ADP1031 Noise in FFT Spectrum

Hi bahadirdonmez,

Thanks for reaching out again. Can you remind me of the application your are doing using the ADP1031? What is the load for the ADP1031 and do you have any information about how much is the noise target for each of the outputs? - so that we can design a suitable filter for each output. 

Yes, I noticed that the output spectrum for the three regulators looks the same and do not follow the default switching frequency of the ADP1031 (250Khz for Flyback and 125kHz for Buck and Inverter). May I know what instrument you are using to measure this spectral noise? I'm suspecting that maybe the 50 ohm termination of the Spectrum Analyzer is loading the outputs - do you use a blocking capacitor when you do the measurement?

Thanks and regards,

Jefferson

Hi Jefferson,

Thank you very much for your response. I am using ADP1031 for two purposes:

1) Generating isolated output voltages of 15V, -15V and 5.15V from a 24V input. 

2) Isolating SPI signals that are used to send data to a DAC (AD5791)

The 15V output of ADP1031 is connected to an LT3045 LDO to generate 13.5V. Similarly, -15V is used to generate -13.5V using LT3093 LDO. These outputs have much less noise compared to 15V and -15V:

The spectrum of 13.5V looks like this:

The  spectrum of -13.5V looks like this:

I am using the 13.5V and -13.5V as analog supply connection (VDD and VSS) to the DAC and several OPAMPS I have on the PCB. 

The 5.15V output of ADP1031 is directly connected to the DAC as digital supply (VCC).

The spectrum of the DAC output looks also quite noisy:

Since the 13.5V and -13.5V connected to the DAC, don't have a very noisy spectrum, do you think these peaks might be caused by the 5.15V output of ADP1031 which is connected directly to the VCC pin of DAC and has the following spectrum?

I am using PicoScope 4262 16 bit high resolution oscilloscope and HAMEG HZ 154 probe in 1X mode.

The input impedances of both the oscilloscope and the probe are 1MegaOhm.

To avoid large ground loops, I am using the tip-and-barrel method described in this article:

https://www.analog.com/en/analog-dialogue/articles/understanding-switching-reg-output-artifacts.html

My probe looks like this:

I did not insert an additional series capacitor to block the DC voltage but I am using the oscilloscope in AC coupling.

Do you think this probe is fine for the measurements or do you have another suggestion?

Thank you very much for your help.

Best regards.

As an additional test, I removed the noisy 5.15V output of ADP1031 going to the VCC pin of DAC and I let the VCC pin of DAC be floating. Then I measured the output of DAC again. For comparison:

Spectrum with 5.15V connected to DAC:

Spectrum without 5.15V connected to DAC:

So the noise at the output of DAC is reduced quite a bit when I disconnect 5.15V output of ADP1031 from DAC.

I discussed with my supervisor about our noise target. Since we have 20 bit DAC and our output span is 20V, we would like our noise at the DAC output to be around 20V/(2^20) = 19uV (~-90dBm). So, I guess in this case the supply voltages should have even less noise, maybe around 2uV to 5uV (~-110dBm to -100dBm). What type of filter would you suggest us to use in this case?

Thank you very much for your help.

Best regards,

 Hi bahadirdonmez,

thank you for providing the information about how you are measuring the noise - it's fine to use the scope and the probe but you will not get the cleanest spectral output using this method. you might want to try the 50 ohm termination on the scope + a blocking cap or use an spectrum analyzer + a blocking cap. but for now we can use that for figuring out how to improve the noise performance of the DAC powered from ADP1031. 

Seems like the VCC is the pain point based from the data we have because it is directly connected to VCC - have you tried powering the VCC with the clean supply and check the output of the DAC? Maybe coming from a bench power supply? This will also give us information if additional filtering is needed for the +/-13.5V rails.

I've seen from your schematic that was shown on the other thread (VOUT3 voltage not correct) that you have a 3.3V LDO, perhaps we can use that to filter the VCC supply of the AD5791 and relocate the connection of the SVDDs of the ADP1031 to the 3.3V also. In this case we have a cleaner VCC supply. Then we can try to add an LC filter to provide additional filtering if still necessary.

On you noise requirement of 2uV to 5uV, i think that was so tight and even an LDO might struggle to achieve that. Also take note of the inherent PSRR on the supply rails but unfortunately, there is no PSRR vs frequency information presented in the AD5791 datasheet for us to use. 

Just a question, what is the bandwidth of interest for output of AD5791? Maybe the output noise will already output of interest and will be filtered.

Thanks and regards,

Jefferson

PS.

If it is okay, can you send me the full schematic for the ADP1031 + AD5791 including the reference supplies in my Email jefferson.eco@analog.com for me to review. 

Hi Jefferson,

Thank you very much for your response and detailed comments. 

1) At the moment I am repeating the measurements using a 50 ohm coaxial cable and a 50 ohm resistor added in series, similar to how it is shown in the image below:

My oscilloscope has a fixed input impedance of 1 MOhm. Is it actually better to have a T-connector at the oscilloscope input and insert a 50 Ohm termination like in the image below:

Does adding a blocking cap give a better measurement than setting the oscilloscope to AC-coupled mode?

2) I was also considering using the 3.3V that I already have on the board to supply digital voltage to ADP1031 and AD5791. I saw in the datasheet that the SPI signals of ADP1031 have a higher propagation delay but lower jitter if I use 3.3V instead of 5.15V for the SVDD pin.

Considering that the overall current consumption will also be lower (=less heat dissipation on the board) if I use 3.3V instead of 5.15V, do you think 3.3V is actually a better option here? The only disadvantage seems to be the increase in maximum propagation delay by 3 ns, but I am not sure if it a very important disadvantage.

Another thing I was considering is the timing characteristing of AD5791. In the datasheet, there is a table showing the timing characteristics as in the screenshot below:

At the moment, I am generating my SPI signals in accordance with the timing characteristics given on the third column of the table above since my digital supply is 5.15V. Do you think I can still keep using the same SPI signals if I drop the digital supply to 3.3V? It is a bit confusing because 3.3V is at the limit on both ranges given the table.

3) Thanks a lot for commenting on the noise requirement. Do I understand correctly that due to the inherent noise attenuation at the DAC power supply rails, I can actually achieve the 19uV DAC analog output noise target even if I have a noise at the LDO outputs higher than 5uV?

4) We would like to use the DAC as an arbitrary waveform generator with 1MHz sampling frequency and with 10V to -10V voltage span. At the moment we have a buffer OPAMP in voltage follower configuration at the DAC output but no filter yet. The OPAMP has a 20V/1us slew rate. What kind of filtering do you think we should use at the DAC output to eliminate the noise we discussed above but still be able to make use of 20V/1us slew rate?

5) I will senn the schematics per email for your review

Thank you so much for you time and assistance so far.

Best regards.