Below is a recent question submitted by a Power Electronics Reader, answered by Steve.
Question: I need to measure PSRR of a linear regulator circuit that powers an oscillator for an ADC. How can I make this measurement using my oscilloscope?
ANSWER: Whether it is possible to measure PSRR using your oscilloscope depends on several independent measurement characteristics.
INPUT SIGNAL LEVEL
Since the load on this linear regulator is an oscillator, which is very noise sensitive, the signal levels at the oscillator are presumably very small. Though it has not been stated, the input power to the linear regulator is likely a switching regulator and the typical ripple level is generally in tens of mVpp. There are many PSRR requirements that I see requiring the PSRR verification test to be performed at 100mVpp over a wide frequency range, and primarily for use following switching regulators. Almost any oscilloscope will easily measure this input signal.
REGULATOR BEING MEASURED
While many older linear regulators, such as the very popular LM317 provide a PSRR of approximately 60dB at low frequency, many newer devices have PSRR in excess of 120dB. Some quick math indicates that if we were to present a 100mVpp signal at the input to the regulator and the regulator provides an attenuation of 60dB then the signal at the output of the regulator would be 1mVpp and if the regulator provides attenuation of 120dB then the output signal will be 1uVpp.
In order to obtain reasonable measurement accuracy we would like the measurement noise floor at least 10dB (roughly three times) better than the signal level or 300uV for a 60dB regulator and 300nVpp for a 120dB regulator. This limits the measurement uncertainty to roughly 10% or 1dB.
WHAT TYPE OF SCOPE ARE YOU USING
The oscilloscope is not generally the instrument of choice for the PSRR measurement due to the low dynamic range resulting from the low bit count combined with the high noise floor associated with the wide measurement bandwidth an oscilloscope provides. While oscilloscopes typically have an 8-12 bit ADC most Vector Network and Spectrum Analyzers have 24 bit ADC’s (or more) and a narrow measurement bandwidth resulting in a much lower noise floor.
While I am not aware of any oscilloscopes that can accurately measure 300uVpp and certainly not 300nVpp in the time domain, it is still sometimes possible to make a reasonable PSRR measurement.
Many oscilloscopes today offer spectrum analyzer capabilities. The spectrum analyzer, being a narrow band instrument, offers a much better signal to noise ratio than the oscilloscopes time domain measurement even within the same instrument, so it is preferable to use this measurement mode.
The example in figure 1 shows a Lecroy 640Zi measuring a 100mVpp signal through a 60dB attenuator with a resulting signal level of -76dBm. The dBm can be converted to volts as
This shows that the oscilloscope using spectrum analyzer mode can properly measure a 100uV peak signal approximately 10 to 15dB above the noise level.
Figure 1 Lecroy 640Zi can measure 100uV signals without difficulty using the spectrum capability. This plot is 5dB/div with nearly 18dB noise margin.
Other oscilloscopes that include spectrum analyzer capabilities, such as the Tektronix MDO can also measure low level signals in this mode. Figure 2 shows that the MDO can measure 39.8dBuV (100uVpk) with approximately 20dB margin to the noise floor. Reducing the margin from 20dB to 10dB, these devices can measure approximately 35uV-50uV signals.
Figure 2 Tektronix MDO can also measure 100uVpp signals without difficulty using the spectrum capability. This plot is 10dB/div.
This level of capability is sufficient to measure an ordinary linear regulator, such as the LM317, and possibly regulators approaching 80dB PSRR. In order to measure state of the art regulators, a 60dB measurement improvement is required. The addition of an ultra-low noise 20dB preamplifier is shown (In this measurement a Picotest J2180A) in figure 3. In this measurement a 20mVpk 5MHz signal is connected through a 70dB attenuator and a 20dB amplifier to the Tektronix MDO using the spectrum analyzer capability. The signal measurement display is not corrected for the 20dB amplifier gain and so the corrected measured signal level is 15.5dBuV (6.3uVpk) and shows more than 10dB margin to the noise floor.
Figure 3 The addition of a 20dB ultra-low noise preamplifier allows smaller signals to be measured. In this case a 12.6uVpp signal is being measured using the Tektronix MDO.
With the addition of the preamplifier it is theoretically possible to measure PSRR up to approximately 100dB using the spectrum analysis mode in a modern oscilloscope. Additional amplifiers might improve this further, though the noise levels due to the preamplifier will also increase, so I would think it is not feasible to measure PSRR above 100dB using this method.
If you will only need to make such measurements occasionally, and at only a few frequencies, the oscilloscope may be a viable method for making such a measurement. If you need to make more frequent measurements a VNA is much simpler and much faster, while easily being able to measure 100dB PSRR. Figure 4 shows the measurement of a linear regulator providing 110dB PSRR with a 15dB noise floor margin.
Figure 4 A VNA offers a much simpler way to measure PSRR than using an oscilloscope. The upper dashed trace is the noise floor and the bottom solid trace is the PSRR measurement showing more than the minimum 10dB separation.
Regardless of how you make this measurement care must be taken in the connection of the linear regulator to the test equipment and you should ALWAYS measure the noise floor to assure the minimum 10dB attenuation is achieved.
Here are a few tips you might find helpful for this measurement:
- ALWAYS measure the noise floor in order to assure the measurement setup has the necessary fidelity
- Avoid using attenuating probes, at least on the output side of the regulator, as this will degrade the available PSRR.
- Avoid long ground wires on scope probes, and if possible it is best to measure the output using an AC coupled 50 Ohm scope input with a 50 Ohm shielded or double shielded coaxial cable.
- Common mode errors can also be substantial if you have ground loops in your test equipment. In this case the addition of coaxial common mode transformers will help a great deal and without the attenuation that would come with the use of a differential probe for this purpose, especially at higher frequencies.
For setup diagrams and additional PSRR application and measurement information, please visit www.picotest.com/blog and search PSRR.
If you would to submit your questions to Steve anonymously, please use the form below. You may also post any questions you have in the comments section below.