All too often signal integrity aberrations are visible in digital signals displayed on an oscilloscope. Fortunately signal fidelity can be significantly improved by a simple homebrew 3D-printed accessory that can be retrofitted to any standard scope probe. The “workhorse” scope probe usually supplied with scopes is the high-impedance 10:1 passive scope probe. While very useful in many circumstances, like all scope probes it has its limitations and can introduce distortions into the displayed waveform. This post complements the scope probe reference material and simply:
- concentrates on one problem that is frequently encountered in medium speed digital circuits
- illustrates commercial accessories used by high-end passive probes to avoid the problem
- describes a simple homebrew 3D-printed accessory
- demonstrates improved signal fidelity
The Problem and Fidelity Improvement
The examples below use an HP10074C scope probe (150MHz, risetime <2.33ns, 15pF, 6inch/15cm ground lead) and Tektronix TDS340 100MHz digital scope.
With the unmodified ground lead, a fast (<1ns) digital edge incorrectly appears to “ring” with a half-period of 5.5ns (91MHz), 20% overshoot, 33% peak-peak amplitude, and 10s of nanoseconds duration. With the homebrew spear ground the ringing is removed, revealing the undistorted waveform.
There is a little ringing at 270MHz, 0.5% overshoot, 4% peak-peak amplitude and <5ns duration. Some pre-transition ringing is visible (even using a 1.5GHz HP10020A probe), but it is not visible on an equivalent analogue scope. Hence that ringing is probably an artefact of the scope itself.
The updated results with a higher frequency oscilloscope and higher frequency probe are even more impressive.
Theory and Commercial Solutions
The cause of the ringing is the interaction between the probe tip’s capacitance and the ground lead’s inductance. An LC circuit will ring at its resonant frequency of . Wires have an inductance of ~0.8nH/mm, so the normal 6 inch (15cm) ground lead and croc clip’s inductance is around 150nH. The probe’s tip capacitance is 15pF. Hence the theoretical resonant frequency is ~106MHz, which corresponds acceptably with the observed value (91MHz). To increase signal fidelity, the objective is to increase above the probe’s or scope’s bandwidth, so that the ringing won’t distort the trace so much. This can be done by reducing the capacitance or inductance.
The only practical way to reduce the capacitance is to use a different probe. There are some high-quality high impedance probes with slightly lower capacitance, e.g. the 6.5pF HP10431A, but the usual technique is to use an expensive, fragile active probe, or a low impedance Z0 probe. Now there are many advantages to Z0 probes, but it is probably easier to make a homebrew Z0 probe than it is to buy one commercially; see scope probe references for examples.
Fortunately there are three straightforward methods for reducing the inductance, which are applicable to any high impedance 10:1 probe.
The first technique is to use the small wire spring clips supplied with the probe (if they haven’t been lost!), but they need a ground connection at the exactly the right distance. That is often possible for analogue circuits without a soldermask, but is often difficult with digital circuits.
The second technique is to solder a testpoint into the circuit, either a commercial product or more likely homebrew wire coils or paper clips. This is justifiably a favoured technique when breadboarding analogue circuits, but is more problematic with densely populated digital circuits
The third technique is supplied with some high-end probes: a short pivoting spear or blade which allows a variable distance between the probe tip and the ground. Examples are HP10431A probes (6.5pF ~500MHz), HP10020A probes (<0.7pF, 1.5GHz), Keysight N2878A probe accessory (2pF, 1.5GHz). This note describes a homebrew version of that technique.
Homebrew 3D-Printed Ground Accessory
The 3D-printed probe ground accessory shown below was inspired by the HP10020A. Although specifically designed for an HP10074 probe, it also fits similar available probes, e.g. an Elditest GE1521.
There are four parts to the accessory:
- the body, designed in OpenSCAD, fabricated in three different materials described below
- the spear, made from 0.8mm piano wire
- the ground connection, etched from 0.2mm phosphor bronze, with the fingers cut using a jeweller’s fretsaw
- the hinge, a 1.7mm/1.5mm spring pin in a 1.6mm hole
The phosphor bronze ground connection fits into the slot and protrudes into the central cavity. It is then deformed so that when the probe is inserted it rubs against the probe’s ground sleeve. Finally the piano wire spear is inserted into the slot against the phosphor bronze, and held in place by the spring pin.
Three versions of the body have been fabricated:
- the nylon version described in this post, fabbed by Shapeways in their strong and flexible plastic nylon material (£13.50 for two including postage)
- a duplicate of the nylon version, but fabricated by Dangerous Prototypes using their SLA 3D white photopolymer process (£3.50 for two, including postage)
- an earlier PLA version made on a local RepRap 3D printer
The experience gained with the PLA version was a useful proof-of-concept, but the professionally produced nylon body is significantly better:
- the Shapeways sintering process means the dimensions are much more accurate; no post-manufacture filing was necessary
- the nylon does not scratch the probe body
- there is a much more satisfactory grip between the body and the probe. The body contacts the probe along the length of the four internal nubs. The distance between the nubs is 0.1mm smaller than the probe body’s diameter. The natural elasticity of the nylon allows the body to deform slightly when the probe is inserted, so that the interference fit is not too tight.
The photopolymer version was made simply to test the new material. Compared with the nylon body, the photopolymer is smoother, marginally less flexible, and might or might not scratch the probe body. It is much cheaper than the nylon version, and is probably just as functional.
Thanks to Russell Dicken and Ian Stratford for help with the RepRap, and Bristol Hackspace for the use of their RepRap.