Is it possible to reduce noise after ADC

[FvM]

I agree, that the suggested differential ADC configuration can considerably reduce the problem of VREF noise. Anyway I would at least provide the minimum ADC reference bypass means suggested in the LM4050 or ADC datasheets. Wheter "gain modulation" creates a considerably noisew contribution depends on the input signal, but it's surely much lower than the previously reported noise levels.

Finally, it's not clear how the external geophone cabling looks like. Exposing the VREF node to the outer world without suffcient bypass capacitors can cause interference susceptibility as well.

 

[nadabro]

Hi.

Thanks for the answers. The MAX7401 indeed is not good enough, the internal ADC filter do a nice job, the noise with MAX7401 is more than 2 times without the external filter. If the results improve, i can replace the amp and the filter with a better one, the amp is a MSOP and the filter is PDIP.

I did run some tests with a couple of LIPO battery's with different voltages (i cant get better than this batteries to run some tests)..11.1V 850mAh and 7.4V, both have almost the same results with the power supply. The noise without the amp is around +/- 80 and with amp +/-800 (without the sensor). So, two different power supplies, two different batteries, almost same results..


But i did connect the 4 pin (Channel 2 with amp, IN -) to GND of the microcontroller, the noise drop to the same results of the channel without amp or filter and the results start at 8 million. What does this mean? It looks that improve a lot the results, with or without the sensor i have noise count at +-800 and the results start at 4 million. Can i connect one pin of the sensor to IN+ and the other pin to GND?


About the cable to connect the sensor, its a simple 15cm jumper cable not shielded..but i isolated it with isolator tape (worse wont do..).

 

[dave9000]

But i did connect the 4 pin (Channel 2 with amp, IN -) to GND of the microcontroller [...] Can i connect one pin of the sensor to IN+ and the other pin to GND?

I am not sure to undenstand correctly: do you mean pin 4 of JP3 to GND?
In this case the opamp is saturated near VCC, and what you are measuring is just the ADC+circuit noise.

By the way: I found (here) that reducing the damping resistor enables you to trade-off sensitivity vs corner frequency.
So, in the end, your original circuit with 0 ohm input impedance (I/V converter) is quite smart ;-)

About the cable to connect the sensor, its a simple 15cm jumper cable not shielded..but i isolated it with isolator tape (worse wont do..).

The best cable for this connection is a twisted shielded pair, but given the short distance it really doesn't matter that much; if you want to improve in this area just twist yourself your wires.

 

[mtwieg]

Several people have asked this question, but I haven't seen it get a response: what is the output impedance of the sensor? Are you connecting directly to the coils in the geophone, or is there some kind of internal conditioning circuitry?

If you're feeding directly to a transimpedance amp, that may be good for gain and bandwidth, but depending on the output characteristics it may be terrible from a noise perspective.

 

[nadabro]

I am not sure to undenstand correctly: do you mean pin 4 of JP3 to GND?
In this case the opamp is saturated near VCC, and what you are measuring is just the ADC+circuit noise.

By the way: I found (here) that reducing the damping resistor enables you to trade-off sensitivity vs corner frequency.
So, in the end, your original circuit with 0 ohm input impedance (I/V converter) is quite smart ;-)



The best cable for this connection is a twisted shielded pair, but given the short distance it really doesn't matter that much; if you want to improve in this area just twist yourself your wires.

Not having anything connected to CH2 or connecting the sensor to CH2, the results start at 4million count, but if i put the pin 4 of JP3 to GND..the results start at 8million an i have much lower noise (+/-30). So in my rude knowledge in electronics:

-PIN 3 JP3 - one pin of the geophone
-PIN 4 JP3 - GND
-the other pin of the geophone to GND

I should have more amplification and less noise...

That part of the circuit is based on a schematic for the same geophone (same specifications) that i´m using, and that schematic has good results with a older amp (AD8628) and MAX7401 in a 16bit ADC..but i thought on getting a better amp, use 23bit and not using the filter i would get better results..the results are good, the noise dont. He can recover much better the lower frequencies, because he has lower noise, and since lower frequencies usually comes in weaker readings, those readings are mixed with noise and i cant recover it even using a 32 pole low pass filter on the software.

I have another 10Hz geophone with a different brand, and this one doesnt need amplification, at 23bits have great results..the 4.5Hz geophones with 23bits i can only get strong movements.

The original schematic:
http://www.infiltec.com/seismo/inf-qmsd.gif

Several people have asked this question, but I haven't seen it get a response: what is the output impedance of the sensor? Are you connecting directly to the coils in the geophone, or is there some kind of internal conditioning circuitry?

If you're feeding directly to a transimpedance amp, that may be good for gain and bandwidth, but depending on the output characteristics it may be terrible from a noise perspective.

I only have this specifications:
**broken link removed**.


i´ve the cable soldered directly in the geophone. The geophone had a 3k ohm resistor soldered in the pins, but i removed because i´m using 2.7k ohm in the AMP..do you think replacing the 2.7k ohm resistor for the original 3k resistor will do any difference?

 

[mtwieg]

The original schematic:
http://www.infiltec.com/seismo/inf-qmsd.gif

The preamp part of the schematic doesn't really make sense. R1 does absolutely nothing in this circuit. The coil is virtually shorted, so it will be very overdamped.

I feel that a noninverting op amp circuit would be much more suitable, since you could apply a damping resistor properly. And the choice of amplifier is absolutely critical. For example,if the coil resistance is 150ohms, then its noise spectral density will be 1.58nV/sqrt(Hz) at 25C°. So over a 10Hz bandwidth that will be about 5nVrms. The input noise voltage of that LT1677 over a range of 0.1-10Hz is 90nVp-p at best. So your noise is going to be dominated by the preamp.

I only have this specifications:
**broken link removed**.

That link doesn't work, I assume you're referring to this page: **broken link removed**

That's a start anyways... it lists several possible values for the coil resistance; you should measure the resistance of the coil to see what it is. Once you know that you should do some noise calculations to determine whether the noise on the ADC you're seeing can be explained by the thermal noise of the coil, and by preamp noise. You'll should take your noisy ADC data and do a DFT of it and figure out what the spectral noise density is (in V/sqrt(Hz)).

edit: what heck I didn't post again?

 

[FvM]

I assume you're referring to this page: L15B Geophone

Thanks mtwieg, that's helpful information. It clarifies in my opinion, why the preamplifier should not be connected as shown in post #13 or the infiltec circuit. Bot are using an I/V converter configuration that effectively shortens the transducer and doesn't achieve a dampening value according to the manufacturer specification. This point hasn't to do with noise performance but is important for geophone sensitivity near the lower cutoff frequency.

 

[nadabro]

Thanks for fixing the link mtwieg.

That's a start anyways... it lists several possible values for the coil resistance; you should measure the resistance of the coil to see what it is. Once you know that you should do some noise calculations to determine whether the noise on the ADC you're seeing can be explained by the thermal noise of the coil, and by preamp noise. You'll should take your noisy ADC data and do a DFT of it and figure out what the spectral noise density is (in V/sqrt(Hz)).

I measured the resistance on the geophone, from my cheap/non precision multimeter it gave me 374ohm..which is close to the 380ohm from the specifications. About the calculations..i will take some time to do that

So while this circuit works, the choice of components and the circuit design is no suitable for the sensor specifications. Will it be possible to achieve good results with this circuit replacing some components or should i start looking a new one? Available schematics to these applications are so rare and i dont know how to design one..

Some available schematics (usually for ~9k ohm coil):
**broken link removed**
**broken link removed**

 

[mtwieg]

Thanks for fixing the link mtwieg.


I measured the resistance on the geophone, from my cheap/non precision multimeter it gave me 374ohm..which is close to the 380ohm from the specifications. About the calculations..i will take some time to do that

It's not so complicated. Are you able to at least able to find out the DFT of the ADC data?

So while this circuit works, the choice of components and the circuit design is no suitable for the sensor specifications. Will it be possible to achieve good results with this circuit replacing some components or should i start looking a new one? Available schematics to these applications are so rare and i dont know how to design one..

Some available schematics (usually for ~9k ohm coil):
**broken link removed**
**broken link removed**

Both of these circuits use a noninverting amplifier as a first stage as I suggested, so I bet they'll work better. But still the most important part of the circuit will likely be the very first amplifier stage. The amplifier should have very low noise in your band of interest; the op amp you have might be fine, but I'm not really experienced in very low frequency amplifiers.

 

[dave9000]

Considering 380 ohm:
SENSOR CONNECTED

gain = 2633
noise seen by the ADC = 428 LSB
​

SENSOR NOT CONNECTED

gain = 371
noise seen by the ADC = 60 LSB​

I considered a band of 0.1-10Hz for the ADA4528-1, a reference of 2.5V and 23 bit for the ADC.

From various posts of yours (including your last test) we see that an input of 2.5V translates into 4M ADC counts: this means that your chain is "calibrated" (no post-processing gain or so). If you are willing to verify this hypothesis, connect a 1.5V battery at the input (you should measure +2.5M LSB).

In this conditions, measuring the noise with the sensor not connected for some 10-20 seconds, you should see a noise of ~60-100 LSB (a longer measurent will return more noise).
So you really have excess noise from the circuit.

I suggest you to insulated pin 3 of IC1 from AGND and to connect it to REF. Being a double layer PCB it should be easy to just cut the ground track that goes to pins 3 and 5; otherwise you have to raise the IC pin (use a desoldering braid).

if i put the pin 4 of JP3 to GND..the results start at 8million and i have much lower noise

When you put JP3 pin 4 to GND you create a DC voltage between the inputs of U$4.
U$4 will "try" to raise the ouput but, since the inverting pin is now locked to GND, the output will stay fix at VCC. That's why you read 8M (upper limit) and that's also why you see very little noise (the output of U$4 can't move).

do you think replacing the 2.7k ohm resistor for the original 3k resistor will do any difference?

Changing the dumping resistor from 2.7K to 3K will make no difference.
The choice of the dumping resistor is not trivial, because the current in the coil generates a feedback force on the sensing mass, and this effect in NOT apparent just looking at the schematic (see this link, look also at the rest of that article for more detailed explanations.)

Reducing the input inpedance from 2.7K (standard passive dumping) to 0 ohm (like in your circuit) or even to negative values (as in some designs) can extend the band to lower frequencies, but you will have a smaller signal, which means an higher noise.

In your case you are probably better off with a standard approch (2.7K) to minimize the noise.
If you want to try this NOW, you can simply insert a 2.7K resistor in series with your sensor.
But I don't expect that this will really improve things because the gain of U$4 will drop off, but the noise will not! The signal/noise ratio won't change.

 

[FvM]

To achieve a sensor load according to the datasheet specification in the present (inverting) amplifier configuration, you can connect the 2.7k resistor in series with the sensor instead of parallel to it. The amplifier gain will of course reduce respectively.

As previously mentioned, I completely agree with Dave about the differential ADC operation suggestion. In addition, minimal bypassing (e.g. 10 uF) should be added for VREF.

 

[nadabro]

Thanks for the advices.

What i need to do is cut the line in PIN 3 to redirect to REF, something like this?
**broken link removed**

About the resistor in series, is one 2.7K resistor soldered in the pins of the geophone?


Sorry for asking for confirmation, but i had bad examples in the past involving electronic experiences

 

[dave9000]

Yes, the modification is exactly as you draw it .
As for the resistor: unconnect the wire from JP3 pin 4 and solder the resistor there (between the wire and JP3 pin 4).

P.S.

Sorry for asking for confirmation, but i had bad examples in the past involving electronic experiences

don't worry: everybody does mistakes, but are the professionals those who make the most expensive ones ;-)

 

[nadabro]

Yes, the modification is exactly as you draw it .
As for the resistor: unconnect the wire from JP3 pin 4 and solder the resistor there (between the wire and JP3 pin 4).

P.S. don't worry: everybody does mistakes, but are the professionals those who make the most expensive ones ;-)

Nice one lol

I will try to make that changes tomorrow..once i test i will report here the results.

Thanks again for the help.