Why use a 6.2V Zener Over Other Available Zener Voltages in a CCS?

Ver 0 before 10-Jan-2005

In the table below, I marked what I felt were good performance values in bold and acceptable values in italics. These data points were taken from the "Q3/1988 Motorola "Rectifiers and Zener Diodes Data" book.

In the two Families of zeners shown below, Family one is a 1/4W family and the other is a 1/2W family.

Nominal Zener voltage	Family_1 1/4 watt	Family_2 1/2 watt	Family_1 Vr test	Family_2 Vr test	Family_1 uV/sqrt(Hz) at 250 uA (Table)	Family_2 uV/sqrt(Hz) at 250 uA (Curve)	Family_1 Zzt @ 250 uA (Table)	Family_2 Zzt @ 250 uA (Table)	Family_2 Zzt @ 1 mA (Curve)	Issues
5.1	1N4625	1N5231B	3.0	2.0	2	1.5	1500	1600	200	Poor Regulation
5.6	1N4626	1N5132B	4.0	3.0	4	4	1400	1600	100	Poor Regulation
6.0	N/A	1N5133B	N/A	3.5	N/A	50?	N/A	1600	15	Unknown Regulation / Availability
6.2	1N4627	1N5234B	5.0	4.0	5	150?	1200	1000	9	Best compromise
6.8	1N4099	1N5235B	5.2	5.0	40	200?	200	750	9	Fam_1 Noisier by 18 dB!
7.5	1N4100	1N5236B	5.7	6.0	40	400?	200	500	12	Noisy
8.2	1N4101	1N5237B	6.3	6.5	40	700	200	500	15	Noisy
8.7	1N4102	1N5238B	6.7	6.5	40	900	200	600	17	Noisy
9.1	1N4103	1N5239B	7.0	7.0	40	1000	200	600	20	Noisy
10.0	1N4104	1N5240B	7.6	8.0	40	1200	200	600	25	Noisy
11.0	1N4105	1N5241B	8.5	8.4	40	1600	200	600	30	Noisy
12.0	1N4106	1N5242B	9.2	9.1	40	2000	200	600	35	Noisy

In the Family_2 noise curves, it was very difficult to read the values from the curve between 6.0 and 7.0V. This was because the curve is changing fast. Every time I held the calipers against the curve, I got a different value. For Family_1, I used the tabulated data instead of curves. From the Family_1 data, we see that the noise gets bad fast at 6.8V and above. This says we should use a 6.2V zener or lower for noise reasons. Unfortunately, when the zener value drops below 6V, its regulation quickly degrades.

Using a 6.2V zener in a current source, we see at Q1, R1 will have 5.6V across it. 5uV of zener noise out 5.6V is 121 dB down (very respectable). 150 uV out 5.6V is 91 dB down, still not too shabby.

If we used 4V drop for the gate of a FET, 150uV out of 2.2V on R2 is 83 dB down. 83 dB down isn't terrible, but it isn't great either. However, when using a FET we don't want to use a single 6.2V zener and the gate of a FET for a CCS, we use two in series. We don't use one 6.2V zener because of regulation and output impedance and the > +/- 1V gate threshold variation of the FET. +/- 1V is in the datasheets for gate threshold variation. I've seen this much variation occur on the factory floor, but it doesn't occur often. I suspect that FET may have been damaged from "over aggressive" temperature cycling.

The zener voltage can change +/- 0.31V for a 5% part. The gate voltage can change more than +0.5 volt between powering up and stabilizing in voltage and we can get up to +/- 1V from FET to FET (at room temperature.) I'll use +/- 0.5V to be more reasonable, for I've never seen +/- 1V on a FET that wasn't damaged. This means that when using one 6.2V zener, the voltage that sets the current can vary from 1.39V to 3.01V, a 1:2.16 initial variation, and up to 35% temperature variation. This much variation is not acceptable.

With a bipolar junction transistor (BJT) and one 6.2V zener, the current variation is more like 1:1.12 and the temperature variation will be closer to 4.5% (for a 60C change).

With two 6.2V zeners and a FET, the variation is 1:1.30.

SUMMARY:

The typical regulation curves show that zeners 5.1V and below regulate poorly with bias current. This means that the zener voltage changes too much when the zener is used below the rated test current. At 5.6V the regulation curves are starting to behave well, and at 6.2V the regulation curves are well behaved. At 5.1V, personal experience has shown that the regulation can be much worse than the curves in the Motorola Zener data book if not buying Motorola (On-Semi) parts. (The "buy the Motorola part" on the bill of materials really does mean buy from that supplier; don't try to save a 1/10 of a cent. Oh well . . .)

The zener dynamic impedance (AC) is best between 6.0 and 9V.

The zener noise is best between 6 and 7V.

Because there are other circuits where we will want to get a zero temperature coefficient by placing a -2.2mV/C diode in series with the zener, this makes 6.2V at +2mV/C the zener of choice.

From On-Semi HBD854/D

"TVS/Zener Theory and Design Considerations"

Below approximately 5.6V, Zener voltage (DC regulation) degrades.

Below 4.7V to 5V the zener is only at it specified voltage in a narrow range of currents.

In the range of 6 to 9V, Zener AC impedances are at a minimum.

Two 7.5V zeners in series will have half the AC impedance as one 15V.

A diode in series with a 6.2V give 6.7V with a near zero temperature coefficient.

At 6.0 to 7.0V zener noise is at a minimum.

Two 6.2V zeners in series has less than 1/3.5 (-11 dB) the noise as one 12V zener.

The Microsemi datasheet for the 1N5525B series shows that the noise is low below 7V and the regulation is good above 5.5V.
The regulation of the 6.2V zener is good down to 0.01mA.
The regulation of the 5.1V zener isn't good even at 1mA.

Two 6.2V in series for 12.4V would have a noise density of
sqrt( [1nV/sqrtHz]^2 + [1nV/sqrtHz]^2) = 1.4nV/sqrtHz
A 12V zener would have a noise density of 10nV/sqrtHz; two 6.2V zeners are 17 dB quieter.

The BZX84C 6.2V performs better closer to 5mA than 1mA.

Fairchild specifies 3 different test currents for the BZX84C series. At 6.2V, the zener regulates nicely.
Fairchild claims 4.2V minimum on a 5.1V zener at 1mA. The Motorola zeners I characterized (tested) measured better. A third vendor ended up being much worse.

Zener Noise (oscillation?) occurs at low bias currents (100uA) unless special zeners are used.

---

First version before 10-Jan-2005, Last change 10-Jan-2005, updated pictures 8-8-2024.

Fixed Grammar error 17-may-2026.

.
( New 2024 index page.) _( Old 2003 index page.)

_( AMP Second index page.) ( Fancy index page.)

.