Building a Simple Amplifier Circuit With High Slew Rate and Bandwidth

[iqjump123]

Hello all,

I have been building a simple amplifier circuit(diagram attached) that will take in a source voltage and use a resistor divider to divide the voltage down on three different ratios, determined by a manual switch- 1/1, 1/4, and 1/40. It goes through a unity gain configuration through the AD622 (used ad623 on the spice diagram due to lack of availability)
instrumentation amplifier (1.2V/us slew rate and 1MHz bandwidth). I used this component because when I used the INA126 instrumentation amplifier, the low slew rate of the chip started to corrupt the output wave at high frequencies (~100kHz), displaying sharp sawtooth waves when input with an AC wave. I need signals that will give response up to 100khz or so.

However, the frequency response of the circuit reveals that the gain sharply drops from around 10000Hz to about half in 100000Hz range. The chip product itself advertises a 1MHz bandwidth at Gain of 1, but it is obviously not doing that- and the phase response is non linear. As shown in the attached circuit diagram, the chip is being powered on two batteries, which was opened new at the same time. Also, the gain setting is done by a manual switch.

All I need is a 1 gain amplifier, just as long as it can supply for me a noise-free high slew rate high bandwidth. This chip was exactly advertised to do so, yet it is not giving me the result (Tried it with different chip of the same kind).

The output of the circuit is being fed to a data acquisition card. Although advertised to be high impedance, is there a chance that the stray capacitance coming from the acquisition card be the issue? Not sure.

Any help on this matter will be appreciated.

thanks very much!

 

[skeptic2]

Can you sweep the frequency from 10 Hz to 1 MHz and show us the plot?

How much amplitude out do you need because if your gain is slew rate limited, lowing your input amplitude may get you better frequency response.

 

[iqjump123]

Can you sweep the frequency from 10 Hz to 1 MHz and show us the plot?

How much amplitude out do you need because if your gain is slew rate limited, lowing your input amplitude may get you better frequency response.

Thanks for the reply skeptic2.

I am not at my circuit at the moment, but I will attach a frequency response graph as soon as I get a chance.

I am doing some wavelet analysis and the waves that I will have to examine will vary from large to small waves in amplitudes. anything around +- 5 V will suffice.

I actually tried driving the circuit through a small input amplitude, but it still resulted in the same issue in frequency response.

 

[AlephZero]

The chip product itself advertises a 1MHz bandwidth at Gain of 1, but it is obviously not doing that- and the phase response is non linear.

I noticed the data sheet of the AD623 gives the "typical" unity bandwith as 800 kHz, with a supply of +/-5V (lower than your circuit) and an output impedance of 10k (10 times higher than your circuit).

It could be simply that you are pushing it too hard, and the SPICE model "knows" enough to show you that.

I didn't check if the 622 has a different spec from the 623.

 

[iqjump123]

I noticed the data sheet of the AD623 gives the "typical" unity bandwith as 800 kHz, with a supply of +/-5V (lower than your circuit) and an output impedance of 10k (10 times higher than your circuit).

It could be simply that you are pushing it too hard, and the SPICE model "knows" enough to show you that.

I didn't check if the 622 has a different spec from the 623.

Hey AlephZero, thanks for the reply.

The AD622 has the typical unity bandwidth as 1MHz, with a supply of up to +/- 15V. I don't know where I can find the information for output impedance measurement in the datasheet, however.

The problem with the decreasing gain still persists, and from debugging, it seems that the chip itself works very well- therefore, since there are no other added circuitry, the decrease in gain is in fact something from the initial resistor divider set up. I was thinking of putting a buffer stage between the divider circuit output and the AD622 amplifier (using the lt1010 or a typical ua741 opamp and setting it to unity) so that it might isolate the impedances.

Also, if I would like to measure the input and output impedance of the circuit stages, would something like the method shown in the link below
http://www.zen22142.zen.co.uk/Theory/inzoz.htm
be enough? This seems a bit inaccurate and too simple- it is because I have seen RLC-based impedance measurement kits that go for thousands of dollars. Would the method shown in the website be good enough to measure the impedances?

thanks for all your help in advance.

 

[NascentOxygen]

All I need is a 1 gain amplifier, just as long as it can supply for me a noise-free high slew rate high bandwidth.

Your schematic shows an open loop amplifier, there is no unity feedback.

Also, what is that 27.7 volts at the op amp output? Is that DC?

Can you confirm that at 1 MHz the output is still (as best you can judge by eye) nicely sinusoidal?

This is all a simulation is it? You are testing with the low-Z voltage source connected directly to the op amp input?

 

[iqjump123]

Your schematic shows an open loop amplifier, there is no unity feedback.

Also, what is that 27.7 volts at the op amp output? Is that DC?

Can you confirm that at 1 MHz the output is still (as best you can judge by eye) nicely sinusoidal?

This is all a simulation is it? You are testing with the low-Z voltage source connected directly to the op amp input?

Hey nascentOxygen, thanks for the prompt reply.

First of all, I put no resistors on the gain inputs of the AD622, because the amplifier datasheet states that with no resistors on it, the chip gives a default gain value of 1.

The diagram above was to visually show the current circuit I am running on. The input was set as so so that I can go through a frequency sweep analysis on pspice.

In the actual circuit that I built, I connect the output of the waveform generator to the resistor divider and measure the output from the AD622 chip to an oscilloscope. I am doing this for varying frequencies. However, I can't get to more than 300KHz by manual ramping up of the frequency. One way or the other, the drop in gain is obvious from roughly the 10KHZ point.

I don't know what the output of the opamp is showing in the pspice diagram and what it signifies- measuring it experimentally gives values a lot smaller than that. What do you mean by a low Z source, however? I feed the input of a waveform generator directly to the resistor divider in my actual circuit- is this a problem?

Also, can anybody share an opinion on attaching a buffer chip? Would this help at all, or just more real estate and power consuming ?

Thanks again.

 

[NascentOxygen]

In the actual circuit that I built, I connect the output of the waveform generator to the resistor divider and measure the output from the AD622 chip to an oscilloscope.

Right. But you don't need that resistive divider for testing, because the circuit has only a gain of one. The signal generator can supply 1 or 2 volts, and that can be be fed directly to the input of instrumentation amp. Also, you don't need a load at the amplifier output during testing. So if you remove those both, then the amplifier will have the best opportunity to work at its widest bandwidth.

I am doing this for varying frequencies. However, I can't get to more than 300KHz by manual ramping up of the frequency. One way or the other, the drop in gain is obvious from roughly the 10KHZ point.

Yes, I understand that. We can eliminate possible causes one by one.

a low Z source, however? I feed the input of a waveform generator directly to the resistor divider in my actual circuit- is this a problem?

A low impedance source. The divider shouldn't be a problem, but if you can remove it then we can be certain that it is not contributing problems. It is not needed for testing, because you don't need to reduce the input amplitude.

Also, can anybody share an opinion on attaching a buffer chip?

I don't think you are not up to that yet.

So the schematic you attached showing 27.7 volts at the output is not the circuit you are using? That is not a simulation of what you have built? Even so, it's puzzling how such a value could come about. It doesn't inspire confidence in the simulation package.

 

[iqjump123]

Thanks NascentOxygen! I think I need to clarify on a couple things from my end.

Right. But you don't need that resistive divider for testing, because the circuit has only a gain of one. The signal generator can supply 1 or 2 volts, and that can be be fed directly to the input of instrumentation amp. Also, you don't need a load at the amplifier output during testing. So if you remove those both, then the amplifier will have the best opportunity to work at its widest bandwidth.

The test involving isolating the chip to test its performance was already done; with input generator hooked to the input directly and the output of the chip directly measured by the oscilloscope(so no resistor divider). This test resulted in good bandwidth output, up to 1MHz, as advertised.

This is why I am thinking maybe the resistor divider connected to the AD622 is somehow contributing to the impedance to the point where it is causing problems to the overall circuit.
The bypass capacitors I attached does help a little, but it doesn't help it entirely.

So the schematic you attached showing 27.7 volts at the output is not the circuit you are using? That is not a simulation of what you have built? Even so, it's puzzling how such a value could come about. It doesn't inspire confidence in the simulation package.

Ah you won't have to worry about the output voltage you see there- pspice has a separate output voltage simulator that I get the resultant values from- the values you see there are not relevant to the values I need to analyze the circuit at the moment.

Imagine the circuit diagram that I attached as a graphical representation (in terms of the resistor divider and the the AD622 chip connections and caps(used AD623 chip since AD622 part is not available), but not as a "chip for chip" representation, so to say.

Also, the load resistor is not part of the actual design- it just needs a load resistor for simulation to work(because if not the simulator sees the output straight to ground.

Thanks for your continued help!

 

[NascentOxygen]

The test involving isolating the chip to test its performance was already done; with input generator hooked to the input directly and the output of the chip directly measured by the oscilloscope(so no resistor divider). This test resulted in good bandwidth output, up to 1MHz, as advertised.

We should expect an instrumentation amplifier to be a good performer. So I'm a bit worried by two specs that are lacking on your data sheet. No details of output impedance, yet surely that is relevant, and no details of effective input capacitance, and again that is relevant to high frequency response.

It's possible that the input capacitance is forming, together with your potential divider, a first-order lowpass filter. So your intuition may be right. Rather than use a buffer, it should be easy enough to change your potential divider into a lower impedance divider, i.e., just make the resistors in the divider smaller by a factor of 100 or so. Instead of using, say, a 22k and 330k divider, make them 220 Ω and 3.3kΩ (providing the waveform generator can drive this lower impedance).

This change is easy enough to test; you have almost confirmed it as the cause.