Glen - do you mean - for example - 0.75 volts? Or more?

Lutz, 

Right Direction.  Thanks for the comment.

The common attribute for silicon diodes is 0.7 Volts. 

In my circuits I designed limiting voltage levels up to 2.1 Volts. And as low as 10 uVolts

( but noise kills this signal ).

That is the level I thought I required when I did the design

... higher than 0.7 and lower than 3.0 V

for various noise level and signal swing considerations .

Also, I was tinkering with the idea 

and just wanted to see if there was a usable range. 

In General, 

the R(series) adds to the total of :

the Diodes' Internal-Resistance

the Diodes' Dynamic-Impedance.  

Stated this way :

VF = VK log (IF/IS(T)) 

where real world "I"

always = V/(R+Dynamic-Impedance) 

and I am adding to the "R" .  

There are at least two 'curves' involved. 

Other materials in nature have several 'curves' of response. 

A subject I found difficult to grasp back in the early 80's

when I tried to tell my mentor that he might be wrong. 

Check the Bode and you will see incrementally greater levels of limit, with greater linearity on the leading/trailing parts of the curve, while maintaining a log shaped limit below the Vsignal input.   

always the apprentice, 

Glen

Glen,

when I interprete you right, you are discussing the limiting properties of a series connection Diode+resistor, correct?

As you know, this is a very popular method for linearizing the diodes exponential characterstics - applied for example as an amplitude-control mechanism in harmonic oscillators.

The same trick is applied to thermistors for designing various bias circuits...

So Glen, by  connecting various elements in series or/and parallel, we can made composite non-linear elements with desired characteristics?

Lutz

Thanks. Working alone, I had not thought about that. 

I am reminded of the WeinBridge oscillator.  

Your experience is far broader than mine. 

The point is that Diodes can be "linearized"

and used in a variety of limiting applications. 

The point of the "Question" 

is for students to check their thinking. 

I frequently meet radio techs who are having difficulties because they lack an appreciation for the real-world physics underlying our electronic experiences.  

always the apprentice,

Glen

Cyril,

Indeed !  

There is a flexible method to do this

"""composite non-linear elements""". 

Following rules will not produce inventive solutions. 

Certainly we must be aware to properly engineer 

the temperature-compensation and bypassing.  

In my published circuits,   I have Obviously skipped 

the important temp and bypass considerations .  

My criteria is not laboratory use, nor R.F. high frequency, 

but basically the ability of the ear/brain to read morse-code.

I always hope that some student will see

one of my incomplete ideas  

and be set off in a new direction. 

You might be amused to  learn

that I tacked one of Barrie's emails to me 

onto the wall of my shop,

as an honorable reminder that I am

always the apprentice,

Glen

Maybe this idea is further developed in diode functional converters with linear approximation?

Supplementing my previuous answer I like to mention that it is, of course, possible and beneficial not only to use a series connection "diode + resistor" but also a suitable series-parallel combination using two linearizing resistors.

This is, in particular, an important aspect for oscillator circuits (amplitude control) - with respect to the following general rule:

As less non-linearity as possible (for a high-quality signal) but - at the same time - as much non-linearity as necessary for a safe start of oscillations (to cover all uncertainties like parts tolerances, etc.).

Hence, we are required to make a trade-off. Nothing new for electronic design - such trade-offs are quite normal in engineering practices.

Yes Lutz,

compromises are necessary in engineering practice almost always - and I am afraid that in everyday life also :-))

Limiter in an oscillator - with low distortion - it's always an interesting problem.

Josef

Limiter in an oscillator - with low distortion - it's always an interesting problem.

Yes - I think, this problem would even deserve an extra thread.

Josef,

Compromise !  to fit the application. 

Diode Limiters are one tool I use in my process

and their application is part of a different topic.

As the subject is being changed from Diodes/Limiters 

...

On the subject of Compromise and Distortion , 

In my designs for audio-filters,

I introduce more distortion 

in order to better read morse-code messages. 

The narrow audio filters remove all harmonics, 

producing a sine shaped pulse of morse-code. 

A medium speed (12wpm) 'dit' may be 18mS long, 

and the sine slope may be 7mS long, up and down.

At a high speed ( 25wpm ) the only thing we have 

is the sine slope up and the sine slope down 

... very soft pulse with short length.  

What is required is a sharper rise/fall , more 'pulse', 

by the time that the Audio is driving an ear-phone. 

(1) 

I use a "hedge" pot on the front panel,

for the sharpest filter stage ( Fx04) , 

to allow moving ths final sharp filter up in frequency 

against the wider preceeding filter 

so that I can hear and measure more harmonic. 

ie,

f(Fx04) is moved within the passband of Fx03.

Harmonics for Fx04 can be measured

in the audio stage,   

and the pulse sound is 'crisper'

and more easily read by the ear/brain.  

This is then delivered to the audio amp. 

(2) 

In the audio amp ( Class B ) I removed the bias diodes 

so that there is a empty gap

between the upper/lower transistors cycle. 

This adds harmonics back into the audio. 

This sound is 'crisper' to the ear

and and easier to read by the ear/brain. 

ie, 

see "Audio-AmpT" jpg attached . 

This Compromising Design was the result

of years of experience listening to morse-code,  

at ever higher speeds, 10wpm, 20wpm, 30wpm,

and measuring the duration of 'dit' and 'dah' pulses

and observing the change in slope of the pulse.  

( see attached JPG for examples, [borrowed] )

ie, 

I designed sharp filters and had 'soft' pulses, 

then designed more distortion into the final circuits

so that the 'soft' pulses became 'sharper' pulses

which were more accurately read at higher speeds. 

Current subject is Diode-Limiter-Circuits.

Diode Limiters are just one tool I use

in the complex design process in this application. 

Glen

I've attached an LT Spice schematic showing the core circuit for a diode bridge limiter circuit.  The circuit clips signal levels to +/- I/R, where I is the current that is set up in the current sources, and R is the resistance to ground at the output.  The limiting voltages can be set to levels well above or below the forward bias voltage of the diodes - in fact the forward bias voltage of the diodes does not appear in the output signal, given matched diodes in the bridge.

of course, the current sources can be implemented by any conventional current source/sink circuit.  The currents in the source and sink need to be matched to avoid introducing an offset between input and output.

David.

David,

Thank you.  I look forward to reading if you have further discussion 

of alternative approaches to stimulate my thinking. 

Very nice, indeed, and neatly done.  Almost text-book accurate. 

Have not used a BAT54,

but am aware that diodes of many types 

respond to this biasing approach. 

I have been using PartSIM.com with the ngSPICE, for several years, 

and it does not accept the .asc files ( much to my regret ),

but I will key it in and run it through some variations.   

Attached is an experimental series I went through,

in which I did  negative-feed-back loop to the ground-side

of a pair of shunt-diode.  

In this, both diode curves show in the final signal,

and the small phase shifting also showed up. 

My approach is Not Linear ,

just tinkering to see if I could shape the limit-curve-shape

to something useful.   The much larger circuit ( linked below #3 )

resolved the minor issues I was chasing. 

So, I hope there other ideas forthcoming, 

by placing this "Question" up for discussion. 

http://www.geocities.ws/glene77is/bmp_et_limiter/AFX_Limiter-Shunt2-S-140821-b.png

http://www.geocities.ws/glene77is/articles_ET_files/article_ET_AFX_CW_Limiter.html

http://www.geocities.ws/glene77is/articles_ET_files/article_ET_AFX-3RLFADI.html

By the way, we can see the same powerful idea ("passive diode compensation") in the biasing circuit of a complementary emitter follower (e.g., at the non-inverting input of a current-feedback amplifier)...

Cyril,

Thank you for your input. 

Clarify this please.

Glen, I mean the terminal 3 of the circuit above where the two diodes are connected. The other two diodes are the base-emitter junctions of Q5 and Q7.

Cyril

below is a more precise description of CFA.

Josef

Cyril, Josef,

There is indeed a similarity in the way that the Vbe's of the four junctions between the non-inverting and inverting inputs of a CFA and the Vf's for the four diodes in the diode bridge limiter cancel out ... an interesting observation, though I fail to see any utility in this when it comes to developing signal limiting circuits.

Glen,

Regarding the choice of diode: the BAT54 is a schottky diode (free of reverse recovery effects), which allows you to build limiters that operate very cleanly on high-frequency signals.  Diodes with even smaller junctions might be used, and with an appropriate choice of parts, the limiter can operate with signals well into the RF spectrum.  The applications you have mentioned are in the audio spectrum, and I would expect conventional signal diodes (1N4148 or similar) to work just as well.

If I was implementing this circuit using currently available discrete parts, and I was concerned about diode-matching to minimise offset, I might investigate using the collector-base or base-emitter junctions of dual-transistor arrays with specified matching parameters between the transistors (e.g., BCM846BS from NXP semiconductors).  In the diode bridge limiter, you only really need to match the junctions as pairs - the upper two junctions and lower two junctions - it is not necessary to match all four junctions together.

The bridge limiter can be adapted to do things other than symetrical limiting.  I think the circuit is capable of asymetrical limiting and level-shifting too - I would have to play with making the current source / sink currents different, and tying the resistor at the output to a voltage bias other than zero to re-familiarise myself with the possibilities of this elegant circuit.

I haven't looked in detail at the circuits you've posted yet - will do when I find more time.

Regards,

David

Cyril and Josef,

Cyril's diagram is clearer, although simplified, student level.  

Thanks. 

Josef's diagram is more realistic, but diode Q9 and diode Q1 

are lost to me.  

In both cases, I can trace the functions of the Class-B Final Amp

although both appear to have two different types of bias control. 

Everytime I study the answers  I learn something more.  

the apprentice Glen

 David,

Thank you for your expansion on the topic.

You are using constant currents for diode bias.

I have used a Negative Feedback Loop. 

I looked up the Schottky BAT54,  double Diode.

I have been using Germanium, not Silicon. 

My project is titled "AFX", and is a very narrow

audio filter with phase sensitive circuits.

In the AFX Limiter circuit ,

I used Germanium, not silicon, by choice.

I did experiments with Schottky.

At the time I ran the Limiter through,

the ngSPICE did not afford a Germanium model,

so I used the 1N4004 for the Spice work.

Having compared the three types ,

in the same circuit,

and having normalized the bodes,

it appeared that the negative feedback loop

was the major control

over the Vout "roll-over-the-log" effect.

By adjusting the amount of 'resistance' I added

to the diode-dynamic-impedance, I could adjust

the level of the upper cut-off,

and make the diode circuits limit similarly,

and maintain a soft roll-over-the-log.

Actual measurements were

that Schottky and Germanium

had a 'softer' roll-over than silicon.

The "Softer" roll-over , as I understand it,

will cause less Intermodulation Distortion

to be generated, and this IMD would be

passed on, into the audio filters

as supra/sub harmonics .

About the IMD,

I preceeded the Limiter section with Roofing filters 

to produce Symetrical waveforms (in place of All-Pass)

and thus reduce potential IMD. References on Website.

I mention the IMD because it loads into the main Quad Filter

section, and raises the overall noise-floor.

The narrow filtering stages will integrate this

out-of-passband engergy into the base-noise level.

I observed and did some measuring on this effect

of Hi-Q sideband engergy being pushed into the base-band-noise.

An associate of mine (on another project) recorded observations

the same and emailed to me.

The AFX Limiter works usefully from about I(forward) 10uA to 10mA

in this application.

As you suggested, matching the "pairs" is important,

and these were close by 5%, (0.23V vs 0.22V) to minimize offsets.

Offsets at this stage would produce Asymetricl waveforms

which I am trying specifically to minimize due to the potential of IMD.

I like the idea of two matched diodes on the same chip,

and will explore the idea. Components I use must be large enough

for an old man to 'see and touch'.

I will be researching dual-BJT on the same chip ,

an 8-pin-DIP would do well to make two matched diodes.

David, you do well expanding an idea to the next level,

without over-loading the idea.

As Barrie has written to me, I put too much on paper,

and I am the apprentice,

Glen

q9, q10, q11 in Josef's diagram performs the same function as q5, q6 in Cyril's diagram - both are current mirrors.  The difference is, the current mirror in Josef's diagram is a Wilson mirror, an elaboration of the simple mirror featured in Cyril's diagram.  (Barrie Gilbert tells the story about the invention of the Wilson mirror in his chapter about current mirrors in the book 'Analogue IC Design: the current-mode approach'.)

Hope this helps ...

David

Cyril, 

Such good observations about

"diode functional converters with linear approximation" ,

there is more territory to roam in my thinking. 

Far enough down the curve of diode current 

everything appears to be very linear.  

Early on, my mentor Dr. Joe Laughter advised me

to get off the top of the log-curve 

and slide down towards the bottom

where the results will be more linear

( as he said "linear enough to be within 0.001%) . 

Lutz, 

"diode + resistor"  ...

"but also a suitable series-parallel combination using two linearizing resistors" , yes indeed.  

I need to try Limiting that in Spice. 

And it certainly could be used in "oscillator circuits" 

like replacing the old light-bulb in a wein-bridge oscillator. 

David, 

Yes, I traced this as you suggested. 

Most relevant web page is "Limiter" below, 

skip down towards bottom of that web page. 

So, 

I wonder where this 'idea' thread can lead? 

(1) Constant current diode bias.

(2) Negative-Feed-Back voltage diode bias. 

(3) Altering series resistance to alter diode bias.

(4) NFB to alter bias to shunt diodes. 

I have tried to "engineer" a project by calculating 

but have more success by Spicing concepts

as new ones present themselves to me.  

Is it true that innovative / "innoventive"  ideas 

present themselves to your mind,  in a moment of time ?

...  as if they have a life of their own. 

To share, 

Barrie has written about an intuitive approach to circuit analysis 

in which he describes "putting on the robes of the electron" 

walking through the circuit

to get the feel of what could conceptually happen. 

I keep his first email to me, tacked on the wall of my shop, 

as a reminder that I am 

just the apprentice, 

Glen

http://www.geocities.ws/glene77is/

http://www.geocities.ws/glene77is/articles_ET_files/article_ET_AFX_CW_Limiter.html

Josef's diagram is more realistic, but diode Q9 and diode Q1

are lost to me.

B-E of the first input transistors