Lowest noise Zener Diodes?

[godfrey]

I'll agree with "best below about 6V". When you said "best are 6.2V", I was expecting to see some sort of minima, like with the dynamic impedance.

 

[SY]

No, not that easy. The real figure of merit is noise voltage divided by reference voltage, which slightly complicates things- very low noise at 0.7V (say) may or may not be better than higher noise at 6.2 (or 5.6)V. Somewhere in the 5.6-6.2 range is the sweet spot.

 

[godfrey]

Fair enough. But more importantly, LEDs are better anyway, as you pointed out earlier.

 

[jcx]

no talk of application specifics? - feeding circuits with any noise rejection at all?

for audio frequencies you may not need low TC or drift, care at all about 1/f - relatively noisy bandgap chips followed by RC filter can be as quiet as needed in many apps

 

[Nikitas]

A cap across the LEDs will lower the AC impedance. And attenuate some of the noise from the quieter LEDs.

What kind of cap?

 

[AndrewT]

Choose where you need most attenuation. @ HF or MF or LF?

 

[jcx]

a cap directly across a refs low Z is very inefficient - add a R and C lowpass, buffer if necessary

again its pointless speculation to ask for "best" without some idea of what the circuits you are feeding really require

http://www.datasheetcatalog.org/datasheet/Cirrus_Logic/mXyrquw.pdf compares a few Vrefs

 

[davidsrsb]

The data in 1.4.2. Look particularly at the first graph, which shows 1-2 orders of magnitude improvement at 5.6 and 6.2V compared to, say, 9 or 20V.

LEDs are still quieter, though- under 0.5nV/rt Hz for a cheap red.

We have this old thread here: http://www.diyaudio.com/forums/parts/35821-some-noise-measurements-leds-zener-diodes.html

Which shows lower noise at 10V

 

[SY]

The manufacturers' data indicate otherwise.

 

[JMFahey]

Can somebody post the schematic (or the relevant section) of the actual circuit where that Zener/Voltage reference will be used?

Otherwise it's shooting in the dark.

 

[davidsrsb]

The manufacturers' data indicate otherwise.

This data sheet is suspicious http://www.aeroflex.com/ams/Metelics/pdfiles/1N4099-1N4135_1N4614-4627.pdf
I do not expect that the diodes suddenly become 40uV//Hz at 6V8 and stay at exactly the same level to 100V. This looks like an arbitrary test limit to me

 

[jackinnj]

This data sheet is suspicious http://www.aeroflex.com/ams/Metelics/pdfiles/1N4099-1N4135_1N4614-4627.pdf
I do not expect that the diodes suddenly become 40uV//Hz at 6V8 and stay at exactly the same level to 100V. This looks like an arbitrary test limit to me

Do you have some empirical data to challenge their assertion?

 

[godfrey]

We have this old thread here: http://www.diyaudio.com/forums/parts/35821-some-noise-measurements-leds-zener-diodes.html

Which shows lower noise at 10V

The manufacturers' data indicate otherwise.

Sure enough, Christer's measurements showed much lower noise for the 12V Zeners than those around 6V, while datasheets suggest the opposite. There's an anomaly worth investigating.

Turns out the answer's fairly straightforward - in a nutshell Christer screwed up the measurements.

What he was trying to do was pass a constant current through the DUT, and measure the noise voltage across it. However if you look at his test rig, you'll see that he used a very noisy current source.

So what he was actually measuring was: measured noise voltage = (noise voltage of DUT) + (current noise from CCS * dynamic resistance of DUT), and that second term was obviously significant. Zeners have much lower dynamic resistance at 12V than at around 6V.

Suddenly it all makes sense.

 

[godfrey]

This data sheet is suspicious http://www.aeroflex.com/ams/Metelics/pdfiles/1N4099-1N4135_1N4614-4627.pdf
I do not expect that the diodes suddenly become 40uV//Hz at 6V8 and stay at exactly the same level to 100V. This looks like an arbitrary test limit to me

Not suspicious at all. It's max, not typical. "less than 40uV//Hz from 6V8 to 100V" does not mean "exactly the same from 6V8 to 100V".

 

[davidsrsb]

Not suspicious at all. It's max, not typical. "less than 40uV//Hz from 6V8 to 100V" does not mean "exactly the same from 6V8 to 100V".

Exactly, it looks like the test rig for diodes >6V5 was set up with an arbitrary pass/fail limit that gave a good yield and no customers came along asking for a better selection

The error in the old report on diyaudio means that zeners run at a reasonably high current are actually fairly low noise

 

[wrenchone]

Well, I can tell you that standard zeners run in starvation mode (low current) are excellent noise sources.That particular problem bit me in the butt one time while working on a HV shunt regulator that used stacked transistors. I thought clamping the bottom transistor with a zener would be nifty, but it turned out a resistor was much (much, much) less noisy due to the low currents involved.

Reverse biased transistor B-E junctions make nice sharp low current 7.5-8V zeners, but I have no idea about how noisy...

 

[kf_tam]

What he was trying to do was pass a constant current through the DUT, and measure the noise voltage across it. However if you look at his test rig, you'll see that he used a very noisy current source.

Why would a current source made from a transistor (BC549) biased at 1.4V with two forward biasing diodes (1N4148) be considered "very noisy"?
The power source is made from 9V batteries with decoupling capacitors. The resistor that feeds the diodes is 3.3K, could be made better with additional RC filtering.

 

[davidsrsb]

A chain of batteries with a series resistor would be quieter.
NiCd batteries are the best, see http://tf.nist.gov/general/pdf/1133.pdf

 

[jackinnj]

A chain of batteries with a series resistor would be quieter.
NiCd batteries are the best, see http://tf.nist.gov/general/pdf/1133.pdf

Batteries may be low noise, but their impedance is higher than a Jung regulator (or Salas). They don't have a regulator's mechanism to distort -- a very good thing:

An externally hosted image should be here but it was not working when we last tested it.

 

[SY]

Interestingly, the data in the NIST paper show that LEDs are as good or better than most batteries (using their units, red LED is about -189dBV/rt Hz at low frequencies, and the Vf is higher than V of AA or D cells).