Back to the index.


Noise in voltage references

Sometimes we need in an electronic circuit a voltage reference, which gives a (more or less) stable output voltage regardless of changes in supply voltage.
In this article you find some measurements on the noise production of such voltage references.
Also a simple test is done to see how well a hum present on the supply voltage is suppressed.

The noise and hum are measured with an audio spectrum analyser.
In front of the spectrum analyser is a 60 dB amplifier as described in  this article .
First we have a look at the noise floor of this test setup, by shorting the input of the amplifier to ground.


Figure 1, measuring the noise floor of the test setup.

This measurement gives the following result:


Figure 2, noise floor of the test setup.
The noise density at the input of the 60 dB amplifier is 8.4 nV/√Hz.

Now we are going to measure the hum and noise out of my power supply.


Figure 3, a picture of the power supply used for powering the test setup.
 

Figure 4, circuit for measuring the hum and noise of the power supply.
The amplifier has a DC blocking capacitor at it's input, so only AC is amplified by 60 dB.
 

Figure 5, this is the output spectrum of the DF1730 power supply I use.
To be precise, this is the spectrum at 12 volt output voltage and 0 mA output current.
Keep in mind that there is a 60 dB amplifier between the power supply output and the spectrum analyser input, so we have to subtract 60 dB from the reading on the dB scale.
The 50 Hz component at the output of the power supply is just below -85 dBV (56 μV).
The level of the 50 Hz component is quite constant when you change the output voltage of this power supply.
However the noise floor (for instance at 20 Hz) reduces considerably at lower output voltages.
 

7809 Regulator as voltage reference

The first "voltage reference" tested is a 7809 voltage regulator.


Figure 6, testing the noise of a 7809 voltage regulator.



Figure 7, this is the output spectrum of the 7809 regulator ( 0 mA output current), placed behind the DF1730 power supply (at 12 volt).
The 50 Hz hum is not visible anymore, but this 7809 regulator seems to produce quite some wide band noise.
 

BZX85C-2V7 Zener diode as voltage reference.

In this test a 2.7 volt zener diode is used as a voltage reference.
The part number is BZX85C-2V7 (datasheet_BZX85C.pdf).
The following test setup is used:

Figure 8, test setup for measuring the noise and hum across the zener diode.
The current through the zener diode is varied by varying the output voltage of the power supply.

You can view the spectrum of all measurements by clicking the links in the table above.
The suppression of the hum is the best at the highest current through the zener diode.



Figure 9, spectrum across a 2.7 V zener diode, at 20 mA DC current (the highest current measured).
The 50 Hz component is suppressed by 30 dB (compared to figure 5).

Next the resistor between the power supply and the zener diode is replaced by a LM334 current source. (datasheet_LM334.pdf)
A current source has a very high dynamic impedance, which will result in much better suppression of the hum.
The current of the LM334 is set with one resistor value, in this case to 2 mA.

Figure 10, the zener diode driven by a LM334 current source.
 

Figure 11, spectrum of a BZX85C-2V7 zener diode driven from a 2 mA current source.
There is no measurable hum anymore in the spectrum.
The only thing we measure is the noise of the zener diode, at a level of 17.7 nV/√Hz.
 

TLCR5800 Red led as voltage reference.

In the next measurement, a TLCR5800 red led is used as a voltage reference (datasheet_TLCR5800.pdf).


Figure 12, test setup for measuring hum and noise across a TLCR5800 red led.
 

In this table you can view the spectrum of the red led at several DC currents.
The hum reduces at higher currents.
But the wideband noise increases at higher currents.



Figure 13, spectrum across a TLCR5800 red led, used as a voltage reference.
The DC current through the led is 2 mA, set by a 470 Ohm series resistor.


And now once again this same red led, but now driven from a 2 mA LM334 current source.

Figure 14, noise measurement of a TLCR5800 red led, driven from a 2 mA LM334 current source.
 

Figure 15, spectrum of a TLCR5800 red led, driven from a LM334 current source at 2 mA.
The only thing we measure is the noise generated by the led, 28.1 nV/√Hz at 1 kHz, and even more at lower frequencies.
 

HSDL4260 Infra Red led as voltage reference.


The next device under test is a HSDL4260 IR (infra red) led  (datasheet_HSDL4260.pdf).


Figure 16, test setup for measuring the noise and hum across a HSDL4260 IR led.
 

In this table you can view the spectrum at the different currents through the led.
The suppression of hum increases at higher led current.
The noise production of this IR led, seems to be very low.



Figure 17, spectrum of the HSDL4260 IR led at 2 mA current.
The suppression of the hum is about the same as with the red led (figure 13) , but the noise level in between the hum components is much lower.

By the way: the led's are placed in the dark during test, however it makes no difference for the results if ambient light can reach the led's.
Only with a light bulb close to the led, I notice some increase in noise.


Now the 470 Ohm series resistor is replaced by a LM334 current source set to 2 mA


Figure 18, noise measurement of a HSDL4260 IR led, driven from a 2 mA LM334 current source.



Figure 19, spectrum of one HSDL4260 IR led driven from a 2 mA current source.
The noise is so low, it can't be measured with this test setup, what we see is the noise floor of the test setup.

In an attempt to make the noise of the IR led visible, I placed 7 of these led's in series, which should increase the noise level by 8.45 dB.
Also I measured the noise at several current settings of the LM334 current source, ranging from 1 to 10 mA.
10 mA is the maximum allowable current through the LM334.

Figure 20, measuring the noise of 7 IR led's in series.

The measurements show that the lowest noise level is reached at 5 mA through these IR led's.

Figure 21, spectrum of 7 HSDL4260 IR led's in series at 5 mA, driven from a LM334 current source.
The voltage across these 7 led's in series is 9 volt.
 

LM336Z-2.5 reference diode as voltage reference.

In this measurement a LM336Z-2.5 reference diode is tested (datasheet_LM336-2.5.pdf).


Figure 22, measuring the noise of a LM336Z-2.5 reference diode.

In this table you can view the spectrum at the different currents through the LM336Z-2.5



Figure 23 , spectrum of a LM336Z-2.5 reference diode at 2 mA driven from a 470 Ohm resistor.

The LM336Z-2.5 has a very low dynamic resistance (according to the datasheet, about 0.2 Ohm), which results in a high suppression of hum.
Even if we use it with a 470 Ohm series resistor to the power supply, we don't measure any hum.
A low dynamic resistance of a device will also cause a low voltage change across the device, when the DC current through it is changed.

The LM336Z-2.5 produces some wideband noise.
The level of the noise is almost not depending on the current through the LM336Z-2.5.

Next the LM336Z-2.5 is driven from a LM334 current source.
In theory this should increase the suppression of hum even more, but I could not verify this because in figure 23 is already no hum visible.

 
Figure 24, spectrum of a LM336Z-2.5 reference diode at 2 mA, driven from a LM334 current source.
The spectrum is about the same as figure 23, the noise level is 125 nV/√Hz at 1 kHz.
 

LM336Z-5.0 reference diode as voltage reference.

The LM336Z-5.0 (datasheet_LM336-5.0.pdf) is another version of the LM336, this one is a 5.0 volt reference diode.



Figure 25, Measuring the noise of a LM336Z-5.0 reference diode.
 

Figure 26, spectrum of a LM336Z-5.0 reference diode at 2 mA, driven from a LM334 current source.
It has about 6 dB more noise then the 2.5 volt version.
Instead of one 5.0 volt device, you could also use two 2.5 volt devices in series, because that would increase the noise only by 3 dB.

Now we have done some measurements on voltage references.
The measurements were done without extra noise filtering.
In this final measurement we will see what happens when we place a low pass filter behind the LM336Z-5.0.



Figure 27, a low pass  RC filter behind the LM336Z-5.0 to reduce the noise.

Figure 28, this is the spectrum of a LM336Z-5.0 reference diode, with a low pass filter added.
The noise is now so low, that we only measure the noise floor of the test setup.
Adding this extra low pass filter can have some drawbacks, the settling time for the reference voltage increases considerably, and the DC resistance of the reference voltage increases by (in this case) 10 kΩ, this can reduce accuracy of the reference voltage if the circuit draws current from the voltage reference.
 

Summary:

For a voltage reference, the best suppression of hum is reached when we drive it from a current source.
The next table gives an overview of the noise of the tested devices.
Also the noise is calculated per volt reference, this is done by dividing the noise density by the square root of the DC voltage.

From this table we see the HSDL4260 IR led is by far the lowest noise voltage reference of the tested devices.




Figure 29, this graph shows how the voltage across the tested voltage references changes at different DC currents.

Back to the index.