Op Amp PSRR problem

Thread Starter

brianmk

Joined Dec 23, 2016
59
Back in the 1970s I built a stereo receiver. It's been sitting unused in the loft for many years.
After taking it out of storage recently I began testing it.
I discovered a design problem with the receiver tuning and AFC.

The receiver IF is based on the famous Nelson Jones design using a CA3189E :-
https://www.worldradiohistory.com/UK/Wireless-World/70s/Wireless-World-1978-09.pdf

It uses a 741 op amp to sum the 5.4V AFC reference and the variable AFC voltage from the CA3189 to provide the tuning voltage for the front end.
In my case the tuner is a EF5803 (Ambit International). It requires a tuning voltage of 14V to reach the top of the FM band (188MHz).
The positive supply for the op amp is an unregulated 20V.
To reduce the effect of noise and ripple on the supply, it is fed via a 680 Ohm resistor and 100uF decoupling cap.
That reduces the op amp supply voltage to around 17V.

The simplified circuit looks like this. (I tried simulating it in ltspice).

1708518194503.png

The problem I am seeing is that the tuning voltage is being modulated by random low frequency noise on the 20V supply.
As the supply is unregulated, this noise could originate from variations in the mains supply.
The noise causes the center zero tuning meter to fluctuate.
If I replace the supply with a stabilised bench PSU, the noise goes away.

The 741 is supposed to have around 80->90 dB PSRR. Is there a better op amp I could try?
I have read that the PSRR gets worse as the common mode voltage approaches the op amp supply voltage.

Could I add a low drop out regulator to provide a stabilised supply for the op amp?
There is quite a bit of ripple on the unregulated 20V supply (around 500mV).

Is there a better solution?
 

Thread Starter

brianmk

Joined Dec 23, 2016
59
Back in the 1970s I built a stereo receiver. It's been sitting unused in the loft for many years.
After taking it out of storage recently I began testing it.
I discovered a design problem with the receiver tuning and AFC.

The receiver IF is based on the famous Nelson Jones design using a CA3189E :-
https://www.worldradiohistory.com/UK/Wireless-World/70s/Wireless-World-1978-09.pdf

It uses a 741 op amp to sum the 5.4V AFC reference and the variable AFC voltage from the CA3189 to provide the tuning voltage for the front end.
In my case the tuner is a EF5803 (Ambit International). It requires a tuning voltage of 14V to reach the top of the FM band (188MHz).
The positive supply for the op amp is an unregulated 20V.
To reduce the effect of noise and ripple on the supply, it is fed via a 680 Ohm resistor and 100uF decoupling cap.
That reduces the op amp supply voltage to around 17V.

The simplified circuit looks like this. (I tried simulating it in ltspice).

View attachment 315799

The problem I am seeing is that the tuning voltage is being modulated by random low frequency noise on the 20V supply.
As the supply is unregulated, this noise could originate from variations in the mains supply.
The noise causes the center zero tuning meter to fluctuate.
If I replace the supply with a stabilised bench PSU, the noise goes away.

The 741 is supposed to have around 80->90 dB PSRR. Is there a better op amp I could try?
I have read that the PSRR gets worse as the common mode voltage approaches the op amp supply voltage.

Could I add a low drop out regulator to provide a stabilised supply for the op amp?
There is quite a bit of ripple on the unregulated 20V supply (around 500mV).

Is there a better solution?
I am starting to think that since the required output is close to the supply voltage, it may need a rail to rail op amp to replace the 741. Any suggestions?
 

crutschow

Joined Mar 14, 2008
34,452
To reduce the effect of noise and ripple on the supply, it is fed via a 680 Ohm resistor and 100uF decoupling cap.
I believe there is an error in your schematic.
The capacitor should be connected at the other side of the resistor from the op amp power pin to ground.
The capacitor may be faulty.
As an initial test you might replace that capacitor with one of 470µF or greater to see if that helps.
I am starting to think that since the required output is close to the supply voltage, it may need a rail to rail op amp to replace the 741.
The 741 should be okay to provide 14V from a 17V supply.
 

Thread Starter

brianmk

Joined Dec 23, 2016
59
I had the same thought about the cap being faulty. I had already removed and tested it on a Marconi LCR bridge. It appears to be ok.
 

Thread Starter

brianmk

Joined Dec 23, 2016
59
Another thought I had is a dry joint somewhere between the transformer and the bridge rectifier for the 20V supply.
I need to check this more thoroughly, although I can see no obvious LF noise when looking at the 20V rail on a 'scope.
I just see the ripple. From my observations it only needs a tiny amount of noise (just a few mV) on the supply to create a problem with the front-end tuning voltage.

I don't think ripple is the problem. There is a 33 Ohm resistor and big 1000uF? cap between the 20V supply and the IF PCB. (Not shown on my ltspice model). That attenuates the ripple significantly before it reaches the IF PCB. The noise that is causing the problem looks like lower frequency fluctuations - Maybe an earthing issue?

The 20V supply includes a large 4700uF 40V smoothing cap that I have also removed and tested because of it's age.
Because it had been is storage for a few years I did take the precaution to reform all the large smoothing caps before initial power on. They all appear to be ok.
 

crutschow

Joined Mar 14, 2008
34,452
I had the same thought about the cap being faulty. I had already removed and tested it on a Marconi LCR bridge. It appears to be ok.
Still suggest you try replacing it with a larger value to see if that helps, before you try anything else more complicated.
 
Last edited:

Thread Starter

brianmk

Joined Dec 23, 2016
59
Assuming I've got it right (and the LM741 model is accurate), the simulation appears to show that the tuning voltage output (Vt) (green) is affected by the supply voltage until the supply voltage (V+) (red) is above 21V. I only have 20V available.
The simulation steps V+ from 10V to 30V in 0.5V steps.

I will try increasing C1 to 470uF and reducing R3 to 150 Ohms to keep the filter time constant about the same.
That should allow the V+ threshold to be reduced to from 21V to 19V.

Update: That last change did not help. I am guessing because V+ is limited to about 18.5V due the additional 33 Ohm series resistor and 1000uF cap. If I reduce the 33 Ohm resistor it will increase the supply ripple which could result in mains hum.

1708546781233.png
 
Last edited:

crutschow

Joined Mar 14, 2008
34,452
I will try increasing C1 to 470uF and reducing R3 to 150 Ohms to keep the filter time constant about the same.
Why do you want to do that?
A longer time-constant makes the filter more effective, which is what we are trying to do.
The idea is to minimize any noise at the op amps power pin.
 

Thread Starter

brianmk

Joined Dec 23, 2016
59
I think I have a fix:-
In addition to changing R3 from 680 Ohms to 150 Ohms and C1 from 100uF to 470 uF,
I reduced the gain of U1 slightly by changing R2 from 4K7 to 5K6.
That still allows Vt to reach 14V needed for 108MHz at the top of the band.
The threshold for Vt not to be affected by V+ is now about 17V.

1708549482461.png
 

Thread Starter

brianmk

Joined Dec 23, 2016
59
Why do you want to do that?
A longer time-constant makes the filter more effective, which is what we are trying to do.
The idea is to minimize any noise at the op amps power pin.
The reason for reducing R3 is to increase the supply voltage on the op amp. The problem seems to be that as the op amp output approaches the supply voltage, the power supply rejection diminishes to such an extent that changes to the supply appear on the op amp output. The simulation seems to confirm that.

Reducing the gain of U1 helps in that the output voltage is lower relative to the supply voltage. Hence it has more headroom and the PSRR is improved.

The fix also means the AFC is now working as it should. Before the modification, when I pressed the 'AFC' button, the center zero tuning meter hardly moved. Now it is pulled to the center zero position, even when the station is initially some way off tune. The front end is being locked to the station as it should. The op amp is unable to handle common mode input signals close to the supply voltage.
 
Last edited:
Top