Question(s) about CFP.

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Why don't consider a CFP like an emitter follower were the first transistor set up the voltage and the second supply the current needed by the load. In this case the "timing" is correct assuming that the second transitor suck_out the current need by the base of the second BJT to the collector of the first one. Is it clear ? I need help to convert Microcap schematics to jpg or gif files.
 
With a schematic, it will be clearer.
What i would say, by "timing" is that for me it doesn't look like feedback, but maybe i am wrong ?
 

Attachments

  • cfp.gif
    cfp.gif
    3.1 KB · Views: 1,309
Thanks PMA, for schematics coversion.

The question about CFP is that this configuration is very linear and except "degeneration" in R1, i don't understand why this is called Complementary FEEDBACK Pair ? But maybe i am wrong and i would like more explanations. Even in the case of it will be true, i don' think that this "feedback" pair has the drawbacks associated with feedback, to the fidelity of the reproduced sound. Has anyone tested (listen), this configuration (input, VAS,output)?
 
Justcallmedad said:
Thanks PMA, for schematics coversion.

The question about CFP is that this configuration is very linear and except "degeneration" in R1, i don't understand why this is called Complementary FEEDBACK Pair ? But maybe i am wrong and i would like more explanations. Even in the case of it will be true, i don' think that this "feedback" pair has the drawbacks associated with feedback, to the fidelity of the reproduced sound. Has anyone tested (listen), this configuration (input, VAS,output)?

D Selfs books cover using the configuration in the input and output.

The two transistors are in a very tight feedback loop, the
"amplified" output of Q1 across R1 drives Q2s base, Q2s
output being the reference for Q1 input, so 100% feedback.

This sort of feedback is known as "local" feedback
and should not be confused with feedback loops.

R1 does not "degenerate" Q1 or Q2.
R1s current is only approximately constant
in that its wired across Q2s base emitter.

:) sreten.
 
These are my thoughts :

CFP behaves in different ways depending on how it's driven :

- If it's driven from a current source or voltage source connected between Q1:B and Q1:E then it works as an open loop circuit

- If it's driven from a current source connected between Q1:B and the other side of Rload then it works also as an open loop circuit

- If it's driven from a voltage source connected between Q1:B and the other side of Rload then it works as a closed loop circuit with current feedback

The lower the drive impedance the more feedback it has, and when driven from high impedances [like the output of a VAS at audio frequencies] it behaves just like an obfuscated darlington but with much worse turn-off characteristics than an actual darlington [bigger current tail and phase shift at high frequencies so it's prone to blown devices and parasitistic oscillation]
 
Muchas gracias Eva por las esplicaciones.

I Will continue in English for others people interested in this forum .

Could you tell me more :

- About the turn-off time, why ?, and is it really a problem for audio amplification ? about wich magnitude vs darlington.
- Even in the case when driven by a low impedance source the collector of Q2 act immediately before the base of Q2 (for me feedback is a problem at the moment where you have a delay, when you "compare" input signal and output (no ?).
- I would like more explanations about phase shift and parasitic oscillations.

I am really interested in working with this circuit, and i didn't find a lot of literature on it, even D. Self (like suggested by sreten)don't explain very much in detail this configuration.

For the moment in simulation it works nice, but real world is "crual" especially in audio ! I would learn more on it, before design the complete circuit, so if people has worked on it, tested, listen, they are welcome.
 
Bipolar transistors are not ideal switching devices. If you abruptly increase Ib, there will be some delay [100s of nanoseconds] before Ic starts to increase and it will take up to some microseconds to reach its steady state value

Also, when there is an Ic and an Ib flowing in steady state, if Ib is abruptly removed, there will be a delay in the order of 1-10uS before Ic starts to decrease and it will take some more microseconds to fall completely

These timings are orientative numbers for power bipolar transistors

In a CFP, the 'current boost' transistor suffers from all those delays and thus the current it provides may be more than 180º out of phase from the driving signal at high frequencies. This causes cross-conduction and unstability problems

Turn-off times and delays may be dramatically reduced by providing a negative base current to the power device at turn-off [for switching circuits optimum value is Ic/2], but the CFP circuit is very poor on that aspect since R1 usually provides a very small current

To see this obscure behavior at high frequencies, just get an oscilloscope, build some CFPs with standard transistors and test them at 100Khz or even at 10Khz of square wave. Simulation software is not very precise simulating bipolar transistors behavior at high frequencies since charge storage and removal phenomena [the thing that causes those delays] are hard to model and depend on the construction and the structure of the die of each particular device
 
Eva said:
Turn-off times and delays may be dramatically reduced by providing a negative base current to the power device at turn-off [for switching circuits optimum value is Ic/2], but the CFP circuit is very poor on that aspect since R1 usually provides a very small current.


This is really very interesting ! You seem to know a lot on transistors switching !
Could you explain further this part ? and particularly the value "Ic/2", I suppose for Q1 ? not yet totally clear for me.
In simulation, this point is not very accurate.

In my design (simulation), i voluntarily "work" both BJT'S at the same Ic current, or near the same (sometimes 2 or 3 times more for Q2), i don't want to "stress" the first transistor.

Here in France is the "Music day" it's means that today a lot of people (amateur, professionals) plays music in the streets and in all places, very sympathetic, maybe it's the same in Spain, well it's the same ALL the days in Spain, always FIESTA !

Well tonight or tomorrow, i will build and test this small circuit, to see his real possibilities and "tune" it.
 
"pair has the drawbacks associated with feedback, to the fidelity of the reproduced sound."

Probably an over generalization. I've got an amp using an output section configured that way. (A Slone design - which uses global feedback as well.) I find no fault in the sound. However, I found it troublesome to construct without getting oscillations. The best solution was to take the zobel and the output inductor off the PCB and locate them at the binding posts.

I think one should be cautious with over-generalizing the faults or benefits of particular competeing topologies. It is better to recognize that each has different problems so solve. EF output sections have difficulties as well. Just because they have been poorly implemented in the past does not mean they should be dismissed out of hand.

This applies to the whole "feedback thing" as well. Just because there were poorly implemented designs using lots of negative feedback in the past does not mean the concept is bogus. In fact, o9ne of the worst sounding but expensive "high-end" amps I ever listened to in a shop had copius literature explain that its sound was the result of the absense of negative feedback. Probably true but not in the way it was intended! I'm sure that with good implementaion, however, the concept could have resulted in a better sounding unit.
 
No my idea is to run the first BJT at constant current and to "feed" the second BJT in current mode, this pair can act as a super transistor. The "Beta" of the pair is high, linearity is very good and like i said, the stress in current changes, sees by the first BJT is is quite negligible, Q2 only needs 1/(Beta Q2) the current from Q1, from where the changes in "caracteristics" are the most determinant for audio quality, (what would you do with a function that changes with the variable (variable is the term in French, i am not sure it's the same in English, well if you have f(x)=Ax, x is the variable.) and like i said Q1/Q2 are in the same range of iddle current. So i think that it can be a good circuit for audio applications.

What would you mean by : Bypass cap placement is tricky.
 
sam9 said:
I've got an amp using an output section configured that way. (A Slone design - which uses global feedback as well.) I find no fault in the sound. However, I found it troublesome to construct without getting oscillations. The best solution was to take the zobel and the output inductor off the PCB and locate them at the binding posts.

Have you a schematic of the Slone design ? or an idea of it. Can you attach it on next post. It can help me to start a first design.

I am not very prone to use Zobel networks, i think they "hide" the real problem to the bench test (scope, distortion meter, fourier etc...)

About feedback, for the moment is just that all amps or preamps that i have listen and i like. - (I like them because they "live", well infact i can feel musicians behind the speakers, not only good sound) - are non feedback designs.
 
Agree with you sam9, but for the design phase, i would'like the output to be as close as possible to a voltage source (like all of us) maybe a lot of output devices... or maybe... the opposite side, drive speakers by current source ? Test the CFP at the output stage, too... But i am not yet at this point. I will begin, by input stage and VAS, and test sound with headphones, after bench test. So, i think i will need more help from experienced designers during real tests.
Probably you are right, it wiill be more safe, with inductor but i hope, without zobel network.
For the global design i would like a very good (truth) sounding amplifier without feedback, and maybe after i will apply some feedback, of course i will need to modify the input stage gain, to correctly apply feedback, and probably not from the output devices (back emf, with not perfect voltage source).
 
Tricky means......

The silly things are prone to oscillation problems. The solution depends on how you are using them.

For a power amp..........

I used them for years, and I thought that they sounded good (after they were stabilised), but then I tried a diamond buffer. Everyone who heard that version.......including all the customers who forked over $$$ to have their amps retrofitted....... agreed.

In power supplies........

They seem better suited there. Current demand is usually constant, you can put caps everywhere to stop them from singing, and the output impedance is flat up to around the alpha cutoff frequency...usually around 100 kHz or so.

Jocko
 
Re: Tricky means......

Jocko Homo said:
The silly things are prone to oscillation problems. The solution depends on how you are using them.

For a power amp..........

I used them for years, and I thought that they sounded good (after they were stabilised), but then I tried a diamond buffer. Everyone who heard that version.......including all the customers who forked over $$$ to have their amps retrofitted....... agreed.

In power supplies........

They seem better suited there. Current demand is usually constant, you can put caps everywhere to stop them from singing, and the output impedance is flat up to around the alpha cutoff frequency...usually around 100 kHz or so.

Jocko

Great hint Jocko!
:)
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.