode does internally. The effective "Mu" of the "Apples to apples" circuit will then be about 3 (depending on the tap ratio of the xfmr, 4 Ohms is about 0.7 turns ratio to 8 Ohm)
Output impedance from a cathode, or source in this case, is about 1/gm of the device. Since the speaker output is taken across the xfmr., the output impedance will be transformed to (1/0.3) squared times 1/gm, giving a damping factor of about 8 ohms times gm/(1/0.3)squared, or 0.8 gm. For the IRF510, gm is typically around 2.0, so damping factor will be around 1.6. So the circuit will have output impedance similar to a triode SET.
I rest my case your honor. (putting flame retardant suit on and sneaking quietly out the back door before mayhem breaks loose)
Don
"Um no"
Well, obviously the mosfet by itself is much more like a pentode with high output impedance, giving a poor damping factor without Neg feedback. But lets look at how things work here.
For the triode, current emission depends on the electric field (to the 3/2 power) at the cathode, which for a triode is something like Vgk + Vpk/Mu. (The pentode has the plate voltage Vpk mostly shielded by the screen grid so as to have little effect.) Since the output AC signal on the plate is inverted, the internal feedback field from the plate acts as negative feedback, subtracting 1/Mu of the output Vpk from the Vgk field.
Similarly, for the mosfet, current emission depends on the electric field at the source due to the gate or Vgs (to some power between 2 and 1 depending on current level). The drain voltage is shielded by the channel pinchoff effect so has little effect like for the pentode.
In the "Apples to apples" circuit, the transformer is connected so as to subtract about 30% of the output voltage from the Vgs, acting as negative feedback to current emission in much the same way as the triode does internally. The effective "Mu" of the "Apples to apples" circuit will then be about 3 (depending on the tap ratio of the xfmr, 4 Ohms is about 0.7 turns ratio to 8 Ohm)
Output impedance from a cathode, or source in this case, is about 1/gm of the device. Since the speaker output is taken across the xfmr., the output impedance will be transformed to (1/0.3) squared times 1/gm, giving a damping factor of about 8 ohms times gm/(1/0.3)squared, or 0.8 gm. For the IRF510, gm is typically around 2.0, so damping factor will be around 1.6. So the circuit will have output impedance similar to a triode SET.
I rest my case your honor. (putting flame retardant suit on and sneaking quietly out the back door before mayhem breaks loose)Don
a familiar circuit
By the way, the "Apples to apples", circuit when implimented on the primary side for tubes, goes by the name "partial cathode feedback" and was very popular with N. Crowhurst and Quad and McIntosh too, in a slightly modified form using separate cathode windings for P-P. When used in P-P form (unmodified) it would be called the "Circlotron" if using 50% cathode feedback.
Another descriptive name would be the "Mu changer circuit" since it can be used effectively to make pentodes behave as triodes (without having to connect the screen grid to the plate). The effective Mu is close to 1/(fraction of the total primary winding for the cathode's use). (not much new under the sun)
Don
By the way, the "Apples to apples", circuit when implimented on the primary side for tubes, goes by the name "partial cathode feedback" and was very popular with N. Crowhurst and Quad and McIntosh too, in a slightly modified form using separate cathode windings for P-P. When used in P-P form (unmodified) it would be called the "Circlotron" if using 50% cathode feedback.
Another descriptive name would be the "Mu changer circuit" since it can be used effectively to make pentodes behave as triodes (without having to connect the screen grid to the plate). The effective Mu is close to 1/(fraction of the total primary winding for the cathode's use). (not much new under the sun)
Don
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.