FPGA based UNIPI system

[moodz]

Hi Moodz,

Given the price of the AD7760, I've been having a look for suitable alternatives for direct sampling and think the following may be quite decent - I've got one on order to play with, just need to knock up a test PCB.

https://www.akm.com/global/en/produc...-adc/ak5578en/

Quite reasonably priced (Around £16 from Digikey), it has the facility to combine all 8 channels and improve SNR and dynamic range accordingly.
There are also 4 and 2 channel versions available at lower prices.

The sample rate is up to 768kHz at 32-bits, so not as high as the 2.5MHz of the AD7760 but the claimed dynamic range of 130dB (in 8 to 1 mode) is mighty impressive. Whether that is achievable at 768kHz sample rate remains to be seen.

Cheers

Looks like a nice chip ... the AD7760 is strange ... the Eval board all built up and tested is only a few dollars more than the one off chip price ??? Be interesting to see what you can make of the AK5578.
I was tempted to move away from the FPGA to a CPU ... but will stick with the FPGA now as it can deal better with weird timing issues that seem to crop up with the ADCs. Besides the FPGA makes for a nice multichannel realtime builtin oscillscope and logic probe which is invaluable for debugging the DSP and ADC logic. I am ordering some FPGA boards from china to replace my obsolete boards but the VHDL code will be the same.
If you are making a board for the AK5578 then I would also be interested as the 2.5 MSPS that the AS7760 offers is nice ... but prob not necessary.

moodz.

[Olly]

Looks like a nice chip ... the AD7760 is strange ... the Eval board all built up and tested is only a few dollars more than the one off chip price ??? Be interesting to see what you can make of the AK5578.
I was tempted to move away from the FPGA to a CPU ... but will stick with the FPGA now as it can deal better with weird timing issues that seem to crop up with the ADCs. Besides the FPGA makes for a nice multichannel realtime builtin oscillscope and logic probe which is invaluable for debugging the DSP and ADC logic. I am ordering some FPGA boards from china to replace my obsolete boards but the VHDL code will be the same.
If you are making a board for the AK5578 then I would also be interested as the 2.5 MSPS that the AS7760 offers is nice ... but prob not necessary.

moodz.

I'll probably use the Raspberry Pi Pico 2 as the controller for the AK5578 and try to get the I2S interface to the chip implemented in its programmable PIO so that I can leave the cores free for processing.

Busy creating the Kicad part symbol for the AK5578 now, then will put a schematic and PCB together with it and the Pico so I can evaluate it. Will let you know how I get on.

I kind of agree with you about the FPGA, it's really nice for development with all the great debug features and I'm glad I've got one to use in the lab - it's just a bit pricey to use in a production model.

Olly

[Olly]

Hi Moodz,

I've made a PCB based on Carl's AMX TX implementation of your CCPI circuit but made provision for the resonant capacitor and using IPB95R310PFD7 FETS. These are 950 V devices.

So far, so good - all seems to work OK so far, but I'm now deciding which path to follow, the original constant current PI idea or using it in VLF mode with the resonant capacitor. You've tried out both approaches, would you say that the VLF mode should be able to deliver superior results re: depth of detection and more importantly - discrimination?

Thanks

[moodz]

At the moment the VLF mode is winning ... The ccpi did work but there was more noise. I suspect the additional harmonics present in the system each contribute more noise than the more frequency limited VLF.
I am waiting on boards from China to replace the FPGA.
The transmit board could be made reconfigurable so you can switch modes. I just use manual switches to add / remove caps.

[Olly]

Thanks for the feedback

[moodz]

The Chinese boards arrived and the AD7760 board will be plug and play on the two FPGA boards .. no wiring yay ! ... I only have to remap the pins in the VHDL source code.
There are two FPGA boards for testing.
The first has Spartan 6 FPGA and VGA video , external RAM, pushbuttons, LEDs, Beeper, SD card slot, Camera interface, RTC, USB etc etc. Cost = $60 AUD.
The second just has the Spartan 6 FPGA and USB interface. Cost = $20 AUD.

For fun I also bought a STM32750 core board that has on board 3 x 16 bit ADCs at 3.6 MSPS. Cost = $ 19 AUD.

[JoyJo]

I, after all, can't understand how the ccpi will work in vlf mode with the addition of only one capacitor. Or do you need a mandatory addition of a coordinating transformer?

[moodz]

I, after all, can't understand how the ccpi will work in vlf mode with the addition of only one capacitor. Or do you need a mandatory addition of a coordinating transformer?

Can you explain what you mean by "coordinating transformer" ? I do use a transformer in the output so I can isolate the single ended coils ( ie grounded ) of a DD coil.

Obviously you cant just connect a CCPI cct to a grounded coil unless you float / isolate the power supply.

[JoyJo]

On your previous circuits, a disconnecting transformer was used to connect the TX coil. It was also mentioned that the transition from PI mode to VLF can be made only with one toggle switch. I understand that the TX coil with one of the leads is grounded cannot be used with the CCPI. But I haven't figured out how to get the VLF mode yet.

[moodz]

Its fairly easy ... see diagram below. Also note there is another important aspect of this circuit.

The transmit waveform is substantially unaffected by variation of the DC power supply ( ie constant AC output ).
This means you dont need a regulator for the DC supply ie run off battery direct. ( The TX current waveform will not change from say 12 volts to 8 volts at the battery )
I patented that idea also. The circuit is shown set up for VLF transmission of 20 Khz.

It draws 150 mA average from the DC supply for 2.6 Amps peak to peak in the Tx coil L7. L5 and L4 are the transformer. C2 is obvioously the cap.( selected on base TX frequency ).

The transformer isolation also prevents any small DC biases that might occur in your typical Halfbridge due to component differences .... this bias will cause an error in the RX due to offset modulation.


moodz.

[JoyJo]

Now I understand. Thank you. But the introduction of an additional transformer significantly increases the size of the product.

[NigelLeeming]

For fun I also bought a STM32750 core board that has on board 3 x 16 bit ADCs at 3.6 MSPS. Cost = $ 19 AUD.

That's quite a spec and price. Do you have a link?

[moodz]

https://de.aliexpress.com/item/10050...yAdapt=glo2deu

In hindsight I advise to get the STM32H743 though .... its got way more flash memory for program storage. ( the 743 fits on the same PCB ) ... its a few dollars more.

[moodz]

Now I understand. Thank you. But the introduction of an additional transformer significantly increases the size of the product.

The real question you need to ask yourself then is are you designing a metal detector or a hearing aid ?

[moodz]

The real question you need to ask yourself then is are you designing a metal detector or a hearing aid ?

But seriously .... its not a mains transformer ... its a high frequency transformer similiar to this ....

about 1cm x 1 cm x 0.6 cm. You are joking that it is going to increase the size ?