Looking for suggestions on simple, yet high-quality DSP processing

[mrkva]

Hello everyone,

I am looking for a solution to this situation:
I have signal ranging from 0 to 300kHz which I'd like to multiply with adjustable sine wave oscillator (which should ideally origin from the chip), do some filtering on it or do a frequency division.
Since my aim is for high fidelity, it should be 16 bit or better processing.
Do you guys maybe have a suggestion for an easy to use DSP chip (preferably with ADC/DAC) which would handle this?
I guess FPGA would be overkill for such simple task.

Thank you!

[mrkva]

So far I have found this one:
http://www.analog.com/en/products/processors-dsp/sigmadsp-audio-processors/adau1761.html

It would be great, but the sampling rate is only 96kHz which means maximum input freq 48kHz.

[danadak]

Take a look at Cypress PSOC, 5LP family. It is a GP processor with a
separate DSP engine in it for filtering work. People have also implemented
FFT on it.

Your sampling rate and precision might be a challenge in PSOC, depends
on what you want to do to the digitized signal. The SAR, 12 bits, can do 1
Mhz, the DelSig can do 16 bits at 48 Ksps. The DSP engine is DMA driven,
so can handle basic filtering in background due to buss architecture and
spoke engine.

www.cypress,com, PSOC 5 family, ,many
videos for training, etc..

Attached an example project, one of many, that are available in  IDE
for examination, use. Note PSOC has I2S module that can be used for
audio work.

Regards, Dana.

[bktemp]

A FPGA is probably the best solution, because 300kHz requires at least 600kHz samplerate. Filtering such a high datarate requires quite a powerful DSP. Even a small FPGA should be able to do it. And you can use an arbitrary bitwidth to fit your quality requirements.

[danadak]

I focused on audio as opposed to the 300 Khz you are asking for. My error.
PSOC cannot handle that at high filter orders.

Attached an example of a 128 Tap FIR that PSOC can do.

Regards, Dana.

[f5r5e5d]

your requirements really need multichip solutions today

you will need pretty good industrial/medical imaging marketed ADC chips for 16 bits and near MHz sampling - Analog Devices PulSAR are one family of ADC with the specs

significant analog support circuitry, separate DSP, DAC chips...

[bson]

Use an external digital downconverter (DDC) and sample the IF, or use a DDC chip with a quality differential ADC built in, add a decent voltage reference and use the DSP to crunch the output.

[Howardlong]

That's do-able on M4(F) if you are smart with your filtering and NCO/mixer algorithms, in particular using polyphase filters. LPC4300 series would certainly do it, I made 2MSa/s direct sampling SDR with one, they are 204MHz multi core devices although I only used one core. The on chip ADCs won't be enough, you'll need an external ADC if you want 16 bit. If you can settle for a 12 bit ADC as I did, you can do it all on chip with an LPC4370.

Without knowing the application it's difficult to give you more direction.

[mach665]

TI's Delfino MCU F28377 should be able to do that. It has 16bit 1.1MSPS ADCs, and powerful floating point engines. Also check F28377s Delfino launchpad.

[sjd.aliyan]

You didnot mentioned your application.
But if it's simple you could use ARM Cortex-m4.Cortex-As are also used fore more powerful DSP processing

[Cerebus]

Or how about staying partially analogue? There are plenty of wide bandwidth multiplying DACs out there. For example the AD5545 is 16 bit has a 2 us settling time and 6.9 MHz bandwidth on the reference input. It's literally the first datasheet off the pile here and was picked just as an example. You'd have to pick through the available parts to find one that suits your application.

[uncle_bob]

Hi

If you are going to do full processing, you will probably need I/Q signal separation. That gets your processing sample rate up to 1.2 Msps in some cases. If you have a "noisy" environment, you likely need a front end filter ahead of your ADC. You can go with a very complex analog filter with modest impact on the sample rate (maybe only 1.5X). If you go with a simple filter (and there's a lot of noise) it will significantly impact your sample rate to get to -96 db (16 bits). The net result could push you into many mega samples per second.

Are there a lot of assumptions and guesses in the stuff above? Sure there are. Without really knowing what you are trying to do, there has to be.

Bob