One or two op-amps for a high-gain amp

[f95toli]

I need to design a "front end" that will connect a power sensing diode to some other measurement instrumentation.

• The gain should be abour x2500, the input signal will be of the order of 1mV
• I need to be able to vary the input impedance from 100 ohms to a few kOhm by changing components
• low-drift since it will be part of a stabilisation circuit
• The PCB desing should be able to handle bandwidths of between 100 Hz to 10kHz, depending on component values
• Reasonably low noise
• The complete circuit also needs to have offset adjustment
• Cost is not an issue (this is for work, and my time is much more expensive than the components)

There is nothing "special" about this circuit and the low BW means that a single non-inverting op-amp (with resistor to ground to change the input impedance) for the gain and then second one for adjusting the offset (summing the signal with the voltage from a precision voltage reference) would in principle work since the GBW is quite low even for x2500 gain.

However, I've seen a few people recommend a two-state two op-amp circuit whenever using gains as high as this, mainly because of problems with oscillations. I am aware that a cascade is sometimes used to increase the input impedance, but that is not an issue here.

Hence, one alternative would be to cascade two inverting op-amps and "divide" the gain between them (say x250 in the first stage), which would also make it easy to change the input impedance.

Does anyone have any practical experienc with this? Also, if using two op-amps could I use a dual op-amp such as a OPA2188 or something similar?

 

[Carl Pugh]

Does front end have to have bandwidth of 100 Hz to 10 kHz or DC to 10 kHz?
This will make a lot of difference in front end design.

Would it be better to have two op-amps with first op-amp fixed gain and second op-amp variable gain?
This would reduce the noise problem.

The OPA2188 should work OK in this application.

 

[f95toli]

It is DC to 100 Hz, or DC to 10 kHz

Sorry if I was unclear.

It is is basically a DC amplifier, but it need to be able respond to slow(ish) changes in the signal from the detector. It is going in front of a PI regulator which will have a slow time-constant, so there is no need for wide BW. The application is basically automatic gain control which will be used to keep the power level constant in an experiment.

The OPA2188 looks attractive, I could even use a second OPA2188 for the offset adjustment, one op-amp could be used for summing and the other as an inverter. The latter would allow me to generate a negative voltage from the (positive) voltage reference, making bipolar offset adjustment possible. Although I guess a dual chopper stabilized op-amp would be better.

 

[berkeman]

It is DC to 100 Hz, or DC to 10 kHz

Sorry if I was unclear.

It is is basically a DC amplifier, but it need to be able respond to slow(ish) changes in the signal from the detector. It is going in front of a PI regulator which will have a slow time-constant, so there is no need for wide BW. The application is basically automatic gain control which will be used to keep the power level constant in an experiment.

The OPA2188 looks attractive, I could even use a second OPA2188 for the offset adjustment, one op-amp could be used for summing and the other as an inverter. The latter would allow me to generate a negative voltage from the (positive) voltage reference, making bipolar offset adjustment possible. Although I guess a dual chopper stabilized op-amp would be better.

Sorry, I'm not tracking very well on this one. What is a "power sensing diode"? And what is the slow(ish) signal? Is 10kHz slow(ish)? It certainly is in some contexts, but if this component is somehow sensing the thermal changes in a linear voltage regulator (is that what you meant by Pi regulator?), then the bandwidth could just be a few Hz, it would seem.

Could you maybe post a piece of the schematic to show what the regulator and diode are doing?

 

[f95toli]

Sorry, I'm not tracking very well on this one. What is a "power sensing diode"? And what is the slow(ish) signal? Is 10kHz slow(ish)? It certainly is in some contexts, but if this component is somehow sensing the thermal changes in a linear voltage regulator (is that what you meant by Pi regulator?), then the bandwidth could just be a few Hz, it would seem.

Could you maybe post a piece of the schematic to show what the regulator and diode are doing?

The diode is a square-law detector that outputs a negative voltage proportional to the power of an incoming microwave signal. The optimum operating point for the type of diode we are using is to have the power be around -30 dBm, which results in the output dc voltage being about -1mV. The idea of the automatic gain control is to use a PI (proportional-integrator) controller to regulate the power in such a way that the average microwave signal stays at that level*

For the gain control we are implementing now we will only need to compensate for changes that happen at a few tens of Hz at most. However, in the future we might want to use the same design to stabilize the microwave power in another part of the setup, and there we need higher BW. 10 kHz is about the BW of our measurment.

The whole setup/circuit is quite complicated so I'd rather not go into that here (PM me if you want a link to an article we published a year ago that describes the experiment)However, from a practical point of view all I need is to make a preamplifier with a gain of about x2500 since this will make the signal about -2.5V, I also need to be able to offset the output of the preamp to translate signal into the the right range for the PI controller (which is +-1 V or 0-3.3V depending on which PI controller we end up using).
I've built a few amplfiers, but I've never had to build anything with a gain this high that also needs to be stable. I am therefore not sure whether go for 1 op-amp or 2, *The measurement signal we are looking it is at about 1MHz (we have a 6 GHz MW signal phase modulated at 1 MHz), so the signal from the diode will be split with the DC going into tthe regulator and the 1 MHz signal going into other equipment).
Also, the signal from the PI controller will be going into a variable attentuator.

 

[Carl Pugh]

Some things that should be taken into consideration before deciding on the circuit are:
What accuracy is required?
What sort of stability is required?
What frequency response is required?
___Is a low pass filter required
Is there a phase shift requirement?
Can the output signal be inverted 180 degrees or does it have to be in phase?
Operating voltage?
___+/- 5 VDC supplied by other equipment
___115 V AC
___Wall wart
Is a filter to keep RF out required?
Is small physical size required?
Is low power desired?
Does OPA2188 have to be used?
Can through hole components be used?
Is this one of a kind or do you plan on making 100’s of them?
Is ease of making changes/adjustments important?
___Gain
___Bandwidth
___Offset
___Input impedance
Are there other requirements?
___Lead free
___Operating temperature
___Does you company have an approved parts list

Sounds like a fun job
Good Luck
Carl

 

[Mike_In_Plano]

To get consistent performance at 10kHz, you'll need at least two, wide bandwidth op amps. Even though they have tremendous gain at DC, their open loop gain will be much lower at 10kHz. I'd suggest a part with at least 5 or 10 MHz gain bandwidth.

As for the noise, 1 mV is actually a fairly large signal for a 10kHz bandwidth. Vn-wise, there are plenty of op amps under 20nv/rtHz that will be fine.

It also sounds like you have a fairly low input impedance, 2k-ohm or less. That being the case, you can get away with most instrument quality bipolar op amps. This will do wonders to keep your output stable.

This doesn't sound like a hard design for stability as long as you have a ground plane. I'd just make each stage a gain of 50 and use two separate single channel op amps. Bypass each with your Plain-Jane .1uF bypass on each supply. If you really want to be safe, place 100 ohm decoupling resistors in series with the supplies to the first op amp. That will help ensure that the first op amp doesn't communicate back through the power lines.

Since your probably working with RF, you want to be VERY careful to decouple any RF that may leak back through the detector to the op amp. This will induce offset and have you chasing your tail. For this purpose, I typically put in a T filter on all lines to RF areas.

One last word of advice. Many good instruments are ruined when the op amp goes to drive a capacitive load or switching ADC. The easiest fix is to place a series resistance at the output of the op amp before running it to the cable. Typically, a 47 ohm will help with stability, and be small enough that your high impedance ADC won't interact with it.

Best Luck,

Mike

 

[f95toli]

Thanks for the replies

I think I will just go for two cascades single op-amps on the input stage(with x50 gain in each), maybe LT1028 or something similar (I have an article where they've tested different op-amps as amplifiers for detector diodes, and LT1028 is one of them). Drift will hopefully not be an issue.
This also makes things a bit easier for me, since I have some experience building simple op-amp based amplifiers, i.e. I know how to decouple them etc. Although I would have been fun to try something new.

However, I now know that one "shouldn't" use a dual op-amp instad of two single op-amps if one want to cascade two op-amps but I still don't know WHY(?)
I did a quick search, and all that came up was the rather obvious problem of thermal management for high-power applications of dual op-amp. I guess another issue could be decoupling etc. since the two ampifiers share voltage rails. But on the other hand one ends up with a compact design which is usually a good thing froman EMI point of view.

Does anyone else have any thoughts about this? Not for my application in particular but just in general?