[SOLVED] Question about op-Amp in control circuit

[Hawaslsh]

I am using a voltage controlled attenuator to try and achieve a specific power output from a VCO. The output from the attenuator is split and fed into a Schottky diode detection circuit. A non-inverting amplifier (U1) adds a gain of 10, and the output goes to another op-amp (U2) where there detected voltage is “compared” to a reference I provide through a potentiometer.

Initially I did not have the LPF network in the dashed box. When I probed the control voltage I saw U2’s output oscillating around some DC value at 15 KHz. Knowing the control voltage pin draws nearly no current, a very simple RC filter fixed the output to that DC value. Now the circuit produces a stable output at the expected level. However, there are still a few lingering issues.

The main issue being the voltage control pin (Vcont) on the variable attenuator can not go negative. If I am operating U2 with a dual supply, there are instances where U2 will place its output at the negative rail. The easiest solution is to operate U2 on a single supply, and that works ok. However, it significantly increases the time it takes to decrease Vcont and discharge the 10uF cap. I could easily change the LFP to avoid the large 10K resistor, but is there an alternate way to ensure the output of U2 never goes negative but keep U2 operating on a dual supply?

I’ve also been trying to understand the practical limit of how small a voltage U1 can amplify before its buried in noise. Assuming the output from the diode detection circuit is noise free, what's the limited factor U1’s performance? The datasheet for U1 has 3 noise specs at the bottom of page 4, I’m not too sure how to interpret them. I’m also not sure if that matters and if U1’s offset voltage would have more of an influence on this spec. Using test measurement equipment, I can measure changes in the diode detection circuit down to the 100uV range before it become indistinguishable to the presence of RF power.

Thanks in advance

 

[KlausST]

Hi,

A few comments.
I do it in signal flow beginning with the splitter.

* next is the left side diode. It does nothing than short circuiting the negative waves from the input side. If you remove it the output will be about the same.
... I can only guess that there is a series C missing.

* Diode, RC filter, all fine

* U1: Missing power supply (and apacitors- all OPAMPs). You talked about noise. So if you want to reduce noise you should consider an additional capacitor at the feedback.

* REF: Rather unknown. Again: if you want to reduce noise, then add a capacitor.

* U2: What? Low noise, low drift, precision... Why? It has no feedback!, thus it´s output is expected to (randomly) switch between the rails. It creates noise as high as can be.
In my eyes it misses to be designed at all. It calls for trouble. Usually one wants a defined gain and phase behaviour for a regulation to be precise, fast and stable.
As you wrote its used as a "comparator" but if you (addressing the circuit designer) want it to be a comparator, then I recommend to use a comparator, not an OPAMP.
My recommendation: add a useful feedack .. to make it a clean amplifer with the function of a "regulator"

* Filter: To cure the problems caused by U2. I personaly would remove the root cause ...
.. but for a "bad fix" ... one could add a diode across the R to speed up discharge of the C.
But with the U2 problems ...it probably does not work at all.

Another "bad fix" is to use dual supply and short circuit negative voltages with a diode across the C.

is there an alternate way to ensure the output of U2 never goes negative

Don´t focus on the output of U2 ... it can be anything. You need to focus on the input of the attenuator. This node must not go negative (according your infromation)

I’ve also been trying to understand the practical limit of how small a voltage U1 can amplify before its buried in noise.

The noise of U1 may be in the microvolts. But isn´t the job of the circuit to (precisely) regulate the output voltage? If you think about "errors" like noise in the microvolts ... you also need to have a look at the error of your "rectifier". It maybe starts working when the output voltage is at 500,000 microvolts (to stay with the same unit) .. and it will thermally drift by about 2,000 microvolts per °C. If you want to improve precision .. you need to reduce the "worst" errors. Improving U1 noise will have no impact on overall performance.
Also mind the U2 problems...

Assuming the output from the diode detection circuit is noise free,

I´m the wrong person for such unrealistic assumptions. Input amplitude noise, switching noise, thermal drift, filter ripple ... is (I guess) more than a factor of 10,000 higher.
(for sure in detail it depends on VCO, VCO supply, diode type, capacitor type, thermal conditions, PCB layout .....)

noise specs at the bottom of page 4

Do a search for "linear technology design note15" / "dn015f"

and if U1’s offset voltage would have more of an influence

I recommend to first decide your application goals.
Then find out wich part is resonsible for wich errors and in what magnitude. Then focus on the biggest errors.

And "offset" in your circuit causes no error at all, since your reference (pot output) is a "relative value" adjusted by hand.
"Offset" is an accuracy error.
--> "offset drift" is a precision error. This is what matters in your circuit.

Now the circuit produces a stable output at the expected level.

You would be surprised if you do an FFT plot / spectral analysis of output signal.
I don´t know what youre design goals are .. but to me the circuit give a lot options for improvement.

Klaus

 

[Hawaslsh]

* It has no feedback

I saw your note about trying to make U2 more of a regulator topology. Clearly I am misunderstanding something. In U2's case, the output isn't a fixed relationship to the reference as would be in a regulator.

This is my misunderstanding again, as I currently have the circuit, isn't U2's positive input fed back though through all the other circuity? Instead of a simple resistive divider as in the case of a regulator circuit, I have the detector/ amplifier path. The gain of U2 is by some crazy function of the [Variable attenuator response * Diode detector response * gain of U1]?

Thanks for taking the time to help

 

[KlausST]

isn't U2's positive input fed back though through all the other circuity?

The OPA2189 has an open loop gain of 170dB, this is about 300,000,000
Let´s imagine you have a 12V supply .. then (calculating back) it needs a tiny of 38nV of input voltage to make the output switch from one rail to the other.

True, there is a feedback .. that you have marked in red. But it is slow ... far too slow (phase shift) to prevent the OPAMP to go into output saturation.
Thus the OPAMP needs a local feedback to make the whole regulation loop smooth, controlled, precise...

In your case the U2 output extremely over-exaggerates ... it bangs form one extreme to the other .. this is not how you can fast and precisely control a signal.

*****
Just to translate it to a more common regulation loop:
Car speed control. There is a speed setpoint. There is the actual speed. And let´s assume your car´s top speed is 120mph.
Then by a tiny difference of 0.000004 mph your throttle goes from 0% to 100% and back...
And this 0% to 100% happens in the fastest time possible. Your passengers will get seasick.
And so does your output signal. For sure you "artificially" smooth it down with your filter.
Compared to the car it means: you attach a trailer to your car with tons of load .. just to make the ups and downs of your accelaration smoother.
And now (in post#1) you complain that the car needs a long time to get stopped.
Yes... this is quite expectable with a that huge load.
.. or with your filter....

Instead of making it more precise . your filter just makes it slow. It seems to become smoother ... but indeed this makes the regulation loop slow to correct for errors .. it becomes unprecise. It over-exaggerates ... overshoots .. detects it´s too fast .. and over-exaggerates to slow down ....
If my guessing it right .. it may have several "oscillation frequencies" ... the lowest is somewhere in the low Hz region.

Klaus

 

[Hawaslsh]

Just to translate it to a more common regulation loop:

Thus the OPAMP needs a local feedback to make the whole regulation loop smooth, controlled, precise..

Thanks for the explanation, it makes a lot of sense. I didn't know where/how to place local feedback around U2 but I found an ADi note on topic that suggested a feedback capacitor. The article describes the capacitor as an integrator tracking changes in current from U1. It works well and reacts quickly with no need for an output LPF. It is fair to say the topology of U2 is the classic inverting amplifier at AC? Setting the gain of AC signals to (- impedance C_F) / R_F?

Thanks again for the time and help. Much appricated.