I've used this VAS in several production amplifiers around 1995-1997 and thought it would be useful to share this idea and to see if that arrangement have been used before - perhaps somebody would remember. The idea is to connect a common drain into a common-base stage, effectively creating a voltage to current convertor. Advantages are - good speed and linearity, completely symmetrical voltage and current limiting on the output, high OL gain in a 2 stage amplifier. On the first diagram here is a simplified configuration, on the second - a complete and known to work circuit with BJT output. However I've used this VAS with MOSFET output stage as well.
x-pro
P.S. - transistor types on the second diagram I had to choose from available models to get the simulation to work. However BC550/BC556 for small signal, BD139/140 (Philips) in place of MJE15030/31 and 2SC2922/2SA1216 in the output were originally used. P-channel MOSFET could be ZVP3306, ZVP3310 or similar.
x-pro
P.S. - transistor types on the second diagram I had to choose from available models to get the simulation to work. However BC550/BC556 for small signal, BD139/140 (Philips) in place of MJE15030/31 and 2SC2922/2SA1216 in the output were originally used. P-channel MOSFET could be ZVP3306, ZVP3310 or similar.
Attachments
AKSA said:
Pmos gives the VAS an extremely high OL gain - over 120 dB into 10K load without much complications, and very good linearity, as it is essentially a voltage to current convertor working on a small part of the MOSFET transfer curve. As the gain in the first stage is high, the distortion are mainly 2-nd and lower 3-rd harmonics from the MOSFET, and even these are at a very low level.
AKSA said:
It sounds good
. I've used this VAS configuration in 5250 integrated amp and A52/A52SE power amplifier from Creek Audio (thought in A52 I've made the input diff. stage from a matched and cascoded JFET pair and the output was N-ch MOSFETs on both) and another good example of this VAS was A3i amplifier from Cambridge Audio which I've designed about the same time, using BJT output transistors (from Sanken). All these amps had a very good sound (IMHO).Cheers
Alex
AKSA said:
Yes and no
. It looks like an LTP, however the input impedance of the BJT in CB configuration is much lower that the output impedance of MOSFET in CD at about the same current. Because of this imbalance the output impedance of the BJT goes higher, the transfer curve looks like a single-ended FET stage and as the voltages on the FET are not changing much - quite a linear stage too.Cheers
Alex
Greetings Alex,
You are a sleeper - an ex-Creek designer, I had not realised.
No bloody wonder.....
I always try to design the VAS with a naturally falling FR to help with compensation. To this end bootstraps are useful, so is the common emitter configuration. Your use of common base, driven in common source at a very high impedance gate (yielding very high gain for the input stage), would appear to give stratospheric top end, very little phase shift (good), and 120dB OLG into 10K (current dependent of course) is also very good.
What distortion is introduced would be presumably low order, particularly H2/H3 from the mosfet as its current would vary very slightly.
What more can be said? It's very good..... congratulations!
Cheers,
Hugh
You are a sleeper - an ex-Creek designer, I had not realised.
No bloody wonder.....
I always try to design the VAS with a naturally falling FR to help with compensation. To this end bootstraps are useful, so is the common emitter configuration. Your use of common base, driven in common source at a very high impedance gate (yielding very high gain for the input stage), would appear to give stratospheric top end, very little phase shift (good), and 120dB OLG into 10K (current dependent of course) is also very good.
What distortion is introduced would be presumably low order, particularly H2/H3 from the mosfet as its current would vary very slightly.
What more can be said? It's very good..... congratulations!
Cheers,
Hugh
Alex,
looking at your circuit, I guess there are probably a few other ways to compensate this amp - did you consider omitting the 100pf caps and setting the front end response with a larger value of C6? I imagine at high frequencies those 100pf caps (=circa 200pf to gnd) become quite a load.
looking at your circuit, I guess there are probably a few other ways to compensate this amp - did you consider omitting the 100pf caps and setting the front end response with a larger value of C6? I imagine at high frequencies those 100pf caps (=circa 200pf to gnd) become quite a load.
AndrewT said:
C8 is just a decoupling for a zener D3, and 100nF is enough there from my experience.
AndrewT said:
There could be no Miller cap around M1 as it is a common drain stage - i.e. drain is AC grounded and it is non-inverting. I would not call it a source follower as it loads into a very low impedance of the following common base stage and can not "follow" anywhere
. C2 is again just a cap across D2, providing noise filtering and AC decoupling.AndrewT said:
There is (almost) no signal voltage across Q4, however there is across Q5 and C5 somewhat improves the stability.
Cheers,
Alex
Bonsai said:
The need for this caps is proven in practice
. I 've shown a real-world circuit. As I've said earlier the output stage input capacitance is about the same order of magnitude. With 10mA available and 300 pF load it makes 30V/us which is enough. Real value of these caps could vary from 33 to 100pF depending on the transistors used. For BD140/139 in the VAS and driver stage 68 pF each would be enough.Cheers
Alex
Alex
I always liked the way you use protection diodes.
Here, it seems that without D1 (1n4148) p-mosfet would 'go to heaven' at negative clipping or slewing, right?
How about anti saturation diodes of VAS? Such designs do not suffer from phase reversal, but are you sure the ati saturation diodes are not needed?
regards
Adam
I always liked the way you use protection diodes.
Here, it seems that without D1 (1n4148) p-mosfet would 'go to heaven' at negative clipping or slewing, right?
How about anti saturation diodes of VAS? Such designs do not suffer from phase reversal, but are you sure the ati saturation diodes are not needed?
regards
Adam
Here is a simulation result of what this circuit possibly capable of without a (relatively) slow output stage. I've added a class A biased PP output. FFT result for the circuit below is for 20kHz 70V p-p output into 1K load (levels are relative to 0dB at 20kHz). I suspect that in real life the distortion would be much higher but even if the simulation is off by 20 dB it would be a fairly good result . At 1 kHz 70V p-p simulated distortions are -138dB 2nd and -128dB 3rd.
It looks like a good opamp type circuit for a preamp stage. Perhaps I should bulid one just for fun... .
Cheers
Alex
. At 1 kHz 70V p-p simulated distortions are -138dB 2nd and -128dB 3rd. It looks like a good opamp type circuit for a preamp stage. Perhaps I should bulid one just for fun... .
Cheers
Alex
Attachments
Alex,
Impressive....
I've just had a thought concerning this comment:
Zin of the BJT would be close to re, that is around 2.5R, and Zout of the mosfet would be recip gm, 40R for the ZVP1320F which has gm of 25mS.
Alex, since we are opeating at such low resistances where noise is no issue, would it be possible to use a PNP bipolar in place of the fet and simply interpose a 33R resistor between the two emitters?
With the superior gm of the PNP we would have very little error signal generated between its base and emitter, less than the fet, and the 33R resistor would swamp the non-ohmic Zout (again, 2.5R at 10mA) yielding far superior linearity.
While there would be a 70% voltage division with the 33R/82R resistors, the advantages would be similar OLG but superior linearity - and likely with some intrinsic damping making it easier to compensate.
Your thoughts? Nyet?
Hugh
Andrew - I always give praise where it is due! It is a privilege to see and talk with some of the great minds in this forum, though these days I'm not exactly didactic in light of the obsession I see with math and measure.....
Impressive....
I've just had a thought concerning this comment:
Zin of the BJT would be close to re, that is around 2.5R, and Zout of the mosfet would be recip gm, 40R for the ZVP1320F which has gm of 25mS.
Alex, since we are opeating at such low resistances where noise is no issue, would it be possible to use a PNP bipolar in place of the fet and simply interpose a 33R resistor between the two emitters?
With the superior gm of the PNP we would have very little error signal generated between its base and emitter, less than the fet, and the 33R resistor would swamp the non-ohmic Zout (again, 2.5R at 10mA) yielding far superior linearity.
While there would be a 70% voltage division with the 33R/82R resistors, the advantages would be similar OLG but superior linearity - and likely with some intrinsic damping making it easier to compensate.
Your thoughts? Nyet?
Hugh
Andrew - I always give praise where it is due! It is a privilege to see and talk with some of the great minds in this forum, though these days I'm not exactly didactic in light of the obsession I see with math and measure.....
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.