Intermediate Frequency Amplifier

[James Kutia]

Hi all, I am working on designing an IF amplifier using opamps and just need some guidance as to whether I am on the right track as this is not one of my areas of expertise.

The input signal is from a frequency mixer, fed by frequencies of approximately 2.45GHz. I am only interested in measuring the frequency difference, which will be between 5 KHz and 6 KHz. The goal of this amp is to receive as small a signal as possible without distorting it so much that the frequency can't be measured so noise performance is probably the biggest issue, but please correct me if I am wrong.

My first step was to create a low pass filter with a cutoff freq of approximately 10 KHz to eliminate the RF signal, this can be seen in the schematic.

Secondly, I have used 3 cascaded LM833 opamps in inverting amp configuration to amplify the signal. Note that I have used 50k pots as the input resistors and adjusted them to get the best signal.

The output will eventually be connected to a DAQ card for Fourier analysis etc on a PC.

I have tested this circuit with the output connected to a scope and it can measure the 5KHz waveform with an input of -60dBm, any lower and it becomes too noisy.

My main question is what can I do to still get a measurable output with a smaller input signal?
- Will a dual supply configuration work better?
- The mixer output impedence is 50 ohm. Do I need a buffer stage before the LP filter to provide 50 ohm matching input impedence to maxmise power transfer?
- Does using 3 stages introduce too much noise - Am I better off using just 1 or 2 amplifier stages?

Any ideas would be much appreciated thanks!

 

[Audioguru]

The junction of R4 and R5 will produce much less noise if it has a filtering capacitor to ground.

You do not need a buffer at the input of the simple filter. This circuit uses signal voltage, not signal power. If you match impedances then the signal level drops to half, which will double the noise level.

 

[James Kutia]

The junction of R4 and R5 will produce much less noise if it has a filtering capacitor to ground.

You do not need a buffer at the input of the simple filter. This circuit uses signal voltage, not signal power. If you match impedances then the signal level drops to half, which will double the noise level.

Ah sorry I forgot to include that in the schematic. I have a 10u cap to ground at that junction. And a 10u and 100n between V+ and ground.

Thanks!

 

[jiripolivka]

I have designed a similar video amp for a Doppler radar. I would suggest to separate the Vcc/2 circuits for each stage to prevent unwanted feedback, and if possible, use a dual power supply. Filtering is essential in such high-gain amplifiers. Check mixer specification for LO/IF leakage, maybe filtering RF is not needed. Use 10 uF nd 0.1 uF parallel capacitors on all DC lines and enclose the amplifier in a metal enclosure to prevent external-signal interference.

 

[Audioguru]

I have designed a similar video amp for a Doppler radar. I would suggest to separate the Vcc/2 circuits for each stage to prevent unwanted feedback.

One Vcc/2 divider will bias hundreds of opamps. It is an INPUT, not an OUTPUT so there is no feedback.

 

[jiripolivka]

Do as you want. I have found this common line a problem in my design.

 

[Audioguru]

The amplifier in this thread has its last opamp with its (+) input not connected so it probably does not work. If that input is biased then the amplifier has a voltage gain of 50 x 50 x 50= 125000 so of course it might oscillate and is very noisy.

 

[James Kutia]

The amplifier in this thread has its last opamp with its (+) input not connected so it probably does not work. If that input is biased then the amplifier has a voltage gain of 50 x 50 x 50= 125000 so of course it might oscillate and is very noisy.

Sorry, it is connected, I was a bit slack with the schematic. You're right, when I adjust the pots so the input resistor is too low they do begin to oscillate. Am I better off having a small gain for the first stage then increasing the gain in subsequent stages? Or will the noise introduction be the same regardless?

How about having a secondary band pass filter after the last amplifier to remove the noise?

Cheers

 

[chuckey]

For the best sig/noise ratio you need as much gain as close to the front end as possible. The reason is that suppose an op-amp is generating 1mV of thermal noise (yes, OK its too much, but bear with me) and your wanted signal is 3 mV, then amplifying them by X 3 will give 3mV of noise and 9 mV of signal. if the next stage adds its own 1mV of noise and has X 100 gain then at the output you will have (3 +1) X 100 = 400 mV of noise and 900 mV of signal. Doing it the other way around gives 1 X 100 = 100 mV of noise add in the extra noise = 101 mV X3 = 303 mV noise at the outpot, while the signal goes 3 mV X 100 = 300 X 3 for the second opamp = 900 mV. So putting all the gain earlier on means you can ignore the noise from subsequent stages . the calculation shown is only an example, noise is added power wise not linearly but the principle holds.
So going back to your U1, its in band gain (low frequencies) is unity (3k in series, 3 K in feedback loop), change this and make that chip work for a living , say a gain of 100 (1000) times or so. Also the series 3K will generate a bit of noise. Its best to preceed the op amp , with a common gate FET amplifier, this will make the input impedance sensiblely close to 50 ohms and put in X 20 gain or so with excellent linearity, this could then feed your filtered opamp circuit.
Frank

 

[James Kutia]

For the best sig/noise ratio you need as much gain as close to the front end as possible. The reason is that suppose an op-amp is generating 1mV of thermal noise (yes, OK its too much, but bear with me) and your wanted signal is 3 mV, then amplifying them by X 3 will give 3mV of noise and 9 mV of signal. if the next stage adds its own 1mV of noise and has X 100 gain then at the output you will have (3 +1) X 100 = 400 mV of noise and 900 mV of signal. Doing it the other way around gives 1 X 100 = 100 mV of noise add in the extra noise = 101 mV X3 = 303 mV noise at the outpot, while the signal goes 3 mV X 100 = 300 X 3 for the second opamp = 900 mV. So putting all the gain earlier on means you can ignore the noise from subsequent stages . the calculation shown is only an example, noise is added power wise not linearly but the principle holds.
So going back to your U1, its in band gain (low frequencies) is unity (3k in series, 3 K in feedback loop), change this and make that chip work for a living , say a gain of 100 (1000) times or so. Also the series 3K will generate a bit of noise. Its best to preceed the op amp , with a common gate FET amplifier, this will make the input impedance sensiblely close to 50 ohms and put in X 20 gain or so with excellent linearity, this could then feed your filtered opamp circuit.
Frank

Thank you for the explanation Frank, my thought process was similar to what you've just said but I wasn't sure enough to try it,

James