This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

LM27761: R2 Requirements

Part Number: LM27761
Other Parts Discussed in Thread: OPA209

Hello: 

The data sheet for the LM27761 gives 50k as the minimum value for R2.  This is one of the two resistors used to set the output voltage of the internal linear regulator in this inverting charge pump switcher.  Usually, this type of feedback resistor is given a maximum value so that the current in the resistor is large compared to leakage current into the feedback pin, so that voltage accuracy is not affected by the leakage current. 

Is it a typo that R2 is specified as a minimum of 50k?

I do note that many of the performance graphs do show R2 as 500k, but I was wondering if this happened as a result of an initial typo calling for R2 to be greater than 50k.  Because, in addition to leakage error with larger resistors, there are two more problems with such large values.  One is that the part has an internal RC filter on the feedback pin, and larger resistors will be more noticeable with respect to that internal resistance, and will thus move the filter corner to higher frequency and thus sharply reduce the phase margin of the internal loop.  The other problem is the higher thermal noise of the large resistors.  The part is specified as 20uVrms error in a 100kHz bandwidth, which is 63nV per root Hz.  But, 500k has 89nV of thermal noise, and when this is multiplied to the output voltage this noise is gained up.  For example, at -3.3V the thermal noise of R2 alone will be gained to 152nV per root Hz.  Then, R1 adds still more thermal noise--in the case of -3.3V it contributes about 117nV.  So, these large values are causing large increases in the noise floor of the regulator. 

Thanks,

Farron

  • Hi Farron,

    According to my understanding, feedback pin is internally connected with the filter which will avoid the noise gain up. But I will double-check it and get back to you.

    Best regards,

    Tanvee

  • Hello Tanvee:

    If the resistors were used to gain up the reference voltage, which was then unity gain followed to the output, then the resistors and gain would have only second order effect on noise (mostly on close in 1/f noise). 

    But, the block diagram shows a standard regulator architecture where a bandgap is noise filtered and presented to one op amp input as the reference.  Then, the output is divided down to the reference.  That architecture thus adds resistor noise, gains that resistor noise, and gains the bandgap noise and input referred op amp noise. 

    Usually in that architecture, the resistors are kept fairly small so that their thermal noise is less than the bandgap and op amp noise.  But, the use of 500k R2 in the data sheet, and the specification of at least 50k for R2, is the opposite of that normal design practice.  The part certainly won't meet its noise spec with those large resistors, and may have voltage error introduced by leakage through those large resistors also. 

    That's why I was wondering if the spec really meant to say R2 < 50k instead of R2 > 50k.  That may be an old error that has slipped through till now. 

    Regards,

    Farron

  • Hi Farron,

    The device is designed for this value and characterized for this value. I am trying to understand the concern regarding the feedback divider impact in your application. Could you explain more about your application?

    Best regards,

    Tanvee

  • Hi Tanvee:

    Thanks for these quick responses. 

    The application is a biomedical sensor product that is very sensitive to supply noise.

    It may make the issue more clear to review a few basic numbers, and the general behavior of this type of classic "gained" regulator with current pump output.  The physics of noise here are architecturally nearly identical to that of resistors used to set gain of low noise op amp circuits.  The basic relations are given in the Texas Instruments OPA209 data sheet (attached), in Fig 34 on page 13. 

    The part's specified noise is 20uVrms from 10Hz to 100kHz..  The output voltage and the R1 and R2 values for this noise are not given, nor is the spectral noise density over frequency.  But, we can easily calculate that the average spectral noise density over the 100kHz as 63.2nV per root Hz. 

    If R2 is 50k (the specified minimum) and the output is to be -3.3V, then R1 is 85.6k.  The thermal noise in R2 is 28.3nV per root Hz.  The gain to the output of this noise is R1/R2 = 1.71, so this noise is gained to 48.4nV.  The noise of R1 is 37nV.  This is not gained, but is rms added to the R2 noise to give a resistor noise contribution of 61nV per root Hz in the output. 

    This is indicating that IF the data sheet spec of 63nV per root Hz is measured with a 50k R2, then most of the noise is due to the resistors,and the part is really capable of considerably lower noise.  If so, Texas Instruments is missing the chance to promote this part as having a much lower noise. 

    And, why should R2 be limited to being 50K or higher?  The part can provide as much as 250mA output current, so the current drawn by the feedback resistors is negligible down to quite low values.  If R2 were 1k, then at -3.3V R1 would be 1.7k, and the current drawn in the resistor string would be only 1.2mA.  This is no problem for this part, and would have much lower noise. 

    Now, at the 500k value for R2 used in most of the performance graphs, at -3.3V the R1 value is 856k.  The noise of R2 is 89.5nV per root Hz, which is gained to 153nV.  The noise of R1 is 117nV, so the total resistor noise is 193nV per root Hz.  Over 100kHz this is 61uVrms, so it is blowing the 20uVrms noise spec way up. 

    The large resistor values are ruining the noise of the part, and there is no current limiting reason to use them. 

    Also, the low pass noise filter from the resistors back to the op amp input (block diagram on page 9) is a strange architectural feature.  The negative feedback of the part introduces 180 degrees of phase shift.  The standard current pump driving a capacitor output structure introduces 90 degrees more, leaving at most 90 degrees of phase margin in the control loop.  A single pole low pass here will be cutting into that phase margin, as well as limiting the bandwidth of the loop.  Being in the feedback path, it does not help with noise suppression, and it adds still more resistor thermal noise.  The circuit provides better noise suppression without it, as well as having better phase margin.

    Clearly, some odd things are going on in this data sheet.  The only logical reason to have larger resistors for R1 and R2 is for them to act as part of that strange feedback noise filter.  But, varying them around by an order of magnitude or more as the data sheet shows, they would likely be playing havoc with the phase margin of the regulator control loop.

    The strangeness of the situation thus leads to my original question as to whether the data sheet meant to say that R2 should be LESS than 50k.  It would certainly be lower noise, and it would not pull in the 3dB frequency of that feedback noise filter to where the phase margin of the loop is degraded.  

    Perhaps the designers of this part could be asked for their input on the situation.  opa209.pdf

    Regards,

    Farron 

  • Hi Farron,

    I will send you an email for further discussion of your application and close this thread.

    Best regards,

    Tanvee