5. Sampling-bridge-driver

Experiments with different buffers.

DIP Socket are not good for high frequency, but ok for fast changing parts.
Drilled and filed. Two X7R capacitors soldered for power supply bypassing.


Top GND plane with a massive connection to the low impedance point.
Some paralled wires, filled with solder.

Power Supply with different types of ceramic capacitors soldered direct on the star-gnd.
Positive and negative supply by wires and polarity protection schottkys.
It's not my first time doing a mistake plug-in another IC.

Power comes with a old 75 ohms coaxial cable, series schottky, 4.7µF/25V capacitor X7R as cable buffer.
Small pieces of PCB as soldering terminal.

Output PIN 6 a small soldering terminal with a eyelet for the probe.
Power supply bypass with 4.7µF and in series 1.5 ohms with 22nF.
Capacitor grounds soldered to one point.

PIN2 non-inverting input coaxial connector.


used equipment in this experiment :
7A26 vertical amplifier
7904 mainframe
7B92A time-base
7M13 readout plug-in
1917A pulse generator

Plug-in amplifiers square wave response from pulse-generator, using a 1:10 probe.
Number 566 and 595 indicates plug-in serial number.
I choosed from my plug-ins two 7A26 channels showing almost the same response.

Input and output coaxial.
Input terminated with 50 ohms.
Probes conncected without alligator crimps.

Circuit powered by my transferable 10V reference, including some low noise power supplies with different output voltages.

Tested OP-amps

I've tried with the fast amplifiers  I have in my sample box at home.

This is a CFA Type xxxx and this type should work with a gain of Av = +1.
Readout indicates 1kohm  CFA feedback resistor Rf.  
1917A indicates input signal.
Datasheet specifies for Av=+1 and Rf=750 ohms.
May be the sample don't work. Possible, I've used this sample many times in the past.

This CFA Type xxxx don't work with Av = +1 gain.
Datasheet do not recommend this amp as non-inverting, large signal unity gain buffer.

Here a fast VFA Type xxxx with N-jFET input.
Don't work good with Av = +1 gain and a fast slewing input.
Using the 1k Rf for a VFA was a mistake.

Here a CFA Type xxxx, don't work with Av = +1 gain. Datasheet do not recommend non-inverting unity gain with high amplitudes.
Non-inverting input should use a 100 ohms in series (I forgot).

In general many CFA amplifiers work better with  a inverting configuration with a gain > -1.
Unfortunately a inverting configuration would load the settle-node.

Here a CFA Type xxxx with very fast slewing.
But I don't like the distorted edge on the rising slope.
May be some experiements necessary with another Rf and voltage steps.
Datasheet notes faster settling for inverting gain configuration, a possible settling problem for many CFA's?

Type xxxx, very good DC specs, low input bias current won't load settle node much, specified settling time.
Rg=0 ohm


Rising voltage curve of the Type xxxx shows almost a straight line for small and large signal,
may be this is a key for a good settling time?, also the fast settling Type xxxx comes with a very straight rising curve.
Some other faster slewing Op amps with exponential rising curve character having slower settling times!
These are things I want to discover when the settling time instrument works.

Here a CFA Type xxxx specified for Av = +1, works good.
Faster AC Specs than Type xxxx, not so good DC-Spec and higher input bias current.
Responses nice.

Type xxxx vs. probes and alligator crimps.

I used the Type xxxx for some experiments with different probe connections.

Input 50 ohms terminated, porbes using no alligator crimps, output probe grounded via a self-made ground spring.

Alligator crimps are bad with high frequency signals.

Using only one alligator crimp for two probe makes the disaster perfect.

Layout Improvements

Parasitic capacitance on the  input's is bad.
Especially on the inverting input a small parasitic capacitance forms small pole at high frequencies,
resulting in higher overshooting and amplitude response.
Removing copper around the inputs with a small DIY milling-machine.

Replacing the Dale THT resistor (old military stuff) with a lower inductance SMD  (102) 1000 ohms resistor.
The wide copper stripe from PIN6 to the resistor decrease also inductance.
The resistor is placed close to the inverting input, the low impedance amplifier's output can easily drive the distance to the resistor.
It's always a good idea place the resistors very close to inputs and leave the low impedance signals larger.

Removing the 15cm cable (used before) from the input signal. Direct connecting to the pulsegenerator's output.

Replacing the 50 ohms terminator resistance by two parallel connected 100 ohms resistors.
In general 100 ohms resistors keep their resistive behaviour over a wide range of frequencies.
Some commercial 50 ohms terminators are made of bad quality, e.g. long wires inside - I can't open mine.
Now the transient 50 ohmcurrent flows a little closer to the clean analog circuit ground
- it can be dangerous -
fast transient current spread wide inside the ground plane.
There are also some notes in the AN-74 about this problem.

Input signal got a small piece of cooper reducing inductance.

Comparison of Layout improvements

Photo A (left) like before with 15cm cable. Photo B (right), inputs with reduced parasitic capacitance.
Lower feedback inductance and without cable direct on pulse-generator,
decreased ringing. Removing the cable showed the most influence.

Photo C (left) without cable and with selfmade 50 ohms termination better than photo B, shows less ringing.
Photo D (right) with small copper on the input - no change in performance.

CFA vs. different Rf

The ideal CFA has a bandwidth independend from closed loop gain over a wide range of loop gains.
Bandwidth changes also with Rf.

Experiments with decreasing Rf:

Rf = 1000 ohms                                                                                                               Rf = 750 ohms

Rf = 470 ohms                                                                                                            Rf = 300 ohms

Decreasing Rf increases bandwidth and slew rate but but also overshooting and ringing.

Experiments with increasing Rf:

Rf = 1200 ohms                                                                                                           Rf = 1350 ohms

Rf = 1600 ohms

I choosed Rf = 1350 ohms for further experiments:

7 volt step                                                                                                                    10 volt step

2.5V step                                                                                                                  600mV step

250mV step                                                                                                                   250mV step

100mV step

Rf=1350 ohms is a good choice for starting, seems a very little overdamped with 1k35, no problem.

Type xxxx vs. capacitive load

Ceramic capacitors  close on the output.
The 1350 ohms are paralled by two 2700 ohms.

1nF   200mV step                                                                                                           10nF    200mV step

10nF  6V step                                                                                                                10nF 10V step

10nF 10V step

I don't know how these amplifiers work as sampling-bridge-driver, we will see. Makes fun to evaluate all this things.
There are many question on my list to discover, this settling time test equipment will be a very powerfull tool to see all these hidden things.


Impressum und Haftungsausschluss