Not shipped to the audio industry, i'll be bound?
The list? Shipped to any industry.
Michael
How much open loop gain at say 100Hz and 20Khz would satisfy your typical needs?
Thanks
-Antonio
There is no basis for this observation in fact.
Note that pole splitting does not occur with shunt compensation used in so-called "CFAs". Therefore, well designed VFAs with Miller compensation will not only have their available loop gain enhanced by pole splitting, but will also possess the advantage that their loop gain is not needlessly reduced by the feedback components reducing their forward path gain.
Moreover, loop gain enhancing double pole compensation is not an option with so-called "CFAs".
To reiterate, a well designed VFA will possess greater useable loop gain than a so-called "CFA" for comparable stability margins.
Last edited:
Antonio, you can read the pdf on this link about linearity on the 2 types, page 1173.
Michaelkiwanuka is only comparing a single transistor cfa but have a look at what happens when you use the typical diamond circuit common with modern CFA. The diamond circuit is superior to LTP. VFA are only more linear because they have higher loop gain but quite high loop gains can be had with CFAs too.
See this link, it seems to me you are way off, page 1174.
http://bgaudioclub.org/gallery/albums/userpics/12018/SmithCFBamp.pdf
Actually that link, which I read a long time ago, discusses a LTP without a current mirror which is required for balance and for doubling of its transadmittance gain.
A poorly balanced and inadequetely degenerated LTP can potentially generate as much distortion as a complimentary common emitter stage used in so-called "CFAs"
And, no, I wasn't "...only comparing a single transistor cfa..."
VFAs are more linear not only because they possess more useable loop gain, but because they have, potentially, a more linear forward path, if well designed, due to degeneration in the first stage and minor loop feedback about the second stage.
Note that although push-pull action in a LTP augments odd order distortion therein, this is reduced to insignificance by degeneration. Moreover, as I have stated above, a properly balanced LTP generates far less even order distortion than the input stage of a so-called "CFA".
Last edited:
Antonio asked which of the two was more linear, that is the answer. There was no mention of current mirrors etc but then again current mirrors wont make much difference as the two circuits were simulated with perfect parts so there is no inbalance and the gm for two is exactly equal. This means that the diamond circuit is more linear, its as simple as that.
Ok, let me say the diamond circuit in isolation is more linear. Is this correct ??
If you evidence on the contrary please provide some.
A simple simulation will do.
The complimenatary common emitter input stage (that you call diamond) is less linear than a LTP. I found this out some 15 years ago to my satisfaction; I will leave the simulation to you as I haven't much time.
In addition, the second level shifting stage of a so-called "CFA" is more non-linear than the second stage of a VFA because it necesarilly uses shunt compensation and not the linearising minor loop feedback of a VFA.
In addition, the second level shifting stage of a so-called "CFA" is more non-linear than the second stage of a VFA because it necesarilly uses shunt compensation and not the linearising minor loop feedback of a VFA.
Last edited:
I am sorry, Claude, but Control Theory was one of my majors, and I am completely satisfied that the error signal driving the amplifier's forward path in a so-called "CFA" is Vin-Vfeedback because the quantity being sampled by the feedback network is the ouput voltage and NOT the output current. This is elementary.
I'e already covered the sampled output quantity. That quantity being I or V, or something like temp, strain, etc., is not under discussion. A VFA op amp can operate in a network where the output current is sampled and fed back. To do so requires either a Hall sensor or a low value resistor to translate I to V. Then the VFA processes this feedback signal in V form. The nature of the output quantity determines whether the network controls V or I. But the op amp servo or error signal is always V for VFA, and I for a CFA.
You claim the error signal in a CFA is really Vin - Vfb? Please show us using the schematic I posted how that is so. You keep asserting the same, but do not prove anything. You "majored" in control theory? What level of education do you have. I have been developing analog electronics for 35 years, went back to school and am now in the final stage of Ph.D. Controls was my specialty from BE through Ph.D. Anybody in the know understands that when the output of a CFA deviates from the input times (1 + R2/R1), that the current in the - node changes from near zero in response. Maybe I need to sketch an example to illustrate.
Of course there are 2 EF at work here. THe input EF buffers the input presenting a low Z node at the - input. THe output EF buffers the output. So now we have a fb resistor across the outputs of 2 EF stages. THe - input also has a resistor R1 to ground, the EDF sources current into this R1. As long as the currents in R2 and R2 are equal, the - input has near zero current. But the Vbe's on the EF devices is non-zero as they are dc biased.
When the output changes, or the input, so that the 1+R2/R1 gain relation between in and out is perturbed, how does the servo loop correct? The signal that changes is In, the current at the - input. KCL (Kirchoff current law) states that the - input current is the difference in R2 and R1 currents. Refer to my schematic. A change in - input current is mirrored over to the cap which sources/sinks current changing its voltage, then output voltage tracks this change via EF buffer.
Of course, a change in - input current, In, is accompanied by a change in Vin - Vfb. You seem to suggest that In is merely an artifact of Vin - Vfb, which it is not. A change in Vn, the - input voltage takes place as a result of currents in R2 & R1 becoming unbalanced. Also, your theory does not account for why the servo correction is faster w/ lower Rfb, thus higher Ifb. If the speed of the loop is controlled by Vin-Vfb, subject to lower Rfb, then it is current feedback.
Vin - Vfb alone does not account for the fundamental difference between VFA & CFA. Lowering Rfb results in higher Ifb and faster servo speed. You suggest that the gain of the CE input is being raised, but this 1st stage feeds a current mirror which conveys said stage current, not voltage.
Your explanation is very lacking.
Claude
Claude Abraham,
you talk here about a schematic.
I have searched this thread for such a posted schematic.
You know what schematic you refer to.
Please link or repost that referenced circuit.
Thanks.
Claude Abraham,
I judge your point of view more credible than our MichaelKiwanuka's view.
Because you seem to be able to back up your view with some facts, numbers.
A schematic, a picture, would make us all a little bit wiser, I think.
No.
The error signal is the difference between the input signal (voltage) and the fraction of the output voltage that is applied to the invering input. This description does not require further elucidation by means of your schematic; you'd know this if you had a visceral understanding of electronics and feedback control theory.
Yes. Among other things.
You have my sympathy. Good luck with that.
Kirchoff's current law is moot; The current in those resistors has nothing to do with how the circuit functions.
It's straightfoward: you sample a voltage with a voltage divider and apply a fraction of that voltage to inverting input. The quantity of interest, therefore, is a voltage NOT a current. Elementary circuit theory.
As I noted previously, if the inverting input were a virtual ground, you wouldn't be able to vary the amplifier's closed loop gain by changing the value of the grounded feedback resistor because the voltage across it would be zero. So this "current feedback" stuff is just rubbish.
Last edited:
Nothings free,
It would seem reasonable to have as much loop gain around the most nonlinear stage(s). This would be the essence of the design process, some points being made are these configurations have different starting points in terms of linearity and available gain.
Input degeneration and Miller compensation both increase local linearity at some expense or trade off.
Just as for a vfa there are techniques to increase the open loop gain for a cfa type input.
Thanks
-Antonio
I'd give up now.
Find your basic text where the two npn "CFA" is broken down into a forward gain and feedback with H parameters. As I have already said with this you can use ONLY the current into the - input and the current out of the feedback as forced equal and you get the SAME answer. The alternative analysis is to me counter-intuitive, as obviously the voltage across the input is a stand-in for the current because it has nowhere else to go.
An interesting (and correct) observation. I went back to my little test sim with the AD844 CFA, using a feedback network of 900 and 100 ohms.
Input current into the inverting input 470nA, gain = 9.97. Then I put a 1k resistor in series with the inverting input. Gain now 9.92.
However, the BW dropped from 25MHz to 2.5MHz...
(An NE5534 with the same feedback network has a gain of 9.97 as well, but of course much less current into the inverting input).
jan
Last edited:
I would be most interested to know these techniques in respect of the so-called "CFA".
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.