Member
Joined 2009
Paid Member
I perceive this as a generalization and I have not the experience to know whether it is correct ? I have been simulating a CFA vs VFA version of the SKA GB150 and found no inherent benefit of one over the other with regards OLG (amplitude and phase) so why would the error signals be different ?
Last edited:
Andrew, I find that a very succint and clear statement. It also points directly to the high slew rate you can get with CFAs.
Thanks,
Jan
Now may i steer you to look at the FET/Bipolar OPS difference in your 'vas isolation test'? More interesting info to be shown/documented.
[Sorry for appearing to be so lazy by asking you/others to do this in your SIM. When I say something from past tests and experience - it isnt accepted as well as your SIM data/graphs being displayed... the black and white data for all to consider. I never kept mine... even though I have MicroCap 10.0. But this forum is dispelling so many audio myths about CFA and similar circuits derived from such that this will help other designers with a place to see the actual results.... in one place.] Thanks goes out to Bonsai for recognizing and opening the subject in its own place.
Thx-RNMarsh
[Sorry for appearing to be so lazy by asking you/others to do this in your SIM. When I say something from past tests and experience - it isnt accepted as well as your SIM data/graphs being displayed... the black and white data for all to consider. I never kept mine... even though I have MicroCap 10.0. But this forum is dispelling so many audio myths about CFA and similar circuits derived from such that this will help other designers with a place to see the actual results.... in one place.] Thanks goes out to Bonsai for recognizing and opening the subject in its own place.
Thx-RNMarsh
Last edited:
This is the only argument that maybe holds water. Almost everything else you said is questionable, if not flat wrong.
- Peak current at the VAS input is irrelevant. If you think about the current available to charge the compensation cap (with the gazillion of V/uS result) this is also irrelevant, if you choose an apple to apple VFA-CFA comparison. "Current on demand" can easily be implemented in a VFA, and probably the simplest example is the Miller Input Inclusive compensation, easily making for hundreds of V/uS slew rates.
- "the loop is faster", I would expect from somebody having an engineering background to avoid using such a fake argument.
- The diamond input buffer is obviously more linear, simply because it's an emitter follower, hence having local feedback. But then you just traded the input stage gain (in a diff input VFA) for linearity (in a diamond buffer CFA), with the usual null game overall result.
Indeed. But, as i did not build power amps with BJTs in OPS any more since decades, i don't have even models for them.
May-be Bonsai could do this or Dadod, when he'll be back ?
Did-you expect BJTs to offer a lower open-loop impedance ?
Forgive me if I misunderstand your point. You appear to be stating that the amplifier eliminates back emf of a speaker.
Attached is a scope capture of the voltage (orange) across a woofer and the voltage across a shunt (blue) measuring the current. Up to about 2A RMS, the back emf is not discernable on the scope. Above 2A RMS, the in-phase current can grow no larger. Increasing voltage only succeeds in creating an ever increasing out-of-phase current.
The back emf is not visible in the voltage delivered to the speaker. Back emf affects the speaker current and the amplifier feedback corrects the voltage across the speaker. The current in the speaker is not corrected by feedback.
I have experimented by placing an op-amp in front of the amplifier and and feeding back the voltage across the shunt. The configuration eliminates the current distortion and transfers the distortion to the voltage waveform. The lash-up was finicky since I had no DC offset correction etc. It was a one-afternoon endeavor.
I performed a similar experiment with a Keithley 2651A and created a current pulse into this woofer. The current pulse was beautiful and the voltage was all over the place as the feedback loop of the sourcemeter forced the current to be as close to perfect as possible.
Back emf will cause either current or voltage distortion depending on whether the amplifier is a current source or a voltage source. As far as I know there are no production model current source amplifiers other than some Pass Labs low-power class A units.
Member
Joined 2009
Paid Member
Esperado,
I was just reading your website article referenced above. It mirrors much of what I have been looking at with the GB150 so it was most interesting to read. I agree with the option of using low impedance feedback network with VFA if it is symmetrical - I have found some stability benefits from this in the GB150 - wider phase margin.
However, you say
"As expected, the CFA is here 5 time faster than the VFA. And this is not due to the miller cap, but the topology: the way the feedback signals are substrated from the input ones.
This is due to the two input stage topologies transconductances: The Long-tail Pair is "compressive" while the CFA is "expansive" (© Richeard Marsh)."
With 100dB of OLG you are operating on the linear part of the error amplifier I-V curve, the error signal is tiny and there isn't going to be any difference between them in terms of their 'compressive' nature or otherwise. Perhaps there is a difference with TIM - which can be avoided using what has already been learned.
Also, I don't understand your description of the CFA being 5 times 'faster' due to the way signals are subtracted ?????? do you mean the low impedance feedback network working with low input capacitance has high bandwidth ? - which we can achieve independent of CFA or VFA ?
I was just reading your website article referenced above. It mirrors much of what I have been looking at with the GB150 so it was most interesting to read. I agree with the option of using low impedance feedback network with VFA if it is symmetrical - I have found some stability benefits from this in the GB150 - wider phase margin.
However, you say
"As expected, the CFA is here 5 time faster than the VFA. And this is not due to the miller cap, but the topology: the way the feedback signals are substrated from the input ones.
This is due to the two input stage topologies transconductances: The Long-tail Pair is "compressive" while the CFA is "expansive" (© Richeard Marsh)."
With 100dB of OLG you are operating on the linear part of the error amplifier I-V curve, the error signal is tiny and there isn't going to be any difference between them in terms of their 'compressive' nature or otherwise. Perhaps there is a difference with TIM - which can be avoided using what has already been learned.
Also, I don't understand your description of the CFA being 5 times 'faster' due to the way signals are subtracted ?????? do you mean the low impedance feedback network working with low input capacitance has high bandwidth ? - which we can achieve independent of CFA or VFA ?
Last edited:
Of course. My sims were only about voltages. The energy from the speaker has to be dissipated somewhere ;-)
My point was just to show how emf/back emf affect the internal signal of a NFB amplifier. And why i find interesting to look at VAS input signals.
At the price of very low feedback impedance (lot of heat) in VFA, because base capacitive impedance is higher than emitter one ?
At the price of offset problems if the input impedances on the two sides of the FTP are so different ?
'Current on demand' is what i had in mind.
Even with a lower feedback impedance, and even with a larger closed bandwidth in VFA than in CFA, hard to achieve so high slew rates in VFAs as well as the same phase margins ;-)
Bigun, my purpose was not to bring *an other* article analyzing deeply the theories of the two topologies. So many good papers available, my English is too poor, and i'm not so good in this. Just to figure-out what was the reality of legends about "more distortion", "more noise" etc. in evils CFAs, with a practical example.
My purpose was not either to pretend VFAs are evils, so many damn good VFAs amps available.
How do-you want me to do them ? 19 kHz and 20 kHz same level ?
Last edited:
The relation between an amplifier and its loading loudspeaker (sorry, it this is in french)
Calcul des filtres par les polynômes
Abstract :
1 ) L'ampli est une source de tension parfaite proportionnelle à la tension d'entrée en série avec son impédance interne.
2 ) La BM, dès qu'elle bouge est aussi une source de tension parfaite, proportionnelle à sa vitesse et en série avec son impédance interne (dont une résistance importante). Tu l'appelles Fem ou Fcem, comme tu veux.
Quand tu les relies entre eux, directement ou via des filtres, tu constitues simplement un circuit RLC (plusieurs) attaqué par deux sources. Leurs mouvements sont régis par la loi d'Ohm (généralisé). Point barre.
Il n'y a pas à se soucier de savoir qui est générateur ou récepteur, ces notions ne jouent que sur le signe des variables.
A good way to analyze the effects on the behaviour of an NFB amp of the voltage generated by its loading driver (by back EMF and microphonicity) is to short the input and to load the amp output either by an other amp through an RLC network or by a driver facing a second identical driver (à la Isobarik for tight coupling) excited by another amp with all kinds of signals.
What will be shown ? Nothing special else than what the fedback amps theory predicts and nothing in relation of the CFA or VFA aspects of the design on test.
A low impedance output is required from the amp to damp the main resonance of a driver but it does nothing to tame the other resonances, cone break-up and microphonicity, all effects which modulate the current (i) through the voice coil and then the force (F = B.l.i) applied to the cone.
For more details, I suggest the reading of this book :
http://www.diyaudio.com/forums/vend...ook-current-drive-loudspeakers-published.html
A current output amp based on a TDA2040 with regulated supplies is described as well as simple PCBs to implement it.
Calcul des filtres par les polynômes
Abstract :
1 ) L'ampli est une source de tension parfaite proportionnelle à la tension d'entrée en série avec son impédance interne.
2 ) La BM, dès qu'elle bouge est aussi une source de tension parfaite, proportionnelle à sa vitesse et en série avec son impédance interne (dont une résistance importante). Tu l'appelles Fem ou Fcem, comme tu veux.
Quand tu les relies entre eux, directement ou via des filtres, tu constitues simplement un circuit RLC (plusieurs) attaqué par deux sources. Leurs mouvements sont régis par la loi d'Ohm (généralisé). Point barre.
Il n'y a pas à se soucier de savoir qui est générateur ou récepteur, ces notions ne jouent que sur le signe des variables.
A good way to analyze the effects on the behaviour of an NFB amp of the voltage generated by its loading driver (by back EMF and microphonicity) is to short the input and to load the amp output either by an other amp through an RLC network or by a driver facing a second identical driver (à la Isobarik for tight coupling) excited by another amp with all kinds of signals.
What will be shown ? Nothing special else than what the fedback amps theory predicts and nothing in relation of the CFA or VFA aspects of the design on test.
A low impedance output is required from the amp to damp the main resonance of a driver but it does nothing to tame the other resonances, cone break-up and microphonicity, all effects which modulate the current (i) through the voice coil and then the force (F = B.l.i) applied to the cone.
For more details, I suggest the reading of this book :
http://www.diyaudio.com/forums/vend...ook-current-drive-loudspeakers-published.html
A current output amp based on a TDA2040 with regulated supplies is described as well as simple PCBs to implement it.
Waly, I'm just enumerating myths that are now busted. I did not really intend to compare CFA/VFA .. only to point out some of the common misconceptions about the difference.
But for the two you seem to have expert knowledge about ..
Noise of simple CFAs better than VFAs
If anyone (especially Waly) has a VFA design you think particularly LN and can post a *.ASC, I will endeavour to better its noise performance by at least 3dB and better or retain any performance characteristics you may like to enumerate ... and probably simplify it in the process.
Not too complex please or my small brain won't cope. Alas, this will be SPICE world only. I'm a complete novice at SPICE noise so please be patient .. but I do have some small 'real world' experience with LN design from Rs = 2R to 2G which I trust will cover your 'real world' applications.
simple CFAs better than VFAs in all da usual performance parameters
I'll issue a similar challenge with a small proviso that the VFA be limited to 15 active devices or less. This is simply cos I think I can do 1ppm THD20k with my #499 VFA but am still working towards this with simple CFAs.
Wahab, 10 devices on the IPS is rather more than I'm comfortable with so outside my Challenge #2. But if you'll post a *.ASC of your circuit, I shall attempt what you ask.wahab said:
It probably won't be simple .. just simpler than your original.
May I ask you to do the SPICE IMD as I'm a SPICE newbie? I take it you mean 1ppm THD20k. You may like to enumerate the other performance parameters you'd like to retain. Thanks for conceding PSRR. 🙂
_______________
And I beg yus gurus forgiveness for the use of evil electrolytics in the above 😱 I shall use only those hand carved from solid Unobtainium and filled with Taipan oil squeezed by virgins from Marshy.
_______________
PS Waly, I've already conceded on several occasions that this thread is really about 'alternative i/p stages to LTP that might be loosely called CFAs'.
Waly
The input current to the TAS/TIS in VFA is set by the LTP current.
In a CFA the peak current is set by the feedback resistor and the peak output voltage.
Whether you use miller, MIC or anything else on a VFA, you are still not going to be as fast as a CFA.
Further, since a buffer is used in a CFA (and yes, I know its unity gain) you have at least 1 less pole to deal with in THD front end.
Hence my statement that THD loop is faster, and the error output voltage of the front end buffer thus lower in a CFA.
The input current to the TAS/TIS in VFA is set by the LTP current.
In a CFA the peak current is set by the feedback resistor and the peak output voltage.
Whether you use miller, MIC or anything else on a VFA, you are still not going to be as fast as a CFA.
Further, since a buffer is used in a CFA (and yes, I know its unity gain) you have at least 1 less pole to deal with in THD front end.
Hence my statement that THD loop is faster, and the error output voltage of the front end buffer thus lower in a CFA.
Well, may-be for some special use, like intrumentation. But for music, indeed.
By the way, some seems to have difficulties to understand things about speakers, or have a lack of basic knowledge about them, or consider their amps as a tool designed to reproduce continuous sinus waves.
May i make remember some points ?
- Music is made of transients.
- See: Impulse & Step Responses in http://guillaume.perrin74.free.fr/ChalmersMT2012/Papers/MeasuringLoudspeakers_JohnAtkinson.pdf
This is what i find difficult to model and have an effect in the closed loop because everything different from the original signal you can see in acoustic step measurements is a coil movement and will produce a current, affecting the signal in the closed loop as previously shown.
Basic for most of us, i presume.
Last edited:
And yet an amplifier with 1 ppm THD + IMD and 10V/us SR
is undoubtly better than one with 100ppm THD and 1000V/us SR.
Also on the subject of noise , even if you can have good linearity
be carefull that thoses great numbers are not completely ruined
by averagely rejected PSU ripples and other potential perturbations...
The subject of this topic is CFA topology Audio amplifier.
Could-we see your 1 ppm THD + IMD and 10V/us SR CFA ?
Please, stick to technical arguments.
http://www.diyaudio.com/forums/solid-state/240712-cfa-topology-audio-amplifiers-90.html#post3680416
Loudspeaker load effects on amps with feedback belong to a theory which is common to VFA and CFA.
So just a few words because it's out of topic
If measurements with transient signals on loudspeakers were modified by the usual voltage amps driving them, they would be a new tool to evaluate amps performances. They are not.
The aim of voltage feedback is to stabilize the instantaneous voltage gain (a simple muliplication operation) of the amp whatever is the load. In this aim, each amp stage has to provide enough voltage and current with good linearity but this does imply that internal waveforms are to be preserved, that's what they become at the output which is fundamental.
Current driving amps I mentionned are often stated as beneficial for the general linearity of the drivers. Why is rarely explained. The cause is nevertheless apparent with a simple investigation of the basic analysis and a look of what is known as the (inherent) instability of loudspeakers. It's an aspect which could not be disregarded if wanting to discuss loudspeaker time response (in another thread, please).
Last edited: