2
\$\begingroup\$

I've been working on an at home project recently that requires the use of a speech amplifier. My ultimate goal is to perform signal processing via a dsPIC on speech signals to identify key words or at the very least, detect frequency of speech. I've taken a break on the signal processing to work on the speech amplifier, but I have run into some difficulties.

My experience in hardware is limited to classes and a few simple at home projects, so I am not very familiar will the non-idealities of circuitry. I'm wanting to make a single supply amplifier using a LM318. The circuit would ideally be powered by 5V (cutting it close to the LM318's supply range limits), and would output an amplified speech signal over the entire voltage range to an ADC. In the non-inverting amplifier, I'm also wanting to include a band pass filter to keep DC amplification at unity and to prevent aliasing at the dsPIC ADC. I've modeled the microphone with a 2.5V biased voltage source with an impedance of 2.2kohms (really it is a microphone sandwiched by 2.2kohm resistors at power and ground). Not included are decoupling capacitors (any bonus advice on how to choose values for those would be appreciated).

enter image description here

I'm using TINA as a simulator. The AC transfer function of the circuit is what I expected with a pass-band from roughly 10-5000 Hz and a gain of 10 (Stack Exchange will only let me post 2 images).

These transfer characteristics are what I was expecting from this circuit, but it is the transient analysis that gave unexpected results. I input a 500 Hz 0.1 V amplitude sine wave. What I expected out was a 500 Hz 1 V sine wave, but instead, the simulation returned the following.enter image description here

I've tested the amplifier at other frequencies, and it looks like the less gain there is at a frequency, the less distortion there will be. I'm at a loss for what is causing this. Any ideas?

I'm also open for any feedback on any other improvements that could be made to this amplifier.

This is not my field of study in EE anymore, but I definitely want to keep working on projects and learning about hardware. Any recommendations for readings or books on topics like these (non-idealities of circuitry/amplifiers) would be much appreciated.

\$\endgroup\$
2
\$\begingroup\$

The circuit would ideally be powered by 5V (cutting it close to the LM318's supply range limits)

Actually, the LM318's minimum supply is ±5 V (i.e. 10 V), according to its datasheet:

Supply Voltage Range . . . ±5 V to ±20 V

You're 5 V short of the minimum with only a single 5 V supply.

You need the extra supply voltage to allow the output to swing to your desired amplitude. With ±15 V supplies the output is only guaranteed to swing to within 3 V of the supply rails:

enter image description here

Either use a supply that is 10 V or more (and bias the input to half that supply voltage) or use an op amp that can be supplied with only 5 V and can swing as close to the rails as you need for your output signal.

\$\endgroup\$
  • \$\begingroup\$ Thanks. I was wanting to use a 9V battery and step it down to 5V,because that's what the dsPIC is supplied by, but I'll think of something different. Is the 3V within supply rails range fairly unanimous for general purpose op amps? \$\endgroup\$ – Paul Bunyan Mar 16 '17 at 16:39
  • \$\begingroup\$ @PaulBunyan An op amp that does not have a rail-to-rail output can only swing a couple of volts close to the rails. It might be 3 V or 2 V or something similar, but that's the general range. A rail-to-rail output op amp can usually swing within 100 mV of the rails. \$\endgroup\$ – Null Mar 16 '17 at 16:42
  • \$\begingroup\$ if you try rail-to-rail, observe load R must be higher. digikey.com/products/en/integrated-circuits-ics/… \$\endgroup\$ – Sunnyskyguy EE75 Mar 16 '17 at 16:43
  • \$\begingroup\$ @PaulBunyan If you've got a 9 V battery the best solution would probably be to find another op amp that can operate on 9 V or less (and has a low quiescent current, so the battery lasts longer). You can probably power it directly off the 9 V, and then you can bias the input to 4.5 V (half the battery voltage). If your output doesn't need to be amplified too much you can possibly avoid a rail-to-rail output op amp. \$\endgroup\$ – Null Mar 16 '17 at 16:44
2
\$\begingroup\$

5 V is not cutting it close, it is half of the minimum. The minimum specified is +/- 5 V split rail, or 10 V single supply. What you need is a good rail to rail op amp that will give you good output swing at 5 V.

\$\endgroup\$
  • \$\begingroup\$ Thanks. Should've read more carefully. Is it ever smart to trust an op amp to not distort signals near the rails. Should I go for a rail to rail op amp powered at 5 V, or would it be much better to increase the supply voltage? \$\endgroup\$ – Paul Bunyan Mar 16 '17 at 16:37
  • \$\begingroup\$ The closer you get you the rails, the more likely you are to get distortion, but some op amps are better than others. What's the output swing you need? 0 to 5 volts? \$\endgroup\$ – alex.forencich Mar 16 '17 at 16:44
  • \$\begingroup\$ 0 to 5 volts would be ideal. That is the range of the controller's ADC. The higher the signal's amplitude the less quantization error there will be from the ADC. \$\endgroup\$ – Paul Bunyan Mar 16 '17 at 17:03
1
\$\begingroup\$

Here is a plot of how close to rails you can get[from MCP651 datasheet]; the typical short-circuit current is 100 mA: at 50% of that, you will be 0.5 volts above GND or below VDD. enter image description here

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.