Scope Probe Enhancing Performance of Circuit

[Robert Baer]

I have a circuit that acts stangely enough when I connect a ground
connection of a scope probe to it.

The circuit has a TI DSP 280x Series Processor, an Isolated Switching
Power Supply All the AD inputs and GPIO are isolated to the AC Line. In
my evaluation of the performance of the circuit I found that the
systems is able to perform accordingly when the scope probe ground is
connected to the circuit. I don't even need to have the probe connected
to the circuit, just the ground. The probe must be connected to the
scope in order for this to happen. If the this connection is not made
the device sometimes delays to react to the input accordingly. I'm
using a portable scope (THS720 Tektronix with isolated ground
connections). The behavior also happens with benchtop scopes that are
not isolated.

I have all my unsuded GPIO set as inputs and with internal pullups
enabled to avoid any parasitic effects.

This is a multi-layered board with Separated Power Planes (3.3V and
1.8V) and a single ground plane.

Any suggestions on where to start?
Thanks for all your help on this in advance.

gt5513e

Look for RF: external sources like radio or TV transmitters, as well
as generators in the circuitry (nevermind they are not supposed to exist).

 

[Robert Baer]

I had something like this happen once. We had to ship scope probes with
the product.

[snip]

I had a circuit that would only work when my middle finger was
touching a particular point.

So I bridged my finger and added that much capacitance ;-)

...Jim Thompson

....and then the circuit gave you the middle finger?

 

[Phil Hobbs]

I've seen PLL designs that would work better with the application of a
finger. From memory, one would reduce its phase noise by about 10dB
when touched. It turned out that the leakage current would cause a
steady state phase shift that moved the phase detector away from its
dead zone.

I hate the 4046.

There are lots of situations in instrument design where you want to avoid
some singular point, such as exactly zero phase error or exactly 100% duty
cycle. In my business, for instance, making a truly dark-field optical
measurement will lose you many decibels of S/N compared to a *nearly*
dark-field measurement.

The dead band on a 4046 is one of those minor warts that make design more
interesting...like the weak pulldowns on some bipolar single-supply op amp
outputs. It's easily fixed with one resistor, as you point out, and once
that's done, you still have all the other virtues of the 4046, such as that
very nice wide range VCO.

Metal-gate 4046s have about a 1000:1 range, but all the HC ones I've tried
crap out on the low end, limiting them to a range of at a bit more than 10:1.
Are there any quick 4046s nowadays that preserve the nice wide range?

Cheers,

Phil Hobbs

 

[Allan Herriman]

There are lots of situations in instrument design where you want to avoid
some singular point, such as exactly zero phase error or exactly 100% duty
cycle. In my business, for instance, making a truly dark-field optical
measurement will lose you many decibels of S/N compared to a *nearly*
dark-field measurement.

The dead band on a 4046 is one of those minor warts that make design more
interesting...like the weak pulldowns on some bipolar single-supply op amp
outputs. It's easily fixed with one resistor, as you point out, and once
that's done, you still have all the other virtues of the 4046, such as that
very nice wide range VCO.

Metal-gate 4046s have about a 1000:1 range, but all the HC ones I've tried
crap out on the low end, limiting them to a range of at a bit more than 10:1.
Are there any quick 4046s nowadays that preserve the nice wide range?

I'm not sure about the VCO, but I use a '9046 in place of a '4046 for
those rare occasions when I need a stand-alone analog phase detector.
It has a charge pump (i.e. current source and sink) output, and avoids
most of the pitfalls of the 4046 PD2 design.

Unfortunately, it's only (readily) available in HCT, so I have to find
a +5V rail someone in amongst all low voltage stuff (most of my last
board was 1.2V, 1.5V and 2.5V). I used caps to couple 2.5V LVCMOS
outputs to the PD inputs on the '9046.

Regards,
Allan

 

[Jim Thompson]

I'm not sure about the VCO, but I use a '9046 in place of a '4046 for
those rare occasions when I need a stand-alone analog phase detector.
It has a charge pump (i.e. current source and sink) output, and avoids
most of the pitfalls of the 4046 PD2 design.

Unfortunately, it's only (readily) available in HCT, so I have to find
a +5V rail someone in amongst all low voltage stuff (most of my last
board was 1.2V, 1.5V and 2.5V). I used caps to couple 2.5V LVCMOS
outputs to the PD inputs on the '9046.

Regards,
Allan

If it's a PFD you want you can roll your own... 1 pkg dual-D, 1 pkg
quad-2-IN-NAND

...Jim Thompson

 

[Tim Williams]

If it's a PFD you want you can roll your own... 1 pkg dual-D, 1 pkg
quad-2-IN-NAND

I made a phase detector with a single 4001.. not the fastest nor the widest
range (maybe 5-355 degrees) but it works for me.

Tim

 

[Jim Thompson]

I made a phase detector with a single 4001.. not the fastest nor the widest
range (maybe 5-355 degrees) but it works for me.

Tim

A PFD is an edge-matcher.

...Jim Thompson

 

[Fred Bloggs]

I have a circuit that acts stangely enough when I connect a ground
connection of a scope probe to it.

The circuit has a TI DSP 280x Series Processor, an Isolated Switching
Power Supply All the AD inputs and GPIO are isolated to the AC Line. In
my evaluation of the performance of the circuit I found that the
systems is able to perform accordingly when the scope probe ground is
connected to the circuit. I don't even need to have the probe connected
to the circuit, just the ground. The probe must be connected to the
scope in order for this to happen.

Your board construction is crap and when you alligator clip the probe
GND lead, you are joining the otherwise open connection between top and
bottom GND planes.

 

[Allan Herriman]

If it's a PFD you want you can roll your own... 1 pkg dual-D, 1 pkg
quad-2-IN-NAND

My "2.5V LVCMOS" source is actually a large FPGA, so I could put the
PFD logic in that, but I still need the external charge pump (or at
least external CMOS buffers (to avoid the noisy Voh and Vol of the
FPGA) followed by an opamp integrator with differential inputs).

The '9046 works and gives me the minimum IC count, so the design is
already optimal in some sense.

It gave 8ps RMS jitter, from memory. (That was three of them in
series. I had difficultly measuring just one.)

Regards,
Allan

 

[Paul Hovnanian P.E.]

I had something like this happen once. We had to ship scope probes with
the product.

[snip]

I had a circuit that would only work when my middle finger was
touching a particular point.

So I bridged my finger and added that much capacitance ;-)

Sometimes its more than just capacitance. I have a portable TV set that
receives perfectly in a marginal reception area if I touch the chassis.
I wish I had a good RF equivalent circuit for the human body.

 

[James T. White]

I have a circuit that acts stangely enough when I connect a ground
connection of a scope probe to it.

The circuit has a TI DSP 280x Series Processor, an Isolated Switching
Power Supply All the AD inputs and GPIO are isolated to the AC Line.
In my evaluation of the performance of the circuit I found that the
systems is able to perform accordingly when the scope probe ground is
connected to the circuit. I don't even need to have the probe
connected to the circuit, just the ground. The probe must be
connected to the scope in order for this to happen. If the this
connection is not made the device sometimes delays to react to the
input accordingly. I'm using a portable scope (THS720 Tektronix with
isolated ground connections). The behavior also happens with benchtop
scopes that are not isolated.

I have all my unsuded GPIO set as inputs and with internal pullups
enabled to avoid any parasitic effects.

This is a multi-layered board with Separated Power Planes (3.3V and
1.8V) and a single ground plane.

Any suggestions on where to start?
Thanks for all your help on this in advance.

gt5513e

Do you have a capacitor between your isolated ground plane and your
supply input ground? A cap in the 1nF - 0.01uf range is often connected
between the isolated ground output of a switching supply and the ground
of the input side.

 

[Joseph2k]

A PFD is an edge-matcher.

...Jim Thompson

Ah ha, i get it, clock distribution systems.