LMX2572LP: VCO4 weird behavior

Boris,

There might be differences in the loop bandwidth between VCO4 and VCO3. Maybe VCO4 has a tuning voltage close to the rail. Be sure you have your calibration settings correct. In fact, try setting CAL_CLK_DIV=3 and see if this helps. It slows down the VCO calibration.

Regards,
Dean

Hello Dean.

Thank you for your answer.

I've tried your suggestion about CAL_CLK_DIV=3, but there are no effect.

Also I've checked other frequencies from my frequency range. All frequencies that has fallen into VCO4 (145 ... 162 MHz) has the same high spurs and noise floor. I can use other VCO's (VCO3 and VCO5 which work perfectly) near the borders of the VCO4 range, however wide number of frequency channels can be synthesized with VCO4 only. 

Maybe you have another ideas which I could try?

And please notice, that the same picture can be seen even if FSK_EN = 0 while host I2S remains active (see the pic. 8 in my first post).

It seems like the VCO4 cannot work properly with I2S FSK.

Can you check this, please?

With sinserely respect

Boris

 

Boris,

From lmx2572 spur characterization, the spurs between VCO3 and VCO4 are very similar;  within 2 dB.

Are you using TICSPro by chance?  If so, could you save the file and post so that we an try to replicate this?

If not, could you give your input reference frequency and register settings (preferrably hex format)?

Regards,
Dean

Hello Dean

Yes, I use TICSPro, the file is attached.

But I have to mention, that this weird VCO4 behavior can be noticed only when FSK I2S is used.

If I disconnect or shut down host I2S pins - noise picture looks perfect with each VCO used, includes VCO4.

But if those I2S pins are connected and active - noise and spurs become just awful (only when VCO4 is used, regardless FSK_EN bit).

I use TICSPro from the file with modulation parameters: one bit per symbol, symbol rate = 9600, sampling rate = 230.4 kHz (24 samples per symbol), deviation = 3.6kHz, data pattern = PN9 with Gaussian shaping (BT=0.45).

With sinserely respect

Boris

 P.S. I've attached .ZIP file as the board doesn't allow to attach TICSPro file (.TCS) :)

l2572lp.zip

Boris,

I got the file and locked it up and it looks OK, but as you say, I need to hook up the I2S pins. I still have to figure out how to do this.

Regards,
Dean

Hello Dean.

You can see the difference between VCO3 and VCO4 if you just use signal generator for generating square wave on I2S pin FS (SysRefReq).

I've attached picture with this difference, had found in my experiments (I used square wave about 192 kHz,  2.6V p-p to I2S pin FS)

FSK_EN bit doesn't affects these spurs.

Actually, you can connect this signal to any I2S pin for the same effect.

With sincerely respect

Boris

Boris,

We were able to replicate this issue on our EVM. Our spurs were -60 dBc with an input level of 8 dBm.

I did not find any programmable settings that could really impact this for the better. If I tinkered with INPIN_FMT and set to LVDS, I could make this 10 dB worse, but found no way to make it better.

The only thing that I saw is that decreasing this amplitude had a direct impact on the spur; if I decreased the amplitude by 5 dB of the 192 kHz input signal, the spur went down 5 dB.

I could switch between VCO3 and VCO4 and the spur would appear for VCO4 but not VCO3.

I can only speculate as to an explanation, but I would not be surprised if VCO core 3 inductor is closer to the SysRefReq pin on the chip layout. That's just a speculation though.

Regards,
Dean

Hello Dean.

Thank you for your help.

According datasheet I2S signals must be between 1.8V and 3.3V, so the spurs will be too high anyway.

If I understand correctly you answer - my problem can't be resolved.
If yes - therefore it means that LMX2572LP cannot be used, when frequency range fallen in VCO4, and FSK I2S planned to be used.

What would you suggest me?
Would TI solve this problem, and will issue "rev. B" sometime?
Should I wait for this, or I should search another product?

With sincerely respect
Boris

Boris,

I'm not the complany expert on the I2S and FSK modulation for the LMX2572LP, and several people were out last week.
If we are lucky, someone has a brilliant idea that I didn't think of, since I do not think we have any plans for a rev B of this silicon.

Regards,
Dean

Hello Dean.

Ok, lets wait for a new idea, I really hope those people will be able to find some workaround solution, when they return.
Thank you for letting them know about the problem.

With sincerely respect
Boris