Cirrus Codec CS4270

I am considering using the Cirrus CS4270 codec in a new design and
wanted to ask if anyone had used this before and if it had any gotchas
that weren't described in the errata sheet.

I was going to use an older AK4552 because of the tiny TSSOP16 package
(without going to a leadless part). But I realized that it isn't ROHS
compliant, so its out with yesterday's fish.

The only other codecs I have used are TI parts and they proved to be a
relative nightmare from the software side. I think there are
something like 100 registers to setup. The CS4270 can be used in a
stand alone mode and still have control over the sample rate using a
couple of pins.

One thing that I am not clear about on both of these parts is the
single/double/quad speed "modes". The AKM part seems to not actually
sample above 50 kHz since it looks like it is saying that in double or
quad speed mode the input/output data is decimated so the higher
sample rate doesn't buy you much. The CS4270 seems to actually
provide the full sample rate in double and quad speed modes and so
should provide a wider frequency response.

I don't need the wider response for this application, but I have been
asked before about ultrasonic audio and this looks like a nice part
for that too.

Anyone used the CS4270 before and can tell me about your experience?

"rickman" <> wrote in message
news:f3e60e4b-b29e-49db-a108-...
> I am considering using the Cirrus CS4270 codec in a new design and
> wanted to ask if anyone had used this before and if it had any gotchas
> that weren't described in the errata sheet.

I didn't use 4270 however I have used many other similar codecs from Currus.
Here are my impressions:

0. Cirrus codecs can be put to work however it may take a week or two of
effort.
1. Cirrus codecs have hardware bugs, especially if you are using the
features like gain control, etc. So I would advise not using the features.
2. Currus datasheets have bad mistakes and inconsistensies, like wrong
registers and/or incorrect I2C addresses.
3. Wrong programming makes Cirrus codecs to do really weird things like the
input only pins become the outputs. There is absolutely no information about
that.
4. The performance is not outstanding by any of the parameters.
5. Cirrus support is good and accurate, but their typical response time is
2-3 weeks.

> The only other codecs I have used are TI parts and they proved to be a
> relative nightmare from the software side. I think there are
> something like 100 registers to setup.

Well, not as many as 100, but something like 32 registers for sure, and the
TI documentation is not an easy reading. However the TI codecs work exactly
as described.

> The CS4270 can be used in a
> stand alone mode and still have control over the sample rate using a
> couple of pins.

You may be unpleasantly surprised. I advise you to use the jumpers to select
the different config options as well as connect I2C or SPI to it.

> One thing that I am not clear about on both of these parts is the
> single/double/quad speed "modes". The AKM part seems to not actually
> sample above 50 kHz since it looks like it is saying that in double or
> quad speed mode the input/output data is decimated so the higher
> sample rate doesn't buy you much. The CS4270 seems to actually
> provide the full sample rate in double and quad speed modes and so
> should provide a wider frequency response.

They all have the response to 85%...90% of Nyquist of the selected sample
rate, however the higher sample rate doesn't buy any improvement in the SNR.

> I don't need the wider response for this application, but I have been
> asked before about ultrasonic audio and this looks like a nice part
> for that too.

What is the application?
What are the performance parameters?

Vladimir Vassilevsky
DSP and Mixed Signal Consultant
www.abvolt.com

On Dec 22, 4:33 pm, "Vladimir Vassilevsky"
<antispam_bo...@hotmail.com> wrote:
> "rickman" <gnu...@gmail.com> wrote in message
>
> news:f3e60e4b-b29e-49db-a108-...
>
> > I am considering using the Cirrus CS4270 codec in a new design and
> > wanted to ask if anyone had used this before and if it had any gotchas
> > that weren't described in the errata sheet.
>
> I didn't use 4270 however I have used many other similar codecs from Currus.
> Here are my impressions:
>
> 0. Cirrus codecs can be put to work however it may take a week or two of
> effort.
> 1. Cirrus codecs have hardware bugs, especially if you are using the
> features like gain control, etc. So I would advise not using the features.
> 2. Currus datasheets have bad mistakes and inconsistensies, like wrong
> registers and/or incorrect I2C addresses.
> 3. Wrong programming makes Cirrus codecs to do really weird things like the
> input only pins become the outputs. There is absolutely no information about
> that.
> 4. The performance is not outstanding by any of the parameters.
> 5. Cirrus support is good and accurate, but their typical response time is
> 2-3 weeks.

This is all interesting and scary to hear. Maybe I should avoid the
Cirrus parts. Mainly I picked them because they came up as in stock
at Digikey and the package fits my needs. I prefer a leaded part and
don't have a lot of board space. Originally I was going to use the
AKM part as it comes in a 16 pin TSSOP which fits nicely, but it is
old enough that it is not ROHS. A lot of the TI parts are in larger
QFPs or leadless packages.

The requirement for this effort is just standard audio freq, 8 or 48
kHz SPS (two modes). It would also be nice to have 16 and 32 kSPS
just in case 8 kSPS in case a third mode is needed. Extreme audio
fidelity is not required, 80+ dB of SNR is good enough. A lot of
converters seem to give you 24 bits, but I'm not clear on what the
extra bits buy you in parts that are worse than 96 dB of SNR or
dynamic range.


> > The only other codecs I have used are TI parts and they proved to be a
> > relative nightmare from the software side. I think there are
> > something like 100 registers to setup.
>
> Well, not as many as 100, but something like 32 registers for sure, and the
> TI documentation is not an easy reading. However the TI codecs work exactly
> as described.

The part I used had close to 100 registers! The software guy was
pretty ticked at me about it so I had to help him set it up. Turns
out you can ignore most of them as they default to typical settings.
But you have to read them all to figure out which to ignore and which
to set.


> > The CS4270 can be used in a
> > stand alone mode and still have control over the sample rate using a
> > couple of pins.
>
> You may be unpleasantly surprised. I advise you to use the jumpers to select
> the different config options as well as connect I2C or SPI to it.

From what you have said, I think I want to avoid the Cirrus parts.
That was why I asked about it here. Thanks...

The control signals all go to an FPGA, so they can be used either
way. But it is much simpler to use the stand alone mode over I2C and
I don't want to push things into the software. That just spreads the
"fun" around too much.


> > One thing that I am not clear about on both of these parts is the
> > single/double/quad speed "modes". The AKM part seems to not actually
> > sample above 50 kHz since it looks like it is saying that in double or
> > quad speed mode the input/output data is decimated so the higher
> > sample rate doesn't buy you much. The CS4270 seems to actually
> > provide the full sample rate in double and quad speed modes and so
> > should provide a wider frequency response.
>
> They all have the response to 85%...90% of Nyquist of the selected sample
> rate, however the higher sample rate doesn't buy any improvement in the SNR.

But that depends on what Nyquist is compared to what they are calling
the "sample" rate. If they internally interpolate before sending to
the DAC, you are still bandwidth limited to the data rate you can pump
into the part. The AKM data sheet is very terse and it looks to me
like in double speed mode you can only use half the "samples" and in
quad speed mode you can only use a quarter of them. So no extra
bandwidth.


> > I don't need the wider response for this application, but I have been
> > asked before about ultrasonic audio and this looks like a nice part
> > for that too.
>
> What is the application?
> What are the performance parameters?

Strictly audio 20 Hz to 20 kHz with nothing special in SNR or dynamic
range. I figure 80+ dB is ok. But I would like to get familiar with
a part that can sample at 96 or 192 kSPS with corresponding increases
in bandwidth.

"rickman" <> wrote in message
news:a237043d-32f4-4bb1-954c-...
> On Dec 22, 4:33 pm, "Vladimir Vassilevsky"
> <antispam_bo...@hotmail.com> wrote:

> > 0. Cirrus codecs can be put to work however it may take a week or two of
> > effort.
> > 1. Cirrus codecs have hardware bugs, especially if you are using the
> > features like gain control, etc. So I would advise not using the
features.
> > 2. Currus datasheets have bad mistakes and inconsistensies, like wrong
> > registers and/or incorrect I2C addresses.
> > 3. Wrong programming makes Cirrus codecs to do really weird things like
the
> > input only pins become the outputs. There is absolutely no information
about
> > that.
> > 4. The performance is not outstanding by any of the parameters.
> > 5. Cirrus support is good and accurate, but their typical response time
is
> > 2-3 weeks.
>
> This is all interesting and scary to hear.

Some more notes:

6. The EMC is terrible. Needs really good layout and heavy decoupling. Do
not separate the planes.
7. The ESD can hang up the codec. The hardware reset is required to get out
of that.
8. The power dissipation is quite high especially in 96 and 192kHz modes.

> Maybe I should avoid the
> Cirrus parts.

Well. Cirrus codecs can be put to work, they are available and not too
expensive.

> The requirement for this effort is just standard audio freq, 8 or 48
> kHz SPS (two modes). It would also be nice to have 16 and 32 kSPS
> just in case 8 kSPS in case a third mode is needed. Extreme audio
> fidelity is not required, 80+ dB of SNR is good enough.

Looks like fairly modest requirements. A cheap codec from NXP would probably
do.

> A lot of
> converters seem to give you 24 bits, but I'm not clear on what the
> extra bits buy you in parts that are worse than 96 dB of SNR or
> dynamic range.

I agree. This is mainly for the compatibility with I2S modes.

> The control signals all go to an FPGA, so they can be used either
> way. But it is much simpler to use the stand alone mode over I2C and
> I don't want to push things into the software. That just spreads the
> "fun" around too much.

It is always nice to hear the voice of experience


Vladimir Vassilevsky
DSP and Mixed Signal Consultant
www.abvolt.com


> > > The CS4270 seems to actually
> > > provide the full sample rate in double and quad speed modes and so
> > > should provide a wider frequency response.

Yes, this is the correct assertion.