Softness of diode knees

[Prehistoricman]

I started testing some germanium and silicon diodes on the Magnatone topic so I've copied that here and tested a bunch of other diodes too:
3.6V Zener, 5.1V Zener, white, red, blue, green, yellow, IR LEDs

Here are the V-I plots for 1-4x silicon and 1-4x germanium D9E diodes in series.
X: 0.5V per divison, right-most border is zero volts
Y: ~2mA per division

Here's a closeup of the two diodes:



The diodes are vintage, I think. They came from a pack that looks like this:



germ x1:



germ x2:



germ x3:



germ x4:



silicon x1:



silicon x2:



silicon x3:



silicon x4:



3.6V Zener:



5.1V Zener:



White LED:



Red LED:



Blue LED:



Green LED:



Yellow LED:



IR LED:




I was surprised at how sharp the LEDs are in general. If you want super sharp, go IR. The zeners are pretty soft at about the same gradient as 3x germanium.

JFETs:
0V is now in the centre of the display. Same V/div.

J113:



J202:





The J113 looks almost like a resistor! It's very slightly curved but you need a lot of voltage on it. The J202 is completely different. You can see a constant current region in the negative voltage and an interesting 2-resistance characteristic.

[Paul Marossy]

I don't have any MOSFETs but I could test a couple of JFETs if anybody's interested. If so, how should they be connected?

It's gate and connect the drain-source together to get a diode.

[DaveLT]

Are the zeners forward biased or reverse biased?
when they are reverse biased is when the Zener effect takes place.

I also have some very special avalanche diodes (but those aren't particularly cheap and were for my hifi projects so they are supremely expensive compared to most diodes) and big fat diodes I can test with  (and germ transistors that i can tie as diodes  )

[Prehistoricman]

I don't have any MOSFETs but I could test a couple of JFETs if anybody's interested. If so, how should they be connected?

It's gate and connect the drain-source together to get a diode.

Updated

Are the zeners forward biased or reverse biased?
when they are reverse biased is when the Zener effect takes place.

I also have some very special avalanche diodes (but those aren't particularly cheap and were for my hifi projects so they are supremely expensive compared to most diodes) and big fat diodes I can test with  (and germ transistors that i can tie as diodes  )

Reverse, of course Their forward-bias is very similar to typical silicon.

Cool! I've looked at strange diodes before and wondered how they would sound in audio but they tend to be huge or expensive.

[DaveLT]

I don't have any MOSFETs but I could test a couple of JFETs if anybody's interested. If so, how should they be connected?

It's gate and connect the drain-source together to get a diode.

Updated

Are the zeners forward biased or reverse biased?
when they are reverse biased is when the Zener effect takes place.

I also have some very special avalanche diodes (but those aren't particularly cheap and were for my hifi projects so they are supremely expensive compared to most diodes) and big fat diodes I can test with  (and germ transistors that i can tie as diodes  )

Reverse, of course Their forward-bias is very similar to typical silicon.

Cool! I've looked at strange diodes before and wondered how they would sound in audio but they tend to be huge or expensive.

What sort of diodes? The diode I'm talking about is a BYV26 and those fat diodes are MUR and MBR series. the higher current MUR series have higher forward voltages so if less clipping is the aim?
on the other hand MBR diodes being schoktty diodes have higher Fv if you go for the higher voltage versions 

I have sources of really cheap diodes these days for those mentioned above 
Also, I'll try P MOSFETs as diode clippers and P JFETs... I have a bag of them leftover from audio because they don't meet spec and are really just rebadged cheap 2SJ103s so I can play with those too

[R.G.]

Softness requires a sense of scale. To a 10mV signal, a diode is a variable resistor. To a 100V signal, a diode is an off-on switch.

Can you annotate your scans with the voltage/current scales?

[Prehistoricman]

What sort of diodes? The diode I'm talking about is a BYV26 and those fat diodes are MUR and MBR series. the higher current MUR series have higher forward voltages so if less clipping is the aim?
on the other hand MBR diodes being schoktty diodes have higher Fv if you go for the higher voltage versions 

I have sources of really cheap diodes these days for those mentioned above 
Also, I'll try P MOSFETs as diode clippers and P JFETs... I have a bag of them leftover from audio because they don't meet spec and are really just rebadged cheap 2SJ103s so I can play with those too

All the weird diodes on Wikipedia basically. Like PIN diode, tunnel diode, gunn diode. Some you can't find on ebay and the ones you can find, as you say, tend to be pricey.
As far as I understand, some zeners diode are actually avalanche diodes?

Softness requires a sense of scale. To a 10mV signal, a diode is a variable resistor. To a 100V signal, a diode is an off-on switch.

Can you annotate your scans with the voltage/current scales?

I did consider annotating them but this would take some time. I mentioned the scale at the top of the thread:
X: 0.5V per divison, right-most border is zero volts
Y: ~2mA per division

[DaveLT]

What sort of diodes? The diode I'm talking about is a BYV26 and those fat diodes are MUR and MBR series. the higher current MUR series have higher forward voltages so if less clipping is the aim?
on the other hand MBR diodes being schoktty diodes have higher Fv if you go for the higher voltage versions 

I have sources of really cheap diodes these days for those mentioned above 
Also, I'll try P MOSFETs as diode clippers and P JFETs... I have a bag of them leftover from audio because they don't meet spec and are really just rebadged cheap 2SJ103s so I can play with those too

All the weird diodes on Wikipedia basically. Like PIN diode, tunnel diode, gunn diode. Some you can't find on ebay and the ones you can find, as you say, tend to be pricey.
As far as I understand, some zeners diode are actually avalanche diodes?

Zeners are not necessarily avalanche diodes, their avalanche effect occurs *after* the zener effect although in some of them they happen at the same time.

[Rob Strand]

FWIW, I did a stack of measurements in 2001 and 2009.

There's also an issue comparing apples to apples with different diodes.

With a diode there's two aspects which are completely different things:  voltage drop and softness.
Voltage drop can be defined as the diode voltage at a given diode current.   
Softness is defined as the how much the voltage *changes* when the current changes. 

The voltage of normal diode might change about 33mV when the current is doubled.  The voltage of a hard diode would not change at all.   The voltage of a soft diode might change 50mV.

To see the difference between voltage and softness imagine putting two of the *same diode* in series.   The point here is the same diode is used so the combined diode has the same softness as a single diode. For the combined diode:
- The voltage drop will increase from say 600mV to 1.2V.   
- When the current is doubled, the change in voltage across the two is now 2x33mV = 66mV.

The change in voltage is now doubled compared to the single diode yet we know the softeness should be the same as the single diode because they are the same diodes.   This highlights the problem of not comparing apples with apples.   To fix the problem imaging putting a gain of 1/2 stage after a two diode clipper.  Now the scaled output is (1/2) * 1.2V = 600mV and more importantly the the change in the scaled output when the current is doubled is (1/2) * 66mV  = 33mV.    So now when we compared the scaled output with the single diode we get the same conclusion about the softness.

The way to generalize this idea is suppose we have a reference current through a diode.  The voltage at this reference current is the nominal diode voltage drop.   If we now measurement the diode characteristic at different currents and divide by the nominal diode voltage  to produce a normalize diode voltage (Vd / Vd @ Iref) then all the diode characteristics will be normalized.   They will all pass through 1 at the reference current.  The main point is the change in voltage at different current, which represents the true softness, can now be compare for diodes with different voltage drops.

If you take a single diode and two of the same diode in series and perform the normalization you will conclude the softness is the same.

So after normalizing stack of different diodes you get this graph,



Notice:
- The ge diodes are softer so the change in voltage is more for a given current change.
- The various silicon diodes are actually quite close after normalization.

What is not included in these graphs is:
- the effect of diode capacitance
- the fact opamps may clip with an LED clipper but not with a silicon diode clipper.   So if you took five silicon diodes in series they might sound different to a single diode because the opamp is now clipping.
- circuits like the tube screamer leaks some of the clean signal in with the clipped signal.  When you add more diodes in series it waters down the clean sound relative to the clipped sound.

[Prehistoricman]

To see the difference between voltage and softness imagine putting two of the *same diode* in series.   The point here is the same diode is used so the combined diode has the same softness as a single diode.

This is a very good point! But it all depends on that definition of softness. My definition was:

At the knee, the rate of change of resistance with respect to voltage

Is that a different definition? I'm not sure actually.
I'm trying to express what we were talking about in the Magnatone thread. To use the diode as a variable resistor, the signal has to be small. It seems to me that putting several diodes in series would allow the signal to be bigger than with a single one, right?

[R.G.]

I did consider annotating them but this would take some time. I mentioned the scale at the top of the thread:
X: 0.5V per divison, right-most border is zero volts
Y: ~2mA per division

Thanks, I apparently missed that in my leap to the pictures. That's what I was looking for - were the pictures all to the same scale.

For audio modulation purposes the softness of a diode knee or any other curve depends on how what you measure it against, at least in my mind. Measuring the rate of change of slope in the knee versus input voltage is a valid way to look at it, but the issue of signal size gets confounded with that in my mind. A curve is after all a curve, and only has a slope that's definiable at one point. Even the calculus definition involves taking a tiny difference in one variable versus a tiny difference in the other variable, then seeing what happens as those "tiny" things go to zero size. A similar view happens here, except that we have to look at the "tiny difference" slopes with a yardstick that is the peal to peak voltage of the signal we're counting on to be resisted.

The knee being a curve, only a zero-amplitude signal will be undistorted when passing through the resistance, because the resistance itself changes along the signal's amplitude.  So the signal size matters to variable "knee" resistors.

One thing I've seen a lot is to pick some acceptable distortion amount, often 1%, and then measuring what signal amplitude gives that amount of distortion when passed through that-amount-curved knee. I have read things over the years that places that amount of acceptable distortion at about 1% for a nominal silicon diode curve and a 25 to 50mV signal.

In my mind, knee sharpness can be measured different ways. One is the rate of change of slope per volt of change across the device. Another that is popular for zeners is the change in voltage per change in current, this being a measure of how low a current you can bias your zeners to and still get small change in zener voltage with small current changes. I think for audio purposes, the size of the signal that can pass through the "resistance" at some small distortion is the most useful.

After all, we pass signals through diodes to get huge distortions all the time. Making the signal pass through with little distortion as in a variable resistor is trickier.

[Rob Strand]

This is a very good point! But it all depends on that definition of softness. My definition was:
At the knee, the rate of change of resistance with respect to voltage

There's nothing wrong with that definition.  You can also say rate of change of resistance with respect to current.  What is ambiguous is the meaning of resistance.   If you plot the V/I curve of a diode, then pick a point along that curve you can define the "large signal" resistance R = V_d / I_d, or, you can define the small-signal resistance, which is the slope at that point r = dV_d/ dI_d. 

The problem with the first definition is you can draw many V-I curves through that point, in the simplest case you would have a real resistors R = V/I, at the other extreme is a hard knee which just happens to pass through chosen VI point.

The problem with the second definition is it doesn't take into account the size of V compared to the resistance.  A given slope r at 1V or going to be more non-linear than the same slope at 100V.    It's like the effect of squashing the top of sine wave looks more clipped in the 1V case.

For the case of two diodes in series the knee voltage doubles and the resistance r doubles.  We know the softness is the same and the know the knee voltage double.  That case inspires the solution a method of comparing softness.   We divide the slope r = dV_d/ dI_d by the knee Voltage.

If you go back to the exponential diode equation it is possible to calculate the slope,

r_d  = n V_T / Id
V_T = k T / q   ~ 25mV to 26mV ;
k= Boltzman's constant, T=temperature in Kelvin, q= electron charge

The factor n is a way modelling diode curves and the diode characteristic which determines the softness.  Transistors wired as diodes tends to have n=1 and silicon diodes have n around 2.

Is that a different definition? I'm not sure actually.
I'm trying to express what we were talking about in the Magnatone thread. To use the diode as a variable resistor, the signal has to be small. It seems to me that putting several diodes in series would allow the signal to be bigger than with a single one, right?

The diode curve (ie. the exponential diode curve) is non-linear so when you deviate the current the voltage does not follow.  That produces distortion.   For a single diode typically you don't want more than about 10mV p-p.  So putting diodes in series lets you increase the allowed voltage swing for the same amount of distortion.

[Rob Strand]

Here's an example showing how the diode which has n=2 and hence a softer knee produces lower distortion for the same output level.

**** This is not a clipper so the DC voltage across the device is not important.  We are concern with outright non-linearity about the operating point.

Note that to produce the same output level,

    rd  = n VT / Id

must be the same between the two test cases. 

Since the diode has n=2 and the transistor n=1 the diode must operate at about twice the current for the same rd.

In this example the output voltage swing is somewhat under the maximum.

Right hand side shows DC offset.


VOQ and VOD2 produce the same output swing:


Compare relative distortions only (voltage level is a little off).
Transistor has more distortion than the diode.

[bool]

You can tweak the bjt with adding a small resistor between b and c. For more dist in a single stage...

[Eb7+9]

to complete the set you could add
curves for real Ge diodes in there

meaning the BE junction of a Ge transistor
(Collector tied to Base, or not - no difference)

The curves should show a sharp knee,
similar to Si diodes, starting at around 200mV

—

a few comments if I may ...

(i) if we normalize the data of either the whisker
or zener curves
as the AN102 app Note does on jFETs
then we see a similar asymptotic 1/x type function
(ignoring offsets)
so, the horizontal and vertical values values don't matter
since external circuit values simply change accordingly
the same way as we do, again,
w jFETs with varying Vp/Idss values

the Zener and Whisker diodes both have similar
soft-knee profiles ...

I didn't realize they were this close before now (thx!)

(ii) years ago I worked on a rare Ampeg 435-S amplifier /
a glorious invention
that carries a very "Vibe-ish" sounding phase shift circuit ...
I spoke to Jess Oliver
about it a few years before he passed

Jess said only about 25 were made and he kept referring
to it as an harmonica amp ... killer on gtr of course ...

Anyway, my point, the circuit features a Phase shift
lattice with
the output being continuously modulated by a
whisker-diode bridge
ie,. 1n34a whisker diodes

The effect sounded great, but also slightly limited
compared to a Vibe
at least that's how it was setup there

https://ampeg.com/support/files/Schematics/Misc/635S,SN%20(Zephyr)/435S,SN%20-%20635S,SN.pdf

(iii) i remember somewhere coming across a laboratory
grade signal limiter (Honeywell?)
based on whisker diodes ...

[Rob Strand]

meaning the BE junction of a Ge transistor
(Collector tied to Base, or not - no difference)

The curves should show a sharp knee,
similar to Si diodes, starting at around 200mV

So basically n = 1 (sharper knee) for transistors.

the Zener and Whisker diodes both have similar
soft-knee profiles ...

For Zeners operating in breakdown (ie. "Zenering") the curves are soft at low voltages and hard at high voltages.   They become hard around 3.3V.    For zeners operating as silicon diodes they tend to be close to transistors, so harder than normal silicon diodes.

[DaveLT]

meaning the BE junction of a Ge transistor
(Collector tied to Base, or not - no difference)

The curves should show a sharp knee,
similar to Si diodes, starting at around 200mV

So basically n = 1 (sharper knee) for transistors.

the Zener and Whisker diodes both have similar
soft-knee profiles ...

For Zeners operating in breakdown (ie. "Zenering") the curves are soft at low voltages and hard at high voltages.   They become hard around 3.3V.    For zeners operating as silicon diodes they tend to be close to transistors, so harder than normal silicon diodes.

Apparently beyond 3.3v as zeners normally operating they are avalanche effect rather than zener so that might explain the hardness as the onset of avalanche is very rapid and sharp

[Fancy Lime]

meaning the BE junction of a Ge transistor
(Collector tied to Base, or not - no difference)

The curves should show a sharp knee,
similar to Si diodes, starting at around 200mV

So basically n = 1 (sharper knee) for transistors.

the Zener and Whisker diodes both have similar
soft-knee profiles ...

For Zeners operating in breakdown (ie. "Zenering") the curves are soft at low voltages and hard at high voltages.   They become hard around 3.3V.    For zeners operating as silicon diodes they tend to be close to transistors, so harder than normal silicon diodes.

Apparently beyond 3.3v as zeners normally operating they are avalanche effect rather than zener so that might explain the hardness as the onset of avalanche is very rapid and sharp

So, do I get this right? Zener effect = soft knee; avalanche effect = hard knee. Wikipedia tells me that the crossover is around 5.6V, not 3.3, but be that as it may. Either way, we won't normally get pure avalanche breakdown clipping in a 9V pedal without extra shenanigans. An interesting question that this raises is: where do I get backward diodes, aka. tunnel-diodes? These are Zeners, where the Zener effect sets in at lower voltages than the forward Si diode. A 0.4V Zener might be interesting as a replacement for a Ge diode. Since the Zener effect is soft, I would not expect a huge difference (beyond the variability within Ge diodes), but may be interesting to try, no? Ge diodes are not going to be around forever. Also, I would imagine that the backward diodes may be available in SMD, Ge probably not so much. I dimly remember having seen a huge datasheet with a range of backward and Zener diodes starting at 0.1V or up to 100V so but I cannot remember the names. Anyone know what I'm talking about?

Cheers,
Andy

[Rob Strand]

So, do I get this right? Zener effect = soft knee; avalanche effect = hard knee. Wikipedia tells me that the crossover is around 5.6V, not 3.3, but be that as it may. Either way, we won't normally get pure avalanche breakdown clipping in a 9V pedal without extra shenanigans. An interesting question that this raises is: where do I get backward diodes, aka. tunnel-diodes? These are Zeners, where the Zener effect sets in at lower voltages than the forward Si diode. A 0.4V Zener might be interesting as a replacement for a Ge diode. Since the Zener effect is soft, I would not expect a huge difference (beyond the variability within Ge diodes), but may be interesting to try, no? Ge diodes are not going to be around forever. Also, I would imagine that the backward diodes may be available in SMD, Ge probably not so much. I dimly remember having seen a huge datasheet with a range of backward and Zener diodes starting at 0.1V or up to 100V so but I cannot remember the names. Anyone know what I'm talking about?

A better way to say it is this,

Zener < 8V
Avalanche > 5V
Between 5V and 8V both mechanisms active.

So < 5V is pretty much all Zener, and > 8V pretty much all Avalanche.

Avalance has a positive temperature coefficient.
Zener has a negative temperature coefficient.
In the region around 4.7V to 5.6V the two mechanisms
cancel and give a zero temperature coefficient.

For zeners below about 3.9V their softness is very much like silicon diodes.

I've never seen zeners below about 1.8V.   I suspect some problems come up with doping the material.  Besides the characteristic of such a zener is pretty much like connecting two normal diodes back to back so there's no point making them.

[notTHEwillienelson]

Sorry to jump on an old thread, but my question is directly related, I believe. Let's say youre using Zener diodes in series with Si diodes (cathode to cathode and anode to anode) as hard clipping, how would the knee look? I guess the question is, when you combine hard and soft knee diodes, what is the overall effect on clipping?