Yes, it is patent #5,783,970
This is a grey-area kind of question
With the same carefull execution (for both CCT's), I 'think' I still prefer the ordinary common emitor for the 2nd stage. The answer heavily depends on the quality of output stage being used. If I'm not mistaken, JCarr somewhere here also wrote from his listening/comparison test, he prefers the conventional 2nd stage (don't know where the thread is). PMA, which is better from your experience (for sonic performance), a floating cascode (Roender's CCT) or a fixed cascode (voltage divider between ground and rail)?
PMA said:
A very good one, yes... I like it more than a DC servo. Should I use an OPA amp instead a good capacitor?
I don't use/like an input cap.
Symasym, a very good amp, especially if you use a hot output stage (500...700mA), but this one is much better.
I must admit that a lot of good stuff I have learned from MikeB and you in symasym project
Unipolar Input/VAS with wide open loop bandwidth
Below is an example of my input/VAS circuit configured and compensated for use in an amplifier with wide open loop bandwidth, for those who prefer such a design.
The design provides 34 dB of NFB up to an open loop bandwidth of just over 20 kHz. As shown, it is configured for a closed loop gain of 26 dB, and it will have a closed-loop bandwidth of about 1 MHz. There are no places in the circuit where open-loop bandwidth is less than 20 kHz, even inside the feedback compensation loop. Note that low frequency open loop gain is set by a local feedback path (R10), rather than loading down the output of the VAS with a resistive load that would increase distortion.
The cascodes that would normally be used with this circuit have been omitted for clarity. Therefore, the circuit as shown with the transistors as shown is not in fact capable of operating with the +/- 64V rails shown.
Cheers,
Bob
Below is an example of my input/VAS circuit configured and compensated for use in an amplifier with wide open loop bandwidth, for those who prefer such a design.
The design provides 34 dB of NFB up to an open loop bandwidth of just over 20 kHz. As shown, it is configured for a closed loop gain of 26 dB, and it will have a closed-loop bandwidth of about 1 MHz. There are no places in the circuit where open-loop bandwidth is less than 20 kHz, even inside the feedback compensation loop. Note that low frequency open loop gain is set by a local feedback path (R10), rather than loading down the output of the VAS with a resistive load that would increase distortion.
The cascodes that would normally be used with this circuit have been omitted for clarity. Therefore, the circuit as shown with the transistors as shown is not in fact capable of operating with the +/- 64V rails shown.
Cheers,
Bob
Attachments
john curl said:
Hi John,
This is a fair question. I think the simple answer is that it would be about the same as for an ordinary current mirror load using the same amount of emitter degeneration.
When I did this design in 1983 I don't recall trying super hard to get the noise down, but if noise injection were a concern, it could be reduced by using larger emitter degeneration resistors and just dropping a little more voltage. All of this scales nicely, so this can work well.
Of course, an S/N of 108 dB, A-weighted with respect to 1 watt into 8 ohms is very good, and is better than most amplifiers that Stereophile has ever reviewed. For example, I recall the Boulder being very good at 105 dB A weighted S/N.
Cheers,
Bob
lumanauw said:Nice design
Mr. Cordell, R10 here, is it has the same fuction with 2x1Mohm in JC3 amp?
What is R3 for? How does it work?
Thanks!
Yes, R10 performs essentially the same function as the pair of 1 M resistors in the JC-3; it sets the low-frequency open loop gain to a fixed and finite value. By doing so, and given the gain crossover frequency of 1 MHz, it also sets the open-loop bandwidth to a value of about 22 kHz.
R3 sets the open-loop gain of the first stage to a fixed and finite value, and establishes an inner-loop open loop bandwidth that is about 80 kHz.
Cheers,
Bob
For the record, fellow engineers: The S/N of the JC-1 amplifier has nothing to do with the input topology. It has to do with making the amplifier able to easily accept either single ended or balanced drive with a minimum of added active components, YET have significant input common mode rejection.
For example, I had to add a 7.15K ohm resistor at the (+) input in order to balance the gain of the inputs of the amp, so that common mode rejection is significant.
Other options would have added more active devices in series with the signal.
I spoke to Parasound yesterday about it. Perhaps, a future upgrade will incorporate a different approach, although no one has complained about it up to this time. You must remember, that we do not usually amplify high sensitivity loudspeakers with a 400W power amp. A small power amp of 50W should be more appropriate for horn speakers, etc.
The problem with active loads will not be very important with power amps, BUT it would be very important for a phono or microphone input stage.
Doesn't anyone know about this? Doesn't SPICE evaluate this.
I know that I saw the SPICE analysis of the 741 back in 1971, and it showed it clearly. This was one of the FIRST times that I saw SPICE show a problem that was ignored by the designer.
For example, I had to add a 7.15K ohm resistor at the (+) input in order to balance the gain of the inputs of the amp, so that common mode rejection is significant.
Other options would have added more active devices in series with the signal.
I spoke to Parasound yesterday about it. Perhaps, a future upgrade will incorporate a different approach, although no one has complained about it up to this time. You must remember, that we do not usually amplify high sensitivity loudspeakers with a 400W power amp. A small power amp of 50W should be more appropriate for horn speakers, etc.
The problem with active loads will not be very important with power amps, BUT it would be very important for a phono or microphone input stage.
Doesn't anyone know about this? Doesn't SPICE evaluate this.
I know that I saw the SPICE analysis of the 741 back in 1971, and it showed it clearly. This was one of the FIRST times that I saw SPICE show a problem that was ignored by the designer.
I don't use spice' noise analysis, usually hand calculation with intelligent approximations/circuit simplifications gives more insight, particularly when input noise specs for the active devices are often given as large ranges, esp for fets
the Vishay 2N5564 noise plot shows 1/f character over the audio range - likely very batch dependent
http://www.vishay.com/docs/70254/70254.pdf
the easiest noise improvement for Bob's front end would be to eliminate (or inductive bypass) the fet degeneration
a use for spice would be the addition of "noise source" sine source at appropriate circuit point to find the output (or input referred) noise sensitivity from that point/device in the circuit
the Vishay 2N5564 noise plot shows 1/f character over the audio range - likely very batch dependent
http://www.vishay.com/docs/70254/70254.pdf
the easiest noise improvement for Bob's front end would be to eliminate (or inductive bypass) the fet degeneration
a use for spice would be the addition of "noise source" sine source at appropriate circuit point to find the output (or input referred) noise sensitivity from that point/device in the circuit
You are missing the point, it is true that inductor bypass would help, BUT the real POTENTIAL problem is the active load, itself. I am trying to teach you something important, that seems to be dismissed on this website. I don't use active loads, so I don't have this potential problem, but many of you do!
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.