Common Analog Signal Frequencies

I'm trying to scope out what frequencies different common analog signals are. The DPScope design specs provides a list of common lower (sub 1 MHz) frequencies (copied below).

• audio (20 kHz)
• infrared remote control signals (38 kHz)
• ultrasound (200 kHz)
• servo signals (a few kHz)
• bio signals, medical instruments (< 100 Hz)
• I2C (1 MHz)
• RS-232 (115 kHz)
• one-wire
• SPI (as long as <= 1 MHz)

However, some of those listed are used for transferring digital data (for example, RS-232).

Now my questions:

Under what situations would it be necessary to have a bandwidth capable of measuring these digital lines as analog signals?

What important analog signals (or digital signals that should be tested as an analog signal) have higher frequencies than those listed above?

• The classic one that dictated 'scope design was NTSC (or PAL) video, needing apparently about 5 MHz. But its been a while since I looked at an NTSC signal, and its importance is fading fast. – The Photon Jul 20 '12 at 21:30
• Ultrasound can easily get up to 10MHz in practical technology. – DarenW Mar 1 '13 at 5:17

I assume "as an analog signal" means on an oscilloscope as opposed to a logic analyser.

For a digital signal, it is important to be able to check the signal integrity and see whether it is subject to problems such as ringing, crosstalk, reflections, jitter, attenuation, etc.

This can only be done with a scope with a bandwidth > the frequencies present in the signal - remember with a digital signal there are frequencies much higher than the fundamental present, how high is dictated by the rise time of the signal. For a 1MHz digital signal you would generally want at least a 5MHz bandwidth, preferably much higher.
For debugging a typical small microcontroller (e.g. PIC, Atmel AVR, Arduino, etc) a scope bandwidth of at least 50MHz is preferable. This should be capable of handling just about all situations you might encounter.

There are many signals above 1MHz that need checking, most microcontroller clock signals are > 1MHz, SPI is often > 1MHz, USB, etc. FPGA designs may run at 100s of MHz, high speed ADCs and DACs, etc.

On a logic analyser all you can see is whether it is above a certain level or below a certain level (like a 1-bit scope) so while useful in other ways they are not suitable for checking signal integrity.

The image below (taken on an MSO - Mixed Signal Oscillscope, a combination of a scope and logic analyser) is a good example of crosstalk causing problems and why a scope is needed to see what's really happening. Notice the waveforms are quite a way from the idea of a "perfect" digital signal:

For the leftmost red arrow the second trace down is the transmitting trace, and the top trace down is the "victim" (receiving trace) and the right hand pulse they are reversed. We can see on the rise of the "transmitting" signal it causes a spike in the receiving trace, resulting in a unwanted glitch on the logic display, which is what the digital receiver would "see".

In this image at the top we can see signal degradation caused by an incorrectly terminated trace, causing reflections. At the bottom we can see the same signal after it has been correctly terminated:

On the logic analyser, both signals may work, but there is no way of knowing how marginal the first signal is without checking with a scope. The incorrectly terminated trace may only cause problems intermittently, so it's important to be able to check it's integrity.

Looking at your link to the DPScope design, I see it's dsPIC based. It won't be comparable to anything you can buy (you can get a 20MHz analogue scope for << £50 nowadays, and a 5-10MHz DSO for similar)
However, it would be a great project for educational purposes, and you will get something perfectly useable for low frequency (e.g. audio, UART, PWM) purposes. Plus you'll have fun building it. If your thinking of doing so, I'd say go for it, just don't expect it to take care of all your debugging needs. If your budget is limited, get a cheap analogue scope - you will generally get the highest bandwidth for your money.
Remember the chicken and egg problem - you need a scope in order to build and test a scope ;-)

A rough rule of thumb is you want about 10 times the bandwidth as the frequency of the digital signal you want to see on a scope. This is assuming you want to see the wave shape, not just when it goes high or low.

Modern scopes have two issues related to bandwidth: The bandwidth of the analog electronics to receive your signal and present it to the internal A/D, and the sample rate of that A/D. Nyquist says you need at least 2x the sample rate of the highest frequency of interest. However, that is the absolute minimum theoretical. In practise you need usually a few times more than that at least.

Sampling aliasing of a scope gets annoying, and sometimes it can give you completely the wrong idea or really confuse you if you're not noticing that's what is happening. A sampling rate 10-20x the highest frequency the scope can handle is usually about right. Keep in mind that 2x is the guaranteed to fail limit. For example, a 60 MHz scope might have a sample rate of 1 GHz or so.

Q Under what situations would it be necessary to have a bandwidth capable of measuring these digital lines as analog signals?

A Debugging phantom errors in the presence of RF noise or EMI such power line transients, ESD events, RF power bursts and last but not least an EMP. Crosstalk is another source of interference and common to bundles of RS232 signals on multi-wire pairs.

So far all the other answers give correct info as well.

External data communication with unshielded twisted pair wires (UTP) is good but insome cases not enough to protect data from transient errors. Shielding can provide 40~60dB of reduction and filtering with magnetics can also reduce common mode noise, but you never eliminate analog noise all together. So it all digital signals are analog with sufficient margin that we treat them perhaps as binary signals most of the time.

Q What important analog signals (or digital signals that should be tested as an analog signal)

A I can list hundreds of communication serials signals that are considered digital that must be scoped as analog; but in general; Disk drives read / write data from heads, Telemetry receiver , I2C signals that fail in a shared access mode, Ethernet signals that cause high error rates, T1 repeaters, all optical data receiver links, DVD error rate debug testing. etc etc. Another is logic "race" conditions in which case a signal changes state just when it is being sampled by a clock.

Q have higher frequencies than those listed above?

A The fastest pulse I have seen is from a human finger discharging 10kV to a door knob. The rise time was 20 pS using an extremely fast scope 20 yrs ago. Plasma arcs arc caused by the tunnelling of electrons in a plasma state created when air breaks down by exceeding its dielectric withstanding voltage of usually 1kV per mm.

Recent photo from EOS/ESD experts.. still does not have enough bandwidth to capture the event, but is not bad at 336ps rise time.

http://www.analog.com/static/imported-files/quality_assurance/eos_esd_chapter_091400b_103.pdf

• To see a 20 ps (lowercase "s") rise time you need at least 500 GHz bandwidth. I'm not sure there were many scopes that fast 20 years ago. Also I doubt that you can attenuate 10 kV to the scope's input level without decreasing rise time. 10 kV in 20 ps is 500 000 000 V/$\mu$s. Are you sure your memory isn't failing you? – stevenvh Jul 21 '12 at 6:17
• You don't have to be derogatorive. And don't try to impress me with all your fancy stuff, I told you before it won't work! You'd better try to make an impression with a really good answer. (And don't complain that users are too simple to understand your answers, they're not. It's just that most of your answers lack the quality to be good answers.) – stevenvh Jul 21 '12 at 13:16
• Electric arcs thru air are not a tunneling phenomena. When the air breaks down it becomes a plasma, which has free electrons, which makes it a conductor. There is no tunneling, just ordinary conduction. Tunneling works over very short distances only, or more precisely falls off much much faster than linearly as distances in increase. Lightning would be impossible if it relied only on tunneling. – Olin Lathrop Jul 21 '12 at 14:54
• @Tony - An answer is no use if nobody can make heads or tails of it, no matter how correct it may be. I don't think you get downvoted because people think you're wrong, but because the answers aren't clear, and then the downvote means "not useful". And believe it or not, you can also be wrong. This is not a good answer because you don't seem to mind an AVR sinking 40 mA, which may damage it. Try some modesty, other users will appreciate it. – stevenvh Jul 21 '12 at 16:58
• @TonyStewart The problem is that your answers are often incorrect and usually communicated poorly. – W5VO Jul 21 '12 at 23:53

I propose to you, that all digital signals are analog in nature and to fully understand issues of reliability, you need to understand all the analog equivalent circuits of digital signals. Recent google search just now showed the analog equivalent circuit I discussed in comments of an HDA part of a HDD with EOS discharge model from tribo-electric effect on moving disks under recording heads for magnetic recording.

We all think of disks as digital storage medium. Now neglecting the magnetic properties and just focusing on the OES/ESD properties of pico second discharge pulses, this is the equivalent circuit , some researchers have come up with. I was ware of this model in my mind back in the early 80's when I was Test Engineering Mgr at Burroughs/Unisys and we refurbished about 100 different old peripherals, built new and tested OEM HDD's for systems integration. So we had several Class 100 Clean rooms and many servo writers where we had to resolve these issues, such as the one modelled for Electro-Static-Discharge (ESD) in this photo.

this is old news to me... but a very good article. which reinforces the need for ANALOG SCOPES of digital signals that are very common to everyday life. (ESD)

• no comments means closed minded when -point added .. ok? I provided relevant proof. YOu are free to disagree without. ( in philosphy, may be considered bad form ) – Tony Stewart Sunnyskyguy EE75 Jul 21 '12 at 15:56
• It is suggested that users leave a comment but no requirement exists. I would say the less likely someone believes a user to take advice directly affects the chance that they leave a comment.) – Kortuk Jul 22 '12 at 6:14