cmos inverters and clipping diodes problem

[deafbutpicky]

Hi guys,
I need some help understanding what is going on in diode clipped inverter stages.

I've messed with the clipping stages of a ROG 22/7 lately, trying different clipping diodes and symmetric vs asymmetric clipping,
and after ending up with some switchabel settings with what I thought was asym clipping, I was curious what it would look like
on a scope. I used a software oszi (Xoscope) and fed the pedal with a 1000Hz sine wave (180mV) testing the different stages with
an audio probe into a clean preamp into the soundcard (esi juli@).

The problem is:
if the gain is low the waveform shown is asymmetrical as expected but the more the gain is turned up, the more the signal rises
up too (wich it shouldn't above a certain level, as I understand clipping) and the more it gets symmetrical, wich shouldn't
happen either?! (The sound is awesome though, the pedal acts like it should sound wise and the Vfs measured are in the right directions
and voltage range) . So what is happening there? Why does the asymmetry gets lost with gain?

The two modes I tried where:
1st stage: yellow led + to - , oa1154+bat41 - to + (measured Vfs: 1,6V on the pos. signal swing / 0,7V for the neg. part)
2nd stage: no diode , oa1154+bat41 + to - (app. 2,4V for the neg. signal swing and again 0,7V now for the pos. side)

1st stage: 1n4148 + to - , oa1154+bat41 - to + (0,6V to 0,7V)
2nd stage: 1n34a - to + , oa1154+bat41 + to - (0,3V to 0,7V)

[deafbutpicky]

I added some pics here:

http://forum.musikding.de/cpg/thumbnails.php?album=661

and I made some more tests and found this:

-the OA+BAT combination seems to conduct much earlier than the 4148, despite the higher
Vf measured with an DMM, (low clipping threshold low gain stage 1 pic shows this). I expected
the positive swing to be a tad smaller than the negative, but the input signal has to be increased
about the double to make this true. I'm getting a feeling the gain in these stages is not what it's
supposed to be...

-inserting DC-blocking caps before the diodes (like in the Big Muff) in the fb-loop increases
that asym to sym problem. Now even the nice asymmetry in the low clipping lowgain stage 2 pics
gets lost completely

-after more diode swapping I also found that the Vf of the diodes directly effects the stages gain.
If a 1n34 is put btb with the led both signal swings go down massively.

-I guessed it might have to do with current draw from the ic, but changing to 12V doesn't help.


Maybe all these observations/questions lead to the same problem, but I can't get my head around it,
so it's just found more questions, yay

[deafbutpicky]

Happy new year everyone!

New year new luck, so I'll try another bump on this as I'm quite sure someone
more capable has messed with this topology before. I'm not exactly a layman in signal processing
but I'm no real EE either and seem to lack some fundamental knowledge here...

I'll try to keep it down to one question:
is there a possibility those inverters are not delivering enough current to let the diodes go into
conduction at higher gain levels in this setup?

Hints for this would be the rising amplitude with rising gain and the very round clipping curves...?!

[ashcat_lt]

I don't really know for sure that I've got the answers that you're looking for, but I'll throw a couple things in here.

1) There really isn't any such thing as an assymetrical square wave.  As the signal unclipped signal gets much larger than the clipping threshold, the difference between the top and bottom threshold becomes insignificant.  If the signal is trying to swing to +-5V, and you clip one side to 1.4V, that's clipping off 3.6V, 72% of the peak.  If you clip the other at 0.7V, you're losing 4.3V, or 86% of the swing.  If OTOH, the signal is trying to go to +-50V with the same clippers, you're loosing 97.2% of the swing one way and 98.6% the other way.  As gain increases, the assymetry of the waveform becomes less meaningful.

2) Sine waves a rounded to begin, so it's a little tough to extract the conduction curve of the diodes from the curviness of the original signal.  In my recent experiments with programming I've been using triangles as input, because then what I'm looking at is closer to the actual transfer curve.  Also - and especially in analog - rounded corners may indicate a bandwidth limitation.  Those sharp corners involve a lot of very high frequencies and will be affected by any low-passing or slew rate limitations in the circuit.

3) Assymetrical clipping essentially creates a DC offset.  This kind of goes back to #1.  Given that you've clipped it far enough to undo most of the assymetry, a square wave swinging 1.4V one way and 0.7V the other is really just a 2.1V P2P wave biased up to 0.7V above whatever you're calling 0.  But the DC coupling caps won't allow that to happen.  The diodes end of the cap will float up and the signal will tend to center itself between the thresholds on either side and what comes out is symmetrical.  I've got this problem in my Rat clone where I added a fancy switching arrangement that allows a few different assym settings...except they're not...and I haven't put as much time into fixing it as I probably should...

[Quackzed]

you could try one 1n34a on the first + to - stage and no - to + diode... then youll get some assymetry at the start , then use that yellow led
on the second stage - to + to keep the highest headroom on one side( - to +) and another 1n34a +to- in the second stage, this  will keep the second stage very assym till the yellow led starts to clip ,the yellow led should clip any harshness from first stage rail clipping in the second stage, and between the 2 second stage diodes maintain more assym till you start to square off.
but as mentioned in the previous post, at some point your gonna amplify and clip so much of the signal that assymetry is gonna start to shrink relative to the clipping amount, just because so much of the signal will be over the clipping thresholds both ways that the dc bias will be less significant ; the signal will be squared on both positive and negative side heavily and the dc shift wont really matter that much... like ashcat explained ( better than i could lol)...
this way will increase clipping on one half + to - . and lessen clipping on the other half -to +, for more assymetry before squaring. the clipping will affect smaller signals on one polarity of signal sooner(more distorted) and do so assymetrically longer before the second stage led kicks in on the other polarity - to + and clips it , to prevent hitting the rails... so more clipping on one half and less on the other... basically, before too much gain squares it all away.<bad pun 

[Quackzed]

  i just looked at the schem again, and thats alot of gain man. because theres so much amplification going on in each stage, the changes or unevenness of the positive / negative thresholds is likely to be noticeable only at the very lowest gain settings. you might or might not like the more distortion on one half/less on the other, though itll be minimal as this thing has ALOT of gain, and it wont change the high gain sound much at all... just the onset of distortion character, and that'll go  into 'DistorTION' / mostly symmetrical zone
pretty fast, maybee too fast to be very noticeable. you could raise those 33k resistors / or lower the 1M's a bit to get the gain down for more noticeable assym before dist, but it looks like this thing was designed for high gain, and too low and it might not be what your after... but hey, if it SOUNDS good, it IS good!

[deafbutpicky]

Thanks for your reply,

to 1)
I stubbornly expected to see the 50% asymmetry in the scope, thinking of the clipping thresholds
as fixed cutoff values for a non floating midpoint, like +1.4 to -0.7 at the output stays the same, regardless of what
amplitude the original signal tries to reach.

to 2)
I didn't expect bandwidth limitations to be a clipping problem, well up to the datasheet;)

to 3)

3) Assymetrical clipping essentially creates a DC offset...
...The diodes end of the cap will float up and the signal will tend to center itself between the thresholds on either side and what comes out is symmetrical.

So the signals asymmetry is no longer recognizeable at it's swings but it's different signal peak cutoff forms, right?
I'll have to give this some more thoughts,
thanks

[deafbutpicky]

i just looked at the schem again, and thats alot of gain man. because theres so much amplification going on in each stage, the changes or unevenness of the positive / negative thresholds is likely to be noticeable only at the very lowest gain settings. you might or might not like the more distortion on one half/less on the other, though itll be minimal as this thing has ALOT of gain, and it wont change the high gain sound much at all... just the onset of distortion character, and that'll go  into 'DistorTION' / mostly symmetrical zone
pretty fast, maybee too fast to be very noticeable. you could raise those 33k resistors / or lower the 1M's a bit to get the gain down for more noticeable assym before dist, but it looks like this thing was designed for high gain, and too low and it might not be what your after... but hey, if it SOUNDS good, it IS good!

Yeah, it really does sound different, and it sure sounds good;)
It's just that I expected to see something different. I looked at some
other pedals (Thor, Prof. tweed, 18, or my own Thrice) with various
gain settings and they clearly show an asymmetrical swing
(Or ass-symmetrical as I'll call it from now on )

[ashcat_lt]

So the signals asymmetry is no longer recognizeable at it's swings but it's different signal peak cutoff forms, right?
I'll have to give this some more thoughts,
thanks

If you're deep enough into clipping on both sides and you ignore (or strip out) the DC offset, the assymetry will show itself as a difference in duty cycle (is that the right term?) - one side will stay flat a bit longer than the other.  As the "attempted" signal swing gets much larger, that difference decreases.

[deafbutpicky]

duty cycle was the word I needed to understand this part