Do low noise zeners exist?

[RuudNL]

I have been experimenting with the classical Schoeps circuit:

I noticed that there was a lot of difference in self noise between circuits, even with noise selected transistors and FETs.
After some searching, I noticed that the noise was caused by a zener diode.  (D5)
In my case the zener is 7.5 V / 500 mW.
I tried to increase the value of C10 to 470 uF, but still there was a difference...
Strange that even 470 uF doesn't completely 'kill' the zener noise.
To get proof of the fact that the noise was effectively generated by the zener, I exchanged this part from the 'quiet' circuit to the circuit under test.
And indeed... now the test circuit was as quiet as quiet can be!
So my question: is there such a thing as 'low nosie zener diodes'? And if so: where do I get them?
(I saw the datasheet of the low noise 1N4100, but I can't find where they sell them!)
I already tried to test 100 'standard' 7.5 V / 500 mW zeners, but they were all noisy as hell! Who knows more about this?

 

[Samuel Groner]

Zener noise is a very complex function of Zener voltage and bias current. Generally speaking, low voltage and high current keeps noise down. In particular there is a significant increase in noise for diodes above 6.2 V--it looks like you've just crossed this border. This behaviour is related to two different breakdown mechanisms. Low-voltage Zener diodes break down with the true Zener mechanism, while parts above 6.2 V follow the avalanche breakdown mechanism. I have not found much more detail in the literature beyond this broad generic statements. There are surely also differences between different parts (and even specimen) of the same nominal voltage (e.g. I have found 1N829 parts to be considerably better than standard non-compensated 6.2 V diodes), but the big step in getting low noise is using parts no higher than 6.2 V. Two 5.1 V in series are much lower noise than one 10 V, and three 3.3 V probably even better.

Samuel

 

[RuudNL]

OK, thank you for this reply!
I will do some experiments with for example 4.7V + 2.7V in series (=7.4V) and see if this is any better.
Also I read that LEDs would produce a lower noise value, when used as zener diodes.
This might also be worth do experiments with. First thing to do in this case, is find the forward voltage of different types of LEDs.
If I remember correctly, the white and blue LEDs have the highest forward voltage?

 

[Samuel Groner]

Also I read that LEDs would produce a lower noise value, when used as Zener diodes.

You mean forward-biased? I'm not sure if LEDs can be reliably zenered (i.e. reverse-biased until breakdown). About any forward-biased diode will have much lower noise than a Zener, but the impedance is usually higher and voltage is lower (i.e. you need more in series, which increases impedance even more, and also brings noise up). Tempco is also a consideration.

Samuel

 

[RuudNL]

Well, it doesn't cost much to give it a try! (I've seen LEDs as reference zeners in power amplifiers.)
The worst thing that can happen, is that is doesn't work!

 

[RuudNL]

Just did some tests...
I used three 2.7V zeners in series to (more or less...) substitute a 7.5 V zener.
The noise I measured was lower than the lowest noise I ever measured with a 7.5 V zener.
So this proves that the theory is correct that zeners with a lower voltage produce less noise!

Also I tested three yellow LEDs with an ordinary 1N4148 in series to produce the right voltage.
Noise was in the same order as with the three 2.7 V zeners in series.
Conclusion: problem solved!

Samuel: many thanks for your help!!! 

 

[henk]

Do think that leds are the better option..........

 

[JohnRoberts]

Zeners are inherently noisy. (Sam gave more detail about this than I, I thought they were all in avalanche mode.) In fact zeners have been used as noise sources to make noise.

I don't know how much room you have but you could use a Vbe multiplier (transistor with two resistor divider feeding base) that could be lower noise, if using a quiet transistor junction to multiply up. The Vbe noise is multiplied by this multiplication but a shunt cap from collector to base can roll off that gain.

JR

.

 

[Gus]

the Schoeps circuit is clever what is the TC of the stock zener?  What is the TC of three 2.7VDC?
IIRC I have an older zener diode Motorola book I should find that goes into detail about zeners

 

[ricardo]

This was analysed to death in Yahoo MicBuilders many moons ago.  The answer is to add another stage of RC decoupling.  Caps across the Zener have to be huge to have any effect cos the Lo Z of the Zener.

Zapnspark's DIYgenericElectretMic.pdf in his MicBuilders directory has details. (You need to join)  His R1 & C1 do the trick.  Note you don't need a cap across the Zener if you do this.  You need a slightly bigger Zener.

 

[RuudNL]

Thanks Ricardo!
Yes, that is indeed a simple solution to keep the zener noise down.
The only problem is where to place the extra components on the existing PCB... (Space is limited!)
I have tested some 6.8 V zeners yesterday and I noticed that they produced far less noise than the 7.5 V zeners.
Does the small difference in voltage make such a big difference in noise, or did I just get a bad (noisy) batch of zeners?
Schoeps has shown that the original concept works. And some of my zeners do too, be it 3 out of 100 or so...

 

[Samuel Groner]

Some numbers I've measured:

3.3 V (1 mA): 7 nV/rtHz
3.3 V (5 mA): 5 nV/rtHz
5.6 V (5 mA): 25 nV/rtHz
6.2 V (1 mA): 150 nV/rtHz
6.2 V (5 mA): 40 nV/rtHz
1N829A (1 mA): 25 nV/rtHz
1N829A (5 mA): 15 nV/rtHz

These are for wideband noise, I've also found 1/f and popcorn noise-like behaviour to various degrees (the 1N829A seems to be excellent in this respect too, which probably is related to precess purity).

Motchenbacher-Fitchen show a graph for voltage noise vs. Zener voltage, and above 5 V all Zeners are around 30 uV/rtHz (i.e. roughly 1000x higher than below 5 V). This is at a low Zener current of 250 uA though; I believe that more typical currents of 1-5 mA move both the knee where high noise starts upwards in voltage, and the absolute noise levels downwards. But I think that the noise difference between 6.8 V and 7.5 V you've observed is surely to some extent attributable to fundamental behaviour.

BTW, what's the Zener current in your application?

Samuel

 

[RuudNL]

Interesting figures!
I am sure I could make a complete study of zener behaviour and become a 'zenerologist'. 
But that is not my intention!
The fact that the three 2.7 zeners work (and in fact better and with lower noise than the best 7.5 V zener I have tested) is good enough for me now...
No idea about the temperature effects, but in this application a little drift doesn't cause problems I think.
Anyway: from now on I will always remember that a lower zener voltage means lower noise!

 

[Kingston]

This was analysed to death in Yahoo MicBuilders many moons ago.  The answer is to add another stage of RC decoupling.  Caps across the Zener have to be huge to have any effect cos the Lo Z of the Zener.

Zapnspark's DIYgenericElectretMic.pdf in his MicBuilders directory has details. (You need to join)  His R1 & C1 do the trick.  Note you don't need a cap across the Zener if you do this.  You need a slightly bigger Zener.

Or upgrade it further with a pass transistor, perfectly clean RC filtered zener reference to transistor base. Goodbye zener noise. Overall current increase can be kept negligible.

 

[Sredna]

8)

 

[Kingston]

8)

Exactly!  ;D



And to get back to the question of the thread title, there actually are very low noise zener-like devices. This thread simply went off on a tangent, an important one but still.

Have a look at these two devices for example.

http://www.ti.com/product/lm431
http://www.ti.com/product/lm329

 

[Samuel Groner]

Have a look at these two devices for example.

http://www.ti.com/product/lm431
http://www.ti.com/product/lm329

Neither of which is significantly better than the figures I've quoted for an ordinary 10 cent discrete 6.2 V diode, and much poorer than the 1N829A or low voltage discrete Zeners... In fact not even the $$$ LTZ1000 is any better.

Samuel

 

[Kingston]

Neither of which is significantly better than the figures I've quoted for an ordinary 10 cent discrete 6.2 V diode, and much poorer than the 1N829A or low voltage discrete Zeners... In fact not even the $$$ LTZ1000 is any better.

Huh?

You're showing:
6.2 V (1 mA): 150 nV/rtHz
6.2 V (5 mA): 40 nV/rtHz

LM329 quotes 7μV wideband noise figure. Since when was in order of magnitude or two not a significant improvement?

This is an expensive device, more than all of the parts of the above pass transistor set up. I'm not exactly an advocate.

 

[JohnRoberts]

Yup, I didn't see a noise spec for either and they will be some multiple of the internal voltage reference that doesn't appear to be bypassed. The designs look concerned about voltage stability and impedance, not noise.

I still like the Vbe multiplier (using a low noise transistor), but a simple added RC can do wonders with a simple zener if the circuit doesn't require very low Z. If you have to add a transistor buffer the Vbe multiplier could be less parts.

JR

 

[JohnRoberts]

Neither of which is significantly better than the figures I've quoted for an ordinary 10 cent discrete 6.2 V diode, and much poorer than the 1N829A or low voltage discrete Zeners... In fact not even the $$$ LTZ1000 is any better.

Huh?

You're showing:
6.2 V (1 mA): 150 nV/rtHz
6.2 V (5 mA): 40 nV/rtHz

LM329 quotes 7μV wideband noise figure. Since when was in order of magnitude or two not a significant improvement?

This is an expensive device, more than all of the parts of the above pass transistor set up..

rt Hz and wide band noise is apples and oranges, while 7 uV wide band is not horrible.

JR