-
Categories
-
Platforms
-
Content
Audio amplifier to amplify a load cell signal
- Thread starter macabnv
- Start date
Not easily, unless you can arrange to excite the load cell with audio frequency AC drive. Normally DC is used to excite the load cell and it produces a DC output, which an audio amplifier would not be able to handle.
Even then, you would have to somehow deal with the fact that most load cells output a differential signal, which is not suited for input to an audio amplifier.
Easier to just use an instrumentation amplifier, which is designed for the purpose.
- Joined
- Jun 21, 2012
- Messages
- 4,880
Before embarking on this quest, you should endeavor to learn where your noise originates and what kind of noise it is. Load cells are traditionally excited with DC (usually 10 V DC), and there is a huge reservoir of theory and application examples you should tap into before even attempting to perform load cell signal conditioning. One good place to start is this excellent application note authored by Crystal, a Cirrus Logic Company, for digital applications. They do toot their own horn a bit in the applications descriptions, but there is good technical theory presented too. Off-the-shelf load cell signal conditioners are readily available and should be used in lieu of a "home brew" configuration, unless you are highly skilled and experienced with low-level signal conditioning.
If you are a novice at load cell signal conditioning, please let us know.
Load cells typically have full-scale sensitivities of as much as 5 mV/V of excitation. Many are much less because of derating to prevent overload, or because the applied tare subtracts from the available full-scale range. So, if you have a 1000 pound force (Lbf) load cell, and you apply 10 V DC excitation, and the load cell sensitivity is 2 mV/V, then the full-scale output voltage will be 20 mV or 20,000 μV. An instrumentation amplifier gain of 500 will give you 10 V output at full scale. It should be possible to digitize the output to at least four significant figures (1 part in 10,000) so a display reading from 0 to 9999 represents a load range of 0 to 999.9 Lbf, or a resolution of 0.1 Lb. Not too shabby for a half-ton load. That's without even breaking a sweat, since the common 16-bit ADC digitizes to 1 part in 65,536, with a sensitive of 152 μV for the least-significant bit when full scale is 10 V DC.
Where you run into problems is when the noise in the output reaches that level, making the 16-bit ADC a 15-bit or 14-bit ADC... or worse. So if your output noise must be less than 152 μV and the instrumentation gain is 500, then the input noise must be less than about 305 nV. A load cell with 350 Ω impedance will have about 75 nV of Johnson (thermal) noise in a 1000 Hz bandwidth, so your noise budget is reasonable. If you can't get the signal conditioner output noise below about 300 nV in a 1000 Hz bandwidth, you aren't trying hard enough.
Unless you are trying for 24-bit Delta-Sigma conversion of nanovolt signals, conventional differential-input DC instrumentation amplifiers work fine for load cell signal conditioning. The resistive strain gauges used in most load cells are around 350 Ω and do not contribute a lot of Johnson noise to the output. Neither do the instrumentation amplifiers. The most common source of noise is AC line "noise" coupled into the external strain-gauge wiring and into the front end of the signal conditioning amplifier because of poor layout and construction practices. There have been entire volumes written on how to suppress AC power-line coupled "noise". You should find, read, and understand those tomes.
AC excitation of load cells has advantages as well as disadvantages. The major advantage is the excitation adds information to the load cell signal, namely the phase and frequency of the excitation, by causing the resistance variations of the load cell strain gauges to modulate the excitation voltage. This added information allows AC-coupled amplification followed by synchronous detection (rectification) and filtering to essentially narrow the noise bandwidth of the load cell. If the excitation frequency isn't harmonically related or coherent with the prevailing AC power-line frequency (usually 50, 60 or, for aircraft, 400 Hz) any induced power-line "noise" will integrate out to zero after synchronous detection... assuming everything between input and output is linear. If the noise component overloads an intermediate stage, all bets are off.
The downside to AC excitation (besides the considerably increased expense and complexity) is the load cell response to changes in load takes longer, because of the decreased bandwidth, to appear as a change in output. This has serious consequences if the load cell is being used to measure a dynamic load, although using a higher excitation frequency can ameliorate that to some extent. Unfortunately, most practical load cells use strain gauges cemented to a metal beam and this construction introduces considerable parasitic capacitance to the bridge circuit. The parasitic capacitance must be carefully "balanced out" and that introduces the possibility of drift. It also attenuates the load cell signal, both at the load cell and in the connecting wires, and the attenuation gets worse as the excitation frequency is increased.
Still, it is sometimes worth the effort to use AC excitation for signal conditioning, and there are load-cell instrumentation amplifiers available "off the shelf" for this purpose. I would explore that route first instead of trying to use "an audio amplifier" with your load cell. Note that whether AC or DC excited, the load cell sensitivity (calibration) is proportional to the excitation amplitude. You either need a very stable excitation source or a ratiometric means to ratio the load cell output against its excitation voltage.
Soooo... please tell us what you are trying to DO, not a proposal on how you think it should be done.
Provide us with specifications. What is the full-scale range and sensitivity in mV/V of the load cell and what is both the recommended and maximum excitation voltages. Does the load cell operate in both tension and compression, or just one of those? What percentage resolution of the full-scale output do you need? At what accuracy? What signal bandwidth do you need? How far is the load cell from the signal conditioning electronics? Are you using a 4-wire or a 6-wire connection to the load cell? How will you calibrate the output of the load cell? If I've left anything important out of this list, other EP members should jump in here with additions and corrections. We are all here to help, but we need more information.
Hop
- Joined
- Jun 21, 2012
- Messages
- 4,880
Yep. That's the way it's usually done... unless the load cell and signal conditioning wiring is so poorly done that induced "hum" from the ever-present electromagnetic field produced by power-line wiring overwhelms the front-end of the instrumentation amplifier. Even operating from batteries to avoid ground loops will not solve poor implementation techniques.
At the very least, foil-shielded twisted-pairs should be used between the load cell and the instrumentation amplifier, shield terminated only at the instrumentation amplifier common to avoid "ground loops". IIRC, some load cell manufacturers provide "pre-conditioning" amplifiers located at the load cell to provide a high-level signal less susceptible to induced noise. For long signal runs and a limited number of load cells it might be worth the extra cost, but good wiring and shielding practices are usually all that is needed, even with several hundred feet of cable between load cell and signal conditioner.
Remote excitation sensing (6-wire foil-shielded 3-pair cabling) is recommended for long cable runs. The cabling is an integral part of the load cell calibration and temperature compensation, so cabling should be considered first, not just "added on" at the last minute.
- Joined
- Jun 21, 2012
- Messages
- 4,880
@WHONOES: Please note (1) this thread is more than two years old; (2) the original poster, @macabnv, only made one uninformative response in this thread after their initial post; and (3) they haven't been seen in this forum since September 8, 2016. Methinks either the thread is dead or the OP's problem was solved.
