Prev: Xilinx Pipelined/Streaming FFT Architecure?
Next: Shannon's upper bound for AWGN channel, but what about Multipath?
From: Vladimir Vassilevsky on 7 Apr 2010 06:40


csilva1975 wrote:


> Also, after the signal is generated it will be upmixed with a 461,538 Khz
> signal to give an output at 455 khz to be filtered by a ceramic filter
> (hello analog world).

And, I guess, the 455 kHz is going to be upconverted somewhere.
Instead of all that, the elegant solution would be generate FM at the
target frequency directly by means of DDS chip controlled by MCU. The
new DDS chips from AD do linear interpolation, so the sample rate does
not have to be very high. You don't have to use DSP, general purpose MCU
would probably do.

Vladimir Vassilevsky
DSP and Mixed Signal Design Consultant
http://www.abvolt.com
From: csilva1975 on 7 Apr 2010 10:19
>
>And, I guess, the 455 kHz is going to be upconverted somewhere.
>Instead of all that, the elegant solution would be generate FM at the
>target frequency directly by means of DDS chip controlled by MCU. The
>new DDS chips from AD do linear interpolation, so the sample rate does
>not have to be very high. You don't have to use DSP, general purpose MCU
>would probably do.
>
>Vladimir Vassilevsky
>DSP and Mixed Signal Design Consultant
>http://www.abvolt.com
>

Hi.

Yup you guessed it. I will upconvert it with a ne602 controlled by the pwm
output of the DSP/MCU.

I have already looked at those chips from AD, and they look great,
unfortunately I don't have any fancy solder machines to prototype a
surface mount boards, so I have to limit myself to use *old* DIP chips, and
the DDS chips don't come in that format.

I believe that TI has a chip with DDS, double balanced mixer and quadrature
input bundled together. I cannot use it because of what I mentioned
earlier, but maybe someone reading has use for them.

I better get back to work (not related with DSP)!

Cheers.
Cesar

From: Clay on 7 Apr 2010 10:28
On Apr 6, 7:05 pm, "csilva1975" <cesard(a)n_o_s_p_a_m.hotmail.com>
wrote:
> >Hello Cesar,
>
> >Both my 1st and 2nd oscillator papers describe a method for easily
> >modulating the oscillator's frequency to generate FM. You set up a
> >quadrature osc at you center frequency (carrier) and then after you do
> >each iteration on the carrier frequency, you then use a 2nd matrix
> >(described) in the 1st paper to advance the same state variables that
> >the carrier's matrix works on. Then finally do a simple amplitude
> >correction. Also described in the papers.
>
> >Happy coding.
> >Clay
>
> Hello everybody!
>
> First let me thank everyone that posted a reply. It really helped!
> I was starting to feel that I was out of options, but after reading your
> posts and links, I will try new things.
>
> Replying in order:
>
> John and Andrea:
> The LUT should be an option to consider. I didn't know the techique to use
> a table to store the values of sine and another to store the diferences.
> Will definitly give it a go. Will also try the Ollie Niemitalo algorithm
> since the microcontroller I'm working has 16 bit fixed point and a single
> clock MAC (Multiply And Accumulate).
>
> HardySpicer:
> I had a really good laugh with your transistor post. Don't feel bad. A
> single DSP chip has millions of transistors in it.
> Also, after the signal is generated it will be upmixed with a 461,538 Khz
> signal to give an output at 455 khz to be filtered by a ceramic filter
> (hello analog world).
>
> Tauno:
> Your post really confused me.
> Since beta=deltaF/Fm, and deltaF=2500 and Fm=3000, beta =0.833(3) hence a
> NBFM.
> Please correct me if I'm wrong. This is just an hobby of mine.
>
> Clay:
> Thank you for your post. I have learn much with your papers.
> I'm using a Coupled-standard quadrature oscillator, as decribed in them, to
> generate the SSB signals.
> If I understand your papers correcly, and since I have a modulating signal
> of 12 bits (or even 8 bits) I would need 1024 (or 256) diferent
> pre-calculated matrixes to advance the state variables, or calculate it in
> real time. I only have 4K of code space so this solution is not viable.
> Is this correct ?
>
> Cheers.
> Cesar

Oh goodness no - you only need two matrices. One is set to the fixed
center frequency. The other matrix has terms that vary according to
your modulation. Depeding on the modulation index you can decide the
accuracy for which these varying terms need to be calculated. For
narrow band FM, usually 1st or 2nd order MacLauren series are needed.

IHTH,
Clay

From: Vladimir Vassilevsky on 7 Apr 2010 10:40


csilva1975 wrote:

>>And, I guess, the 455 kHz is going to be upconverted somewhere.
>>Instead of all that, the elegant solution would be generate FM at the
>>target frequency directly by means of DDS chip controlled by MCU. The
>>new DDS chips from AD do linear interpolation, so the sample rate does
>>not have to be very high. You don't have to use DSP, general purpose MCU
>>would probably do.
>>
> Yup you guessed it. I will upconvert it with a ne602 controlled by the pwm
> output of the DSP/MCU.

OMG. I can only imagine what the output spectrum is going to be.
What is the final goal that you try to accomplish?

> I have already looked at those chips from AD, and they look great,
> unfortunately I donòÀÙt have any fancy solder machines to prototype a
> surface mount boards,

There is no problem to solder/desolder any chip by hand. If you need to
do it only once or twice, there is no need for fancy reflow station.
Also, you can always order assembly work on the side.

> so I have to limit myself to use *old* DIP chips, and
> the DDS chips donòÀÙt come in that format.

:)))))

Vladimir Vassilevsky
DSP and Mixed Signal Design Consultant
http://www.abvolt.com

From: Darol Klawetter on 8 Apr 2010 11:51
On Apr 7, 9:19 am, "csilva1975" <cesard(a)n_o_s_p_a_m.hotmail.com>
wrote:
> >And, I guess, the 455 kHz is going to be upconverted somewhere.
> >Instead of all that, the elegant solution would be generate FM at the
> >target frequency directly by means of DDS chip controlled by MCU. The
> >new DDS chips from AD do linear interpolation, so the sample rate does
> >not have to be very high. You don't have to use DSP, general purpose MCU
> >would probably do.
>
> >Vladimir Vassilevsky
> >DSP and Mixed Signal Design Consultant
> >http://www.abvolt.com
>
> Hi.
>
> Yup you guessed it. I will upconvert it with a ne602 controlled by the pwm
> output of the DSP/MCU.
>
> I have already looked at those chips from AD, and they look great,
> unfortunately I don’t have any fancy solder machines to prototype a
> surface mount boards, so I have to limit myself to use *old* DIP chips, and
> the DDS chips don’t come in that format.
>

For parts with blind solder pads, some resource-poor guys will apply
their own solder paste and use a toaster oven to reflow. Some will
even try to control the toaster oven to match the recommended thermal
profile for reflow. Less cautious folks will use a heat gun.

Expresspcb.com can make PCBs for you at low cost.

Darol Klawetter
First  |  Prev  |  Next  |  Last
Pages: 1 2 3 4
Prev: Xilinx Pipelined/Streaming FFT Architecure?
Next: Shannon's upper bound for AWGN channel, but what about Multipath?