differential line in gain network math

[zamproject]

Hello

I have this line input (pict attached) and want to change the gain setting

do I have to consider the IC working "differential" with:
NFB = - R108/R(TrafoSecondary)
FB = 1+ (R109/R107)

If yes the gain should be equal right ? so If FB gain = (1+(3K9/100) = 40
, Rtrafo is about 107.5 ohm (4K3/40) ?

Best
Zam

 

[JohnRoberts]

without getting into the weeds it looks like scaling the one feedback resistor R108 will change gain without changing differential.

The positive feedback may cause some funny business.

JR

 

[zamproject]

without getting into the weeds it looks like scaling the one feedback resistor R108 will change gain without changing differential.

The positive feedback may cause some funny business.

JR

Hello JR

I try that yesterday (trim pot in place of R108), I got wired low pulse/oscillation/saturation without that many more gain...
That's why I wondering if I need to scale both feedback, I'm not sure about the math around.

I can also change the input pad, (I need about 4 more dB), have to calculate if the impedance don't drop too much.
But be able to change the gain have interest too

Best
Zam

 

[Bo Deadly]

This circuit is non-trivial. The gain depends on the impedance of the source resistance on the negative input which depends on the transformer impedance which depends on R101-R104 and the impedance of the source device. You could reduce R101-R104 but that is dubious because that would increase the dependence on the impedance of the source device. You could try reducing positive feedback by reducing R107 but that seems dubious because it's probably doing some negative impedance converter stuff that I just don't understand. I would have to study this in LTSpice for a few hours before messing with it.

 

[JohnRoberts]

Hello JR

I try that yesterday (trim pot in place of R108), I got wired low pulse/oscillation/saturation without that many more gain...
That's why I wondering if I need to scale both feedback, I'm not sure about the math around.

that would be the funny business...

I can also change the input pad, (I need about 4 more dB), have to calculate if the impedance don't drop too much.
But be able to change the gain have interest too

I don't see pad in that schematic...  1.8k and 10nF looks like a pole at 8.8kHz unless I am reading the values wrong.

Best
Zam

I need to tap out... I don't have any good suggestions.

JR

 

[zamproject]

Ok....

Don't bother sim the thing, I just lower 103/104 to 2K and get close to unity form circuit input to IC out.
I'll check freq response, and if it's ok I can leave with that
By the way Rpri=144ohm Rsec=104ohm

Best
Zam

 

[Bo Deadly]

I just lower 103/104 to 2K and get close to unity form circuit input to IC out.
I'll check freq response, and if it's ok I can leave with that

Checking the frequency response may not be enough. The problem is that I think the whole positive feedback business is a "negative impedance converter" which is supposed to get better LF out of a smaller transformer by cancelling winding resistance or some such [1]. However, the whole thing is very delicately balanced. If it's off by even a little bit it can become unstable. You can get an oscillator. Or it just doesn't do what it's supposed to. It's tuned you might say. So you might need to study the LF behavior and look for slow oscillation or "funny business" as has already been stated 3 times now.

By the way Rpri=144ohm Rsec=104ohm

That's DCR. The AC impedance depends on inductance and source / load. I was never good at such things. I wouldn't know spit about Electronics if it were not for LTSpice.

[1] https://sound-au.com/articles/audio-xfmrs.htm section 5

 

[user 37518]

Interesting, what's the purpose of positive feedback via R109 and R107? better stability perhaps?

 

[zamproject]

Checking the frequency response may not be enough. The problem is that I think the whole positive feedback business is a "negative impedance converter" which is supposed to get better LF out of a smaller transformer by cancelling winding resistance or some such [1]. However, the whole thing is very delicately balanced. If it's off by even a little bit it can become unstable. You can get an oscillator. Or it just doesn't do what it's supposed to. It's tuned you might say. So you might need to study the LF behavior and look for slow oscillation or "funny business" as has already been stated 3 times now.
That's DCR. The AC impedance depends on inductance and source / load. I was never good at such things. I wouldn't know spit about Electronics if it were not for LTSpice.

[1] https://sound-au.com/articles/audio-xfmrs.htm section 5

Hello

I tweak R 107/108/109 consecutively in the pas few days...I think I already know the "funny business" it can provide 
There is probably a way with R108/109 simultaneous change, with a constant ratio maybe, or specific math I don't have for this.

I just do FFT plot, after R103/104 change, I have a very little more "wave" in the 10-30Hz area, but HF roll off is better (higher) than before...
I don't hook the scop for the moment, I have to do it at IC out to be sure about low freq oscillation, FFT is trough the whole circuit, with few other stage, coupling caps, and trafo out that might cut things

Have you tried increasing the value of R107 slightly?  Leaving the rest alone.
I'm simulating it right now in LTSpice.

Adam

I do, 200ohm trim pot, change nothing regarding gain IIRC, but suddenly get the funny business

Best
Zam

 

[moamps]

It is a zero field circuit that can be analyzed as a current amplifier with a current transformer at the input. When I designed the line preamplifier with the Lundahl 7101 transformer, I used a simple calculation that I will try to apply here
Input impedance
Zin = sum (R101-R104), should be about 10 to 20kohm
Input (primary) current
Ipr = Vin / Zin
Output current (secondary)
Isec = lpr / N2 * N1
Output voltage
Vout = Isec x Rf

Vout / Vin = (Rf* N 1) / (N 2 * Zin)

Stability conditions:
Rf / Rsec equal or little bit higher than R109 / R107
(Rsec-resistance of the secondary coil)
Adjust R107 for lowest THD at low frequencies, keep R109 / R107 ratio

For Zin = 10.8k, Rf = 4k3, N1: N2 = 1: 0.62
Vout / Vin = 0.63 = -3.8dB

For Zin = 7k
Vout / Vin = 1 = 0 dB
I hope I didn’t make a mistake somewhere in the calculation.

 

[zamproject]

Hello moamps

For Zin = 10.8k, Rf = 4k3, N1: N2 = 1: 0.62
Vout / Vin = 0.63 = -3.8dB

That's exactly what I measure !! (-3.8dB, RMS DMM)
from bal input(float) to opamp out (across 0V)
and that's the 4 dB I was looking to recovery (for reasons after the input stage)

For Zin = 7k
Vout / Vin = 1 = 0 dB
I hope I didn’t make a mistake somewhere in the calculation.

that confirm what I "empirically" do and basic math I came of by changing the two 3k9 to 2k (total 7k in place of 10k8).
Yesterday I measure RMS voltage (1k) at trafo input, where there is attenuation (not a pad as I wrongly say and JR correct this), I deduce the pri "resistance", 145ohm, and measure 144 after... good.... I simply consider the input resistor network (including trafo) as a voltage divider, and do basic math to have to proper RMS voltage at trafo input (not system input) to have my 4 (3.8 ) dB.
Honestly, I don't know how I get this but your math (can you elaborate ?) find same practical result as my dumb speculation
Now I'm good by 0.1dB accuracy ...

So definitely the trafo RESISTANCE (and not Z...) is part of the whole network ?!
This is Studer I should have recognize the topologies I saw many time...like in 0ohm sum/amp ?

I'll see later if I try again (N)FB trick with your explanation

Thanks !
Zam

 

[moamps]

.... I don't know how I get this but your math (can you elaborate ?) ...
So definitely the trafo RESISTANCE (and not Z...) is part of the whole network ?!

This is just an applied calculation for a current transformer. The current transformer has very small reactive components so IMO they can be neglected for the audio spectrum.

This is Studer I should have recognize the topologies I saw many time...like in 0ohm sum/amp ?

This approach (using a current transformer for summing amplifiers) can be found in a lot of design solutions from Studer, Siemens, etc.

I'll see later if I try again (N)FB trick with your explanation

Positive feedback network is used to get zero field inside transformer core by equalizing currents (Isec / Ipr) = (Npr/ Nsec), R107 should be equal to Rsec, and than adjusted a bit to get the best THD figure.