EEVblog #1157 - Transistor Zener Clamp Circuit

[EEVblog]

A look at the dual bipolar transistor zener clamp circuit, and it's usefulness as a fast low leakage overload clamp.
The first part is lifted from video #1000 with 12 minutes of extra commentary.

Interview with Dave Taylor, designer of the famous Fluke 8060A multimeter:
https://theamphour.com/180-an-interview-with-dave-taylor-multi-talented-meter-maker/

[MT]

IMX9/25/26 Vebo are mentioned in the electrical characteristics. High Vebo trannies is used by the audio industry in AC mode to mute popping on outputs when power up/down, base needs to be floating (PNP driver) else it will acts a diode and clip out signal.

[johnlsenchak]

Great informative  video, would like to see more like this  !  Thumbs up!

"Winner Winner chicken  dinner"

[Dr. Frank]

Dave,

using a bipolar transistors reversed  BE diode, or n-JFETs to protect sensitive amplifier inputs, with ultra-low leakage currents ~pA, has been very common in metrology circuits.
Fluke and also Keithley often use that, maybe hp also, but I could not tell, who was the earliest adopter.
Flukes 845A null detector/µV-amplifier is from 1965, but still uses these ultra low leakage diodes, like CD12599, which all cost a fortune.

I think, I've seen that transistor solution already in early 1970ties circuits.

What's missing in your nice video, is a measurement on this leakage current at higher voltages, and a measurement on the zener voltage of the transistors.
In the 121GW, on its 50MOhm and 5 MOhm ranges, it's  important to keep these currents really low, in the low pA range.

If high break-down voltage transistors are used, there may be higher leakage currents than on a 'normal' 5..7V break down-type. 

I've already investigated on that 121GW problem a few months back: https://www.eevblog.com/forum/testgear/eevblog-121gw-multimeter-issues/msg1770224/#msg1770224, and my final solution, using a jelly-bean n-JFET, which D-G or D_S diode withstands up to 30V, has a few pA leakage only:
https://www.eevblog.com/forum/testgear/eevblog-121gw-multimeter-issues/msg1790081/#msg1790081

So this protection, either using n-JFETs, or (leakage - selected) bipolar transistors, will cure the 5M / 50M problem, but requires a re-calibration of the 121GW, as the temperature dependent leakage currents
of the former 1N4001s had been included in the former calibration.

Frank

PS: And yes, I'm glad, that UEI did NOT use TVS in that place!

PPS: Does this circuit now really protect the input of U9?

It looks like it replaces D8 only, that means that R_RLD is clamped at an (unspecified) BE-zener voltage of the transistor of maybe +/-25V, or more.
If IC U9 is operated  at either 4V or 15V, that might damage its input pin15. And this voltage should not go below -0.6V.

I think, that circuit is a complete fail in this application.

[TinyMirrors]

I tried simulating this in lt spice and no dice! Well no clamping anyways. I guess I'll have to build it in real life.

[pc2005]

Hi,

It seems a modified version of the circuit was used in the eastern bloc probably from like late '60 to early '90. In my country it was in every TV made and it was used for stabilization of ~30V tuning voltage (varicap in superheterodyne tuner). The original device seems to be TAA550.

The most common problem was degradation of the part so a tuned channel became really unstable.

[tszaboo]

Anyone knows transistors that are defining the base-emitter voltage not in the "Absolute maximum ratings" part of the datasheet? Or where they define that this is a normal operation of the transistor? I'm asking because I'm designing intrinsically safe circuits ( ATEX ) and it would be nice to use these techniques, for high power (or low leakage) zener circuits. But the notified body would just throw this back that we exceed 2/3 of the maximum rating (requirements for such circuits) so we are not compliant. 

[santiall]

Japanese use to mute the outputs of professional audio devices with BJTs. It is a bit puzzling the first time you see that and even more once you try the circuit with a normal transistor and doesn't work till you learn about the reverse EB voltage

For example Yamaha has been using 2SD1915 (25V) BJTs in many products and lately the INC2002AM1 (50V). I guess no wonder that those 'muting' specific transistors are mostly Japanese.

[AndersJ]

I don't get it.
How do you mute audio using the reverse EB voltage?

[santiall]

you mute when the transistor is conducting, what you want is a high reverse voltage so when the transistor is off, when not muting, you don't clip half of the audio signal.
I'm attaching the output of the 0V96 digital mixer so you can see how they do it

EDIT, an interesting link with more details: https://www.electroschematics.com/9660/muting-transistor-attenuator-circuits-2sc2878/

[mrpackethead]

Anyone knows transistors that are defining the base-emitter voltage not in the "Absolute maximum ratings" part of the datasheet? Or where they define that this is a normal operation of the transistor? I'm asking because I'm designing intrinsically safe circuits ( ATEX ) and it would be nice to use these techniques, for high power (or low leakage) zener circuits. But the notified body would just throw this back that we exceed 2/3 of the maximum rating (requirements for such circuits) so we are not compliant.

Not sure about how those rules work, but are you saying that if you had a 15V zener you would only be able to use it up to 10V?

[Kleinstein]

There is another downside on using a transistor as a zener: the tolerance of the break down voltage. The absolute maximum specs are kin dof minimum values. The actual value can be quite a bit higher.

The maximum limits rule applies to the absolute maximum ratings only, as exceeding these could damage the part. So if a 15 V zener would have a maximum rating of lets say 18 V you are stuck. It's stupid rules because it's just a question of having the numbers there or not.
Using the transistor as a zener is not an intended function - so no normal specs on this.

[ggchab]

Very interesting video 
Happy to see again this type of content 

[tszaboo]

Anyone knows transistors that are defining the base-emitter voltage not in the "Absolute maximum ratings" part of the datasheet? Or where they define that this is a normal operation of the transistor? I'm asking because I'm designing intrinsically safe circuits ( ATEX ) and it would be nice to use these techniques, for high power (or low leakage) zener circuits. But the notified body would just throw this back that we exceed 2/3 of the maximum rating (requirements for such circuits) so we are not compliant.

Not sure about how those rules work, but are you saying that if you had a 15V zener you would only be able to use it up to 10V?

I need to de-rate anything for current, voltage and power (amongst others) to 2/3 of it's original rating. That is for stuff which is in the "absolute maximum rating" part, and Vbe is there. So "normally" i wouldn't be able to use these transistors in this configuration, because it is exceeding that rating.

For a zener, I only need to investigate for current and power of course, because the zener voltage is not a rating, it is "electrical characteristics".

Basically I'm asking if there are transistors specifically designed for this, with reliably doing this breakdown, with the manufacturer stating that it is made for it.

There is another downside on using a transistor as a zener: the tolerance of the break down voltage. The absolute maximum specs are kin dof minimum values. The actual value can be quite a bit higher.

That doesnt sound too good for my application.

[T3sl4co1l]

I tried simulating this in lt spice and no dice! Well no clamping anyways. I guess I'll have to build it in real life.

SPICE doesn't model breakdown of BJTs, you need to put a zener in parallel with the respective junctions.  (The zener can be an ideal type, i.e., N = 20 and CJO=0 so it doesn't interfere with normal operation, and RS, BV and IBV set to model breakdown.)

I'm not sure anyone has modeled punch-through breakdown, i.e., the effect that gives rise to relaxation oscillators using E-B or C-B breakdown (the latter of which is famous for giving sub-nanosecond edges).  Extreme nonlinearities are not easy to model, or to simulate in SPICE.

Tim

[artag]

HP1740A .. that was the pride of my bosses' workshop back in 1980.

[iMo]

I'm not sure anyone has modeled punch-through breakdown, i.e., the effect that gives rise to relaxation oscillators using E-B or C-B breakdown (the latter of which is famous for giving sub-nanosecond edges).  Extreme nonlinearities are not easy to model, or to simulate in SPICE.

There is such a model with the 2N2369A for LTSpice in "Bordodynov library"

[SiliconWizard]

I tried simulating this in lt spice and no dice! Well no clamping anyways. I guess I'll have to build it in real life.

Had to dig up a little on this one, but it turns out LTSpice does model the collector-base breakdown voltage and base-emitter BV (BVcbo and BVbe parameters) in its BJT model, but doesn't model the collector-emitter BV as far as I've seen (look up "Q. Bipolar Transistor" in LTSpice's help.) Some other Spice-based simulators may include breakdown voltage parameters as well, but as far as I've seen, ngspice doesn't (unless its doc. is not up to date).

Unfortunately, the base transistor models shipped with LTSpice haven't been updated to make use of those parameters. These models can be found in the "standard.bjt" file in LTSpice's "lib\cmp" subdirectory.

I added the following model (modified 2N3904):

Code: [Select]

.model 2N3904_BD NPN(IS=1E-14 VAF=100 Bf=300 IKF=0.4 XTB=1.5 BR=4 CJC=4E-12 CJE=8E-12 RB=20 RC=0.1 RE=0.1 TR=250E-9 TF=350E-12 ITF=1 VTF=2 XTF=3 Vceo=40 Icrating=200m BVcbo=60 BVbe=6)
(values taken from a Fairchild 2N3904 datasheet)

With the attached LTSpice schematic, the voltage gets clamped at about +/- 8.3V.

[TinyMirrors]

I added the following model (modified 2N3904):

Code: [Select]

.model 2N3904_BD NPN(IS=1E-14 VAF=100 Bf=300 IKF=0.4 XTB=1.5 BR=4 CJC=4E-12 CJE=8E-12 RB=20 RC=0.1 RE=0.1 TR=250E-9 TF=350E-12 ITF=1 VTF=2 XTF=3 Vceo=40 Icrating=200m BVcbo=60 BVbe=6)

This worked perfectly! I'm gunna dig up that datasheet and take a look. Great post!

[David Hess]

Dave, you asked about the history of these transistors.

They were what was used before JFETs became available for chopper and sampling circuits and they often had completely symmetrical construction so the Vcb and Veb were identical.  This yielded poor gain compared to an asymmetrical device however it improves the saturation voltage and I think the Early voltage also.

As mentioned above, they were used in audio muting applications and this was done even after JFETs became available because they can operate at very low voltages.  Note that when the emitter is forward biased, the transistor will conduct in either direction making them useful in AC applications.  This is where they are used as shunt choppers in audio muting applications and the Rigol DS1000Z schematics you made show the same configuration for the bandwidth limiting circuits which I have seen in some other oscilloscopes.  A normal transistor works for this but a chopper transistor works better.

Some analog circuits use the reverse connection of a standard bipolar transistor for its low saturation voltage to do things like reset an integration capacitor but earlier a chopper transistor would have been used.  Early low offset and offset drift chopper amplifiers used them.

Below is an except from the GE Transistor Manual discussing chopper transistors and their circuits.  I have this in paperback but copied it out of an existing PDF available online which shows they had a 50 microvolt matched device in the 1960s.

Tektronix used the dual series transistor chopper configuration twice in their 7T11 sampling sweep as shown below but I am not sure why since JFETs were readily available at that time.  My guess is that the transformer coupled bipolar chopper has less charge injection.  The chopper output drives a hold capacitor which is buffered by a JFET source follower or a very simple JFET operational amplifier.  In other places in the 7T11, a JFET transistor was used for chopping instead so Tektronix was certainly aware of their capability.

Q400 and Q402 are just 2N3904s demonstrating that there is nothing special required to use this configuration.  Q446 and Q448 use the 2N3563 is a 600-1500MHz 20-200hfe 12Vce 50maIc part designed as an RF amplifier but it is not gold doped; they might have used it for less feedthrough capacitance.

I'm not sure anyone has modeled punch-through breakdown, i.e., the effect that gives rise to relaxation oscillators using E-B or C-B breakdown (the latter of which is famous for giving sub-nanosecond edges).  Extreme nonlinearities are not easy to model, or to simulate in SPICE.

The GE book linked above mentions that this behavior is what makes avalanche switching work.  I have seen it several times on my curve tracer without realizing its significance which suggests an easy way to look for suitable transistors.

[floobydust]

Audio muting transistors i.e. 2SC2878 have a high reverse-hFE rating and higher VEBO 25V.

A big issue is clamp leakage current verse temperature, you can measure a nice low leakage current at room temp but I usually find it's 10X higher with a 10°C temperature increase.

[SilverSolder]

I thought the BE junction gets damaged over time, with a reverse current?

[David Hess]

I thought the BE junction gets damaged over time, with a reverse current?

It does but this does not matter for a low leakage clamp or Vbe zener diode.

[EEVblog]

you mute when the transistor is conducting, what you want is a high reverse voltage so when the transistor is off, when not muting, you don't clip half of the audio signal.
I'm attaching the output of the 0V96 digital mixer so you can see how they do it

EDIT, an interesting link with more details: https://www.electroschematics.com/9660/muting-transistor-attenuator-circuits-2sc2878/

I'd never really given it a thought before, as line level stuff would be fine for a regular tranny. But yeah, higher voltage line signal need a special snowflake.

[santiall]

I'd think that when creating those type of transistors the designers were aiming for true symmetrical BJTs with a good negative hFE and the high reverse voltage is a necessary evil. They are specified as switches and mute transistors, not as clamps so I'd think that'd be their origin rather than the other way around.

As said, they have been extremely popular in Japanese electronics for muting due the very low on-resistance, lower than FETs, and probaly cheaper and more consistent too than FETs.

[T3sl4co1l]

Incidentally, if you do need low switch resistance in a BJT, "low Vce(sat)" types usually have very good saturation resistance -- including high inverted hFE (though they aren't rated for it -- do measure this yourself on whatever parts you get!).  Veb is not very high though.

Tim

[free_electron]

You really shouldn't be doing this. If you drive a B-e junction into reverse you create hot electrons that collide with atoms outside the the depletion zone. They can create new hole-electorn pairs or can collide with the oxide round the junction. ( making it resistive) . This causes permanent, irreversible damage. They can penetrate the oxide and become trapped. IF they have enough energy they create something called an 'interface state' in the silicon/silicondioxide contact layer. This has an effect on the b-e junction. It increases the recombination speed of the charge carriers in the depletion layer and this creates additional base current... collector current remains unchanged. Ic/Vbe remains stable , but at low currents the current gain collapses (ib/ic) due to the added base current caused by the trapped charge carriers.

Problem two is that this is an avalanche behavior. Every electron-hole pair will generate another pair and things avalanche very fast.
The trapped electrons collapse the hfe.

Problem three is that this stuff increases the noise. you get effects like popcorn noise, RTN ( random telegraph noise) cause by the trapped stuff.

You can partially reverse this by heating the transistor to 300+ degree centigrade which will 'heal the interface states' but the traps are lost forever...

Certain RF transistors can be damaged with as little as 2 volts reverse across their b-e junction.

In situations where a b-e junction can see reverse voltage you will find that the designers put two silicon diodes (two diodes in series ) in antiparallel with the b-e junction. That way the reverse voltage can not go above 2vbe.
There exists a special diode stack for that functions.

[nctnico]

But wouldn't this effect depend very much on the type of transistor process used and maximum current? It even seems some transistors are made for this purpose.

Years ago I experimented with trying to get BJT transistors to avalanche. The more modern ones just didn't want to avalanche at all.

[drtaylor]

Almost everyone is missing the point of this circuit. It is a protection circuit for the ohms function of a DMM. In normal operation, no current flows through the transistors. As long as the normal ohms excitation voltage is present, the bipolar clamp is not conducting. The main selling point is the leakage current is so low, that in parallel with the unknown resistor, it does not cause any errors. If you used standard zeners, you'd never get a 20Megohm circuit to be accurate!

When an external voltage is applied to a DMM in the ohms mode, this is a fault that can damage the excitation circuitry, which has to be fairly low impedance. So a PTC device is placed in series with the excitation source. The dual transistor clamp will zener at around 6-7V (it does not matter how accurate it is) from the externally applied fault voltage. The clamp circuit will heat up, but so will the PTC. Once the PTC switches to its high impedance state, the clamp is safe, the DMM is safe. The clamp circuit does not have to dissipate power for long because the PTC switches.

This circuit, built with 3904 transistors, is proven reliable, and can protect the ohms circuitry for thousands of events without failure. There is no noise generated in normal operation. Noise generation does not matter when the circuit clamps, because the ohms function will overload. If you use two back to back, it is a bipolar clamp that protects the ohms circuitry from the accidental application of AC line voltage.

All that being said, you can use 3904s and similar transistors as low leakage diodes in a lot of circuits. Using just the BC junction, you can select devices that leak in the low pA range. This is very advantageous in circuits like sample and holds where even a low leakage diode would not suffice.

I do not suggest or recommend that you would ever use a forward biased BE junction as a zener. It would not be capable of much power. But given sufficient resistance in front of it, it is a great clamp...a low leakage diode in one direction, and a zener in the other. Emphasize clamp! Not a full time conducting zener.

A major advantage of this circuit is that it is dirt cheap!

[tszaboo]

I thought the BE junction gets damaged over time, with a reverse current?

It does but this does not matter for a low leakage clamp or Vbe zener diode.

Well, there goes my idea of using it.

[Dr. Frank]

After reviewing the circuit diagram of the 121GW, i come to this concluion:

The usage of this clamping circuit as protection of the 121GW  O/B/D/C modes is a complete fail.

It does not protect the Y port of the 4053 MUX, as the clamping / zener voltage is too high or un-specified, as the circuit clamps in both directions at > +/-25V only.

So it's letting Y / pin 15 go far beyond its supply voltage (+4 or +15V), if positive over voltage occurs, but also lets the  Y / pin 15 go far below -0.5V if negative voltages to the input jacks are applied, so damaging the 4053 in both cases.
Especially in 'normal' mode, when NOT using the 15V supply, but +4V only, the 4053 is completely unprotected now.

To have low leakage protection, you should only use the BE diode of a bipolar transistor in diode mode, but not in zenering mode, as zenering might damage the BE junction.

In other words, for an effective protection, you should have replaced both 1N4007 of the original circuit by two separate BE diodes of transistors (or n-JFETs) with > 20V reverse voltage specification.

Frank

[Dr. Frank]

OK,
here's the analysis schematics.

The DUT Rx receives either a constant current or test voltage from the MUX, pin15 (or 15V from the SMPSU, U10, not displayed here) via SW 36-37, and over current protection R16 / PTC 3.
The generated voltage over Rx is routed to voltage input, PB0, of the HY3131.



In the original circuit, in case of (over) voltage to V/Ohm jack. D7 will clamp positive voltages beyond VDD into pin 15 of the MUX U9, HEF4053, and D8 will clamp negative over voltages  below VSS, or GND.

As can be seen in Dave's video #1158, 15:46 min onwards, D7 is removed, and D8 is replaced by Dave's zenering circuit, as shown in the changed circuit diagram.
This assembly will zener at >+25V or < -25V only, so exceeding the absolute limits of pin15 in any case of over voltage on the Ohm input jacks.



Effectively, there is now no protection any more.

Frank

[mjs]

I did some Vebo damage tests 2 months ago with HP 4142B I bought. Here's Vebo change https://plot.ly/~msyrjala/11 and hFE damage https://plot.ly/~msyrjala/9 from one of the tests.

I think I'll find my scripts and check if the leakage current is affected by the damage, too!

[Kleinstein]

The protection with a kind of +-25 V zener clamp in deed look odd for an input that might withstand something like -1 V to + 16 V.  For a positive over-voltage the  chip internal "diodes" would direct some current to rise the supply and depending on the version some 15 or maybe 18 V are still OK, though the digital signals would be no longer valid - so temporary male-function cold happen.
With to the negative side even -7 V would be too much, as there is no negative supply.

Too much current to the chip internal protection "diodes" could cause latchup and thus permanent damage if the supply is strong enough. So a weak supply could safe the chip.
 

It is known that reverse (zener) current to the BE junction can cause some more or less permanent damage.  A similar hot electron effect is also suggested to happening in normal zener diodes and possible process to produce 1/f and popcorn noise.

If this damage also causes an increased leakage current is a good question - it at least seems possible, though it may no effect PNP and NPN transistors in the same way.

[nctnico]

I probably missed this but shouldn't the HEF4053 have internal ESD protection diodes? Why aren't these working? There is enough resistance in front of it to limit the current. I'd just remove the 1N4007s. Who thought those where a good idea anyway? And clamping to the supply voltage?

[HKJ]

I probably missed this but shouldn't the HEF4053 have internal ESD protection diodes? Why aren't these working? There is enough resistance in front of it to limit the current. I'd just remove the 1N4007s. Who thought those where a good idea anyway? And clamping to the supply voltage?

The transistor clamp may have to handle 0.5A until PTC heats up, that is probably too much for the internal diodes. The smart way to do it is to have a resistor between the transistor clamp and the chip, then you can control how much current the internal diodes must handle.

[nctnico]

I probably missed this but shouldn't the HEF4053 have internal ESD protection diodes? Why aren't these working? There is enough resistance in front of it to limit the current. I'd just remove the 1N4007s. Who thought those where a good idea anyway? And clamping to the supply voltage?

The transistor clamp may have to handle 0.5A until PTC heats up, that is probably too much for the internal diodes. The smart way to do it is to have a resistor between the transistor clamp and the chip, then you can control how much current the internal diodes must handle.

Such a current will likely kill a transistor clamp as well. The HEF4053 from NXP is specified to have a maximum clamping current of 10mA.

[HKJ]

The transistor clamp may have to handle 0.5A until PTC heats up, that is probably too much for the internal diodes. The smart way to do it is to have a resistor between the transistor clamp and the chip, then you can control how much current the internal diodes must handle.

Such a current will likely kill a transistor clamp as well. The HEF4053 from NXP is specified to have a maximum clamping current of 10mA.

I do not know it, but they can handle about 200mA, that is about what they get when you plug a typically cheap meter into 230V mains when in ohm range and then usual survive that.

[David Hess]

But wouldn't this effect depend very much on the type of transistor process used and maximum current? It even seems some transistors are made for this purpose.

A lot of transistors are made for this purpose; they are called zener diodes.

Gold doped and high power transistors are largely immune because they already suffer from low minority carrier lifetime.  Reverse breakdown of the base-emitter junction most seriously affects low current gain.  At high currents, the loss of gain from the junction damage becomes insignificant.

Years ago I experimented with trying to get BJT transistors to avalanche. The more modern ones just didn't want to avalanche at all.

Avalanche operation is a completely different thing where the base-emitter junction never goes into breakdown.  It relies on the negative resistance of the Vces curve (shown in my third link above) at low currents which not all transistors display.

If this damage also causes an increased leakage current is a good question - it at least seems possible, though it may no effect PNP and NPN transistors in the same way.

I have never noticed that it affects leakage unless of course the transistor gets destroyed and then it first becomes a short.  This is how zener zapping works.

[julian1]

I think it should be possible to use two b2b npns with base and collector tied in series (eg. 4x bjts) for twice the drop.

So, if it clamps at 8V (per the video), that would give you 16V, which would be about right for a 15V range diode tester circuit.

The advantage being the choice of a better characterized part, albeit maybe higher BOM cost.

I just experimented with a couple of 2n3904 in b2b config with base and collector tied.  Element 14 part number gives manufacturer as On semi, and On datasheet specified V(EBO) at 6V.  I find they clamp at 9V for 10mA. So even with a 0.6V drop on the other emitter, I think it must be a very soft knee. 
 

Edit.

I think the 6V V(EBO) figure is for 1uA leakage.  At least that matches what I measure.

[hatte]

Reviving an OLD thread to provide some information to passers by.

If you desire a higher "zener" clamp voltage than the typical BVbeo + PN of 8V (7 + 0.7), you can use a pair of darlington BJTs and obtain double that, typically 15V (14 + 1.4V) in an otherwise unmodified configuration.

With regards to the accuracy of the BVbeo voltage, it is typically mentioned in either Absolute Maximum Ratings (indicating the low end of the BVbeo range) or as a Minimum value in the Electrical Specifications, as in the Diodes FZ605 Darlington NPN Datasheet (attached)