Is the CFB topology superior, and why?

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I don't understand why you are taking in consideration the harmonic distortion beeing the most important .Personaly I don't make a difference under 0.06 % .
As I said already there are others to consider ...
Nobody want's to talk about the phase or about the negative feedback speed ?
 
Is the speed of the feedback circuit before the substracting point .To have a good feedback correction this needs to be more faster than the speed of the corrected signal circuit or the speed of the signal circuit .
Can somebody post a open loop gain&phase response for VFB and CFB ?

cfb.png


VFB.png
 
Last edited:
Elvee ,I think that all 3 "preconceptions" are theoretical correct .
Please say where I am wrong in these theories .
Let us try, or at least begin to ...

Fact 1:
Nobody talks about transfer functions; about that we have a first pole at low freq in VFB and the first pole in CFB is more more higher .This will shift the phase for VFB below 20khz with -90 degrees ,but in CFB ,because the pole is at a freq very high ,we have no phase shifting in all audio band !
So even the gain is high at VFB at low freq ,the phase is shifting from 0 to -90 in audio band ;It starts with 180 in DC to very low freq ,then at about 10-200Hz we have 45 degrees(the pole) ,and after we have 90degres at 1-3 khz.
All these in open loop .

So if the designers consider the open loop gain for thd why they don't consider also the open loop phase ?
1/ Absolute phase is irrelevant, it simply amounts to transmission delay.
2/ Variation of phase with frequency is not only innocuous, it is required to minimize group delay (the relevant parameter): if it is not proportional to frequency, it will cause phase distortion.
3/ Anyway, what counts are the residual errors (group delay, THD, etc) once feedback has been applied. Otherwise, comparisons are meaningless: if you have an open-loop group delay of 1ns and a loop gain of 20dB, this will be ten times worst than an OL group delay of 10ns but 60dB loop gain
Why accuphase amps with CFB sounds better ?
I would call that a personal opinion, not exactly a fact

Also phase for negative feedback input in VFB is turned by 2 poles but in CFB is turned only once and this is at high freq.If in VFB we have 2 common emitter stages for the negative feedback ,in CFB we have a common emitter and a common base .

So logical is to choose the circuit with less transformation of the negative feedback signals .Because the non-linear distortion is important .
I fail to see any logic

So the CFB at this hour has high speed ,high bandwith ,excelent phase response .
The VFB has speed but not as much as CFB ,low bandwith ,poor phase response .
That is a purely anecdotal and qualitative appreciation (and not a very clear one)

Fact 2 :
What about the THD ?
Well , we can say that the THD is almost the same with VFB and CFB ,but at low freq .
Here again, what counts is the end result: if an amplifier shows a 20x increase in distortion from 1KHz to 20KHz, it will still be better than a flat one having an higher overall THD
Is more important the THD at high freq than at low freq ?
I am thinking only at the fact that the low frequencies in music is at least 3-5 times higher in amplitude than the high freq and that the SNR is also higher at low F and low at high F.
So to have a good separation between the instruments we need more low thd at high freq compared to the thd at low freq .The small amplitude signals in music are the high freq signals majority !

When we listen music we are searching to have good separation of instruments ,we are searching also to hear the complex pieces clear ,clear means that the high amplitude signals don't distort the small signals .The brain recognize the high amplitude easy but the low amplitudes (which is the high frequencies ) is a little more hard to understand .Also the small amplitude signals(HF) have a smaller SNR .
So then we need more accuracy at high frequency .
Not very clear either, but it goes against all accepted psychoacoustic data and evidences: the ear gradually looses its ability to detect distortions outside the 400 to 4KHz range, is less effective at low levels and is sensitive to masking


Fact 3.

Negative feedback circuit should have a high speed ?
What speed will have the negative feedback signal to compensate the output signal in CFB compared to VFB ?
In order to make meaningful and fair comparisons, post the .asc of your CFB:

cfb.png


I will convert it into its exact VFB-dual image
 
Last edited:
Wow, Catlin, the VFB graph is just wickedly bad compensation overdo, and the CFB is very slightly undercompensated but your RF filter or feedback cap could trim off that slight excess. At least with the CFB there is a tiny excess instead of a severe shortage.

In the case of the CFB, the amp is running faster than the outputs, and the simple fix to align matters is set the amp to match those two curves perfectly, or use a slightly faster output device. With MJL21194, 4mhz, I got Circlophone pushed to 3.5mhz Phase Linear, but any faster amp settings (compensation reductions = faster) made that "exceeding" graph like shown above with more treble in than out. If you don't those two lines on your graph perfectly, treble quality will suffer a bit. The exceeding is a lovely problem to have because the fix is so very simple. Please adjust to fell the amp before exceeding the outputs.

This is an unfair race what with one example having no brakes and the other with the parking brake on. In the case of the VFB, you've got the parking brake on. Some kind of heavy handed compensation overdo mistake on that one. The pre sections are actually trying, but the more it tries to amplify, the worse it gets? I'm not sure what that's for.
 
Wow, Catlin, the VFB graph is just wickedly bad compensation overdo

Can you say what you mean here ? I see that there are 55 degrees reserve .Minimum 45 are ok but not always the phase reserve it says things about the stability .Also it needs study with square wave response .
For example the "helpful " transitional Miller compensation does not add nothing at the phase reserve but will dramaticaly distort the square wave response with capacitive loading (1uF for example)


But here we are just talking about the CFB vs VFB and the pictures are just for the quality comparision .
And I don't understand why the curve with gain must be on the same place like the one with phase .Maybe you want to say another thing .
Thank you .
 
Last edited:
Can you say what you mean here ? I see that there are 55 degrees reserve .Minimum 45 are ok but not always the phase reserve it says things about the stability .Also it needs study with square wave response. For example the "helpful " transitional Miller compensation does not add nothing at the phase reserve but will dramaticaly distort the square wave response with capacitive loading (1uF for example)
Hey, I'm a newbie for this stuff. But I've definitely had problems with "Miller comp versus Gain setting" from playing with the TDA7294. If I increase the gain, I've decreased the need of compensation and then have the ugly situation of having more compensation than needed and the options all sealed shut inside the chip. An additional problem is that the datasheet schematics are counterproductive with every value off the mark running unstable, midfi and hot. Even when revising the obvious errors (for cooler+clearer results), with the wrong compensation still locked in place inside the chip, no matter what I could do with the tone or power filtering, the soundfield (per each speaker even if monophonic) was smushed/shrank. The only solution I could do was adjust both the feedback current and the gain until the amp performed perfectly (which was a nice big room filling surprise, even in monophonic), and then add a preamp. It was a good learning experience about the caveats of Miller comp. Perhaps it accounts for a portion of the the commonplace advice to use low gain and then add a preamp? If the situation was any worse, I'd think they had wanted to make a buffer product.
And I don't understand why the curve with gain must be on the same place like the one with phase. Maybe you want to say another thing. Thank you.
Only this: "Phase Linear"
If you get the two lines on the graph to match/track each other precisely, all the other test also line up too. Sorry that I have forgotten the particulars, but I remember it was because of one of the other tests. I will resume simulator jockey training in the winter and may find out then.
 
Last edited:
And now for some more interesting news on CFB topologies, hope elvee doesnt mind.

For years me and the world have been under the impression that comlinear was the developer of the famous CFB topology that uses the diamond buffer as input. They also were the first to patent it in 1985 if Im correct and hold more than 20 other patents regarding CFBs.

Yesterday I got proof of what I have been was suspecting for a while now, a Japanese company had a amp on sale in 1980 with that topology and quite surprisingly very advanced form of that topology. Take about two years for development and production and Id guess the topology dates from about 1977. This company is Pioneer. One has to wonder where comlinear got the idea whether themselves or ...... In those days circuit topologies were a sacred secret in japan and its very difficult to find schematics of their amps which were sold only in their homeland.

Here is a preview although I hope to get the rest of the schematic into electronic form.
Send me a email and I send you some more info about this pioneer super linear circuit. I also have M-Z1 service manual in PDF.
 
In the end Danielwritesbac you are kidding yourself if you think you can hear it.

In the end all that a cfb op amp is, is just a method of getting more bandwidth from an integrated amplifier with a fixed pole splitting capacitor.

There are many vfb op amps on the market that have plenty enough bandwidth for audio, and do not have the drawbacks of a cfb type, and you really can't hear them.
rcw
 
Hearing the difference between two phase linear amplifiers? No I can't do that. I would think any difference was power circuit.
I'm only asking for one thing--large/unhindered soundstage size in monophonic. That is something I would notice.
If that feature is lacking, small or hindered, then for sure the compensations need a repair.
There are many vfb op amps on the market that have plenty enough bandwidth for audio, and do not have the drawbacks of a cfb type, and you really can't hear them. rcw
Of course either type can be made to perform. One way or the other we're going to get a workout and I suppose that goes differently for each type. But, what were the drawbacks of the CFB types?
 
Last edited:
Any sound stage effect on mono has nothing at all to do with the amplifier.

The disadvantage of a cfb op amp is that its inverting input has a low impedance and is very sensitive to capacitance.

Exactly why you would use an amplifier with these disadvantages to amplify signals when there are perfectly good ones that do not have these I am not sure. You need a gain bandwidth product of 8-10MHz. for audio and the 5534 has been giving this for many years, plus you can set the compensation rather than being stuck with the one the manufacturer chooses.
rcw
 
VFB
ac_VFB.jpg


CFB
ac_cfb.jpg


standard.zip is the model library for the BJT transistors
The duplicate models of 5401 and 5501 in your library seem to be corrupted, and I had to remove them before being able to run the sim.

Anyway, here are the results.

First, your frequency analysis measures the open loop gain under forced VFB conditions (and extreme ones), which is rather ironical from a CFB proponent.
I made the measurement in the less intrusive, normal way. For a VFB amplifier, it would make little difference, but for a CFB it is essential, otherwise it becomes completely meaningless.
I made the loop gain plot of the complete amplifier, as it is the way the amplifier is supposed to be used.
I also included yours, purely open-loop for reference and comparison (red trace), between the output and the (-) input.
It is obvious that the stability is inadequate, to say the least, but that's not our concern here, and we will not attempt to fix it.
We will just compare the CFB circuit with an equivalent VFB one.
To make the conversion to VFB, we need to make the impedance of the (-) node ~=∞, and equalize the gain for a fair comparison.
That is done with a real follower (not a spice version), a common collector stage. Level shifting is included to compensate for the additional Vbe, but it changes nothing dynamically.
Basically, we see that the outlook is the same, the loop gain is slightly lower because the follower is not actually unity gain, but that is not essential.
Here, the VFB circuit has been tailored to be strictly equivalent to its CFB counterpart, and there is no advantage in doing so: it simply demonstrates the very close equivalence (save 1 or 2° of phase margin, which is already negative here anyway).
We could use the VFB stage to gain some advantage: increase the loop gain, like you did in your analysis, increase the impedance, tailor the compensation, etc.
In short, by degenerating VFB, you can always revert to CFB. But if you cleverly use the additional gain, you can increase the impedance levels, increase the loop gain, etc.
The supplement of gain could easily pose stability problems, but clever compensation techniques, like nesting, TMC, etc, can get you the best of both worlds: stability combined with the highest possible loop gain at all frequencies.

Let us now compare the group delays, since it seems to be the strong point of CFB.
These are the next two pics.
Here again, the values (~1.5µs in the audio band) of the model are copied almost unchanged in the VFB version.
This shows that both topologies can be close equivalents if one chooses to make them so, but it is also possible to use their differences to advantage, when needed
 

Attachments

  • CatCfb1.png
    CatCfb1.png
    112.9 KB · Views: 463
  • CatCVfb1.png
    CatCVfb1.png
    95.1 KB · Views: 456
  • Catcfbgd.png
    Catcfbgd.png
    105.4 KB · Views: 205
  • CatCfb1gd.png
    CatCfb1gd.png
    100.4 KB · Views: 112
  • volt_fb_AC.asc
    8.7 KB · Views: 49
"First, your frequency analysis measures the open loop gain under forced VFB conditions (and extreme ones)" Can you explain ?

Elvee please excuse but your explanations are wrong .
Firstly you analyze the OLG with the AC signal between the output and the negative feedback .The reference for the negative input must be the GND beacuse we transform the voltage in current .If the negative reaction is in current then we should add a current generator but spice don't have such thing and then we shoul transform the V in I .

You missplaced wrong the current source for the differentail input in Vfb .This will do other impedances in the nodes and will give small AC modified behaviour which is wrong .

Also I can show you a real amp CFB schematic has linear phase (0 degrees until 1 MHZ) who has also good phase reserve but you can't show me a VFB with the same phase because it doesn't exist .The theory says so !
You need to compensate it !
Always a Vfb will have one pole at very low F ,even there are no discrete capacitors .But the parasitic/internal Miller capacitor will give the look of my OLG VFB picture . If you use the cascode stage like CFB has, than Miller is not matter anymore .
Can you show a vfb compensated with linear phase ?I guess not ..

The last 2 pictures are the same . Group delays will not be the same when you have a compensated VFB .
Thank you
 
Last edited:
...
In the end all that a cfb op amp is, is just a method of getting more bandwidth from an integrated amplifier with a fixed pole splitting capacitor.
...
rcw

rcw,

Just trying to understand the above, are you saying that a Vfb type amp has a fixed pole splitting capacitor and a Cfb does not?

Along the same lines, why can't vfb and cfb have identical circuitry following the input pairs?
I understand the desire for a typical vas stage and it's linearizing feedback capacitor, but I still fail to see a fundamental difference between vfb and cfb if the currents in the input are held below the vfb bias current.
Yes the cfb can supply lots of current to prevent slewing but will one ever enter that mode during audio amplification?

Thanks
-Antonio
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.