|
|
Вот, как то делал подборку по этой теме.
Reference:
http://groups.google.com.ru/groups?hl=ru&lr=&ie=UTF-8&oe=UTF-8&threadm=uEg85.36%24Zp1.2239%40news.get2net.dk&rnum=1&prev=/groups%3Fq%3DAD73322%26hl%3Dru%26lr%3D%26ie%3DUTF-8%26oe%3DUTF-8%26selm%3DuEg85.36%2524Zp1.2239%2540news.get2net.dk%26rnum%3D1
От:Nienke Bloem (bloem@wxs.nl)
Заголовок:AD73322 codec problems
View this article only
Группы новостей:comp.dsp
Число:2000/07/03
Hi there,
I have a problem with the AD73322 codec from analog devices. I use this codec in combination with a ADSP218x dsp and the problem occurs when I continuously send 0x0000 to the codec: it starts to make instationary chirping sounds. Does anybody experienced simular problems?
Thanks
Erik van der Tol
bloem@wxs.nl
Message 2 in thread
От:Theo Gjaltema (t.gjaltema@wxs.nl)
Заголовок:Re: AD73322 codec problems
View this article only
Группы новостей:comp.dsp
Число:2000/07/03
Indeed we had the same problems, it does not even matter what you send to
the codec as long as it is the same value. The chirping sounds even
continue when the DSP is halted by the ICE debugger. We also mentioned that
the frequency components of the chirping sounds go way beyond the sample
frequency used.
We did not find a solution yet. If you find any, please let us know.
Theo Gjaltema
t.gjaltema@planet.nl
Message 4 in thread
От:Jan Rørgård Hansen (jrh@person.dk)
Заголовок:Re: AD73322 codec problems
View this article only
Группы новостей:comp.dsp
Число:2000/07/04
Hi
I have also used the AD73322, this has caused so many problems for my
application, that I do not intend to use it again. If I had read the data
sheet in more detail before choosing the AD73322 I might have seen the
problems before it was necessary to use a lot of time to solve them.
Therefore you might benefit from reading the following if you intend to use
the AD73322:
Analog Devices writes in the datasheet that: "The Ad73322 is particularly
suitable for a variety of applications in the speech and telephony area,
including low bit rate, ....."
In my oppinion this is not the case because in telephony applications it
would be desirable to run the AD73322 at a sample rate of 8000samples/s.
However at this samplerate the attenuation at 3000Hz is 7dB and at 3500Hz it
is 9.5dB. The 3dB attenuation occurs at 2000Hz, which means that the speech
band width is much lower than acceptable.
The datasheet also says on page 28: "In the case of voiceband processing
where 4Khz represents the Nyquist frequency, if the signal to be measured
were externally bandlimited then an 8kHz sampling rate would suffice.
However if it is required to limit the bandwidth using a digital filter,
then it may be more appropriate to use an initial sample rate of 16kHz and
to process this sample stream with a filtering and decimation algorithm to
achieve a 4kHz bandlimited signal at 8kHz rate."
This means: You either have to filter the signal in hardware or in software,
the AD73322 does almost nothing for you.
On page 29 about the on-chip filtering: "The spectral response of a
sinc-cubed filter shows the characteristics nulls at integer intervals of
the sampling frequency. Its passband characteristic (up to Nyquist
frequency) features a roll-off that continues up to the sampling frequency,
where the first null occurs. In many applications this smooth response will
not give sufficient attenuation of frequencies outside the band of interest
therefore it may be necessary to implement a final filter in the DSP which
will equalize the passband rolloff and provide a sharper transition band and
greater stopband attenuation."
The second sentense again says it all, the AD73322 does nothing for you, it
is necessary to do all the filtering in the DSP to avoid high level
frequency components mirrored around the Nyquist frequency.
On top of this the AD73322 in my configuration with 16kHz sampling rate
provides me with a DC component being only 40dB below the level of the
actual signal.
Прочитать остаток сообщения... (еще 23 строк)
Message 5 in thread
От:Jan Johansson (redbaron@speedy.ludd.luth.se)
Заголовок:Re: AD73322 codec problems
View this article only
Группы новостей:comp.dsp
Число:2000-10-02 07:09:04 PST
In a current design i have chosen to use the ad73322 codec Also, how intensive filtering is needed to obtain a decent Any information appreciated Thanks in advance Jan Johansson The AD73322 is very easy to interface with the ADSP218x series and you are The AD73322 performs excellent in terms of low group delay and linear phase I did some testing trying to compensate for the low bandwidth and using a Hope this was a help
E-mail:
"Jan Rorgård Hansen"
: Hi
: I have also used the AD73322, this has caused so many problems for my
: application, that I do not intend to use it again. If I had read the data
: My conclusions are: The area of application for this chip does not include
: voice band telephony, since you have to go through a lot of trouble to
: actually get a decent frequency response in the voice band (300Hz-3400Hz).
Reading this has discouraged me but since the product is
already in the prototype stage i wonder if there are any
real alternatives to the 73322.
Considering price and the wish for easy interfacing to the
ADSP218x series codecs.
freq-response for voiceband telephony applications?
Message 6 in thread
От:Jan Rørgård Hansen (jrh@person.dk)
Заголовок:Re: AD73322 codec problems
View this article only
Группы новостей:comp.dsp
Число:2000-10-03 00:43:20 PST
right it is hard to find alternatives although I have seen that Texas
Instruments are planning to make a DSP codec with a wide range of sample
frequencies and word lengths to choose from.
Actually I have gone through a lot of trouble trying to find another CODEC
matching our demands to variable sample rate, easy interface, low group
delay and cascading of up to 8 CODECs but did not find an alternative. Now
that we are aware of the bad things about the AD73322 we have been able to
take this into account in our new product.
if this is of any interest, but you need to be aware of 2 things:
1) The 3dB cut off frequency is located in the frequency domain at 0.25*fs
both in the DAC and in the ADC, so if you do a loop back you get 6dB
attenuation at 2kHz if using a sample rate of 8kHz.
2) The CODEC does not include a high pass filter so if you plan to do any
filtering you have to make sure that the DC offset does not accumulate
through your filter.
5tap FIR filter increased the 3dB cutoff frequency to 0.365*fs; (0.344*fs in
loop back). This of course implies that you scale down the digital
representation of the signal so that overflow does not occur and then use
analog gain if necessary.
Jan
Ответы
info@telesys.ru
