offset voltage without connecting any signal

[Neo6]

Hello Lynn,

Thanks for your support. your questions has given me some solutions, I corrected one by one like this

1) I checked for proper ground connections

2) I used proper biased resistors ( 68 ohms)

3) I resoldered everything.

4) DC offset in the function generator was set to zero.

Now voltage waveform is good, exactly sinewave and even peak to peak values are exactly matching with function generator.

But in the frequency spectrum, power at the input frequency is very low, it is only -70db and y axis in the power spectrum is in  v square/ Hz rms. 

why the power is still low  and what is this unit  /hz represents ???

Thank you.

[johnsold]

Sorry I did not reply sooner.  I have been ill.

2) 68 ohm bias resistors seem quite low.  As they are effectively in parallel with the 50 ohm load, they result in an attenaution of the signal by about 4.8 dB (assuming a 50 ohm source impedance on the generator).  Try larger resistors, pehaps 6800 ohms.

The spectrum axis values appear to be power density (the square of the voltage of the signal component in a bandwidth of 1 Hz at the specified frequency).  The reference may be 1 V^2/Hz or some other value. Your signal is small. You had been using a large full scale range. You are almost certainly not sampling synchronously. The combination of all these factors may well result in a peak at -70 dB.

Lynn

[johnsold]

You asked how I analyzed your graphs from only the images you posted.  I am replying here so that anyone else who may follow this thread can also benefit.

I have been doing signal analysis work for 40 years, so this is not necessarily something that you will pick up in a few days.

Consider the power spectrum with input.jpg that you posted in message 8 of this thread.  First I look at the graph label and axes labels.  That tells me that I am looking at a spectrum, that the vertical axis is logarithmic (dB) and in power unit, and that the horizontal axis is linear and covers an audio frequency range.  I also noted that you are using a D/A converter with 16-bit resolution.

Next I look at the shape of the spectrum.  It has two prominent peaks and a baseline below -100 dB.  The theoretical signal to noise ratio for a 16-bit converter is 98 dB.  The peak at zero is close to 0 dB.  So anything below -100 dB is completely meaningless. The only other peak well above the noise is at 5 kHz and about -55 dB.  A power ratio of -55 dB represents ~3E-6, so there is very little power at 5 kHz compared to the DC component.  This is compatible with your indication that you have a signal of a few millivolts and DC offsets which can reach volts.

The no signal spectrum image is analyzed the same way.  Again the DC component appears to be on the order of volts and nothing else of significance shows up.

Lynn

[Neo6]

Hello Lyn,

Thanks for the explanation.  But there are few more questions.

1) I am using analog to digitial converter not D/A.

2) From your analysis, can I confirm that noise which is shown on my graph is  quite common  and my system is not picking up any external signal. Can I say my system is EMI free ? 

2)If i understand correctly,  If we try to calculate the voltage value of 5KHz frequency peak using decibels , The denominator in the logarithm ratio is 1 V .. am I right???   Is that what represented on y-axis ( db rms 1V/ sqrt(HZ))  ?????? 

Thanks 

Vaidhin.

[johnsold]

Vaidhin,

1) Sorry.  Transposed the letters.  A/D is correct.

2) It does not appear that you are picking up any significant external signal.  One of the images shows two peaks at -85 to -90 dB, so it is possible that you would see some interference if you got rid of the DC and used a more sensitive input range.

2) The scale is not specified clearly. It claims to be in power units but gives a reference in terms of voltage spectral density (V/sqrt(Hz)).  A voltage alone does not determine the power.  You also need a current or resistance.  Audio frequency systems using a 1 V reference often use 600 ohms as a standard impedance, but it is not clear whether that applies to this case.  Decibels in power units are 10log(power ratio) while in voltage units they are 20log(voltage ratio).  For a fixed impedance the two are the same but you need to know what units you are using.  It does look like the reference voltage is 1 V rms.

Lynn

[Neo6]

Hello Lyn,

I think you have almost cleared 90% of my query..... will keep working on this and will update if I have anything new

Thank you

regards 

Vaidhin.