Limiter

[Tim Williams]

Nobody really makes a Magic-Joerg-Block where the output saturation is
adjustable (and preferably, accurate), so naturally, I made my own.
http://www.seventransistorlabs.com/Images/Limiter.png
Threshold between inputs is, of course, exponential (tanh more precisely,
I suppose). The comparator output crosses threshold over 10s of mV
difference, good enough for most purposes. Put an op-amp or comparator in
if you'd like it sharper?

I find 31mV offset with fresh-out-of-the-bag transistors (say, that's
about a V_TH). So unless you get lucky with a monolithic matched set,
you'll need the offset trimmer, or a fancy op-amp setup, to do better.

Tim

 

[Tim Williams]

The problem with bipolar limiters in general is that they slow down when
you overdrive them, so you get a fair amount of AM-PM conversion. If
your input is near ground, your diff pair will do its limiting by cutoff
rather than saturation, which helps a lot.

When I need a fast logic function, and TinyLogic is unsuitable, I favor
discrete CML or ECL style over RTL/DTL/TTL style saturating stuff.
Of course one of the nice things about CML/ECL is the analog nature.
Hence circuits like this...

IME FETs do a better job of limiting without a lot of AM-PM. Of course,
I'm a big fan of the SA604A FM IF chip because it produces this
amazingly useful logarithmic RSSI output for $2 or so, even though its
AM-PM performance is fairly mediocre.

I'm not doing radio here (as you might guess from the resistors, it'll go
to... oh, probably a few MHz, which isn't bad, but a long way from FM
BCB), so AM-PM isn't a problem.

To which kind of FET limiter are you referring? FETs will suck if used
directly in a circuit like this, of course. The cascade-of-amps is
time-tested, although if your FETs have a lot of Cds at low voltage,
you'll see a lot of sloppiness.

I may investigate a log amp or something like that one day. Perhaps build
the better part of a spectrum analyzer. Those damn things sure jacked up
in price the last couple years.

Tim

 

[Bill Sloman]

Use my LED current sources!

https://dl.dropbox.com/u/53724080/Circuits/Current_Sources/Isrc_LED.JPG

https://dl.dropbox.com/u/53724080/Circuits/Current_Sources/LED_Isrc_d...

https://dl.dropbox.com/u/53724080/Circuits/Current_Sources/LED_Isrc_b...

They glow in the dark, too.

It looks as if a lot of your +90ppm/C temperature drift is in the
temperature dependent current gain of the MMBTH81. You could get more
current gain out of a PNP/NPN complementary Darlington, but IIRR they
have a nasty tendency to oscillate. If you ran one with just 0.5mA in
the MMVTH81 it would be slower, but still fairly quick (if it didn't
oscillate).

The green LED version is unexpectedly horrible - did the MMVTH81 still
have enough Vce left in your test circuit?

 

[Nico Coesel]

Nobody really makes a Magic-Joerg-Block where the output saturation is
adjustable (and preferably, accurate), so naturally, I made my own.
http://www.seventransistorlabs.com/Images/Limiter.png
Threshold between inputs is, of course, exponential (tanh more precisely,
I suppose). The comparator output crosses threshold over 10s of mV
difference, good enough for most purposes. Put an op-amp or comparator in
if you'd like it sharper?

I find 31mV offset with fresh-out-of-the-bag transistors (say, that's
about a V_TH). So unless you get lucky with a monolithic matched set,
you'll need the offset trimmer, or a fancy op-amp setup, to do better.

What is the purpose? You'll just add a lot of distortion this way.

 

[Joerg]

Nobody really makes a Magic-Joerg-Block where the output saturation is
adjustable (and preferably, accurate), ...

Sure they do. For example in the audio world:

http://en.audiofanzine.com/guitar-combo-amplifier/crate/GT65/user_reviews/r.36076.html

Quote "Setting the tones, bass mid and treble, volume for clean and
crunch (+ saturation) of a single knob Tone for channel distortion with
adjustable saturation and volume".

In the ham radio world you have clippers. I made a transistor clipper
sometime in the late 70's, after I got my license.

[...]

 

[Gerhard Hoffmann]

Am 16.03.2013 16:04, schrieb John Larkin:

There are a couple of opamps that have explicit clamp inputs. HFA1130 (fast and
a tad quirky) and the slower but more accurate AD8036/8037.

Or TI OPA698/699. Work nicely here in a slope amplifier in
the Oliver Collins style.

On some OpAmps one can clamp the compensation pin. That limits the drive
to the output followers.

regards, Gerhard

 

[Jamie]

Nobody really makes a Magic-Joerg-Block where the output saturation is
adjustable (and preferably, accurate), so naturally, I made my own.
http://www.seventransistorlabs.com/Images/Limiter.png
Threshold between inputs is, of course, exponential (tanh more precisely,
I suppose). The comparator output crosses threshold over 10s of mV
difference, good enough for most purposes. Put an op-amp or comparator in
if you'd like it sharper?

I find 31mV offset with fresh-out-of-the-bag transistors (say, that's
about a V_TH). So unless you get lucky with a monolithic matched set,
you'll need the offset trimmer, or a fancy op-amp setup, to do better.

Tim

Lets see now. I just quickly picked up the calculator and it appears the
current sink for the high side NPN maybe over driving it a bit..

At a 50 ohm load I would say that is fine however, at a short circuit
output, it looks like it forcing the high side output (NPN) to sit
somewhere in the 200ma+ region.

Maybe you don't plan on output being low Z or shorted but I would be
concerned with the livelihood of the NPN. who knows..

Jamie

 

[Tim Williams]

There are a couple of opamps that have explicit clamp inputs. HFA1130
(fast and
a tad quirky) and the slower but more accurate AD8036/8037.

Ten bucks, you're kidding right?

Tim

 

[Bill Sloman]

The emitter resistor tempco spec is 100 PPM.

Why? 15ppm/C resistors are widely available, if not quite as cheaply
as 100ppm/C parts.

I think it's pretty good. A junction drop has a voltage change of about 4000
PPM/K.

Sure, but you've cancelled that out.

 

[Bill Sloman]

I hadn't seen Collins' work before. Interesting.

We recently designed something sort of similar, but not a ZCD. It's an amplifier
with a photodiode input and two outputs that have x1 and x1000 gain ratios. The
x1 is used to measure a main laser pulse, and the x1000 observes tiny pre- and
post-pulse optical effects, within +-10 ns of the big pulse. We wound up using a
number of cascaded, relatively low gain, clean-clamping amp stages, sort of like
the Collins idea.

This was hard to test. The PD is an ideal floating signal source, but real
signal generators and scopes have high frequency ground loops. With over 100 GHz
of GBW, it really wants to oscillate. We wound up kluging a small floating,
battery powered pulse generator to drive the input, fiberoptic coupled.

It might have been cheaper to decouple the power supplies of each of
the cascaded amplifier stages with a properly damped non-wound
inductor/resistor/capacitor networks. Most amplifiers have very little
common mode rejection at high frequencies, and an output stage with a
releatively high output current can feed quite a lot of signal back
into the input stage of the first amplifier in the chain through the
power supply pins if you aren't very careful.

Regular decoupling capacitors crap out about 300MHz, so you can end up
having to by-pass them with something less inductive. I didn;t have to
go to the trouble of buying some 1nF porcelain capacitors, but I
wasn't looking for your kind of GB product.

 

[Bill Sloman]

Precisely. Well, approximately. Plus, it glows in the dark.

Precisely enough that your crummy emitter resistor or the temperature
dependence of the PNP base current are enough to explain the residual
tempco.

If you'd made them a bit better, you might have a better idea of how
precisely the LED cancells the PNP's Vbe.
And it would still glow in the dark.

 

[Bill Sloman]

Well, do it better and tell us how things work out.

A little difficult. I'm in Christchurch, New Zealand, at the moment,
and my electronic gear - such as it is - is out of reach.

You were interested eniugh to measure the temperature stability of the
circuits you'd put together. Why weren't you interested enough to work
out where the temperature variation was coming from?

 

[Bill Sloman]

It's plain that this circuit will never be a really good current source, for
several reasons, so why bother doing a lot of math? Some of us don't needto do
pages of algebra or a Spice sim to see the obvious

It can't be perfect, but it could have been better than you achieved
with fairly minor modifications, none of which demand pages of algebra
or a Spice sim - though it's often quicker (if less reliable) to run a
Spice sim than it is to solder in the extra/differemt components.

https://dl.dropbox.com/u/53724080/Circuits/Current_Sources/Isrc_LED.JPG

Power supply rejection can be greatly improved, and I bet you can't tell us how,
without looking up my old posts.

The obvious technique is to make it a "ring of two" circuit, and use
the LED voltage to set up a current source to bias the LED at a more
nearly constant current. The traditional way would add three more
transistors - two of them in a current mirror.

<snipped the abuse>

 

[Bill Sloman]

The oscillation was caused by the external test equipment, as noted. The board
doesn't oscillate itself when driven from the floating photodiode.

But ground loops put noise on the power supply rails.

You called it oscillation, rather than electromagnetic interference,
so the implication is that the extra feedback paths through the supply
rails (including the ground connection) were creating the instability.