[sdiy] Oscillator jitter measurements

[Fredrik Carlqvist]

Martin,

Thank you for the answer!

The 1nF cap between + and - inputs was suggested in the LM311 data sheet. It
improved the response somewhat, removing some oscillation. But I agree, it
does not seem like a good idea in combination with the 2M2 resistor. 

Are you also suggesting adding clipping diodes to protect the inputs?
Perhaps that is a good idea, now that you mention it ;-). Does the LM311
have protective diodes on the input?

Exactly what would a small capacitance on the feedback resistor do, and how
large should it be? 10pF?

I do get a clean rising edge on the scope, and I can make repeatable noise
free measurements of reference oscillators that stay within +/-100ns, so
with scope aided trimming and sharp edges the circuit is good enough. 

I will remove the cap between + and - and separate the two grounds, it seems
like a good idea. Also, I have changed the layout so that the signal goes
into the negative input instead of the positive. It feels better to have the
positive feedback on the reference rather than on the signal. 

The power supply is a 9V battery with a 10uF capacitor and a 7805 regulator
and one 100nF at the output. There are also several 100nF decoupling
capacitors that are not shown in the schematic, one for the LM311 and two
for the microcontroller. So I don't think the power is an issue. 

The main difficulty is that it needs scope trimming to be reliable. Perhaps
thresholding the derivate is not a good model? I want to measure the time
between the steep flanks, it seems like the most reliable method. My idea
was to only have the polarity switch and a fixed reference voltage, but that
has proven overly optimistic. 

It is often fairly obvious when the measurements fail and when they succeed,
looking at the distributions. I will hopefully do some more measurements on
Wednesday, on a Moog Source and a Sequential Pro-One among others. I have
some preliminary results on the 4069VCO and CEM3340, which are very
interesting, but the presentation will have to wait until I am sure of the
method.


Fredrik C

-----Original Message-----
From: Czech Martin [mailto:Martin.Czech at Micronas.com] 
Sent: Monday, June 21, 2004 10:44
To: Fredrik Carlqvist; synth-diy at dropmix.xs4all.nl
Subject: RE: [sdiy] Oscillator jitter measurements

I assume there are two design problems in this circuit.

1.) the designer of the 311 took great care to seperate the output
open collector current from the differential input stage.
Otherwise pin 4 could
be connected to pin 1 internally. He did not do this, and you should
not attempt to do it, either.
The 311 app notes will tell you
about that. After all, a comparator is a very high gain circuit
without any feedback for gain reduction, thus very susceptible
to oscillations that would never be seen with a normal op amp.
At the same time it produces fast transients, especially with
the output stage, an "interesting" mixture.
A good start would be a good groundplane board.
Or copper foil. You need this also for the reference point
of the input stage supply decoupling capacitor.

Althoug it sounds simple, it is not so simple. Perhaps some
experiments have to be made in order to set up things properly.
This is because of the sometimes unexpected parasitics
of your components and the board. Even the way the PSU and
cables are made could have an influence, so I'm sorry that I can not
give you a 100% solution.

The 20MHz bandwidth limit of the scope can be an obstacle
when observing glitches, and those can turn your measurement
into rubbish. I think that during such a measurement a 50 ohm
scope output for monitoring and adjustment is a must.

There is nothing worse then a computer aided measurment which
runs merily with bogus signals, producing a lot of bogus
data (which is perhaps not so obvious in the first days of
thinking, wondering and analysis).

2.) What is the 1nF between + and - doing?
You have added a 2M2 for hysteresis, but you slow it down
with this large capacitive load. Delete this cap.
Better give the 2M2 a few pF in parallel in order to 
overcome the + input parasitics. The faster the input, the less
time in linear mode where all this nasty oscillating things can happen.
If you want to reduce the bandwidth of the input (why?)
you should do it at the input.
And watch out for the maximum differential mode input specs.

The basic problem is that you and me are used to the slow, closed loop
behaviour of a 741, 071 or similar. I.e. bandwidth about 3MHz or lower. 
But comparators are different. They need to be very fast
and have in some cases 10x bandwidth or more. If we use our slow linear
op amp habits with comparators chips, we are in trouble.

m.c.

-----Original Message-----
From: owner-synth-diy at dropmix.xs4all.nl
[mailto:owner-synth-diy at dropmix.xs4all.nl]On Behalf Of Fredrik Carlqvist
Sent: Freitag, 18. Juni 2004 13:26
To: synth-diy at dropmix.xs4all.nl
Subject: [sdiy] Oscillator jitter measurements


Hi all!


I have constructed a jitter measurement device (called the jitterbug,
actually ;-) with a resolution down to 100ns. I have done some initial
measurements on the synths I have at home. No conclusive results yet. But
the jitterbug works very well. My plan is to measure all different VCO
topologies I can get my hands on and present the results on a web page with
statistical analysis, distribution diagrams and everything. 
 
However, I need your help with the trigger circuit. It uses a LM311. There
is an image of the current design here:
 
http://www.minod.com/jitter/schemo.gif
 
The problem is that it has to be trimmed in with an oscilloscope before each
measurement as to remove oscillation and false triggers. My scope goes up to
20MHz and has no memory, so it is very difficult to see when the trig pulse
is ok. It is not impossible though. 
 
I want the trigger circuit to be useable without using a scope and without a
very pure input waveform if possible. Right now I have to hack into the VCO
core itself, since the slopes at the audio output of the synth often is not
very sharp. Most synths cannot open the filter enough it seems. 
 
So, for the benefit of science and our future VCO constructions please help
me with the trigger circuit!
 
 
Fredrik C