The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Design >> Analog Design >> LDO flick noise on LCVCO
https://designers-guide.org/forum/YaBB.pl?num=1330735868

Message started by lunren on Mar 2^nd, 2012, 4:51pm

Title: LDO flick noise on LCVCO
Post by lunren on Mar 2^nd, 2012, 4:51pm

Hi guys:

I am working on a LCVCO design. A on-chip LDO will provide power supply for LCVCO. One big problem is that the LDO flick noise affect LCVCO's phase noise a lot. Without LDO, the phase noise is very good, with LDO, the phase noise drops almost by 20dB at the interest frequency range. I am wondering if there is a good solution for this ( I am still prefer to use LDO), or how to beat flick noise?

By the way, according to my AC simulation (LCVCO as load), the integrated noise at the LDO output from 1 to 10GHz is only 64uV. From the noise summary, I could see that thermal noise dominates. However, the pnoise noise summary says that flick noise from diff pair of error amplifier inside LDO dominates. Any explanation for this?

Thanks,

Lunren

Title: Re: LDO flick noise on LCVCO
Post by loose-electron on Mar 2^nd, 2012, 5:50pm

The differential pair will often dominate the flicker noise line-up.

You can push the W x L size up
of the diff pair to reduce the flicker noise.

You can also consider chopping methods
to take the flicker noise and bring it to a
higher frequency.

(essentially up conversion mixing the noise)

If you frequency plan that right, then
your decoupling will absorb the HF noise.

There are some folks here who have
done this quite a bit.

Title: Re: LDO flick noise on LCVCO
Post by raja.cedt on Mar 3^rd, 2012, 3:41am

hello,
20dB is too much. When you have VCO as load, how simulator finds operating point (because vco can't oscillate), so don't believe those results instead use PAC. In case of vco or some rf block integrated noise doesn't matter, only spot noise matter because that noise will fold into the band of interest. As people suggested try to decrease the flicker noise by increasing the are. Please post the results (i am so interested to see how 20Db difference between those two case )

Thanks,
raj.

Title: Re: LDO flick noise on LCVCO
Post by lunren on Mar 5^th, 2012, 5:01pm

Hi Raj,

I think you are right. Here is a link from TI talking about LDO noise.
http://www.ti.com/lit/an/slaa412/slaa412.pdf
But is there any rule of thumb about how low the PSD of noise needs to be? Or which spot noise should we pay attention to? I guess we still need to verify it through simulation.

Here is my simulation plots. My description was wrong in my previous post. The 64uV integrated noise is based on ideal reference voltage. If I added bandgap as reference, the integrated noise will up to 200uV and the VCO phase noise becomes terrible (unacceptable).

I am wondering how to design a good regulator for PLL, or any specific spec in terms of self-generated noise? PSR of this ldo is good (not great due to area limitation), >20dB across all frequency. Any recommendations/papers are welcome.

Thanks,

Lunren

Title: Re: LDO flick noise on LCVCO
Post by weber8722 on Mar 6^th, 2012, 4:12am

Indeed low-power bandgaps are bad on noise. As discrete IC some manufacturer are offering real low-noise LDOs for just this purpose.
Power-up your BG and check BG flicker noise, consider additional filtering.

Bye Stephan

Title: Re: LDO flick noise on LCVCO
Post by lunren on Mar 6^th, 2012, 11:11am

weber8722 wrote on Mar 6^th, 2012, 4:12am:

We don't have room to put a big LPF at the output of bandgap. But some LPF doesn't help to reduce noise.
Chopping is the only solution in this case?

Title: Re: LDO flick noise on LCVCO
Post by RobG on Mar 6^th, 2012, 12:06pm

lunren wrote on Mar 6^th, 2012, 11:11am:

weber8722 wrote on Mar 6^th, 2012, 4:12am:

We don't have room to put a big LPF at the output of bandgap. But some LPF doesn't help to reduce noise.
Chopping is the only solution in this case?

Lunren, I agree with what the other people have mentioned in regards to chopping since your problem is low frequency noise which is impractical to filter on-chip. Looks like you will have to chop the bandgap too based on that sim. Make sure 1/f is modeled in the high-res resistors if you are using them.

Hopefully your LDO cap will offer enough filtering to remove the chopping noise of the LDO. I expect this will also act as a filter for the BG chopping noise so that might not need to be directly filtered.

You might also look in the last few years of JSSC for chopped amplifiers. They do an internal notch filter that removes the chopping noise. I think they've been using the idea in temp sensors too so do a search on those.

Title: Re: LDO flick noise on LCVCO
Post by RobG on Mar 6^th, 2012, 12:11pm

Also, you'll probably have to be careful about the value of the chopping frequency as I'm sure it will interact with the VCO and give some sort of spurs.

Title: Re: LDO flick noise on LCVCO
Post by lunren on Mar 6^th, 2012, 1:55pm

Hi Rob,
Yes, I think chopping is a good option to deal with low frequency noise and you pointed out some important issues. I am hoping we have bandwidth to design a chopped bandgap and LDO. The schedule is very challenge at the moment ...

Thanks,

Lunren