It looks like using R8 for a filter doesn't work very well. Meh. It was good exercise figuring that out.

 

Have you told us what your required gain range is?

Not really. Unfortunately, I don't have a really clear-cut idea of what the gain needs to be. I'm using this circuit to amplify 0.1 mV - 10 mV signals picked up from a ferrite-core antenna and output these to a speaker. The whole system will be used to locate ground-loops in an acoustically loud (hence the need for a loud speaker) lab environment. Because of the application, the speaker may need to have the gain turned up while there is only noise being picked up, hence my need to eliminate all noise but the desired signal.

It looks like using R8 for a filter doesn't work very well. Meh. It was good exercise figuring that out.

Right. It's just good to hear that I wasn't totally missing something. I'll go ahead and try to find another way around it. I'll see if I can turn the first stage into a tight band-pass filter and just add some tighter hi-passes in the circuit. Thanks for the help anyway. Please don't hesitate to let me know if you think of something.

 

Are you using a fixed frequency in the transmitter? IIRC, you said it was near 7kHz.

 

Yes; A fixed 7kHz with amplitude modulation on the transmitter.

 

Yes; A fixed 7kHz with amplitude modulation on the transmitter.

What is the modulation frequency? Is it sinusoidal?

 

The modulation waveform is a square wave going between 0% and 100% at 1 Hz or 0.5 Hz. I've basically been assuming that I have constant 7kHz since the modulation frequency is so, so low compared to the audio.

 

The modulation waveform is a square wave going between 0% and 100% at 1 Hz or 0.5 Hz. I've basically been assuming that I have constant 7kHz since the modulation frequency is so, so low compared to the audio.

I'm curious why you chose 7kHz, but here's an idea that I think will get rid of most of your noise:

There are chips available called switched-capacitor filters. One of them is LTC1060. The cool thing about these filters is, they are clocked. The cutoff frequency, or center frequency in the case of a bandpass filter (BPF), tracks the clock frequency. Change the clock frequency, and the center frequency tracks it. The ratio of clock/center freq. is 100 or 50, your choice.

So here is how you would use it: Generate a 700kHz clock (clk), divide it by 100 to get 7kHz, apply your modulation, and apply that output to the transmitter.
Make your receiver amplifier with op amps, using simple RC low pass and high pass elements in it to get rid of low frequency and high frequency noise, as you have been attempting to do. Apply the amplifier output to the switched-capacitor BPF, setting the Q to some reasonably high value, e.g., 10. Use the 700kHz clk as the BPF clock input.
You will have a very narrow band output that automatically tracks the clock frequency. If you decide that you want to try 5kHz, change Fclk to 500kHz, and shazam!, your filter will be centered at 5kHz.

BTW, I referenced LTC1060. LTC1059 may be a better choice, but the 1060 datasheet has more applications info. Also, other manufacturers make switched-capacitor filters.

 

Hey Ron,

Honestly, I chose 7kHz because it couples well through the transmitter and it is annoyingly high-pitched enough to easily hear from the speaker.

I do think that I will try to use the LTC1059 (which seems to simply be a high-performance version of the LTC1060, correct?). I'm thinking of putting this filter after U2 and before U4: That way, the filter does not have to output a large-voltage signal. Or would you recommend putting it after the LPF at the input? I could see advantages to either. My main worry is that the 2.5kHz is being generated somewhere else in the circuit and if I put the filter too close to the input, I won't filter it out well.

I have two more questions about the implementation of the LTC1059, if you don't mind:

1) I don't currently have a clock in my system. My 7kHz is simply given to me by a function generator. Am I alright to go head and just use a CMOS 555 timer to generate 700 kHz? I have done a bit of research and it looks like trashy 555's can have problems above 100's of kHz, but the CMOS 555 is supposed to be better. Any comments?

2) Page 17 of the LTC1060's datasheet shows a 4th-order band-pass filter that I plan to emulate -- or something like it, maybe configured so that the FClk:FCenter is closer to 1:10, rather than 1:200 as shown. In such a configuration, however, I do not understand the gain relationship for the external op-amp. Is the gain RG / (RL2 // RH2)? The input series resistance looks difficult to calculate there... Any ideas? I may just need to slam my head on my desk for a while and figure it out...

Thanks, Ron; Great idea!

 

Hey Ron,

Honestly, I chose 7kHz because it couples well through the transmitter and it is annoyingly high-pitched enough to easily hear from the speaker.

I do think that I will try to use the LTC1059 (which seems to simply be a high-performance version of the LTC1060, correct?). I'm thinking of putting this filter after U2 and before U4: That way, the filter does not have to output a large-voltage signal. Or would you recommend putting it after the LPF at the input? I could see advantages to either. My main worry is that the 2.5kHz is being generated somewhere else in the circuit and if I put the filter too close to the input, I won't filter it out well.

I have two more questions about the implementation of the LTC1059, if you don't mind:

1) I don't currently have a clock in my system. My 7kHz is simply given to me by a function generator. Am I alright to go head and just use a CMOS 555 timer to generate 700 kHz? I have done a bit of research and it looks like trashy 555's can have problems above 100's of kHz, but the CMOS 555 is supposed to be better. Any comments?

2) Page 17 of the LTC1060's datasheet shows a 4th-order band-pass filter that I plan to emulate -- or something like it, maybe configured so that the FClk:FCenter is closer to 1:10, rather than 1:200 as shown. In such a configuration, however, I do not understand the gain relationship for the external op-amp. Is the gain RG / (RL2 // RH2)? The input series resistance looks difficult to calculate there... Any ideas? I may just need to slam my head on my desk for a while and figure it out...

Thanks, Ron; Great idea!

AFAIK, you have two choices for Fclk: 50x or 100x Fcenter. 10x is not an option.
Secondly, the BIG advantage of using a switched capacitor filter, and a 100x freq divider for your signal, is that the center frequency will always be the same frequency as your transmitted signal. With a high-Q BPF, you cannot afford drift between the signal frequency and the center frequency. Using a function generator for the signal, and a 555 for the clock, is a BAD idea. It negates the entire purpose of the switched-capacitor filter.
WHen you start with a 700kHz clock, and divide it by 100 to get your signal frequency, you are guaranteed that the center freq will be equal to your signal freq.
I don't see why you would need a 4th order filter. You can get a very narrow band filter with a single 2nd order chip, and with a 1Hz modulation frequency, your sidebands are so close to the carrier that you should not need a flat top on the BPF response curve. You will probably get slow output risetimes on the modulation envelope, and some delay, but with a fox and hound (sounds to me like that is what you are trying to make), those should not be an issue.
BTW, couldn't you just buy one, or is this a homework problem?

I don't want to comment on where to put the BPF in the amplifier chain, because I wouldn't do it the way you did.

 

Hey Ron,

It seems like the Fox and Hound Signal Tracer is for finding live wires. I'm trying to make something more like a ground loop detector. For that reason, I don't have a physical connection between the transmitter and receiver and cannot directly share a clock. I haven't made my own transmitter because this is intended to be a small project (if there is such a thing). This is a project for work; Not a homework problem.

Thanks for the brilliant chip suggestion. I do think I'll go ahead and use that or try to design my own tight band-pass filter in my amplifier chain.

 

Hey Ron,

It seems like the Fox and Hound Signal Tracer is for finding live wires. I'm trying to make something more like a ground loop detector. For that reason, I don't have a physical connection between the transmitter and receiver and cannot directly share a clock. I haven't made my own transmitter because this is intended to be a small project (if there is such a thing). This is a project for work; Not a homework problem.

Thanks for the brilliant chip suggestion. I do think I'll go ahead and use that or try to design my own tight band-pass filter in my amplifier chain.

You might want to consider a Sallen & Key bandpass filter.

If you need extreme stability, you can use the switched-capacitor filter, and generate the clock for it with a crystal-controlled oscillator.