Someone told me... Carlos, goes making changing in VAS
I made it.... real clear, undoubtly differences were hear... some very good improovements, others not good result.... this one is tuned.... alike on RF amplifier....have to use "the transistor", and the correct condenser.... and "the" correct value....if not...hummmm, no good results.
When reached the point, have to change other parts too.
But the teacher is repeating as a parrot.... i could see this written many times.....(not me).... no one can see the sunligth.
What a crazy world this one!
I remember other things never heard:
The secret is love on each other, and special effort have to be made to love the enemies too.
too much money brings only problems
hapyness is inside us.... alone or not alone.
Nobody will love you as you mother.
Brazilians are wonderfull (hahahahaha)
bye guys... i am happy today, father's day, hearing Johnny Mathis...hummm god singer.... good sound.... i do not here music... other way to use year guys... do not know what he is saying ... this is not important.
Carlos
I made it.... real clear, undoubtly differences were hear... some very good improovements, others not good result.... this one is tuned.... alike on RF amplifier....have to use "the transistor", and the correct condenser.... and "the" correct value....if not...hummmm, no good results.
When reached the point, have to change other parts too.
But the teacher is repeating as a parrot.... i could see this written many times.....(not me).... no one can see the sunligth.
What a crazy world this one!
I remember other things never heard:
The secret is love on each other, and special effort have to be made to love the enemies too.
too much money brings only problems
hapyness is inside us.... alone or not alone.
Nobody will love you as you mother.
Brazilians are wonderfull (hahahahaha)
bye guys... i am happy today, father's day, hearing Johnny Mathis...hummm god singer.... good sound.... i do not here music... other way to use year guys... do not know what he is saying ... this is not important.
Carlos
AKSA said:
please look that the most important "polluter" was the first stage but the problem seems to be solved with a constant power circuit.
What do you think to ommit vas stage?:
To stear the output devices direct from the first stage?
So more current is needed?: For example into the place of higher jfet k246 from the constant power triple simply put IRF610 which can deliver desired power to further power transistors?
Besides I think that a use of more stable thermal packages of the discrete devoces are imposible. Why? Because the thermal effect becomes from the junction, not from the case. And heating the case do not solve this problem too...
Hummm very interesting...cut one stage off
Reducing one stage you will go closer to the wire with gain... this is interesting... i will be waiting his answer, will be interesting to hear what hugh will said.
He will be on line in a matter of.... hummmm ....3 hours
I will be here to read!
Carlos
Reducing one stage you will go closer to the wire with gain... this is interesting... i will be waiting his answer, will be interesting to hear what hugh will said.
He will be on line in a matter of.... hummmm ....3 hours
I will be here to read!
Carlos
Poor old VAS stage - everyone criticises you for things that aren't your fault.
Considering that the VAS is typically biased some 10x the normal dynamic current and the output stage is biased some << 1x the normal dynamic current, I would have thought it better to eliminate the output stage rather than the VAS.
Considering that the VAS is typically biased some 10x the normal dynamic current and the output stage is biased some << 1x the normal dynamic current, I would have thought it better to eliminate the output stage rather than the VAS.
Treading murky waters here since none of these claims are verified and i am not that experienced in this area.
Isn't the problem that over the VAS the voltage Vce varies a lot and the current not so much so.
This gives a signal dependent power over the transistor.
If the Vas is attributed for voltage gain then the varying Vbe(t) will result in ~Vbe(t)*gain error whilst, and now i am comparing to a follower output.
The resulting error is ~Vbe(t)*1
The EF output also has a less voltage across it when there is more current so the effect is less, after a point at least.
Tube VAS maybe?
Conceptually if there is gain in the output stage the effect might be less than a high biased Vas but very dependent on degeneration resistors bias and load.
Isn't the problem that over the VAS the voltage Vce varies a lot and the current not so much so.
This gives a signal dependent power over the transistor.
If the Vas is attributed for voltage gain then the varying Vbe(t) will result in ~Vbe(t)*gain error whilst, and now i am comparing to a follower output.
The resulting error is ~Vbe(t)*1
The EF output also has a less voltage across it when there is more current so the effect is less, after a point at least.
Tube VAS maybe?
Conceptually if there is gain in the output stage the effect might be less than a high biased Vas but very dependent on degeneration resistors bias and load.
The books,and magazines, always informed me, that:
If you can reduce one stage... run fast and cut it out!... one problem less.
In my idea, every stage cutted is good, i do not matter wich one... the reduction idea is great.
Carlos
If you can reduce one stage... run fast and cut it out!... one problem less.
In my idea, every stage cutted is good, i do not matter wich one... the reduction idea is great.
Carlos
The more you cut, the more you are asking of each stage. Otherwise why build anything other than a Zen.
Hjelm says it all....... this is indeed the problem, a simple P = IV issue.
It can thus be shown quite simply that using a single ended, 8mA VAS on the negative rail, with a direct-coupled amp, regardless of the configuration of the output stage if the amp clips to within say 2V of a 50V rail, minimum VAS dissipation at bottom of the waveform is 2 x 0.008 = 16mW, and maximum dissipation at the top of the waveform is 98 x 0.008 = 784mW. This is a 34dB range, quite enough to convince the VAS that it doesn't know if it is coming or going.
In a highly Miller sensitive common emitter configuration this thermal cycling of VAS die is very significant, not least of all because the depletion dielectric coefficient is likely to vary considrably with variations in temperature AND voltage, and likely quite non-linearly.
In any event, the distortion is likely to be higher than H2.
In an emitter follower output stage, however, the chief distortion determinant is variation of Vbe with current and voltage across the output stage. This too is non-linear, but of a lower order, since the EF distortion is always an order of magnitude less than the common emitter configuration.
The global negative feedback loop is expected to pick up these non-linearities and straighten them out. This is, of course, a function of the feedback network itself, the Miller capacitance, the compensation regime used, and the speed of all devices in the loop.
Does this exposition help? I suspect not...... this stuff ain't easy!
Cheers,
Hugh
Re: The books,and magazines, always informed me, that:
ideal is an use of a single device (our trplet of the transistors from memdist) with high gain and current drive abilities. Next only an emmiter or source followers. The amp is ready Do not forget to set the output device's idle current.
destroyer X said:
ideal is an use of a single device (our trplet of the transistors from memdist) with high gain and current drive abilities. Next only an emmiter or source followers. The amp is ready Do not forget to set the output device's idle current.
AKSA said:
Hugh,
Very clear, I agree to your analysis. What about the following solution: use a dual transistor (or possible a transistor array), one transistor for the Vas, and arrange the second transistor in such a way that its Vce is the difference between the Vas transistor Vce and the total supply voltage. That way, the sum of the dissipation of the two will be (largely) constant, and, since they are on the same die, the temp effect should cancel out.
Another option of course is to heavily cascode the Vas as to keep it's Vce small and constant, the P will still vary but much less than 34dB.
Jan Didden
Does silence indicate you are still thinking about this or does it mean you have discarded this comment as insignificant?
go ahead traderbahn... show us your idea.
I supposed, because of your questions, you do not think this is the best solution.[
Can you please, let us know, what is your idea?
Carlos
I supposed, because of your questions, you do not think this is the best solution.[
Can you please, let us know, what is your idea?
Carlos
Jan,
Damn good point! Are you familiar with the diff VAS of the old Hitachi AN amps, commonly used in PA work? These would seem to me to fall into this category, since the two devices are driven in antiphase, and thus their dissipations sum pretty much to a constant (though strictly we need two identical current sources to supply them).
I use antiphase drive for some of my tube circuits, because it permits a very simple power supply since current draw becomes constant, like a diff pair. This saves money and has sonic merit.
The cascode, on the face of it, solves the dissipation variation. However, in a cascode the lower device is operating in constant Vce, so technically merely becomes the current input device. The voltage swing is then presented at the upper device; and it too suffers the same high range of dissipation, so the problem is merely transferred. I have found no sonic benefits with a cascoded VAS, but it sure looks great on the schematic......
Traderbam, you mention an 'ideal voltage source'. If by this you mean an infinite high impedance, voltage only, like a rubbed nylon comb attracting bits of blotting paper on a hot day, then I'm not sure of the usefulness of the concept. I'm not up on this terminology; OTOH it's possible you mean a zero impedance voltage source. Please let me know what you mean. I see the latter as a very useful concept, since the VAS of a power amplifier must drive either gates or bases, which are demonstrably NOT of infinite impedance. This principle mandates use of an emitter follower coupled to the VAS output to drive the output stage. Gates/bases should be shaken not stirred (same is true of grids, in my experience) and this requires a little more horsepower than an infinite resistance voltage source.
On reflection, I believe there are other factors intervening here; I've found huge variations in sonics with only subtle changes in the VAS operating environment, much more than the output stage. In fact, you can take global feedback from the drivers of a conventional Self-type bipolar amplifier and still have a mighty good sound, a little twee, certainly, but very tube-like. This indicates to me that the VAS is most likely the problem, though Padamiecki suggests it is the input stage. I refute this because the input differential stage is operating pretty much at constant Vce and only slight current variation. However, this is my opinion, and YMMV.
Cheers,
Hugh
Damn good point! Are you familiar with the diff VAS of the old Hitachi AN amps, commonly used in PA work? These would seem to me to fall into this category, since the two devices are driven in antiphase, and thus their dissipations sum pretty much to a constant (though strictly we need two identical current sources to supply them).
I use antiphase drive for some of my tube circuits, because it permits a very simple power supply since current draw becomes constant, like a diff pair. This saves money and has sonic merit.
The cascode, on the face of it, solves the dissipation variation. However, in a cascode the lower device is operating in constant Vce, so technically merely becomes the current input device. The voltage swing is then presented at the upper device; and it too suffers the same high range of dissipation, so the problem is merely transferred. I have found no sonic benefits with a cascoded VAS, but it sure looks great on the schematic......
Traderbam, you mention an 'ideal voltage source'. If by this you mean an infinite high impedance, voltage only, like a rubbed nylon comb attracting bits of blotting paper on a hot day, then I'm not sure of the usefulness of the concept. I'm not up on this terminology; OTOH it's possible you mean a zero impedance voltage source. Please let me know what you mean. I see the latter as a very useful concept, since the VAS of a power amplifier must drive either gates or bases, which are demonstrably NOT of infinite impedance. This principle mandates use of an emitter follower coupled to the VAS output to drive the output stage. Gates/bases should be shaken not stirred (same is true of grids, in my experience) and this requires a little more horsepower than an infinite resistance voltage source.
On reflection, I believe there are other factors intervening here; I've found huge variations in sonics with only subtle changes in the VAS operating environment, much more than the output stage. In fact, you can take global feedback from the drivers of a conventional Self-type bipolar amplifier and still have a mighty good sound, a little twee, certainly, but very tube-like. This indicates to me that the VAS is most likely the problem, though Padamiecki suggests it is the input stage. I refute this because the input differential stage is operating pretty much at constant Vce and only slight current variation. However, this is my opinion, and YMMV.
Cheers,
Hugh
AKSA said:
Hugh,
may I tell my point of view.... You wrote, that the "memdist" is not a big problem in the early stages. I think, this is not necessary true. The amount of memdist is not depended directly of the signal level, due the delay of the chip temperature tracking. So You can't tell, that it's -say- 5%. Let's say, that Your input stage produce 1mV distortion. That will be amplified by the VAS, and it could be large. The early stage works with low level signal...
In some of my first amplifiers I used emitter follower between the long tailed pair input stage, and the VAS. I simply connected this emitter follower between the GND, and the negative supply rail. It sounds terrible
, due the very large variation of the dissipation in the emitter follower stage. I tried to use regulated collector voltage for the follower and it was better.I read some good articles in my country about memdist. The final predict was, that not possible to eliminate, but try to minimize! This could be if the stage works with the maximum dissipation without signal, and the signal can reduce it only. This possible if You load Your stages with resistor, and keep the DC voltage across the resistor, and the voltage across the input stage same.
So I set the voltage for my emitter follower 3.3V And there was 1,65V on the transistor, and 1.65V on the emitter resistor (I used local feedback for VAS, which means 1V across the degeneration emitter resistor of it). This gave the best sound🙂
So today I use this solution for the input stage too. I use two more resistor at the collector of the input differential pair, and set the collector voltage half of the positive/negative PSU voltage.
sajti
Please look at the attached schematic. It is rather an attenuator not amp like the simulations showed. R2, R3 is a potentiometer. The Transistors are bc546,556, k246,j103, irf530,9530. What is wrong? Too little power from the jfets?
Ah, I see.
An "ideal voltage source" is an engineering term indicating a source with zero impedance. The perfect battery if you like. The "ideal current source" is one which has infinite impedance. These things don't exist, of course, but are useful for simplifying a circuit to allow a more focussed analysis.
The thing is, at audio frequencies, the so-called VAS is much more like an ideal current source than an ideal voltage source. The argument that the impact of output transistors on "memory" distortion makes the assumption that the VAS is more like an ideal voltage source - hence it argues that the change in transconductance with temperature of the output transistors is an order of magnitude lower than the change in beta of the VAS. But if one treats the VAS as an ideal current source you'll find the beta of the output transistors is the important parameter. The temperature variation of the output transistors is, typically in a class AB, considerably greater than that of the VAS. So pound for pound the output transistors are greater contributors to "memory" distortion.
An "ideal voltage source" is an engineering term indicating a source with zero impedance. The perfect battery if you like. The "ideal current source" is one which has infinite impedance. These things don't exist, of course, but are useful for simplifying a circuit to allow a more focussed analysis.
The thing is, at audio frequencies, the so-called VAS is much more like an ideal current source than an ideal voltage source. The argument that the impact of output transistors on "memory" distortion makes the assumption that the VAS is more like an ideal voltage source - hence it argues that the change in transconductance with temperature of the output transistors is an order of magnitude lower than the change in beta of the VAS. But if one treats the VAS as an ideal current source you'll find the beta of the output transistors is the important parameter. The temperature variation of the output transistors is, typically in a class AB, considerably greater than that of the VAS. So pound for pound the output transistors are greater contributors to "memory" distortion.
A little fast and loose with the math there Hugh,
dB is power ratio so 784 mW / 16 mW is only 17 dB
assuming a rather high 100 degree C / W Rta for the vas device gives:
(300 + 78.4)/(300 + 1.6) = 1.25 absolute T variation or ~ 12% amplitude T variation
simply clamping a vas Q ( in a package with a short thermal path such as TO-126 or 220) to a thick copper heatsink will cut low frequency T modulation an order of magnitude or more when the added thermal capacity is accounted for - the thermal capacity of the metal tab of the package will probably roll off thermal modulation at fairly low audio frequencies by itself
cascode transistor nonideal properties can be reduced 2 orders of magnitude at audio frequencies as was discussed extensively last year in the “hawksford cascode” and “baxendall super pair” threads
i see no reason why the improvement wouldn't apply to thermal modulation as well
Hawksford, Reduction of Transistor Slope Impedance Dependent
Distortion in Large-Signal Amplifiers
Vol. 36, Number 4 pp. 213 (1988)
dB is power ratio so 784 mW / 16 mW is only 17 dB
assuming a rather high 100 degree C / W Rta for the vas device gives:
(300 + 78.4)/(300 + 1.6) = 1.25 absolute T variation or ~ 12% amplitude T variation
simply clamping a vas Q ( in a package with a short thermal path such as TO-126 or 220) to a thick copper heatsink will cut low frequency T modulation an order of magnitude or more when the added thermal capacity is accounted for - the thermal capacity of the metal tab of the package will probably roll off thermal modulation at fairly low audio frequencies by itself
cascode transistor nonideal properties can be reduced 2 orders of magnitude at audio frequencies as was discussed extensively last year in the “hawksford cascode” and “baxendall super pair” threads
i see no reason why the improvement wouldn't apply to thermal modulation as well
Hawksford, Reduction of Transistor Slope Impedance Dependent
Distortion in Large-Signal Amplifiers
Vol. 36, Number 4 pp. 213 (1988)
- Status
- Not open for further replies.