Help: problem of v22bis software modem by Motorola SDK 2.0E

Hello dear all,

Have someone here successfully implemented the v22bis software modem
library probvided by Motorola SDK 2.0E on a 5685x DSP chip?

I tried hard but could not make it talk with a PC Modem smoothly. The
connection was established for a few seconds and some characters have
been transfered, then the software modem received a lot of garbage
strings. After some retries by the answering software modem, Finally
the connection lost.

The AT command that I set the v22bis modem (should be 2400bps ONLY) in
the hyperterminal is,

at\n0
at+ms = v22b,0,2400,2400,2400,2400

Are they correct? Hope to have your advice. Thanks.
Kelvin

Are you using a loop connection, or direct? If direct, you may
need to attenuate to obtain decent receive levels. I have seen
that V.22bis modems I designed years ago would have problems
during testing with high receive levels. Does either modem require
loop current to properly terminate the line? Some designs will not
operate well without loop current.
Also, if you are using your own op-amp duplexer, try to optimize the
impedance match.

regards,
Darcy Roberts
Ag Instruments.

Kelvin wrote:
> Hello dear all,
>
> Have someone here successfully implemented the v22bis software modem
> library probvided by Motorola SDK 2.0E on a 5685x DSP chip?
>
> I tried hard but could not make it talk with a PC Modem smoothly. The
> connection was established for a few seconds and some characters have
> been transfered, then the software modem received a lot of garbage
> strings. After some retries by the answering software modem, Finally
> the connection lost.
>
> The AT command that I set the v22bis modem (should be 2400bps ONLY) in
> the hyperterminal is,
>
> at\n0
> at+ms = v22b,0,2400,2400,2400,2400
>
> Are they correct? Hope to have your advice. Thanks.
> Kelvin

One other thing to check is the accuracy of the clocks. The
software version may be correcting for clock slips, but this can
be an issue. As you know, there is an async-to-sync conversion
done somewhere in the code(and the reverse). I have found if the
clocks are out enough, then you can get bursts of wrong characters
until the modems synchronize again. Note: this is based on experience
with a 73K224 based (hardware) modem design and may not apply
to you.

Can you try connecting at an FSK rate (V.21 or Bell 103) ? This may
help shed some light on the problem...

Can you configure the software to do an analogue loopback? If the
problem goes away, it could be a clock issue.

Is there an AGC block in the software? Can it be disabled?

Is there an adaptive receive filter in the software? Can it be disabled
or locked at a certain set of tap values?

regards,
Darcy Roberts

Kelvin wrote:
> Thanks for your advice.
>
> I am testing the software on real telephone line now. The test is done
> in this way:
>
> DSP <--> CODEC <--> TELEPHONE INTERFACE <--> CENTER OFFICE <--> PC
> MODEM
>
> I have done something such as DTMF and tone detection on the same
> circuit. Seems okay. What I donot understand is that why the software
> modem could receive the correct characters only in the first few
> seconds then garbage characters.
>
> Cheers,
> Kelvin
>
> Darcy Roberts <> wrote in message news:<>...
>
>>Are you using a loop connection, or direct? If direct, you may
>>need to attenuate to obtain decent receive levels. I have seen
>>that V.22bis modems I designed years ago would have problems
>>during testing with high receive levels. Does either modem require
>>loop current to properly terminate the line? Some designs will not
>>operate well without loop current.
>>Also, if you are using your own op-amp duplexer, try to optimize the
>>impedance match.
>>
>>regards,
>>Darcy Roberts
>>Ag Instruments.
>

Hello Darcy and John,

This is my first modem project and thanks for your enolightening
advice. I have tried all the digital and analog loopback tests on my
target board. All passed.

There is only one crystal on my target board, which provides the dsp
with about 60 MHz clock signal and help codec generate a 7200Hz Frame
Sync of the serial communication between dsp and codec. I checked the
clock for dsp and the frame sync signals. Both of them were very
steady and I did not see anything abnormal.

Shall I change the dsp clock crystal to a more accurate one? I need
also have to look at the application code which may cause the timing
corrupt? Thanks.

Kelvin

You should strive for a carrier frequency of 1800 Hz
+/- 1Hz or better. The modulation rate accuracy should
be +/- 0.01% or better.

Does this CPU use a PLL to generate the 60 MHz? Nudging
the crystal a bit might help here to generate a more
accurate carrier.

regards,
Darcy Roberts

Kelvin wrote:
> Hello Darcy and John,
>
> This is my first modem project and thanks for your enolightening
> advice. I have tried all the digital and analog loopback tests on my
> target board. All passed.
>
> There is only one crystal on my target board, which provides the dsp
> with about 60 MHz clock signal and help codec generate a 7200Hz Frame
> Sync of the serial communication between dsp and codec. I checked the
> clock for dsp and the frame sync signals. Both of them were very
> steady and I did not see anything abnormal.
>
> Shall I change the dsp clock crystal to a more accurate one? I need
> also have to look at the application code which may cause the timing
> corrupt? Thanks.
>
> Kelvin

Correction:

The carrier for V.22bis is 1200 Hz +/- 0.5 Hz (low band)
and 2400 Hz +/- 1 Hz (high band). I was thinking of V.32bis
carriers...

The data rate is to be 2400 bit/s or 1200 bit/s +/- 0.01%
with a modulation rate of 600 baud +/- 0.01%

Start off with a very accurate frequency counter, and get
the software to generate the guard tone of 1800 Hz.

regards,
Darcy Roberts

Kelvin wrote:
> Hello Darcy and John,
>
> This is my first modem project and thanks for your enolightening
> advice. I have tried all the digital and analog loopback tests on my
> target board. All passed.
>
> There is only one crystal on my target board, which provides the dsp
> with about 60 MHz clock signal and help codec generate a 7200Hz Frame
> Sync of the serial communication between dsp and codec. I checked the
> clock for dsp and the frame sync signals. Both of them were very
> steady and I did not see anything abnormal.
>
> Shall I change the dsp clock crystal to a more accurate one? I need
> also have to look at the application code which may cause the timing
> corrupt? Thanks.
>
> Kelvin

Sorry a mistake in my message:

The crystal is about 3.6864 MHz. The CPU PLL Module generates about
120 MHz for internal buses and 60 MHz for external. It seems that
V22bis modem needs pretty accurate clocks for the communication. I
will check the spec of the oscillator.

Thanks.