ADAU1701 possibilities regarding FIR calculation

Hi,
I plan to build a project related to a (long) FIR filter implementation [SPDIF In -> DSP for FIR filtering -> SPDIF Out that will feed a separated DAC chip].

Maybe the FreeDSP boards using the ADAU1701 could be a solution, but I have read in this thread, post#30, as a reply to FIR filter calculation, that:
ADAU1701 works with Rephase out of the box so to speak, limited amount of taps available though <500.

and it is unclear if the 500 Taps limit is due to Rephase, to the ADAU1701 or to the FreeDSP boards ??

I don't plan to work with Rephase, but if the ADAU1701 cannot perform more than 500 Taps in a FIR Filtering, I need to know and identify a more powerful DSP.

Thank you for your reply,
 
I've stuck with the 1701 and IIR filters so far. You could go with an RPI 4 and something like CamillaDSP or a PC. You could also look at the miniDSP 2X4 HD. I'd recommend you look at the Rephase thread to see the pros and cons of using a very large number of taps.

PS: As always, it depends on how much you're willing to spend!
 
Hi,
I plan to build a project related to a (long) FIR filter implementation [SPDIF In -> DSP for FIR filtering -> SPDIF Out that will feed a separated DAC chip].

Maybe the FreeDSP boards using the ADAU1701 could be a solution, but I have read in this thread, post#30, as a reply to FIR filter calculation, that:


and it is unclear if the 500 Taps limit is due to Rephase, to the ADAU1701 or to the FreeDSP boards ??

I don't plan to work with Rephase, but if the ADAU1701 cannot perform more than 500 Taps in a FIR Filtering, I need to know and identify a more powerful DSP.

Thank you for your reply,

With ADAU1466 you can have 2 FIR filters with max 10000 taps each.
Or 4 FIR filters with 5000 taps each.

You can also simulate it on Sigma Studio with all available sigmaDSP chips without any board attached.
 
You could go with an RPI 4 and something like CamillaDSP or a PC. You could also look at the miniDSP 2X4 HD
I plan to integrate a DSP in my own designs, so a commercial product such as the miniDSP is useless for me.

I'd recommend you look at the Rephase thread to see the pros and cons of using a very large number of taps.
Can you give me the exact URL ? (because there are various threads related to Rephase).

With ADAU1466 you can have 2 FIR filters with max 10000 taps each.
Or 4 FIR filters with 5000 taps each.

This is a VERY interesting information for me !

10000 Taps for a FIR filter will be enough, and I suppose that I could use 2xAD1466 in //, one for Left channel and the other for the Right channel, in case I need more than 5000 Taps ? or in case I want to separate Left and Right ?

Your reply also let me think that Sigma Studio (which is free) can undertake more powerful DSPs than the ADAU1701 (limited to 500 Taps).

The problem is that ADAU1466 has a 32-LSCSP-VQ package : how to solder it ?? Is there an adaptor to DIP for a Through hole PCB ?

To date, I have already managed to solder a SRC4392 having a 48-PQFP package, but thepins are visible: no idea how to solder the 32-LSCSP-VQ package ?!?.

Is there the list of the supported DSP somewhere ? (in order to identify the powerful DSPs that are supported by Sigma Studio, although the ADAU1466 looks like a excellent candidate).

Maybe you know the most powerful DSPs that are supported by Sigma Studio ?

Thank you very much,
 
Last edited:
Most commercial products will probably use either the ADAU1450/51 or 52.
Which is what the vast majority is using.

I don't see why so many taps are needed. Just use enough to fit the response. That being said the 1701/1401 is a little to small in practice.

Good luck with sourcing any of those. Because of the current chip shortages nothing is in stock. Lead times are around 40-50 weeks.
 
Damned, you are right : I have just checked the availability (Mouser, Farnell, Digikey, RS): the planned delivery date is around April 2024 !!
so we can continue to discuss, but the solder iron is on vacation for more than 1 year !
If you can't get sigma studio chips a year is long enough to learn how to write the algorithms for another DSP from scratch. IIR and FIR filters are very easy and they usually have libraries to help.
 
Damned, you are right : I have just checked the availability (Mouser, Farnell, Digikey, RS): the planned delivery date is around April 2024 !!
so we can continue to discuss, but the solder iron is on vacation for more than 1 year !
Eval Boards are available; they have everything you need in order to test what you want with Sigma Studio.
Board Converters are a bit limited in noise and THD (when compared to CS4398/CS5361 for (old) example).
Once programed, the board could self-boot; no more need for Sigma Studio in that case.

https://pt.mouser.com/ProductDetail/Analog-Devices/EVAL-ADAU1466Z?qs=BZBei1rCqCDrvhRjMpIH0w==
https://www.digikey.com/en/products/detail/analog-devices-inc/EVAL-ADAU1466Z/7915254
 
If you can't get sigma studio chips a year is long enough to learn how to write the algorithms for another DSP from scratch.
I will not learn from scratch because it is not my final target: although I have understood how to program a FIR Filter (I have learnt convolution and etc...) I prefer to have an IDE that will save time on the long run: I will have to evauate different FIR filters, so I prefer to have a software tool allowing to rapidly perform the changes.

But the most important to me is to solve all the connection problems:
-electrically: to get a complete schematic from the SPDIF to the DSP : it is still unclear what is the proper way to connect both on a PCB, and I have not yet found a simple example diagram. I assume that the DSP should have a proper output to be sent to the DAC chip
-software side: how to define these DSP links in Sigma studio, because I have never used Sigma Studio...

Eval Boards are available; they have everything you need in order to test what you want with Sigma Studio.
100% agree Toino ! 👍

it is exactly what I thought about: use the 1 year unavailability period, to work with a DSP evaluation board, in order to understand how it works, and how to implement a FIR filter with Sigma Studio .

If I success to use an evaluation board, then I will be able to design my own PCB based on the evaluation board diagram (the diagram is available I believe), I use to design my own PCBs.
So an ADAU1466 evaluation board will soon be ordered !

By the way, I have found the list of the Sigma Studio supported DSPs here: https://wiki.analog.com/resources/tools-software/sigmastudio/supportedic
but I don't understand the relation between the reference of a DSP and its power ???

For example the ADAU1466 is more powerful than the ADAU1701 : so is there a "method" to evaluate the power from the chip number ?

Thank you !
 
You really need to look through each datasheet to determine the maximum number of FIR taps. For example, the 1701 normally (when operating at 48 KHz sampling rate) has 1024 steps per cycle. So the theoretical maximum number of taps is 1024. The 1466 has 6144 steps, but has a quad core MAC which allows it to do 4 taps per instruction. So it can process approximately 24,000 taps. Here's a LINK that describes it.

Do be aware that all the other components (IIR filters, volume controls, etc.) also use instructions that will in turn reduce the number of available FIR taps. And if you're processing two channels your maximum # of taps is halved.
 
If you can't get sigma studio chips a year is long enough to learn how to write the algorithms for another DSP from scratch. IIR and FIR filters are very easy and they usually have libraries to help.
You're the first person ever that said that Laplace transformations are easy.

I am not saying it's not possible, but if people don't have at least a medium level of knowledge of math, that's gonna be a tough cookie. On top of all the programming one has to do, and more importantly has to figure out.
 
You really need to look through each datasheet to determine the maximum number of FIR taps. For example, the 1701 normally (when operating at 48 KHz sampling rate) has 1024 steps per cycle. So the theoretical maximum number of taps is 1024. The 1466 has 6144 steps, but has a quad core MAC which allows it to do 4 taps per instruction. So it can process approximately 24,000 taps. Here's a LINK that describes it.

Do be aware that all the other components (IIR filters, volume controls, etc.) also use instructions that will in turn reduce the number of available FIR taps. And if you're processing two channels your maximum # of taps is halved.

True. However the Taps limit is imposed by Sigma Studio itself; max limit is 10000-Taps for each FIR.
 
You really need to look through each datasheet to determine the maximum number of FIR taps.
Thank you for the link !
For my first project, I will stay with the 1466 (it is powerful enough) and more important: I am sure to get an evaluation board to begin my project before 2024... 🙂

And if you're processing two channels your maximum # of taps is halved.
I have seen a few designs using 2 identical DSP in //, one for the Left channel and the other for the Right channel: it could be a solution (even if I don't believe I need it).

I don't know your experience level, but designing a proper DSP board is not something you "just" do.
I am more afraid by understanding how to program the DSP using Sigma Studio, understanding how to interface the DSP with its surrounding parts, than to design the PCB: but you are surely right, and I will ask for the main guidelines before proceeding !
I always remain "modest" in order to not be disappointed by myself !😎

If you can't get sigma studio chips a year is long enough to learn how to write the algorithms for another DSP from scratch. IIR and FIR filters are very easy and they usually have libraries to help.
You're the first person ever that said that Laplace transformations are easy.
@b_force: Not do I !! I agree with you, and I don't want to spend my time in such programming (even if my level in mathematics might be enough): I hope that Sigma Studio can do this in a graphic approach with its bundled components ?
I have seen a "FIR filter component" in Sigma Studio: I hope that it does what I imagine...

Can you confirm ?