F5X -- the EUVL Approach - The Build Thread

> Can someone explain the purpose of R21 and R22 which connect between the outputs and ground? Is it for DC stability ?

In a proper F5X with floating X, the feedback resistor network is not referenced to Gnd, unlike the single ended F5X.
Even if the speaker is connected, it is wise to have a reference path to ground to keep absolute DC (as opposed to differential DC) under control.
You will find the same resistor at the output of the Aleph-X.


Patrick
 
The DIL sockets I buy from a local shop. So no part number I am afraid.
But the PRP 1/4 resistors has lead diameter of 0.6mm. This will fit most, if not all DIL sockets.

Failing that you can always solder some 0.8mm solder pins on the PCB at those positions, and then solder the resistor onto the pins.


Patrick
 
> Can someone explain the purpose of R21 and R22 which connect between the outputs and ground? Is it for DC stability ?

In a proper F5X with floating X, the feedback resistor network is not referenced to Gnd, unlike the single ended F5X.
Even if the speaker is connected, it is wise to have a reference path to ground to keep absolute DC (as opposed to differential DC) under control.
You will find the same resistor at the output of the Aleph-X.


Patrick

I have been analysing equations for the X circuit and cannot see any evidence that the reference path to ground is ill-defined. The ground reference is established by the JFET input inputs and is the common-mode voltage of the inputs. The voltages on the X wires will be nearly identical and the same as the common-mode input voltage minus the difference of the P and N JFET Vgs values. There will also be minor voltage differences due to the different distortion currents running through the 2 connections.

I still do not know if there is a significant difference in distortion cancellation ability between the X and the ungrounded balanced topology (ie. shorting the to X wires).

Maybe I am confused or missing something?
 
> Maybe I am confused or missing something?

Yes.

Think what is in the circuit that will stop the outputs from drifting relative to Gnd together (common mode DC) with floating X.
Differential DC is taken care of by the feedback network.

> I still do not know if there is a significant difference in distortion cancellation ability between the X and the ungrounded balanced topology.

In terms of distortion no.
Nelson did published another balanced frontend in his F5Turbo article, which I would call F5H.
By all means try them all. The PCB will support all these configurations.


Patrick
 
> Maybe I am confused or missing something?

Yes.

Think what is in the circuit that will stop the outputs from drifting relative to Gnd together (common mode DC) with floating X.
Differential DC is taken care of by the feedback network.

> I still do not know if there is a significant difference in distortion cancellation ability between the X and the ungrounded balanced topology.

In terms of distortion no.
Nelson did published another balanced frontend in his F5Turbo article, which I would call F5H.
By all means try them all. The PCB will support all these configurations.


Patrick

I see the problem now.

The problem is if the DC voltages of the outputs drift equally inopposite directions, keeping the voltages at the X wires the same. Bad news: the DC bias current increases on one side and decreases on the other side.
 
I retract my previous post, and now totally agree with you about the problem.

The output voltages can drift together causing the X voltages to also drift by the same amount. This is equivalent to the common mode DC input voltage drifting by that value with the X wires grounded. Not a good situation w.r.t. output bias currrents.
 
I retract my previous post, and now totally agree with you about the problem.

The output voltages can drift together causing the X voltages to also drift by the same amount. This is equivalent to the common mode DC input voltage drifting by that value with the X wires grounded. Not a good situation w.r.t. output bias currrents.

I now retract my retraction. Here is a pictorial DC analysis of the X circuit
assuming transconductance amplifiers, and only common mode input.
 

Attachments

  • x-theory-1.jpg
    x-theory-1.jpg
    72.9 KB · Views: 1,561
My line of thinking is slightly different.

The top and bottom halves of the F5X are using different devices which has different tempco.
So suppose during warm up, the lower half bias drifts differently to the upper half.
This you can simulate by artificially introducing an offset voltage (Vo) between In1=In2=0 and U1, U3 +Input.
The question is then whether the OUT1=OUT2=0 (rel. to Gnd)?

OUT1-OUT2 (differential output to the speaker) will be 0.


Patrick
 
My line of thinking is slightly different.

The top and bottom halves of the F5X are using different devices which has different tempco.
So suppose during warm up, the lower half bias drifts differently to the upper half.
This you can simulate by artificially introducing an offset voltage (Vo) between In1=In2=0 and U1, U3 +Input.
The question is then whether the OUT1=OUT2=0 (rel. to Gnd)?

OUT1-OUT2 (differential output to the speaker) will be 0.


Patrick
This might look overwhelming, but describes the common mode output voltage. The calculations were done using Maxima.

g1 and g2 are the open-loop gains of the two transconductance amplifiers.
xc = INcom.

OUTcom = Rload*xc*((g2+g1)*(R2+R1)+2*g1*g2*R2) / (2*(g1+1)*(g2+1)*R2+(g2+g1+2)*R1)

Certainly, differences in the open-loop gains affect to output voltage, but is it well defined. If the output voltage were ill-defined, the system of equations would either have no solution, or there might be a zero in the denominator to the above equation. Since all of the parameters are positive numbers, the denominator in non-zero.

The equations could be further complicated by allowing the resistors to vary, but that will not change to stability of the result.
 
IMHO Rload across differential outputs plays no parts in any common mode analysis, as by definition Out1=Out2.
So there will be no current through Rload, hence equivalent to open circuit.

In case R21, R22 = infinite, there is no other means to drain any bias current difference between top and bottom half,
so that OU1=OUT2 has to be no zero in order to re-balance the DC bias currents of U1 & U2 (as well as U3 & U4).


Patrick
 
> so that OU1=OUT2 has to be no zero

Of course that is fine and does not affect the functioning of the circuit.

One can apply the same argument and remove the 100k resistors at the input, and short the 1ks by a jumper.
In pure circuitry terms they also do not have a real function.
Nothing then connects to Gnd and you can even remove that as well.

But I agree with Nelson that it is a good idea to have some ground reference, both at the input and the output.
So personally I do want R21, R22.

But do as it pleases you.


Patrick
 
IMHO Rload across differential outputs plays no parts in any common mode analysis, as by definition Out1=Out2.
So there will be no current through Rload, hence equivalent to open circuit.

In case R21, R22 = infinite, there is no other means to drain any bias current difference between top and bottom half,
so that OU1=OUT2 has to be no zero in order to re-balance the DC bias currents of U1 & U2 (as well as U3 & U4).


Patrick

I agree that there are errors in my presentation. I need to work out the bugs in my analysis and presentation.

The basic nature of my argument is:

1. Assume that the amplifiers have been adjusted for proper bias currents with the X connections grounded. The nominal output voltages are

Out = gain*In.

2. Unground the X connections.

If the output voltages drift away from gain*In, then the negative feedback voltage to the transconductance amplifiers will also drift
by the amount related to the output voltage drift and the voltage on the X connections (and also the closed-loop gain, ...). This drift in voltage to the negative feedback inputs to the transconductance amplifiers will force to output currents to increase into one of the rails and decrease into the other rail, driving the output voltage towards gain*In.

Explain to me the difference between this situation and the common
non-inverting unity gain opamp circuit where the output is directly fed to the negative input of the opamp. The only difference is the possible influence of the X connections.

BTW: If the currents into the negative terminals of the amplifiers is ingored (or assumed to be identical), the the voltages on the X connections will be identical, resulting in the balanced H topology.
Those voltages will be .5*(out1+out2).
 
I refer to NP's F5 Turbo article, P.15 in which he shows a balanced circuit with input resistors to Gnd.
In that particular schematics he does not use any output resistors to Gnd.

I already mentioned that I agree with you that the feedback network will also control common mode DC.
It is my personal choice to want to use R21, R22, as I like to have a ground reference for the outputs in a balanced circuit.
Just my personal taste, no scientific reasons. So you do not have to use them.


Best regards,
Patrick
 
Last edited:
I refer to NP's F5 Turbo article, P.15 in which he shows a balanced circuit with input resistors to Gnd.
In that particular schematics he does not use any output resistors to Gnd.

I already mentioned that I agree with you that the feedback network will also control common mode DC.
It is my personal choice to want to use R21, R22, as I like to have a ground reference for the outputs in a balanced circuit.
Just my personal taste, no scientific reasons. So you do not have to use them.


Best regards,
Patrick

There are rational reasons for the input resistors: The R1 gate stopper minimizes the chance for JFET oscillation. The R2 resistor to ground establishes a voltage (ground) on the JFET gates when the input is disconnected, or driven by a capacitor coupled output.

I have totally reworked my analysis in terms of SUSY for the next post.
 
I refer to NP's F5 Turbo article, P.15 in which he shows a balanced circuit with input resistors to Gnd.
In that particular schematics he does not use any output resistors to Gnd.

I already mentioned that I agree with you that the feedback network will also control common mode DC.
It is my personal choice to want to use R21, R22, as I like to have a ground reference for the outputs in a balanced circuit.
Just my personal taste, no scientific reasons. So you do not have to use them.


Best regards,
Patrick

Looking at Nelson's SUSY Patent (Amplifier with Gain Stages Coupled for Differential Error Correction), it looks to me that Fig. 7 (Prior Art)
is closest to the circuit topology of balanced F5 of F5X. Nelson claims in the Patent the following:

"The distortion and noise contributions of both amplifiers will show in both outputs with opposite polarity so that they will not be removed differentially at the load".

Assuming that Nelson is correct, then balanced F5 and F5X are not SUSY, but do tolerate common mode on the inputs.