示波器

An oscilloscope is a device that captures an input voltage signal and converts it to a correctly scaled voltage versus time waveform that is displayed on a scaled grid. The oscilloscope has a triggering circuit that defines when the input signal should be captured and displayed, and a variable gain front end that permits (vertical voltage) signal adjustment to accept a wide range of input signal amplitudes. A horizontal (timebase or sweep) adjustment defines the period of time to acquire the signal.

Many will claim to have invented the analog oscilloscope, but Tektronix can rightly claim to have invented the first triggered-sweep (analog) oscilloscope, which vastly improved the usefulness and versatility of the instrument.

Walter LeCroy and his design team at LeCroy Corporation (now Teledyne LeCroy) in 1985 released the first digital storage oscilloscope (DSO, or now just referred to as a digital oscilloscope) – named the Model 9400 – that replicated and improved on the features and capabilities of the analog oscilloscopes in use up to that time. The Model 9400 had bandwidth (125 MHz) equivalent to what was available in an analog oscilloscope (at the time) and could continuously capture a signal for a long period of time using 32,000 sample points (at the time, an amazingly long acquisition record length). A tenuous claim could be made that LeCroy’s WD2000 Waveform Digitizer (launched in 1971) was the first digital storage oscilloscope, but the record length was limited to 20 sample points and the architecture could not easily scale to longer record lengths. Read the full story here https://www.teledynelecroy.com/walter-lecroy.

Walter LeCroy and his design team at LeCroy Corporation (now Teledyne LeCroy) in 1985 released the first digital storage oscilloscope (DSO, or now just referred to as a digital oscilloscope) – named the Model 9400 – that replicated and improved on the features and capabilities of the analog oscilloscopes in use up to that time. The Model 9400 had bandwidth (125 MHz) equivalent to what was available in an analog oscilloscope (at the time) and could continuously capture a signal for a long period of time using 32,000 sample points (at the time, an amazingly long acquisition record length). A tenuous claim could be made that LeCroy’s WD2000 Waveform Digitizer (launched in 1971) was the first digital storage oscilloscope, but the record length was limited to 20 sample points and the architecture could not easily scale to longer record lengths. Read the full story herehttps://www.teledynelecroy.com/walter-lecroy.

Oscilloscopes accept voltage signals as inputs. A probe or sensor must be used to convert a non-voltage signal (e.g., a current signal, a magnetic field signal) into a voltage signal, correctly scaled in the appropriate units. Probes or sensors to measure current are commonly available from oscilloscope manufacturers, and sensors to measure other units are widely available. Most professional-grade oscilloscopes provide support for common rescaling (e.g., from Volts to Amps) and many other units, but if this is an important feature for your requirements, it is best to check the support for rescaling within the oscilloscope before purchase, especially if the sensor has a non-linear input to output ratio.

Reference webinarsPart 7: How Do I Make a Current Measurement with an Oscilloscope?andPart 8: How Do I Measure Current on an Oscilloscope Using a Shunt Resistor?in the 2023 Oscilloscope Coffee Break Webinar Series for other details.

TheIEEE 1057 Standard for Digitizing Waveform Recordersspecifies the analog bandwidth of a digital oscilloscope as the frequency at which the amplitude response is -3 dB (which equates to 70.7%) of the response at the reference frequency (which for an oscilloscope is DC). While it may seem confusing to have an analog bandwidth specification in a digital oscilloscope, the digital oscilloscope has many analog amplifier components prior to the portion that digitizes and stores the signal.

The bandwidth required for capture and measurement of signals depends greatly on the signals to be measured, the types of measurements to be made, and the accuracy desired of the measurements. A rough rule of thumb most engineers use is to have an oscilloscope with three times the bandwidth of the highest frequency signal they wish to measure, though this becomes impractical for very high frequency signals.

Reference the definition for oscilloscope bandwidth in the FAQ (above). Most oscilloscopes approach the -3 dB bandwidth-rated frequency slowly, beginning with a gentle amplitude rolloff at 50% (or so) of the bandwidth frequency rating. This means that if the oscilloscope amplitude response is -1 dB at 70% of rated bandwidth and -2 dB at 85% of rated bandwidth, then the amplitude of the captured pure sinusoid will be approximately 90% (-1 dB) or 80% (-2 dB) and 70% (-3 dB) compared to when the input sinusoid frequency is approaching the bandwidth rating of the oscilloscope. However, most engineers are not measuring pure sinusoids with their oscilloscope. Note that the highest bandwidth oscilloscopes may have a flatter (less amplitude rolloff) or adjustable amplitude response, for a variety of reasons.

More likely, an engineer is measuring a signal that resembles a square wave. In this case, it is known that a square wave can be represented as a Fourier series expansion comprised of the sum of the fundamental frequency and odd harmonics, with the Nth harmonic contributing a 1/N amplitude at that frequency. What this means is that to accurately represent a square wave, you need enough bandwidth to capture the fundamental frequency and enough of the odd harmonics. How many odd harmonics is “enough” (and how much bandwidth is needed) is determined by the engineer’s tolerance for a rise time measurement on the oscilloscope that is slower than the real signal, and the amount of additive overshoot and ringing present on the measured signal. If only the 3rd harmonic is captured, the rise time will be appreciably slower, and the overshoot and ringing will be noticeable compared to if the 99th harmonic is captured (in which case the captured signal will be indistinguishable from the original input signal).

This gets us back to the original answer that is given most often in response to the question of “how much bandwidth is needed?” – about 3x the bandwidth of the highest frequency signal. But what does “highest frequency” mean? In this context, most engineers are thinking of the rise time measurement capability of the oscilloscope (which is related to bandwidth). If an engineer wants to measure a signal with a rise time of 1 ns, they would not choose an oscilloscope with a 1 ns rise time (such an oscilloscope would typically have a bandwidth of 350 MHz) – they would choose an oscilloscope with bandwidth 3x that (or 1 GHz).

Reference webinarPart 2: How Much Bandwidth Do I Need in My Oscilloscope?in the 2023 Oscilloscope Coffee Break Webinar Series for other details.

Resolution is the number of analog-to-digital converter (ADC) quantization levels, with an N-bit ADC having 2^N quantization levels. For instance, an 8-bit oscilloscope has 2⁸ = 256 quantization levels whereas a 12-bit oscilloscope has 2¹² = 4096 quantization levels. Note that the number of bits (quantization levels) in the ADC is no guarantee that the rest of the oscilloscope’s signal path (notably the analog components) will have noise performance worthy of a high resolution ADC. Thus, an advertised high-resolution oscilloscope may perform no differently than a conventional 8-bit resolution oscilloscope. ReferenceComparing High Resolution Oscilloscope Design Approachesfor more details on that tradeoffs that many oscilloscope manufacturers make when designing high-resolution oscilloscopes. Reference webinarPart 1: What is Oscilloscope Resolution?in the 2023 Oscilloscope Coffee Break Webinar Series for other details.

A high resolution oscilloscope is any oscilloscope that is advertised as such and that uses either improved hardware, software filtering (that reduces bandwidth and sample rate), or a combination of both to provide improved resolution and signal-to-noise ratio compared to a conventional 8-bit oscilloscope. A marketing claim of high resolution is no guarantee of real-world performance. Claims of high resolution specific to the ADC, or improvements in baseline noise or signal-to-noise ratio that are only possible at reduced bandwidths, are red flags that the so-called high resolution will not be realistically achieved in all normal oscilloscope operating conditions. ReferenceComparing High Resolution Oscilloscope Design Approachesfor more details.

There is no difference – these are just two ways to express the same thing, though it should be noted that Teledyne LeCroy has a registered trademark on the name High Definition Oscilloscope and the acronym HDO, having been the first oscilloscope company to offer 12-bit high-resolution oscilloscopes that provide 12 bits all the time with no reduction in sample rate or bandwidth.

A mixed-signal oscilloscope (MSO) commonly refers to an oscilloscope that has both analog and digital (logic) input channels. A common configuration is 4 analog input channels plus 16 digital logic input channels. The digital logic input channels can preserve the scarcer (and more expensive) analog input channels for signals that require their capabilities, and the digital logic input channels can be used for simple toggle or logic signals, or low-speed serial data (e.g., I2C, SPI, UART, etc.) signals.

Mixed-domain oscilloscope (MDO) is a marketing term for an oscilloscope that provides some type of radio frequency (RF) input or conversion to capture signals in both the time and frequency domains. If a dedicated RF input is provided, capabilities can be similar to that of a spectrum analyzer. Software fast fourier transform (FFT) techniques can be used to provide similar capabilities without a dedicated (and costly) RF input.

The amplitude accuracy of an oscilloscope is comprised of many different components and will vary depending on the oscilloscope resolution, input path, input frequency content, whether a probe is used, etc. Amplitude accuracy can range from better than 1% for a 12-bit high definition oscilloscope (HDO®) with a cable signal input, to 5% (or more) for an 8-bit oscilloscope operating with an active probe coupled to the oscilloscope via the 50 Ω termination. While these accuracies may seem low compared to a digital voltmeter (DVM), an oscilloscope provides far more capabilities than a DVM.

ReferencePart 1: What Is the Difference Between Oscilloscope Resolution, Accuracy and Sensitivity?in the 2024 Oscilloscope Coffee Break Webinar Series for more details.

Sensitivity is the smallest signal change that can be viewed in the oscilloscope. An oscilloscope with high sensitivity can be used to view smaller signals compared to an oscilloscope with lower sensitivity. Sensitivity adjustment on the oscilloscope is made using the vertical gain setting (volts/division). Note that high sensitivity does not necessarily correlate to high accuracy, and that an analog vertical gain setting indicative of high sensitivity (e.g., 1 or 2 mV/div) may be limited in usefulness by the ADC resolution or noise in the oscilloscope. ReferencePart 1: What Is the Difference Between Oscilloscope Resolution, Accuracy and Sensitivity?in the 2024 Oscilloscope Coffee Break Webinar Series for more details.

Historically, an engineer would consider rise time to be related to bandwidth according to the formula TR(s) = 0.35/Bandwidth (Hz), with TR being the 10-90% rise time (as defined by the IEEE). This formula was (mostly) true in an era when oscilloscope bandwidths were very low (1 GHz or less) and amplitude rolloffs were very gradual. This formula can still hold true for lower bandwidth oscilloscopes.

Today’s higher bandwidth oscilloscopes—or oscilloscopes with more complex, lower-noise signal paths—might adhere to the TR(s) = 0.35/Bandwidth (Hz) formula for models at the lower (bandwidth) end of the product line but adhere to TR(s) = 0.4/Bandwidth (Hz) or perhaps approaching TR(s) = 0.45/Bandwidth (Hz) (or higher, in some cases) for maximum bandwidth models. The reason for the lower numerator in lower bandwidth models is that they are likely using an analog signal path that has more high-frequency headroom for a slower amplitude rolloff compared to the highest bandwidth models. On the highest bandwidth oscilloscope model in a product series, the analog signal path likely has reached a hard, upper limit on amplitude response, and the amplitude response rolls off quickly beyond that, which results in a slower rise time (and higher numerator) due to the highly attenuated high frequency response past the bandwidth rating of the oscilloscope.

Reference webinarPart 3: How Is Rise Time Related to Bandwidth in an Oscilloscope?in the 2023 Oscilloscope Coffee Break Webinar Series for other details.

A digital oscilloscope digitizes signals through analog-to-digital converters (ADCs) that sample and hold voltage values to create discrete sample points. Sample points are recorded at a given frequency (time interval), and the sample rate is referred to as Samples/second.

Reference webinarPart 4: What Is Oscilloscope Sample Rate and How Much Do I Need?in the 2023 Oscilloscope Coffee Break Webinar Series for other details.

The minimum sample rate needed, according to Nyquist theorem, is twice that of the frequency you wish to measure. In a digital oscilloscope, this is commonly interpreted as sample rate and must be a minimum of twice the oscilloscope’s bandwidth rating. However, the oscilloscope doesn’t usually have a brick-wall amplitude response past the bandwidth rating, and it will pass some high frequency content beyond the bandwidth rating. Therefore, most oscilloscopes provide a minimum sample rate to bandwidth ratio of 2.5. This can be considered the minimum to reconstruct a sinewave from digital sample points.

To accurately reconstruct more complex signal shapes from digital sample points, engineers commonly desire 5 or perhaps up to 10 sample points on a rising edge. If an engineer is following the common rule of thumb of selecting an oscilloscope three times faster than the signal they wish to measure (Reference webinarPart 2: How Much Bandwidth Do I Need in My Oscilloscope?in the 2023 Oscilloscope Coffee Break Webinar Series for other details, or the similarly titled FAQ), then 5 to 10 sample points on a rising edge is easily accommodated.

Reference webinarPart 4: What Is Oscilloscope Sample Rate and How Much Do I Need?in the 2023 Oscilloscope Coffee Break Webinar Series for other details.

Acquisition memory is what is used to store the digital oscilloscope sample points for recall to a display or for further processing to make measurements, perform math calculations, etc.

Oscilloscope acquisition memory stores the oscilloscope sample points of the digitized signal, whereas the central processing unit (CPU) that is powering the oscilloscope functions has its own random access memory (RAM) to serve the CPU’s needs.

Memory depth is just another way to describe the total length of the acquisition memory, whether in points (e.g., kilopoints (kpts), megapoints (Mpts), Gigapoints (Gpts)) or in samples (e.g., megasamples (MS)).

More samples (or points) provide more capability to capture very long continuous time intervals before needing to reduce the sample rate. How many samples an engineer needs depends on the bandwidth of the signals an engineer wishes to capture, the time resolution an engineer wishes to capture those signals with, and the amount of continuous time an engineer wishes to acquire.

If an oscilloscope had a sample rate of 10 GS/s and 1 GS (or Gpts) of acquisition memory, then it could acquire 100 ms of time (1 GS / 10 GS/s = 0.1 s, or 100 ms). If it was desired to capture 200 ms with 1 GS of acquisition memory, the sample rate would have to be reduced to 5 GS/s, which may (or may not) be acceptable.

Oscilloscope baseline noise is the measured AC RMS value of an oscilloscope input channel with no signal connected to it. A simple baseline noise test will provide a general indication of noise performance when no signal is present on the input to the oscilloscope. While this test is simple and easy to perform, it is not the most realistic test of oscilloscope performance, because most oscilloscopes are used with input signals connected to them. Nonetheless, noise will not decrease when input signals are added, as the added signal amplitude will only add noise to the measurement later. Thus, baseline noise can be a useful test for roughly assessing overall performance.

Note that in a Teledyne LeCroy oscilloscope, the SDEV measurement equates to AC RMS.

ReferenceComparing High Resolution Oscilloscope Design Approachesfor more details about various types of noise in oscilloscopes.

Signal-to-noise ratio is the calculation of the ratio of full scale range divided by the baseline noise, expressed in volts according to the following formula:

SNR (dB) = 20*log10((V_Full-scale/(2*√2))/V_AC-RMS))

With V_Full-scalebeing the full scale voltage on the oscilloscope (equal to number of vertical divisions * V/div gain setting) and V_AC-RMSbeing the AC RMS value for the baseline signal at a given V/div gain setting.

Note that some oscilloscopes (e.g., Keysight, Teledyne LeCroy) have 8 vertical divisions for full scale whereas others (e.g., Tektronix) have 10 vertical divisions for full scale.

Note that Teledyne LeCroy’s AC RMS measurement is named SDEV, whereas other oscilloscopes typically have an RMS measurement that is selectable as either AC or DC reading. Be sure to use the AC RMS value or the SNR calculation will incorrectly include the effect of any small DC offset errors in the oscilloscope channel.

SNR(dB) = 20*log10( (V/div*8/(2*sqrt(2)))/noise_in_rms)

ReferenceComparing High Resolution Oscilloscope Design Approachesfor more details about various types of noise in oscilloscopes.

Per IEEE Std. 1057 IEEE Standard for Digitizing Waveform Recorders, SINAD is the ratio of root-mean-square (rms) signal to rms (baseline) noise and distortion. SINAD is measured at a specific frequency and amplitude using a sinewave input, and the amplitude at which the measurements are made does impact the distortion and should be specified (90% of full-scale amplitude is typical). SINAD is a more complete measurement of the performance of the oscilloscope in actual operation.

ReferenceComparing High Resolution Oscilloscope Design Approachesfor more details about various types of noise in oscilloscopes.

The best method to reduce noise on signals measured with your oscilloscope is to use a low-noise, high-resolution oscilloscope that provides 12 bits resolution at full bandwidth. But any oscilloscope can have its noise reduced using analog hardware or digital software filters provided that the tradeoff of lower bandwidth in exchange for reduced noise is acceptable.

Hardware filters are usually displayed as a 20 MHz or 200 MHz (or similar) bandwidth limit in the channel menu. These filters tend to have very slow rolloffs, so their noise reduction capability is probably less than that of a digital software filter.

Digital software filters may be math functions, may be high-resolution modes, or may be software filter selections in the channel menu (e.g., Teledyne LeCroy’s Enhanced Resolution (ERes) selection). Mathematically, every halving of the sample rate (and bandwidth) reduces noise by 3 dB (~30%, or 0.5 effective bits). Sometimes the digital software filters interpolate sample points after the mathematical filter operation, but the hardware sample rate has still been reduced.

Be wary of high-resolution modes that promise better performance than what is mathematically possible, or that are the only means of achieving high resolution (and lower noise) in what would otherwise be an 8-bit resolution oscilloscope.

ReferenceComparing High Resolution Oscilloscope Design Approachesfor more details about tradeoffs made to reduce noise in oscilloscopes. Reference webinarPart 6: How Can I Reduce Noise on Signals Measured With an Oscilloscope?in the 2023 Oscilloscope Coffee Break Webinar Series for other details.

 

Oscilloscope ENOB is derived from measurement of the oscilloscope SINAD as follows:

Oscilloscope ENOB= (SINAD-1.76)/6.02

If the front-end amplifier is not the dominant source of noise in the oscilloscope system, the system ENOB will approach the ENOB of the ADC. It is important to understand that the ADC ENOB is an upper bound on the system performance, but the system performance is the critical performance to understand. Realistically, the oscilloscope (system) ENOB will always be less than the ADC ENOB.

If the applied input signal is not 100% of full-scale amplitude, then the ENOB is derived as follows:

Oscilloscope ENOB= (SINAD-1.76+20 log((FullScale Amplitude)/(Input Amplitude)))/6.02

A “rule-of-thumb” of 6 dB SINAD per effective bit can be inferred from this equation. Thus, improvement of half an effective bit equates to 3 dB (30%) reduction in noise, and improvement of a full effective bit equates to a 6 dB (50%) reduction in noise. Small differences in ENOB mean a lot in terms of vertical (voltage amplitude) noise.

ReferenceComparing High Resolution Oscilloscope Design Approachesfor more details about various types of noise and why the ADC rated number of bits isn’t fully achieved when deployed in digitizers or oscilloscopes.

ReferencePart 2: What Are Oscilloscope ADC Effective Bits and ENOB?in the 2024 Oscilloscope Coffee Break Webinar Series for more details.

The ADC ENOB is an upper bound on the oscilloscope ENOB, but the oscilloscope ENOB is the critical performance to understand. Realistically, the oscilloscope ENOB will always be less than the ADC ENOB. If an oscilloscope makes specific claims about the ENOB performance of its ADC, it is probably a red flag that the complete oscilloscope ENOB performance is much less.

ReferenceComparing High Resolution Oscilloscope Design Approachesfor more details about various types of noise and why the ADC rated number of bits isn’t fully achieved when deployed in digitizers or oscilloscopes.

ReferencePart 2: What Are Oscilloscope ADC Effective Bits and ENOB?in the 2024 Oscilloscope Coffee Break Webinar Series for more details.

The Nyquist theorem states that a sinusoid can be reconstructed with no loss of information provided it is digitally sampled at twice (or more) of the frequency of the sinusoid. Typically, this means that the minimum sample rate in a digital oscilloscope is 2.5 times the bandwidth on all channels. 2.5:1 sample rate to bandwidth (SR/BW) is the ratio used (instead of the minimum 2) to take into account that the oscilloscope will not have a perfect brick-wall filter at the rated bandwidth. Less than the 2:1 SR/BW ratio will create the risk of aliasing of the digitally sampled input signal.

If the Nyquist sampling rate requirements are not met, the signal is considered undersampled and cannot be reconstructed with no loss of information. Instead, the reconstruction of the signal will still occur, but it will be an incorrect reconstruction, referred to as aliasing.

ReferencePart 3: What Is Oscilloscope Aliasing?in the 2024 Oscilloscope Coffee Break Webinar Series for more details.

Spurious Free Dynamic Range (SFDR) is the ratio (usually expressed in dB) of the root-mean-square (RMS) amplitude of a fundamental oscilloscope input signal to the RMS amplitude of the next largest spurious signal in the oscilloscope output. SFDR is usually measured in the oscilloscope using an FFT or spectrum analyzer-like amplitude vs. frequency oscilloscope display. The spurious signals could be caused by distortion or other noise components, or could be at a frequency consistent with the core analog-to-digital converter (ADC) sampling frequency.

SFDR is one of the most misunderstood quality checks engineers perform on oscilloscopes. Any ADC is going to exhibit spurs at the sampling frequencies, and these spurs are usually of such low amplitude (compared to the input fundamental) and of such narrow frequency band that the SFDR ratio is well above (not as worse as) the baseline noise signal-to-noise ratio or signal-to-noise-and-disortion (SINAD) ratio for a given input frequency. Occasionally an oscilloscope might exhibit serious distortion components at specific frequencies, which is easily exposed by an SFDR test, but this is not common.

ReferenceComparing High Resolution Oscilloscope Design Approachesfor more details about SFDR in oscilloscopes.

ReferencePart 4: What Is Oscilloscope Spurious Free Dynamic Range (SFDR)?in the 2024 Oscilloscope Coffee Break Webinar Series for more details.

Properly called an equivalent-time sampling oscilloscope, a sampling oscilloscope provides one sample per trigger, with a small time delay added after each trigger so as to reconstruct a repetitive waveform from multiple triggered events. The measurement bandwidth is only limited by the frequency response of the sampler, which can be very high at very low cost. The limitation is that a sampling oscilloscope cannot capture a continuous waveform.

A sampling oscilloscope can only acquire a repetitive signal, whereas a real-time oscilloscope can acquire a continuous time waveform in one continuous sample record.

ReferencePart 6: What Is The Difference Between a Real-time Oscilloscope and a Sampling Oscilloscope?in the 2024 Oscilloscope Coffee Break Webinar Series for more details.

Digital Phosphor Oscilloscope (DPO) is a marketing term used by Tektronix to describe their oscilloscopes that utilize a fast waveform display architecture (more recently marketed as DPX Technology) to mimic the display appearance of a phosphor-beam CRT display used on an analog oscilloscope.

Some other oscilloscope manufacturers have similar features. All of them optimize for display update (refresh) at the expense of storing data, so if an anomaly is viewed during the fast update display, it cannot be saved or retrieved for further inspection. Furthermore, they are still based on digital capture techniques and therefore have large amounts of dead time during which they are not capturing (or displaying) waveforms (or anomalies). Oscilloscopes with fast update are typically usable only on very short acquisitions of repetitive signals, and the update rate degrades at longer (and more useful) time periods, and they are not very useful for viewing more than one signal at a time. In essence, the feature was conceived during a time when analog oscilloscopes were transitioning to digital oscilloscopes, and there is no longer much practical usage of this feature for most customers.

ReferencePart 9: What Is a Digital Phosphor Oscilloscope?in the 2024 Oscilloscope Coffee Break Webinar Series for more details.

A fast update rate display might provide usability and comfort to someone who is used to an analog oscilloscope (though most of these engineers have long since retired). They might also be useful to an engineer who is viewing a very short duration repetitive signal with many obvious anomalies. Engineers who are capturing longer, non-repetitive time intervals will probably find fast update rates to be an interesting feature that gets little use in real-world debugging.

Eye diagrams and eye patterns are display tools that are used to assess the signal quality of a digital signal by overlaying the digital levels for every bit (along with any transitions before or after each bit) to provide a quick visual assessment of the quality of the digital signal. Ideally, the eye diagram/pattern is very open in the middle with a clear top (digital 1 level), base (digital 0 level) and transitions (rising and falling edges of digital level transitions). Multi-level signals, such as PAM-3 or PAM-4, can also be displayed as eye diagrams.

An eye diagram and an eye pattern are two ways to describe the same thing.

ReferencePart 11: What Is an Oscilloscope Eye Diagram?in the 2024 Oscilloscope Coffee Break Webinar Series for more details.

There are two basic methods to display an eye diagram using a digital oscilloscope.

The first method is the most basic but has the most limitations. An edge trigger is used to trigger on the 50% level of a rising or falling digital signal edge, with the timebase of the oscilloscope set to be a little longer than a single bit period, and the oscilloscope trigger point set to be about one quarter from the left edge of the oscilloscope grid. Display persistence is used to capture many short acquisitions of a single bit period, and the triggered signals are overlaid for visual observation. This method is intuitive but doesn’t provide an eye diagram of a continuous signal, doesn’t permit any type of post-processing to determine the cause of any eye diagram anomalies, and is affected by the added trigger jitter of the oscilloscope. It is a good, quick check whether a digital signal has good quality.

The second method is more robust and more widely used, especially with high-speed serial data signals. A long, continuous acquisition is made of a digital signal and the clock is extracted mathematically, with the extracted time period of the clock used to mathematically “slice” the continuous acquisition into bit periods that are overlaid to form the eye diagram. Since the data is continuous, additional mathematical processing may also be performed to simulate the use of a phase-locked loop (PLL) in the clock circuit, calculate jitter, measure various aspects of the eye opening (amplitude, width, etc.), and debug any anomalies present.

A sampling oscilloscope (described in an earlier FAQ) creates an eye diagram through the use of a hardware clock recovery circuit that works with the sampling module to create the eye diagram. This is generally considered an archaic method today and is not widely used unless the high-speed serial data signal can be completely analyzed and assessed with non-continuous (not real-time) data acquisitions. In that case, this method is perfectly satisfactory and is very low cost for the oscilloscope bandwidth provided. However, it does require different hardware anytime the signal has different bit rates or PLL requirements.