ADSP-CM408F and AD1938/1939, SPORT I2S handling

Folks,

Has anyone ever dealt with the new Analog Devices ADSP-CM processor 
series in connection with a codec such as the AD1938 which uses SPORT I2S?

According to our software engineer there isn't much in terms of examples 
available and we are right now mostly trying to get our hands around 
which signal connections we need and then how to best go about creating 
a driver.

We need to operate in full duplex at 48ksps on all four ADC channels and 
on two DACs (but as AFAIU it has to internally run 8ch/8ch for that with 
lots or zeroes).

-- 
Regards, Joerg

http://www.analogconsultants.com/

Joerg:
> Folks,
> 
> Has anyone ever dealt with the new Analog Devices ADSP-CM processor 
> series in connection with a codec such as the AD1938 which uses SPORT
I2S?
> 
> According to our software engineer there isn't much in terms of
examples 
> available and we are right now mostly trying to get our hands around 
> which signal connections we need and then how to best go about
creating 
> a driver.
> 
> We need to operate in full duplex at 48ksps on all four ADC channels
and 
> on two DACs (but as AFAIU it has to internally run 8ch/8ch for that
with 
> lots or zeroes).
> 
> -- 
> Regards, Joerg
> 
> http://www.analogconsultants.com/
Which processor are you using (exact part number) - I've looked at these
for a client but I'm waiting until they have actually got them for sale
from distribution before doing any more work. If you really have to
start design work so early I'd be really tempted to put a little FPGA
between the uP and the AD1938 so that the bridge between AD's actual
silicon and your intepretation of their intentions can be bridged.

MK

Michael Kellett:
> Joerg:
>> Folks,
>>
>> Has anyone ever dealt with the new Analog Devices ADSP-CM processor
>> series in connection with a codec such as the AD1938 which uses
SPORT
> I2S?
>>
>> According to our software engineer there isn't much in terms of
> examples 
>> available and we are right now mostly trying to get our hands around
>> which signal connections we need and then how to best go about
> creating 
>> a driver.
>>
>> We need to operate in full duplex at 48ksps on all four ADC channels
> and 
>> on two DACs (but as AFAIU it has to internally run 8ch/8ch for that
> with 
>> lots or zeroes).
>>
>> --
>> Regards, Joerg
>>
>> http://www.analogconsultants.com/
> Which processor are you using (exact part number) - I've looked at
these
> for a client but I'm waiting until they have actually got them for
sale
> from distribution before doing any more work. If you really have to
> start design work so early I'd be really tempted to put a little FPGA
> between the uP and the AD1938 so that the bridge between AD's actual
> silicon and your intepretation of their intentions can be bridged.
> 
> MK

I meant to say:

so that the gulf between AD's actual 
silicon and your intepretation of their intentions can be bridged

MK

On 2015-04-28 1:21 AM, Michael Kellett wrote:
> Michael Kellett:
>> Joerg:
>>> Folks,
>>>
>>> Has anyone ever dealt with the new Analog Devices ADSP-CM processor
>>> series in connection with a codec such as the AD1938 which uses
> SPORT
>> I2S?
>>>
>>> According to our software engineer there isn't much in terms of
>> examples
>>> available and we are right now mostly trying to get our hands around
>>> which signal connections we need and then how to best go about
>> creating
>>> a driver.
>>>
>>> We need to operate in full duplex at 48ksps on all four ADC channels
>> and
>>> on two DACs (but as AFAIU it has to internally run 8ch/8ch for that
>> with
>>> lots or zeroes).
>>>
>>> --
>>> Regards, Joerg
>>>
>>> http://www.analogconsultants.com/
>> Which processor are you using (exact part number) -


We are using the ADSP-CM408F. Later we might back off to a less powerful 
version but only after we know how heavy the processor workload ends up 
being.


>  ... I've looked at
> these
>> for a client but I'm waiting until they have actually got them for
> sale
>> from distribution before doing any more work. ...


That has me concerned a wee bit as well, considering that it's been more 
months than usual from announcement to devices showing up in the sales 
channels. But when adherig to the CMSIS libraries for the most part it 
should not be too difficult to switch to another brand in case things 
hit the fan.


>>                                        ... If you really have to
>> start design work so early I'd be really tempted to put a little FPGA
>> between the uP and the AD1938 so that the bridge between AD's actual
>> silicon and your intepretation of their intentions can be bridged.
>>

It's a small team, essentially me and a SW engineer. Neither of us is 
very used to FPGA so we'll do the Sauerkraut thing. Lots of non-install 
resistors that can be used as a switchboard. So far it seems we can run 
full duplex on just ABCLK, ALRCLK, ASDATA1 and DSDATA1. The B clocks 
from the ADSP-CM408F will be ferried over via resistors in case the 
shunting inside the ARM processor won't work for some reason (according 
to AD it should).

Right now we are using their EZ-Kit boards to save some NRE and to get 
to a prototype stage early. Later we'll ditch all unwanted stuff and 
make a single clean board. With or without the ADSP, depending on 
product availability. But definitely with the AD1938 because that has 
DC-capability.



>> MK
>
> I meant to say:
>
> so that the gulf between AD's actual
> silicon and your intepretation of their intentions can be bridged
>

:-)

What was a bit disappointing is the scarcity of example code and 
obtaining answers seems to take a long time. Some of the advertised 
features such as USB aren't really available unless one buys an 
expensive RTOS license on top of the fat four digits we paid for the 
compiler. So we'll be using SCADA software via RS232 for the GUI.

-- 
Regards, Joerg

http://www.analogconsultants.com/

Den mandag den 27. april 2015 kl. 22.16.24 UTC+2 skrev Joerg:
> Folks,
> 
> Has anyone ever dealt with the new Analog Devices ADSP-CM processor 
> series in connection with a codec such as the AD1938 which uses SPORT I2S?
> 
> According to our software engineer there isn't much in terms of examples 
> available and we are right now mostly trying to get our hands around 
> which signal connections we need and then how to best go about creating 
> a driver.
> 
> We need to operate in full duplex at 48ksps on all four ADC channels and 
> on two DACs (but as AFAIU it has to internally run 8ch/8ch for that with 
> lots or zeroes).
> 

it wouldn't suprise me if the SPORTs are reuse of something used in 
something like blackfin. So Id look there first

a scope/logicanalyser lots of coffee and it shouldn't be too hard, the 
AD SPORTs are generally very flexible 


-Lasse

On 2015-04-28 4:19 PM, lasselangwadtchristensen@gmail.com wrote:
> Den mandag den 27. april 2015 kl. 22.16.24 UTC+2 skrev Joerg:
>> Folks,
>>
>> Has anyone ever dealt with the new Analog Devices ADSP-CM processor
>> series in connection with a codec such as the AD1938 which uses SPORT I2S?
>>
>> According to our software engineer there isn't much in terms of examples
>> available and we are right now mostly trying to get our hands around
>> which signal connections we need and then how to best go about creating
>> a driver.
>>
>> We need to operate in full duplex at 48ksps on all four ADC channels and
>> on two DACs (but as AFAIU it has to internally run 8ch/8ch for that with
>> lots or zeroes).
>>
>
> it wouldn't suprise me if the SPORTs are reuse of something used in
> something like blackfin. So Id look there first
>

We are essentially copying methods from a Blackfin implementation 
because that seems to be the only available information right now. We'll 
be using TDM.


> a scope/logicanalyser lots of coffee and it shouldn't be too hard, the
> AD SPORTs are generally very flexible
>

And one Guinness in the evening :-)

Also, both at the SW engineer and at my lab we are working under strict 
canine supervision which helps.

-- 
Regards, Joerg

http://www.analogconsultants.com/