Help with Signal Conditioner

[Heher]

<Cross posted in Comm & Signal forum>

All,

I am a newbie Electronics Engineer and am working on my first circuit. It is a simple signal conditioner which is just designed to take an input signal (square or sin wave, 50mV to 10V & 10Hz to 20kHz), clean it up and then output it via a RS-485 driver for dual logic levels.

Eveverything seems to be working just fine as far as the circuit goes except for one part, the Hysteresis filter at the end of the signal path (right before the last component, the RS-485 driver).

Initially the signal is passed through a 10Hz high pass filter and then to an AD620 Instrument amp which is set to amplify at 1000 gain, (because this circuit must accept input from 50mV to 10V) then through a 40kHz low pass Bessel 2-pole filter. The signal coming out of the Bessel filter is then offset 2.5 volts with an offset follower and offset summer. The signal is then passed on to a D1N4148 voltage limiter to establish the signal from ~0 to ~5V. At this point, the signal is very clean and looks great within the required frequency range.

The last stage after the voltage limiter is a hysteresis filter. This is where things go bad. I modeled the hysteresis filter (Schmitt trigger) out of p. 231, The Art of Electronics 2nd ed. I am using a LM2904 op-amp with 5V Vcc and V+, V- is GND. All resistors are 10k (I was able to get the low switch point at ~1.7V and the top of the band at ~3.0V). A 33pF speedup capacitor is used across the feedback resistor. A pullup resistor shouldn't be needed because of the op-amp used here.

The filter works fine as I ramp the frequency up to around 3kHz which is where the output changes from a nice square wave to something with a tiny peak at the top and no pulse width and as I continue to increase the frequency the amplitude of the output finally sinks down to almost nothing above 20kHz. This is causing the RS-485 driver to only detect a HIGH for a very small period of time and thus I am getting a pulse waveform out instead of a nice filtered square wave. The output wave (@ 20kHz) has a pulse width of around 5us which should obviously be around 10 times that for a 20kHz wave.

Thanks for any help and I apologize for the lengthy post.

 

[Lewish]

Go to

http://www.national.com

and download the datasheet for the LM2904. Then look on page 7 at the graph on the left side labeled &quot;Large Signal Frequency Response&quot;. That should answer your questions.

 

[Heher]

DOH!! Proves I'm a newbie! Thanks much for pointing out the obvious to me. Learning every day.

Thanks again Lewish

I guess its back to componenet selection now....

 

[brennaj]

I take it you just want to take the input signal and square it up? May be you should use a comparator instead of an op amp at the output - it will switch a lot faster.

Another option is to AC couple your signal to a bias voltage which is adjusted to the threshold voltage of a SChmitt Trigger (eg MC14584BCP) or any high speed comparator. Although it might be a bit rough working down to 50mV.

 

[Heher]

The main point of this circuit is to take a noisy signal (sine or square wave only) from a noisy source and varying amplitude, clean out the noise and convert it to dual TTL logic levels for output to another device. The Schmitt Trigger part of this circuit is merely used to clean the ripple off the input signals HIGH and LOW and prevent false switching due to that ripple.

Any idea off the top of your head what op-amp series has relatively flat frequency response from 10Hz to 40kHz? Switching speed doesn't appear to be much of an issue at this juncture and I have to keep the op-amp configuration because this circuit has already been laid down on a production level PCB by the old engineer they fired before they brought me in. So basically management is asking me to try to make it work with just switching components, R, Cs, etc. so they can just take the existing models they have in stock and remove the bad components and solder on the new components. This means the current packaging type of SOIC must stay as well along with everything else, etc. etc. Since this particular IC has two op-amps on it and one is used for a different part of the circuit, this is the reason I must stick with an op-amp.

Thanks for all the help!

 

[Lewish]

TL071, TL081, TL072 (dual) TL082 (dual) are good cheap FET input generic Opamps. I use a lot of TL082s in designs. They work well in filters and other general applications.
Last I checked, they were about $0.30 each in 1000 piece quantity.

 

[Heher]

I just have a general question but I decided to post the reply on this thread because it involves this circuit specifically.

This is probably also a silly question but this is the first analog circuit and really even the only one with op-amps that ive worked on since my 1st year in school.

Question is....why does my Hysteresis filter (using the LM2904) work well from 1kHz to 20kHz at 6.1 Vpp with a 0V DC offset, but when I increase the DC offset (even a few hundred mV up to 3V) on the input signal, the output signal goes nuts?

I know this is probably as obvious as it gets and probably why I'm missing the answer to this simple question but could someone slap some sense into me on this one?

Thanks a bunch guys.

 

[Lewish]

Hello Heher, ah the joys of the first job. Don't feel too bad, I have seen veterans, even me, make similar mistakes.

First question, you state in your first post that the LM2904 is powered by +5V and GND. But in your last post, you say the input signal is 6.1 Vpp. Do you see anything wrong with this? Say maybe the fact that your input is exceeding your Opamp's voltage rail?

 

[brennaj]

Good point Lewish. When your input signal exceeds the rail on the opamp, the opamp will tend to saturate low. Also I read you had a gain of 1000. The Dc offset gets amplified as well.

 

[Lewish]

Hi brennaj, yes, sometimes. Sometimes it will oscillate.