Signal Conditioning Circuit

Have you considered a current to voltage converter (for example an op-amp circuit) as the front end to the amplifier rather than just a resistor (which may load the accelerometer sensor signal causing a non-linear response).

_{Image courtesy of Wikipedia}

What accelerometer are you using, what is the output current and the passive resistor resistance? A circuit diagram would help.

I suspect your configuration suffers from insufficient load impedance on Op-Amp before A/D. Shottky reverse-biased is actually a leaky NONLINEAR capacitor so non-linear integration is expected. Try paralleling it with a variable resistance and test if it gets better. It's possible this behavior was acceptable for the application cirquit was designed. By the way how 74HC4053 is driven? Could be a hiccup in timing there.

1. I am not clear what you mean by "quants". Since AD73360 is 15 bit converter (32767 steps from zero to full scale) I suppose you mean that you want the error [50 quants?] to be less than 50/32767 = 0.00153 of full scale = 0.153 %.
2. It is not clear if you mean zero, span or non-linearity error from a "straight line".
   
3. The AD73360 has an internal PGA (Programmable Gain Amplifier) in each A/D channel, which is specified on the data sheet to have a gain error about +/- 0.8 dB (dB = decibels). In linear scale, 0.8 dB is 9 % error approximately - this is far more than the 150 "counts" error you have measured - you could expect 32767 x 9/100 = about 3000 counts error at full scale. You may get only on one measured channel, but other channels can have more error up to 3000 at full scale over full chip temperature.
4. Accelerometers come as standard with 10% error - few do better than 5% absolute error - so why do you think your A/D system needs 0.15% error?? Error of 1% would make little difference to overall measurement error, considering the transducer error is 5% and more.
5. The AD73360 appears from its data sheet to be a speech signal channel converter - it does not need, or have, high overall accuracy against true measurements, just adequate linearity.
6. You have not said what the accelerometer type or current output is supposed to be; nor have you given the resistor values. So I cannot tell if the resistor values give a proper current to voltage conversion. It may only look like a "voltage divider" - the series resistor at the input might be to protect against high current at the input - but it seems wrong.
7. 74HC4053 is a 1 pole 8 way switch - it is not clear why this was needed or what it does.
8. Accelerometers come in many forms, but many operate with a permanent DC voltage at the output. This DC voltage needs to be removed, usually by a capacitor coupling - but you have not mentioned any capacitor.
9. The AD73360 takes both positive and negative inputs - putting a schottky diode across its input, removing one polarity does not seem necessary - it makes a problem!
10. At first thought, it seems you could have a) passed the accelerometer current through a resistor, to get a proportional voltage b) coupled to AD73360 input with a capacitor from that resistor. No need for more amplifiers! effectively

74HC4053 is a Triple 2-channel analog multiplexer/demultiplexer

My mistake. Must have looked at wrong function diagram of data sheet. But it is no more clear what a triple channel multiplexer is doing than a one channel. It is not stated how many multiplex switches are tied to one AD73360 channel. Maybe, it is applying a positive or negative offset, but without circuit this is a guess.

1. by quants i refer to the digital samples i receive at pc

2. unable to understand your second query

5. can you suggest any better adc

6. accelerometer senses 0.12mA/g

7. 74hc4053 is used to remove dc bias from the signal, which might have added up due to amplifier or some other problem

9. what kind of problem schottky have, its 1n5818. switching frequency is 32khz

im glad to see your responses. hope to have further suggetsions and support

1. Now I do not understand. A digital sample is 16 bits - how can an error be 50 of those??
2. Also known as "offset" and "gain" errors, look at http://focus.ti.com/lit/an/slaa013/slaa013.pdf
3. Your item 5 - "Better" ADC depends on what you want and wish to pay... Compare the data sheet for AD7661 device. At "Specifications", DC ACCURACY it gives zero error and full-scale error. It also gives the temperature effect on those errors. The AD73360 does not give temperature effect or full scale error. What is not specified is not controlled or guaranteed. The big difference between AD73360 and AD 7661 is the PGA errors.
4. Your item 6 - Sensitivity 0.12 mA/g is one item of information. But without the current output at "zero g" and the peak g of the measured signal, one cannot design a good interface.
5. Your item 7 - 74HC4053 can switch on a bias signal, it cannot adjust its value - there must be a programmed current value (milliamps) from a D/A converter in your system.
6. Your item 9 - The diode will remove most of one half cycle of your input vibration signal, the part which exceeds about 0.3 volts, by drawing off the signal current. The waveform will be very different to the mechanical vibration input, one polarity will have its top flattened and cut off at a lower value than the other - I do not think that is what you want. The question is why the diode is needed at all, the AD73360 will take both + and - input polarities, the digital value will have a "sign bit".