Definitely!!
And I would add to anyone - always measure and investigate your audio chain as a whole - with signal source, preamp, poweramp, all the cables etc. And preferably with your loudspeaker connected. Of course be careful not to damage the speakers. You might be surprised what your system would "generate".
I think you guys are guessing and somewhat off track as to how Charles makes his volume control. There are only so many degrees of freedom with an IC op amp, and Charles and I design discrete circuits mainly, so we can do things somewhat differently.
Charles did use an AD844 in one of his former preamps, but he is now disappointed in their characteristics today, This is where he probably loaded down the comp pin to reduce the gain significantly in order to remove the need for global feedback.
Charles did use an AD844 in one of his former preamps, but he is now disappointed in their characteristics today, This is where he probably loaded down the comp pin to reduce the gain significantly in order to remove the need for global feedback.
Hi,
First, "loading VAS" refer to the misguided "loading the VAS" on existing designs employing looped feedback, not to designing adjustable loads into open loop ciruitry.
But it will not use the reduction in gain to further linearise the circuit either, it literally just throws away gain. This may of course be just what is needed (in a linestage for example)...
Ciao T
PS, I probably prefer to do this using Tubes AND making both the Anode load and the common cathode resistor adjustable, this way we can maximise linearity AND SNR...
First, "loading VAS" refer to the misguided "loading the VAS" on existing designs employing looped feedback, not to designing adjustable loads into open loop ciruitry.
But it will not use the reduction in gain to further linearise the circuit either, it literally just throws away gain. This may of course be just what is needed (in a linestage for example)...
Ciao T
PS, I probably prefer to do this using Tubes AND making both the Anode load and the common cathode resistor adjustable, this way we can maximise linearity AND SNR...
No, John restated his teams position elsewhere yesterday.
The position is clear, simply moving the 90 degree phase transition of the low frequency breakpoint to a higher frequency by lowering Aol is what they are talking about. This has little to do with the closed loop response except to INCREASE the demand on input stage linearity. As I stated two days ago you can manipulate this break point with very weak feedback to the input stage current balance, even make it -90 degrees. Johns Harris op-amp story is the same, resistor on the VAS.
Hi,
Yes, I think I agree with John that it is desirable to have constant feedback factor with frequency well past the audio band. I would actually like INCREASING feedback factor with frequency, but I have yet to come with a scheme that does that without producing an oscillator...
Instead of resistor loading the VAS I vote (just John did many years back and as JH did) for localised feedback looped around the VAS plus degeneration.
This way we improve things across the board, open loop linearity is (excluding output) is improved by the reduction in open loop gain, bandwidth is increased and therefore the unavoidable signal dependent bandwidth modulation happens at higher frequencies and the output impedance of the VAS is reduced significantly.
In fact, if we have a decently linear output stage and a decently linear input stage any looped global feedback becomes optional, rather than mandatory and thermal memory distortion (thermal tails) impact on the signal is lowered...
Sorry, but from where I stand I can see no negative, at least until our VAS distortion, loaded by our output stage (usually class AB) has been made linear enough that it is no longer the dominant source of distortion in the Amp...
Ciao T
Yes, I think I agree with John that it is desirable to have constant feedback factor with frequency well past the audio band. I would actually like INCREASING feedback factor with frequency, but I have yet to come with a scheme that does that without producing an oscillator...
Instead of resistor loading the VAS I vote (just John did many years back and as JH did) for localised feedback looped around the VAS plus degeneration.
This way we improve things across the board, open loop linearity is (excluding output) is improved by the reduction in open loop gain, bandwidth is increased and therefore the unavoidable signal dependent bandwidth modulation happens at higher frequencies and the output impedance of the VAS is reduced significantly.
In fact, if we have a decently linear output stage and a decently linear input stage any looped global feedback becomes optional, rather than mandatory and thermal memory distortion (thermal tails) impact on the signal is lowered...
Sorry, but from where I stand I can see no negative, at least until our VAS distortion, loaded by our output stage (usually class AB) has been made linear enough that it is no longer the dominant source of distortion in the Amp...
Ciao T
This suggests that incorporating the RIAA in the feedback of an opamp would be a good thing. In a device like the AD797, the open loop gain somewhat approximates the RIAA correction needed. That gives more or less constant feedback factor when so-called "active RIAA" equalization is applied.
A few 10s of ppm in the open loop condition is quite easily achievable in a VAS. See Hawksford for example. A triple EF on the other hand, biased correctly for 26mV/Re is typically 100s of ppm and presents little load on the VAS - and Cdom of course provides local feedback which lowers the o/p impedance as frequency increases, and this improves linearity ( maybe an example of increased feedback with frequency . . . ?)
Last edited:
But they're concerned about the evil PIM in an otherwise ideal amplifier (based on Barrie Gilbert's example where input stage transfer function is the only non-ideality), so again we talk about two different things. Barrie unfortunately used essentially a 1969 era op-amp just so the numbers would not be absurdly low.
This with the lowering of gain with a shunt to GND, can be made by making a current output on a open or local loop circuit..then the shunt resistor will define the gain, and thus the voltage swing into at High impedance buffer..only problem is as you reduce gain, you also reduce the possible voltage headroom...
Hi,
Indeed.
Indeed.
See my mid/late 90's "Analogue Addicts" Phono Preamp...
The VinylZone: Analogue Addicts Phono Pre
This was designed and written at a time when it was easy to get high grade parts in the UK, but most pre-amplifiers came sans Phono and there where no affordable high quality phono stages to buy...
It was posted originally in the UK dominated Analog Addicts E-Mail list, as ASCII Schematic, probably in 95 or 96...
Same later versions (like my 2K7 one) have found their way here.
Ciao T
Indeed.
Indeed.
See my mid/late 90's "Analogue Addicts" Phono Preamp...
The VinylZone: Analogue Addicts Phono Pre
This was designed and written at a time when it was easy to get high grade parts in the UK, but most pre-amplifiers came sans Phono and there where no affordable high quality phono stages to buy...
It was posted originally in the UK dominated Analog Addicts E-Mail list, as ASCII Schematic, probably in 95 or 96...
Same later versions (like my 2K7 one) have found their way here.
Ciao T
If you have a current output, you should get lower distortion as you lower the gain...I have made open loop riaa's this way, and they have lower distortion with higher frequency, because of the lower gain in the higher frequencies, however the circuit is not suitable for line level because of an initial high open loop gain of 77 dB,
We speak about VAS load by a resistor, and the VAS is inside feedback loop. If you load the VAS (though it has output from collectors), you decrease voltage gain inside feedback loop, input stage has to drive VAS harder to get the same resulting CL gain and distortion rises.
Hi,
Not that easy apparently. Very few projects I see here on DIYA come even close to that. Even 0815 cascodes for the VAS are rare, Hawekesford cascodes, I cannot remember seeing them at all...
Equally, EF3 Output stages are rarely seen (as are Fet Drivers on Bipolar Outputs, which, using laterals would be a good compromise) in the designs that are hyped here.
So yes, using EF3 or SFEF2 outputs, hawkesford cascoded VAS and many other measures can make an amplifier that is inherently very linear. However they are very uncommon, probably because they are messy to implement.
Most common are non-cascoded VAS stages driving EF2 outputs. In this case the VAS is the major source of distortion (including thermal).
Actually, it provides an example of DECREASING feedback with frequency, in the outer (main) feedback loop.
One could shift the CDOM takeoff to after the output stage, this would go a long way to address the reducing feedback factor issue. And it is probably the almost sole reason why the Nat Semi LM38XX series of Chip Amp's sounds quite good (the other is the copious application of followers to buffer audio nodes, reducing loading effect distortion drastically).
However, CDOM compensation creates some really nasty PSRR related issues, it is really a subideal way of dealing with things, if implemented the common way. Hence probably the reason why the Nat Semi LM38XX gets THAT MUCH better when given a well regulated and very clean power supply...
Ciao T
Not that easy apparently. Very few projects I see here on DIYA come even close to that. Even 0815 cascodes for the VAS are rare, Hawekesford cascodes, I cannot remember seeing them at all...
Equally, EF3 Output stages are rarely seen (as are Fet Drivers on Bipolar Outputs, which, using laterals would be a good compromise) in the designs that are hyped here.
So yes, using EF3 or SFEF2 outputs, hawkesford cascoded VAS and many other measures can make an amplifier that is inherently very linear. However they are very uncommon, probably because they are messy to implement.
Most common are non-cascoded VAS stages driving EF2 outputs. In this case the VAS is the major source of distortion (including thermal).
Actually, it provides an example of DECREASING feedback with frequency, in the outer (main) feedback loop.
One could shift the CDOM takeoff to after the output stage, this would go a long way to address the reducing feedback factor issue. And it is probably the almost sole reason why the Nat Semi LM38XX series of Chip Amp's sounds quite good (the other is the copious application of followers to buffer audio nodes, reducing loading effect distortion drastically).
However, CDOM compensation creates some really nasty PSRR related issues, it is really a subideal way of dealing with things, if implemented the common way. Hence probably the reason why the Nat Semi LM38XX gets THAT MUCH better when given a well regulated and very clean power supply...
Ciao T
Joachim,
Die ausnahme bestaetigt die regel!
(A german saying which implies that presence of exceptions actually reinforces the rule, for if there where no exceptions, there would be no rule)
Ciao T
Die ausnahme bestaetigt die regel!
(A german saying which implies that presence of exceptions actually reinforces the rule, for if there where no exceptions, there would be no rule)
Ciao T
Also, see my PIM paper on my web site (CordellAudio.com - Home) where it is shown that it is modulation of the CLOSED LOOP BANDWIDTH that results in PIM, and, as such, the generation of PIM has little to do with open-loop bandwidth. The math does not lie.
We have been here many times before, but some just do not accept (or understand) this.
Cheers,
Bob
- Status
- Not open for further replies.