ΣSigmaStudio Ask AI Русский Log in

the right solution for my this audio project

SigmaDSP & SigmaStudio aMartins Jul 27, 2020 00:00 8 replies Original on forum ↗
Subscribe
aMartins Question
Jul 27, 2020 00:00

hi Forum,

I'm an hardware designer for a couple of decades not never got into DSP audio processing, so will need some help as a start point. I've been searching for the best solution for this new development I'll start soon and was just amazed with the possibilities achieved with SigmaStudio!

So, I'm upgrading a costumer product and need at first to choose the technology to work on; so I need your experience to tell is sigmastudio is the way to go, and eventually witch DSP...

on the audio side I need to apply a simple passband filter (audio coming from microphone), but it should be adaptative: I was thinking of scanning a "narrow" part of the audio band where I will work on and if audio level is strong enough, apply the passband filter to the main frequency detected.
is this possible to do with SigmaStudio?

KJBob Reply
Jul 28, 2020 00:00

Hello,

You can tell from my avatar that I'm an old hardware designer too! With no DSP experience prior to SigmaStudio. You can do some amazing things with it, not necessarily limited to the individual function blocks provided. And if I can't help you with your application, there's several ADI support engineers on this forum who can.

Often there's several ways to accomplish a desired function. And you're improving an existing product with its own set of constraints. Both of these influence your choice among the available SigmaDSP chips with their different capabilities. So to narrow things down, allow me to ask some initial questions:

Is the product battery operated (power constrained)?

Do you prefer a one-chip solution?

How much user control over the DSP's functions is desired?

Is a latency or delay (about 1.5 mS) in audio from analog in to analog out tolerable?

These questions relate to chip features such as low power, self-boot, built-in mic preamps & converters, and low latency. For more info, see the SigmaDSP Selection Guide. If you need low latency, the brand-new ADAU1787 may suit.

Now finally to your first question: Can you make an adaptive-frequency filter with SigmaStudio? If I understand correctly, you need to perform two functions -- find the dominant incoming frequency, and adjust a filter to match it. One way to achieve both functions is to build a bank of fixed filters, measure their outputs and gate through those of sufficient amplitude. A search through this forum will uncover several projects of this sort. Turning off frequency bands not found in the signal helps reduce noise -- so if this is your goal, you'll find this method of interest. And it works with any chip depending upon how many frequency bands are needed.

The more advanced DSPs have FFT capability which could theoretically find the frequency -- yet if I'm not mistaken, it still takes some custom third-party programming to use it. And some can calculate their own coefficients (sets of five numbers that determine a filter). I'm afraid I haven't kept up with what can be done with the algorithm blocks available on the higher-end SigmaDSPs since there's no longer much interest in audio applications where I work.

Best regards,

Bob

aMartins Reply
Jul 29, 2020 00:00

Thanks Bob for your extensive answer, I appreciate the time spent.

here goes more info on this proj:
- battery operated, so minimal power consumption is a concern. From what I've been searchin on A.D. audio DSP (ADAU1701 and related), consumption can goes to 100mA depending on algorithm, that would give less than 10h battery life on a 1000mAh LiPo... but again it will depend on algorithm
- onechip solution (only for DSP) is better yes - I already have a constraint using existing uC connected to a smal LCD+bluetooth control, etc
- user control will be limited to volume and eventually manual band selection (instead of adaptive)
- I will work on a limited frequency range, below 2kHz (eventually <1kHz) - with 3 or 4 passbands to be applied on dominant frequency (only one dominant frequency will be present each time)

Best Regards.
AM

KJBob Reply
Jul 29, 2020 00:00

Hello,

The -1701 would likely work except that it's power-hungry and lacks a built-in mic preamp. The ADAU1761 is low-power, has built-in converters and can handle a mic input, but it doesn't self-boot so you'll need to have your uC boot it via I2C or SPI. This does get a bit complex to implement. The ADAU1787 is designed as a one-chip solution for noise-cancelling headphones, thus it has everything -- low power, low latency, self-boot, and built-in converters. It has two cores -- a "normal" one and an additional, super high speed core. I'm not yet familiar with this chip.

Best regards,

Bob

aMartins Reply
Jul 30, 2020 00:00

thanks for the suggestion Bob,

booting from ext EEPROM is not actually a problem.

I've been reading different ADAU options and 1787 seems to be one that meets my requirements, except for the high EvalBoard price :) - will look at it closely!

Ken.M Reply
Jul 30, 2020 00:00

Before you get too far along, you should be aware that the 1787 is only available in wafer-level packaging (WLCSP). It requires a layout with via-in-pad and not all assembly houses can place it. That's part of the reason for the expensive eval board. It is a great part, but it is a little tricky to use in small volumes.

aMartins Reply
Jul 30, 2020 00:00

right,Thanks.
missed that on the first view of the datasheet: I assumed it should have 2/3 different footprints I could choose from...

KJBob Reply
Jul 30, 2020 00:00

Sorry I should have thought about that. I work for a contract manufacturer & have roles besides engineering -- for example X-ray inspection of BGAs. I've seen some tiny ones which we can mount, yet I'm not familiar with this WLCSP. The micro-BGAs I've seen have only nine pads so it's possible to route them on the top layer without vias. What I find really interesting about the ADAU1787 is not its features optimized for noise-cancelling 'phones -- ADI has been in that market for awhile. It's all its HW accelerated log, trig, and complex math functions as if the chip were destined for instrumentation of some sort. For one example -- a member here had an application to measure phase difference between channels of a professional tape deck in order to align the heads. It's quite possible that a method we tried to cram into a -1701 would work splendidly in a -1787. As John Hammond said: "Spare no expense!"

Now for your application, consider the ADAU1761. It's a sweet part with a good selection of SigmaStudio algorithms available as well as programmable gain amplifiers and such. Its eval board is half the cost. With it you can cut your teeth on SigmaStudio and, after making the usual beginner's mistakes as many of us have before you, can build a useful system. It may not do everything you're thinking of (or it may do more!). Either way, it can result in an improved product. Booting it from your uC will be a challenge but not impossible as long as your uC (or an upgraded one) has the needed I2C or SPI bus and sufficient memory. Learning to do it is like a rite of passage for engineers. My partner at work has booted a gaggle of six -1701's from one PIC micro. Once you have a working project, SigmaStudio can export a set of files containing the data which your micro needs to store and then deliver to the -1761 at boot time as a bunch of block writes -- some small and some large. Once booted, your uC can control many DSP functions via the I2C / SPI bus, but I'm getting ahead of myself.

I still wish that ADI would make a low-power 1701, or a self-booting -1761, or something like that.

Best regards.

Bob

aMartins Reply
Aug 4, 2020 00:00

Thanks to all once again.

After lot of research, I still don't have a clear idea of where to go

and I don't think I will have until I start "doing it", so will order 1701 devboard as a start point.

Best Regards!