Oscilloscope input circuit

I am trying to design an oscilloscope/BNC input circuit for a side project and to learn more about filters.

The requirements are:

• 150-200 MHz -3dB bandwidth
• 1 megohm impedance
• 50 ohm termination option (I can omit this if it causes problems.)
• Can be switched between 2x and 20x attenuation
• Measures up to 80Vpk (160Vp-p), tolerates up to 400Vpk (800Vp-p) (5Vpk on 50 ohm mode.)

I designed a neat circuit that allowed switching between 2x and 20x attenuation while keeping 1 Mohm (or thereabouts) attenuation, as below:

                499k        442k
( Input ) -----/\/\/---+---/\/\/---+----- To JFET input buffer
|           |
/           /
499k \           \ 49.9k
/           /
|           |
+---+   +---+
|   |
\       relay to switch
|     attenuation 2x/20x
---
-


(All resistors 1%.)

However, then I learned oscilloscopes have input capacitance, and I'm not sure how to add it. Also, if I want the input to tolerate 400Vp-p swing but on the wrong input range (2x attenuation leaving 200Vp-p to damage the inputs of the JFETs), I couldn't use diodes as the lowest capacitance I found was 0.13pF for an RF diode, which would limit me to about 612kHz for two diodes (one to each rail, reverse biased.)

-

Input capacitance is significant for capturing high frequency content signals. Given that you are using discrete parts and not considering RF parasitics, this system will be restricted to the 100's of kHz for 1% accuracy.

The goal is to maintain constant attenuation and input impedance over the entire frequency range. Assume Cp is 100pF, Rp is 941kΩ, Cs is 1000pF and Rs is 49kΩ (ignore adjustment cap for now) The capacitors' impedance will vary as follows:

• ZCp : 1.6GΩ @ 1Hz; 1.6MΩ @ 1kHz; 16kΩ @ 100kHz
• ZCadj. : 16GΩ @ 1Hz; 16MΩ @ 1kHz; 159kΩ @ 100kHz
• ZCs : 159GΩ @ 1Hz; 159kΩ @ 1kHz; 1.6kΩ @ 100kHz

This results in the following range of impedances:

• Zprobe : 940kΩ @ 1Hz; 591kΩ @ 1kHz; 15kΩ @ 100kHz
• Z'scope : 49kΩ @ 1Hz; 37kΩ @ 1kHz; 1.5kΩ @ 100kHz
• Zinput : 989kΩ @ 1Hz; 628kΩ @ 1kHz; 16.5kΩ @ 100kHz
• approximate attenuation: 20X @ 1Hz; 17X @ 1kHz; 11X @ 100kHz

As you can see, adjustment capacitors are required to tune the input. The grander issue is that this calls for wideband matching networks. Another option, and what is done in addition to matching networks in professional equipment, is characterizing the input impedance and compensating for irregularities in software. Standardizing at 50Ω impedances allows the construction of modular probing hardware while maintaining wideband matching and constant attenuation.

Your requirements for switching impedances and an optional 50Ω input will need some more complication. A fantastic project, and I wish I could partake!

-
I'm sorry - I must be missing something here. I thought the impedance on a scope was pretty much flat at 1 megohm to the rated BW, but you're saying it's not? How do 10X probes work then? I guess I need to do some more testing. Oscilloscopes are far more complicated than I expected - I should know by now! –  Thomas O Jan 9 '11 at 3:42
Yeah, 'scope impedances are fairly flat, and have less capacitance. I just put in values that would be typical using discrete parts, large'ish SMD packages (SOT, 0805), and a few cm of traces. Cs would be ~20-100pF without external caps, but would make numbers even worse. Just do the same calculations over again, or use SPICE sweeps (even MATLAB as the circuit is simple), with different cap values until you get about the same impedances (and attenuation) at all frequencies. Broadband matching is kind of difficult! May edit in more a bit later. –  tyblu Jan 9 '11 at 7:21
Thomas O, please look again at my answer at electronics.stackexchange.com/questions/8712/…. I've put more details there since my first reply. –  zebonaut Jan 9 '11 at 11:05

For attenuation, you may consider looking at what a probe does. The factor is the same (1/10 = 2/20). Just short the additional resistor and capacitor from the probe circuit with a switch, like 1:1 and 1:10 switchable probes do.

-
I want it to be controlled electronically, and with minimum cost. Do you have a better idea- I could use a switch, but it might be more expensive than a DPST relay. –  Thomas O Jan 8 '11 at 21:04
Before looking too deeply into the idea of using a relay, you might want to check its parasitic properties. I have no experience with using relays at high frequencies, but as far as I've heard, RF relays are way expensive and pretty special. It must be possible to use relays, however, because that is what expensive scopes do when switching the input between 1 MOhm and 50 Ohm. CMOS switches might work, too. –  zebonaut Jan 14 '11 at 7:42