Simple Symetrical Amplifier

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
The only thing VFB is better are DC conditions, but that we do not hear through our speakers. You should make one medium complex SSA and you'll get an impression of what we talk about here. LTP's are sounding dull compared to CFB.
Oh, thanks L.C. i was afraid to have been mistaken during 40 years of my life.

Wahab, please, play with my simulations here, Crescendo revisited

You will see the differences between CFB and VFB *on the same amp*.
VFB ---> NFB
Bandwich: 200Khz ---> 3Mhz
Slew rate: 220v/µs ---> 1200v/µs
hd: 0.0023% ---> 0. 0002%
IM: x ---> x /10

As you can see, and as said L.C, improvement is in all the domains (except DC stability). The major improvement is listening pleasure.
 
Last edited:
As a 67 years old electronic and sound engineer, which had dedicated near all his life in music reproduction/creation i would like to add the followings:

This SSA (or CSA) amp, proposed by Lazy Cat is the best DIY amp i ever seen during all my life, and at least equal to the best produced amps, all categories.

Lazy Cat do not play the "Guru" like some on this forum, do not takes poses, do not play with poetic words to promote itself in the light.
He just had designed a fantastic Amp in a wonderful manner.

First there is nothing open to criticism in the schematic. Simple, accurate, using the most clever technology at each stage *with no compromise*, and no 'religion', or 'esoterism'.

Including the so rare but so good sounding Current feedback topology.

We can feel that each part of his amp has been considered during long hours of work with attention, modesty, deep and objective reflexion between 'pro' and 'cons'. Both as individual stages and part in the whole. It is the very first time, looking at a schematic that i had nothing i would try to change ( Apart the protection stages ;). It is just 'Harmony'.

Printed circuit design is just perfect. Care had been taken everywhere to shorten the path lengths and reduce parasitic caps and inductances: A Master piece that can be taken as a reference.
With something more, the beauty of the design, beauty of the realization, as a cherry on the cake.
And all that with always this simple goal in mind: To reproduce music in the most transparent way.

Well, there is some well known names of electronic designers in the hifi hall of fame . Some are present on this forum. Some are very passioated people, innovative and honest, (Nelson Pass is one of them) others are just fake VIPs, experts in the art of 'make believe'.

On my admiration scale, Lazy Cat is at the top. Because he offer us both his great and *deep* knowledge and somethind more, witch is in the domain of the art, beauty and love, with a so modest attitude i can only admire... and love.

At a very disappointing moment of our history, where all the objects proposed by the industry smeel bad, because there are only marketing pieces of junk, with no love inside, neither for the product, neither for the consumer, this work is more than a gift.

Looking at this work, it was one of the rare times in my life i had this kind of emotion you can feel listening to a master piece record. Music is beautiful (Schematic), interpretation full of sensibility and virtuosity (Design) , production just perfect (Assembly).

And this is a single man work ? Wow !
 
Last edited:
AX tech editor
Joined 2002
Paid Member
Hi , Edmond

I was talking effectively of the impossibility for a GNFB loop
comprising several stages to reduce distorsion of a given stage
using available loop gain provided by the following stages.

To simplify , the IPS gain used for GNFB can reduce distorsion of the VAS
as well as the one produced by the OS , while eventual loop gain quantity
that originate from the VAS can reduce distorsion for the VAS and the OS.

If the OS has also some gain that is used for GNFB , this loop gain
quantity could only reduce the distorsion of the OS.

Wahab how can you say that in earnest? Don't you see that the distortion of whatever stage is never changed by feedback? You can only see that the distortion of the whole chain enclosed by the feedback is lowered.
The distortion of the stages within the feedback loop do not change at all!
As I said, each stage inside the loop always works open loop, because there is nothing that changes in that stage. The only thing feedback changes is the effective signal level at the input.

If you have an amp with say three stages with gains 10x, 20x and 0.9 times, for a total open loop gain of 180. With 1V in that would give 180V out (if the amp could do that).

Now you close a feedback loop around it for a loop gain of 18x. How does that work? It works because the feedback puts 1/10 of the output signal on the inverting input. The effective input signal is the difference between the non-inverting and the inverting input, of course.
So now the effective input signal becomes not 1V that you put in, but only 0.1V! Since the amp itself has not been modified it still amplifies 180x, so now the Vout becomes 0.1V (effective Vin) x 180 = 18V...

So you can't say that the gain of one stage only decreases distortion of that stage. The feedback doesn't do anything on that stage, feedback doesn't do anything on any stage, it just fudges Vin to make the output what you want it to be.

jan
 
............. I will not discuss this any further. (these are questions for myth-busters)...........................................

Totally agree; let's hope that the subject is now (more) than adequately dealt with!

This red-herring has played out for far too long when included as part of a thread regarding a specific amplifier design. It has become the tail wagging the dog. If it is considered - by some - of great importance, then I would suggest it deserves a thread of it's own.
 
Oh, thanks L.C. i was afraid to have been mistaken during 40 years of my life.

Wahab, please, play with my simulations here, Crescendo revisited

You will see the differences between CFB and VFB *on the same amp*.
VFB ---> NFB
Bandwich: 200Khz ---> 3Mhz
Slew rate: 220v/µs ---> 1200v/µs
hd: 0.0023% ---> 0. 0002%
IM: x ---> x /10

As you can see, and as said L.C, improvement is in all the domains (except DC stability). The major improvement is listening pleasure.


It is clear that you are in love with this amp, and I can accept that. I just can't accept that this type of amp(current feedback) is the Holy Grail of all amps. You need it for other kind of amplification, but for audio nothing is wrong with voltage feedback.
I do not think that an audio amp needs so wide bandwidth, and so fast slewrate, 220 V/usec is more than enough.
I simulated a bit last LTsice schematic you have sent and here is the result. I changed D1 and D3 only from zener 3.9 V to 4.7 V as I do not have models for your originals.
At output level close to clipping THD20k is still quite low:

Fourier components of V(vin)
DC component:8.60974e-011

Harmonic Frequency Fourier Normalized Phase Normalized
Number [Hz] Component Component [degree] Phase [deg]
1 2.000e+04 2.191e+00 1.000e+00 0.00° 0.00°
2 4.000e+04 1.325e-10 6.048e-11 73.22° 73.22°
3 6.000e+04 2.382e-10 1.087e-10 2.02° 2.02°
4 8.000e+04 7.669e-11 3.500e-11 -1.00° -1.00°
5 1.000e+05 3.325e-10 1.517e-10 -1.38° -1.38°
6 1.200e+05 2.219e-10 1.013e-10 -58.95° -58.95°
7 1.400e+05 3.414e-10 1.558e-10 -6.02° -6.02°
8 1.600e+05 4.316e-10 1.970e-10 -69.82° -69.82°
9 1.800e+05 2.542e-10 1.160e-10 -17.76° -17.76°
Total Harmonic Distortion: 0.000000%

Fourier components of V(vout)
DC component:-0.00715549

Harmonic Frequency Fourier Normalized Phase Normalized
Number [Hz] Component Component [degree] Phase [deg]
1 2.000e+04 5.888e+01 1.000e+00 -0.38° 0.00°
2 4.000e+04 1.153e-03 1.958e-05 -13.89° -13.51°
3 6.000e+04 5.273e-04 8.956e-06 -76.37° -75.99°
4 8.000e+04 2.683e-04 4.557e-06 -175.14° -174.77°
5 1.000e+05 7.603e-04 1.291e-05 -83.66° -83.28°
6 1.200e+05 1.828e-04 3.105e-06 -167.69° -167.31°
7 1.400e+05 2.934e-04 4.982e-06 -77.47° -77.09°
8 1.600e+05 1.352e-04 2.297e-06 -164.91° -164.53°
9 1.800e+05 1.656e-04 2.812e-06 -77.33° -76.95°
Total Harmonic Distortion: 0.002643%

I my humble opinion this amp is in the B. Cordells EC MOSFET amp class and I doubed that any golden ears could distinguish between them.
dado
 

Attachments

  • distortion sim.jpg
    distortion sim.jpg
    259 KB · Views: 454
  • LG.jpg
    LG.jpg
    236.7 KB · Views: 455
I changed D1 and D3 only from zener 3.9 V to 4.7 V as I do not have models for your originals.
At output level close to clipping THD20k is still quite low:
Total Harmonic Distortion: 0.002643%

I think using voltage source of equivalent voltage rating is more accurate than changing zener voltage from 3v9 to 4v7. This topology has issues with stability. You cannot change a value and expect it will be fine. Problem is, there is trade off here between THD and stability. If you reach such a low distortion in simulation (0.0026% seems to be too good to be true), I have a feeling that it is not a stable circuit anymore.

improvement is in all the domains (except DC stability)

So how do you define your minimum stability (requirement) in simulation?

You will see the differences between CFB and VFB *on the same amp*.
VFB ---> NFB
hd: 0.0023% ---> 0. 0002%

Is it simulated performance? What is the real number, not simulated numbers? I believe that it is very difficult to achieve simulated performance without careful and skillful fine tuning because from the nature of this "current" feedback it is easy that the input transistor will have too low Vbe.

I saw that you have trimmed the current source (lower ccs has 3K32, upper ccs has 3K3) and also the base voltage of VAS transistor with 30 Ohm above zener. Could you give a hint how do you do the trimming process?
 
I think using voltage source of equivalent voltage rating is more accurate than changing zener voltage from 3v9 to 4v7. This topology has issues with stability. You cannot change a value and expect it will be fine. Problem is, there is trade off here between THD and stability. If you reach such a low distortion in simulation (0.0026% seems to be too good to be true), I have a feeling that it is not a stable circuit anymore.

When I simulated with voltage source of 3.9 V I got this:

Fourier components of V(vin)
DC component:1.67198e-010

Harmonic Frequency Fourier Normalized Phase Normalized
Number [Hz] Component Component [degree] Phase [deg]
1 2.000e+04 1.183e+00 1.000e+00 -0.00° 0.00°
2 4.000e+04 3.045e-10 2.574e-10 -116.19° -116.19°
3 6.000e+04 1.369e-08 1.157e-08 146.39° 146.39°
4 8.000e+04 2.402e-10 2.030e-10 34.05° 34.05°
5 1.000e+05 8.349e-09 7.056e-09 -54.02° -54.02°
6 1.200e+05 7.537e-11 6.370e-11 -158.86° -158.86°
7 1.400e+05 9.989e-10 8.442e-10 167.09° 167.09°
8 1.600e+05 7.847e-11 6.632e-11 125.45° 125.45°
9 1.800e+05 1.014e-08 8.567e-09 66.91° 66.91°
Total Harmonic Distortion: 0.000002%

Fourier components of V(vout)
DC component:0.00114608

Harmonic Frequency Fourier Normalized Phase Normalized
Number [Hz] Component Component [degree] Phase [deg]
1 2.000e+04 3.176e+01 1.000e+00 -1.47° 0.00°
2 4.000e+04 1.425e-03 4.485e-05 -11.25° -9.78°
3 6.000e+04 1.658e-01 5.221e-03 73.42° 74.88°
4 8.000e+04 1.137e-03 3.581e-05 176.95° 178.42°
5 1.000e+05 8.904e-02 2.804e-03 -119.80° -118.33°
6 1.200e+05 1.617e-04 5.092e-06 -43.20° -41.73°
7 1.400e+05 5.725e-02 1.802e-03 45.59° 47.06°
8 1.600e+05 5.330e-04 1.678e-05 171.37° 172.84°
9 1.800e+05 4.010e-02 1.262e-03 -150.47° -149.00°
Total Harmonic Distortion: 0.632175%

and with voltage source of 4.7 V I got this:


Harmonic Frequency Fourier Normalized Phase Normalized
Number [Hz] Component Component [degree] Phase [deg]
1 2.000e+04 1.183e+00 1.000e+00 0.00° 0.00°
2 4.000e+04 5.483e-12 4.634e-12 0.04° 0.04°
3 6.000e+04 8.223e-12 6.950e-12 -0.07° -0.07°
4 8.000e+04 1.097e-11 9.268e-12 -0.16° -0.16°
5 1.000e+05 1.371e-11 1.158e-11 -0.24° -0.24°
6 1.200e+05 1.645e-11 1.390e-11 -0.31° -0.31°
7 1.400e+05 1.919e-11 1.622e-11 -0.38° -0.38°
8 1.600e+05 2.193e-11 1.854e-11 -0.45° -0.45°
9 1.800e+05 2.467e-11 2.085e-11 -0.52° -0.52°
Total Harmonic Distortion: 0.000000%

Fourier components of V(vout)
DC component:-0.00485397

Harmonic Frequency Fourier Normalized Phase Normalized
Number [Hz] Component Component [degree] Phase [deg]
1 2.000e+04 3.180e+01 1.000e+00 -0.37° 0.00°
2 4.000e+04 1.013e-04 3.184e-06 -37.91° -37.54°
3 6.000e+04 3.877e-04 1.219e-05 -84.20° -83.84°
4 8.000e+04 1.218e-04 3.831e-06 -173.05° -172.68°
5 1.000e+05 1.040e-04 3.272e-06 -82.05° -81.68°
6 1.200e+05 9.281e-05 2.919e-06 -170.45° -170.08°
7 1.400e+05 3.915e-05 1.231e-06 96.16° 96.52°
8 1.600e+05 7.232e-05 2.274e-06 -168.40° -168.03°
9 1.800e+05 9.196e-05 2.892e-06 102.03° 102.39°
Total Harmonic Distortion: 0.001441%

so maybe thia amp works better with 4.7 V zener instead of 3.9 V.
dado
 
Wahab, please, play with my simulations here, Crescendo revisited

You will see the differences between CFB and VFB *on the same amp*.
VFB ---> NFB
Bandwich: 200Khz ---> 3Mhz
Slew rate: 220v/µs ---> 1200v/µs
hd: 0.0023% ---> 0. 0002%
IM: x ---> x /10

As you can see, and as said L.C, improvement is in all the domains (except DC stability). The major improvement is listening pleasure.

Hi , Esperado , i ll give it a try as soon as possible.:)

I'm afraid that according to my understanding (I don't claim to have a perfect understanding of anything in this world) most, if not all of this, is completely wrong and will be confusing.

There are things that are counter intuitive , mind you....;)

So you can't say that the gain of one stage only decreases distortion of that stage.

jan

I said that it cancel distorsion of that stage as well as the one of following stages but not the distorsion of preceding stages...

Say an amplifier with two chained stages with gains of A and B respectively.

Let s call their caracteristic functions F(x) and G(x) respectively,
x being the input signal and a function of time.

The first stage caracteristic function can be summarized as :

F(x) = ax + d , a being the stage gain and d a function of x that is the stage distorsion.

Identicaly , the second stage caracteristic function will be :

G(x) = bx + e , b is the stage gain and e its distorsion.


Let H be the caracteristic of the chained stages , wich can be devellopped as :

H(x) = G(F(x)) , that is G as a function of F(x)

= G(ax + d)

= B(ax + d) + e

= ABx + Bd + e

Now we apply NFB in such a way that the amp has CLG of 1 , that is , we divide
the final caracteristic function by the factor AB wich is the total available gain.


This yield the CLG caracteristic function , let s call it K(x).

K(x) = H(x) / (AB)

= (ABx + Bd + e) / AB

= x + d/A + e/AB

As you can see , the first stage distorsion , d , is reduced only by the
loop gain provided by the first stage , while the second stage distorsion , e ,
is reduced by both the first and second stage loop gains.
 
Last edited:
AX tech editor
Joined 2002
Paid Member
I don't think that is correct. If you have two functions in series F(x) and G(x), the total fransfer function is the product of the two, namely G(x)*F(x), which in your example should come out to (ax+d)*(bx+e), no?

Edit: I do think actually you are right; the error of the last function e is not processed by the first function (ax+d).
Wow! That is really interesting. I need to digest it furtehr.

jan
 
Last edited:
Edit: I do think actually you are right; the error of the last function e is not processed by the first function (ax+d).

It is a very simple Mathematics, Jan. I thought you cared with the understanding/interpretation of the Math, not the Math itself.

The math presented by Wahab is correct (of course), but the interpretation of it can be right can be wrong. Grammatically I disagree with the logic:'

= x + d/A + e/AB

As you can see , the first stage distorsion , d , is reduced only by the
loop gain provided by the first stage , while the second stage distorsion , e ,
is reduced by both the first and second stage loop gains.

The Mathematical expression does not say anything about a reduction of stage distortion, rather, it says that the final distortion is "dominantly" determined by the earlier (here the first) stage distortion. This is logical because first stage distortion will be "augmented" by the gains of stage 2, stage 3, stage 4 and so on.
 
AX tech editor
Joined 2002
Paid Member
Yes I agree with you and of course Wahab is correct in that view - it is a different way of expressing it. We might have a langauge thing here.

Wahab is correct, as you made clear, that the first stage distortion is dominant. I took his statement too literally; of course the stage distortion is not modified as such.
But, to be honest, I didn't realise this dominance of the 1st stage so in that respect I learned something from this!
Thank you and thank you Wahab - you can have that beer now ;)

jan
 
I don't think that is correct. If you have two functions in series F(x) and G(x), the total fransfer function is the product of the two, namely G(x)*F(x), which in your example should come out to (ax+d)*(bx+e), no?

Edit: I do think actually you are right; the error of the last function e is not processed by the first function (ax+d).
Wow! That is really interesting. I need to digest it furtehr.

jan

Hi , Jan

That s not a transfer function but a caracteristic function
wich is in time domain.

The Laplace transform of a caracteristic function is a transfer function ,
that is , it is mapped in a complex frequency domain.

I myself need to look further as it s not a rigorous demonstration
mathematicaly speaking , since i deliberatly simplified the things
by downplaying non dominant terms.

Indeed , IIRC , some people that were in this forum already pointed
what is explained in theses lines.


Thank you and thank you Wahab - you can have that beer now ;)

jan

A Mort Subite or a Jenlain will do it perfectly...;)
 
so maybe thia amp works better with 4.7 V zener instead of 3.9 V.
dado

With the tendency for the output to have positive DC, and with both FB paths have equal 1K resistance, it is the lower input transistor that will have the tendency to get too low Vbe.

There are many ways to increase this Vbe. Easy way is to increase the current thru the 1K resistors (set by the 3K3 of the CCS).

Using higher zener voltage will have the same effect to the CCS (i.e. to increase current).

Increasing the FB resistors (e.g doubling the resistance with 2K||75R) will also solve the issue.

But all of these will put the circuit in worse stability. This is probably the reason why this topology is not popular, despite its good sound character.
 
With the tendency for the output to have positive DC, and with both FB paths have equal 1K resistance, it is the lower input transistor that will have the tendency to get too low Vbe.

There are many ways to increase this Vbe. Easy way is to increase the current thru the 1K resistors (set by the 3K3 of the CCS).

Using higher zener voltage will have the same effect to the CCS (i.e. to increase current).

Increasing the FB resistors (e.g doubling the resistance with 2K||75R) will also solve the issue.

But all of these will put the circuit in worse stability. This is probably the reason why this topology is not popular, despite its good sound character.

Yes, that is correct, but I've got to high distortion 0.3% at 1 kHz and 0.5% at 20 kHz and very bad FFT, when 3.9 V was used. I don't get it, some people like high distortion or what. By the way, with 4.7 V phase marhin I've got is 47 degree. Am I doing something wrong with this simulatio??
dado
 
So how do you define your minimum stability (requirement) in simulation?
For Dc ? i don't care it in simul, but in real life.
Is it simulated performance? What is the real number, not simulated numbers? I believe that it is very difficult to achieve simulated performance without careful and skillful fine tuning because from the nature of this "current" feedback it is easy that the input transistor will have too low Vbe.
I saw that you have trimmed the current source (lower ccs has 3K32, upper ccs has 3K3) and also the base voltage of VAS transistor with 30 Ohm above zener. Could you give a hint how do you do the trimming process?
Yes this distortion is in simul, i don't have powerful enough distortiometer to measure so low levels in real life. In real life, i use crossed fingers when power on, then my ears and an oscilloscope. ;-)
I published-it for educational purpose, to demonstrate the benefit of Current feedback.

But this amp is playng music (fine, so fine) in my real system. I use-it to power a JBL motor in a spherical waves horn made in plain wood, between 1KHz and 20 KHz
The resistance you are talking about is just a fine tuning for 0Vdc at the output. I needed-it because my protection circuit is tuned to be very 'aware' of DC and HF. The real values are not exactly the same on both channels.

I am waiting L.C propose his parts (if i can afford them) to compare on a musical point of view.
 
Last edited:
I don't get it, some people like high distortion or what... Am I doing something wrong with this simulatio??

Hehehe nobody likes high distortion especially when it is third order. The only objection I have with this circuit is the relatively higher 3rd order distortion as compared to 2nd order. I would prefer higher overall distortion but from 2nd order.

You didn't do something wrong I believe. Here is a trick:

The FB resistor (1K) is connecting the input transistor's emitor and the output.

You want the input transistor to have Vbe of at least 0.6v (approximation). Base voltage (Vb) is assumed zero and so is output voltage (Vout). So to get 0.6v at the emitters (Ve) you want a certain current thru the FB resistors (This is what the CCS is for).

V = 0.6V
R = 1K
I = calculate! ---> 0.6mA.

CCS current == current thru FB resistor + current thru transistor

Note that the current thru the transistor is also affected by the collector resistor (1K here, which is a good value). This current is around 0.5mA. So more or less you want the CCS to give out 0.6mA+0.5mA = 1.1mA.

You may want to set both CCS to output 1.2mA. You will find that the higher the current, the lower the distortion. That's why you got ("too") low distortion when you use higher zener voltage which affect the base voltage of the CCS, giving very high current I believe.

But I suspected that it would be better (e.g. for the stability) if top CCS gives different current than the bottom CCS. With positive voltage at the output you want the top CCS to have slightly higher current. You can try 1.2mA for the top and 1.1mA for the bottom CCS.

That's just my guesswork, you will have to troubleshoot the correct setting after you build it (if you intent to). In Christophe's circuit, top CCS has 3K3 and bottom CCS has 3K32. This means that the top CCS has slightly more current (I guess). But unfortunately the base voltage of the ccs also determines the current. I guess that you need to work out the 3K3 resistors on the CCS and the 30R resistor at the base of the VAS transistor.

You owe to yourself to build any of the SSA amp. Better is the more complex one.
 
For Dc ? i don't care it in simul, but in real life.

Actually I meant stability against oscillation, the hallmark of all high speed circuits.

The resistance you are talking about is just a fine tuning for 0Vdc at the output. I needed-it because my protection circuit is tuned to be very 'aware' of DC and HF.

Ah, so it will save the speaker when there is oscillation.

I am waiting L.C propose his parts (if i can afford them) to compare on a musical point of view.

Unfortunately he didn't use lateral output. I guess the CSA will use the BIGBT output. :eek: But hopefully your amp will outperform the CSA so I don't need to build the CSA :rofl:
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.