Post Go back to editing

LT7171RV-1 voltage offset

Category: Hardware
Product Number: LT7171

I'm developing a product with LT7171RV-1 supplying 0.8 V to FPGA core. The voltage output by LT7171RV-1 is 10-11 mV lower than expected, and it's far outside of +/-0.25% accuracy advertised.

On PMBUS side, VOUT_COMMAND is 0.7998 V and READ_VOUT outputs 0.7998-0.8003 V, yet when I measure voltage on output capacitor I get 0.78994 V and similar (checked on two high quality voltage meters). FPGA itself reads about 0.78-0.79 V, but its ADC is not that precise.

FPGA has voltage and ground sense pins that are directly connected to VSENSEP and VSENSEN of LT7171RV-1 routed with closely coupled track pair. These are not easily accessible for measurement or alteration.

When I set VOUT_COMMAND to 0.81 V it will bump capacitor voltage to 0.80, so voltage regulation is working. I've tried internal and external compensation, VOUT_RANGE of 0 and 1 and it changes nothing - always 10 mV short. IgnoreRConfigPins is set. Mode is forced PWM around 1 MHz.

I've checked 3 boards, and it's the same offset on all of them. Current at the moment is not high - around 0.5-1 A. Output voltage looks fine on scope – using 20 MHz bandwidth limit I see less than +/- 2 mV ripple (mostly 1 mV ripple), no oscillations or anything suspicious.

I would expect the voltage to be higher than target to compensate board copper loses, but to have too low voltage is hard to explain. And LT7171 sees voltage as in range and regulated to target.

What should I do next? Is there supposed to be some sort of a filter on VSENSEP?

  • Hi Rafal,

     

    Can you supply a schematic and layout for your board? That'd help me to take a look and see if we could find anything may be causing that ~ 10 mV drop you are seeing. 

     

    Additionally, did you run load transient and bode analysis since getting your boards back? Those could give us some indicators as to why we are seeing this.

     

    If you don't want to share the schematic/layout publicly we have a technical request portal you can use here (Support | Analog Devices), under the “submit a technical support case” button.

     

    Cheers,

     

    John 

  • I've also checked VSENSEP against VSENSEN by unmasking vias near LT7171 - reading is 0,7891 V (while 0,7894 V on output capacitor). Adding filtering capacitor 470pF on VSENSEP and VSENSEN didn't change output voltage.

    I will prepare some materials for support case.

  • Hi Rafal,

     

    Sounds good. In the meantime if you want to continue on engineer zone. Can you supply a scope capture of your output ripple including where you are measuring it? That'd one of the next things to take a look at without sharing your whole schematic. 

     

    I do agree that I'd expected to be how the feedback loop is set up. After conferring with some colleagues, measuring as close as you can to the capacitor between VsenseP and VsenseN. The voltage seen there will be what the feedback loop is "seeing" to regulate the output voltage. 

     

    Cheers,

     

    John

     

    John Park 

    Applications Engineer - Central Applications  

  • I'll try to add some more data.

    We've used higher quality scope for this, with a differential 3.5 GHz probe, as it is way easier to probe unmasked vias tightly placed. The downside is that probe can't be switched to AC mode, so all measurements were in DC mode, 20 MHz bandwidth limit was on.

    First measurement is one of the ceramic output caps of LT7171RV-1. Swtiching is visible as bumps. LT7171RV-1 is in forced PWM mode in light load about 0.4-0.5 A.

    Next is layout of feedback connection:

    LT7171RV-1 is on top. Sense vias were unmasked on bottom side.

    Last one - VSENSEP vs VSENSEN measured on vias - bottom side:

    At the moment of measurements there was no filtering capacitor attached to VSENSEP and VSENSEN vias. However it doesn't seems to make any difference, as it looks like the voltage setpoint is wrong. It looks like analog part is working fine, it just regulates to wrong voltage. But why READ_VOUT show higher voltage than measured on sense pin?

    I've also checked if the VOUT_COMMAND is properly coded and it seems it's ok - register value is 0x3A66 that translates to 0.79980 V. I've programmed LT7171RV-1 with older LTPowerPlay (1.18), but 1.21 reads same values, so I don't think it's a problem.

    I don't have tools and PCB capable of transient measurements or bode plot, however the problem here is steady state without much load, so probably they are not that relevant.

  • I've attached schematic and layout in case CS-538333-K7P4H3

  • I've also tried two more test:

    1) Pulsing 10 A load - not as fast as for looking at transient response, but still useful - voltage drop on output was around 3 mV - so control loop is in regulation

    2) Using hotair I rised temperature of LT7171RV-1 to 70-75 (using READ_TEMPERATURE) - output voltage changed about 0.3 mV, so it's within specification.

    I'm wondering if something could be wrong with internal reference, as booth setpoint and voltage measurement are off for me?

  • Hi Rafal,

     

    Thanks for the extra tests and files. In order to consolidate everything we are discussing here I am going to close out this engineer zone thread and continue on the case you opened (CS-538333-K7P4H3). So, moving forward let's do our communication on that case.

     

    Apologies for the delay, I didn't realize you are a European account which caused a bit of confusion. 

     

    Cheers,

     

    John