Hello,

I'm using an AD7276 single-ended ADC and I'm trying to understand how to couple in a typical line-level audio signal (I'm just using a PC line-out for testing). The AD7276 is a single voltage (3.3V in this case) device, and the input can be 0V to Vdd.

I believe that typical line-level inputs swing about 1V peak-to-peak (+/- 0.5V) around the ground that is provided with the line input.

With nothing hooked up to the ADC input, the output is very quiet. So first I tried the line input connected directly to the ADC input and the line ground connected to the ADC's common. With this I get a lot of hiss and clipped / distorted sounding audio.

So I thought maybe I need to A/C couple the input and bias the ADC's input. So I put a 1uf cap at the input, and a 10K+10K voltage divider between Vdd and ground to bias the input. This eliminated the distortion (the audio sounds rather nice now), but there is still a *lot* of hiss.

Removing the line-input's ground from the circuit reduces the hiss a little, but the noise is still very noticeable.

Finally I just connected a short (about 3") wire to the ADC input and I get a lot of hiss (maybe I just made an antenna). It seems anything that I connect to the ADC input is causing a lot of noise in the circuit.

I'm also not sure what to do with the line-input's ground? Where should it connect. And how is it I'm introducing so much noise? This is really a basic circuit, or so I thought (then again, for me nothing is every so simple with analog electronics.)

The ADC is pretty isolated in a part of the PCB with not much else around it, and I tried to follow the layout rules mentioned in the datasheet. The 3.3Vdd is supplied from an LDO. I have two decoupling caps, 1uf and 0.1uf ceramic, on Vdd near the ADC. The audio input trace is about 13mm long and does run near the LDO, but I'm going to change that in the next rev. I'm not sure if that could introduce all the noise I am hearing? But I don't get any noise until I hook *something* up to the PCB input pin.

I'm really just making limited educated guesses here and was hoping for some help to get me in the right direction. Any insight would be greatly appreciated.

Thanks,
Matthew

 

The dynamic range of human ear ~ 120 db.

The range of that A/D is ~ 70 db.

What is the D/A being used to regenerate the audio ?

What is sample rate of A/D, and DAC ?

Regards, Dana.

 

I'm not sure the poster knows what an ADC is used for, sounds like there is no DAC.
Here is a little blurb to help the OP understand what an ADC does and what it is used for;
https://learn.sparkfun.com/tutorials/analog-to-digital-conversion

 

I am reading ADC samples at about 1.562M samples per second (about half the max rate), i.e. with a 25MHz clock and a 16-bit sample period. This rate is just convenience because I already have the 25MHz clock.

The DAC to regenerate the audio is a simple delta-sigma PWM circuit with a PWM frequency of 100MHz. The PWM is output on a 3.3V pin to an RC low-pass filter and a small speaker. I have used the PWM circuit before with audio data that was generated totally within the FPGA, and in those cases I do not have this noise problem. It is only when I use the 12-bit ADC sampled data that I get the noise.

 

@R.E. Really? What about my post indicates I have no idea what and ADC is or does?

 

@R.E. Really? What about my post indicates I have no idea what and ADC is or does?

No mention of it? It looked to me you were applying an audio signal to a ADC and trying to get audio out, I didn't see DAC until later. R/C Filter on output maybe?
See you later.

 

ADC is great at picking up noise all you need is a short wire and more then 10k resistance

 

Sorry for any confusion, I try to keep my posts as short and specific as possible to the problem. The overall picture is:

audio in -> cap -> input-bias voltage divider -> ADC (spi) -> FPGA -> PWM out -> R/C filter -> small test speaker

The audio input is just a typical line-out from a PC for convenience (it is simple to get audio from a PC), and represents the typical use-case pretty well, i.e. line-level single-ended audio. I'm using the PWM output from the FPGA just as a simple test, since ultimately the sampled digital audio will be part of a bigger system and I wanted to make sure it was a decent conversion.

I have used this simple PWM before for audio output without problems (i.e. no noise like I am experiencing now), so I was quick to rule that out as a possible problem and focused on the ADC and the audio input.

 

This seems to imply inherent response characteristics of DelSig
make it audio application of choice. Unfortunately the excellent
noise discussion in it of DelSig was not discussed in SAR dis-
cussion.

https://www.google.com/url?sa=t&rct...lication.pdf&usg=AOvVaw3Ym5m_WGfO-GQVhycrmtBI

Makes one think probing around with spectrum analyzer might
help.

Regards, Dana.

 

Perusing the data sheets I saw this;
"Large source impedances significantly affect the ac performance of these ADCs and can necessitate the use of an input buffer amplifier."

Page 17 of the datasheet here;
http://www.bdtic.com/datasheet/adi/AD7276_7277_7278.pdf

Computer soundcards are around 10K impedance, with some being 600 ohms.
Explains why when you used the divider the audio sounded better.

Might want a buffer amp with low output impedance.