LF 356 opamp question

[LA_DIY]

I'm designing a little 4 track recorder for a senior design project at USC. Things are going pretty smoothly, but I have one problem that seems so simple to fix, but I haven't been able to solve so far.

The input to the ADC that I'm using for this project has to be between GND and +5V. Ideally the audio would use that whole range so that the full dynamic range of the ADC would be used. The problem is, the opamp circuit I'm using outputs a peak to peak voltage of 5V, but its between -2.5V and +2.5V. I just need to offset the voltage be +2.5V and it will work perfectly.

The opamp I'm using is an LF 356 if that helps.

My professor has said its not something to worry about because this class is more about the embedded system design and not the analog electronics, but I was wondering if anyone here had some insight on how to solve this problem since I'm sure its something that must have come up before somewhere (and because I want my little recorder with a 4kHz sampling rate to sound as amazing as possible :wink: ).

Thanks for the help,
Dave

 

[gyraf]

Wouldn't a simple capacitor do the trick - maybe offset to a point in the mid between 0 and +5?

 

[bcarso]

What are your power supplies?

Offsetting a signal is easy if the new range is within the limits of the hardware's ability to produce it on a given output. It sounds like your 356 must have some + and - supplies already if it is putting out +/- 2.5V.

As usual, it would be helpful to see the circuit to give the best advice. Is the opamp used in an inverting or a non-inverting configuration? Is the signal a.c.-coupled (and if not, why not, since it is, after all, audio)?

If it is a.c coupled and non-inverting, it may be as simple as referring the input resistor after the coupling cap to the appropriate bias voltage to cause the output to sit at 2.5V d.c.

If it is inverting, then another resistor to the inverting input from the negative rail of the power supply should be selected to produce a d.c. voltage of +2.5V on the output of the amp. For example, if the feedback R were 10k and the negative power supply were -15V, the resistor would be 60k. This assumes that the noninverting input is tied to system ground.

 

[LA_DIY]

Thats the design that provided the best gain for the microphone we're using. The feedback resistor will be a pot so we have some volume control, but this is the design that worked the best without the volume control.

Power supplies are +/- 12V, its connected in an inverting configuration.

 

[pstamler]

You may want to look seriously at changing the configuration. Couple of reasons:

First, it'll be noisier than necessary; the 1.5k resistor will be in series with the microphone's impedance, and will add its Johnson (thermal) noise to the microphone's noise. I realize that with a 356 this may be something of a moot point, since it's already far noisier than one wants in an amplifier for a microphone.

Second, because the circuit is DC coupled and the feedback resistor varies for volume control, any DC you put onto the input will be amplified by different amounts, depending on the volume control setting.

You might want to try two opamp stages rather than one: the first could be a fixed-gain non-inverting opamp, a nice quiet one like an NE-5534. That could be AC-coupled into a conventional volume control; the output of the volume control would be AC-coupled into a unity-gain non-inverting opamp, with the resistor that connects to the + input of the second amp (and the coupling cap) connected to a +2.5V supply rather than ground. You can make the +2.5V supply from an LM317 regulator, or use a 2.4V zener diode (buy a dozen and select the one that's actually 2.5V).

Peace,
Paul

 

[bcarso]

Paul is spot on as usual :green:

 

[PRR]

> LF 356 opamp question

NOT an LF356 question; a general opamp question.

> You may want to look seriously at changing the configuration.

Bah. The lesson is on "embedded system design", not Good Audio.

> it'll be noisier than necessary

Yeah, but less noisy than the average EE Lab.

And I have a sneaky feeling he's been given 8-bit ADCs: what analog noise?

But audio life is less insane if you cap-couple a lot. Use an input cap. You are not amplifying DC (so much), so stray DC in the source or op-amp won't throw your ADC's input away from the +2.5V center-point.

I would think to see the problem in two steps. Amplify to some reasonable level like 1Vrms (or 5Vp-p); then offset from our conventional ground-reference to a +2.5V reference. The first problem goes away if you have a mike preamp, and the world is full of better preamps than an inverting LF356 (though it won't suck).

If you assume a 5Vpp signal is available, then use two 100K resistors across the +5V supply to set the ADC input pin near +2.5V. Put a coupling capacitor in front so whatever source does not suck-off the DC level. Done. No, put about 1K in series with the cap so that signals over +5V or under zero V can't flow infinite current into the ADC's input pin. (If it has a 10mA melt-down rating, with 1K it may survive 10V signals.) That resistor does degrade ADC accuracy; total ADC input design is a big headache; this will work good enough to sing Mary Had A Little Lamb in 4-part harmony.

There are other ways. Cap-couple the input of the mike-amp, and connect the non-inverting input to +2.5V. Output will sit at +2.5V (proof is left to the student). Wire the opamp as non-inverting, cap-coupled input, and return the feedback network to a clean +2.5V source instead of to ground.

But ultimately: the lesson IS "embedded system design". If you capture wonderful 2KHz audio, but the system can't digest the data because you used all your lab-time perfecting the audio, you fail. If you capture a buzz that resembles a voice, and process it well, you pass and you learned the rudiments of a commercial design skill. You can always work on the audio later.