Single chip VLF detector?

[sled]

I came across a nice little chip at work, the ADAU1761 (Datasheet). It's basically a 24-bit 48/96kHz stereo audio codec with an integrated 50MIPS floating point DSP which can be programmed graphically.
What caught my attention was this forum post here: https://ez.analog.com/thread/16947

This guy built a speaker impedance measurement tool with a phased locked loop outputting an I and Q signal. (Schematic here)



The cool thing is, the DSP can be programmed visually with Sigma Studio (free, Link) and comes with an extensive Toolbox (Link). Since it has 2 DACs with a headphone amplifier it could create both TX and Audio. Only question is whether the specs are good enough, i.e 48/96kHz with 95dB SNR. At 96kHz we only have half the instructions per sample of course.

What do you guys think?

[Davor]

95dB is a bit thin for a decent detector, so directly and with no oversampling it would not make a great detector, that's for sure. But there is a DSP inside, so why not.

[sled]

Could it be made with a decent preamp with ~40dB gain?

[moodz]

...reading the data sheet reveals 2 x 35.25 db PGA diff amps on the inputs. ;-)

[sled]

...reading the data sheet reveals 2 x 35.25 db PGA diff amps on the inputs. ;-)

Hm... usually those PGA have really bad specs, but it could be worth a try
Also misread the SNR, it's actually 96dB on the line-inputs or 99dB (A-weighted RMS, but don't think this applies to metal detecting)

[Carl-NC]

95dB is a bit thin for a decent detector, so directly and with no oversampling it would not make a great detector, that's for sure. But there is a DSP inside, so why not.

If I recall, the Minelab X-Terras use direct sampling using an audio codec. They're not barn-burners, but do fairly well.

From the photos I've seen of the Equinox PCB, it looks like a similar design.

[Altra]

The Whites Prizm used a PMC1803 codec and a pic24 to control and process. Someone here in the past mentioned that it used 4x sampling.

[Altra]

Typo for codec in above PCM1803 not PMC1803

[sled]

The Whites Prizm used a PMC1803 codec and a pic24 to control and process. Someone here in the past mentioned that it used 4x sampling.

Very interesting, seems like the codec does not have a DAC, so it must be generated separately. Do you know whether they use a DAC or MOSFET driver for the TX? The tricky thing with the 4x sampling and codecs is that you are not free to choose the sampling frequency.

[Altra]

Very interesting, seems like the codec does not have a DAC, so it must be generated separately. Do you know whether they use a DAC or MOSFET driver for the TX? The tricky thing with the 4x sampling and codecs is that you are not free to choose the sampling frequency.

It appears to be driven by a TLC372 "U4" reading the data sheet it is a comparator with a push-pull output, much like a fet driver. As far as the sampling goes you can only guess? Maybe set the codec frequency to 4x or 8x the TX which was 8khz? You can feed the codec clock pin any frequency you want to get it to sync. Carl might remember?

Add: Reading the data sheet and looking at the photo. You can see a 8.3mhz clock feeding the codec. This is divided by 256 inside and results in 32421.875Khz sample rate, divide this by 4 and you get about 8.1khz which is the TX freq. Might be fun to try.

[Carl-NC]

Correct, 6T had 8kHz TX and 32kHz RX sampling.

[sled]

Interesting, the ADAU1761 allows to set the frequency to a number of values, didn't even think about that! By using exactly 4x / 8x oversampling the processing required should be minimal. All could be done inside a single chip, only a simple microcontroller is needed for some controls to change parameters on the fly via I2C

@Carl-NC Can you tell me in what range (dB) the gain of the pre-amp was on the Prizm?

[Carl-NC]

@Carl-NC Can you tell me in what range (dB) the gain of the pre-amp was on the Prizm?

Sorry, don't recall. I think there was a fixed-gain preamp followed by a PGA. The PGA was used for dynamic gain control.