The role of oscilloscope ENOB value in improving accuracy

Experienced oscilloscope users will further evaluate the oscilloscope's update rate, inherent jitter, noise level, and all other specifications that can improve measurement quality. However, when evaluating oscilloscopes with bandwidths in the GHz range, another quality indicator must be considered, which is the effective number of bits (ENOB) to describe the analog-to-digital converter (ADC) characteristics in the oscilloscope. Therefore, how to effectively grasp the measurement accuracy of the oscilloscope, ENOB becomes a crucial indicator.

When examining the quality of front-end/ADC design, the most important thing to look at is ENOB/noise level

In the oscilloscope architecture, the front-end and ADC technology are key elements to improve measurement accuracy, because the oscilloscope front end can condition the signal sampled by the instrument so that the ADC can correctly digitize the signal. Among them, the oscilloscope front-end components include attenuators, preamplifiers, and signal distribution paths. Oscilloscope design engineers often have to work hard to design front-end components to obtain flat frequency response, lower noise, and the required frequency reduction.

Since the requirements of oscilloscopes for ADC vary, oscilloscope manufacturers generally design their own ADC chips. However, each development of a new front-end component or ADC requires considerable investment, so the designed ADC will be used in multiple oscilloscope series and different generations of models. The oscilloscope design team will try to reduce the impact of these circuits on the measurement results of the sampled signal to improve the measurement accuracy.

Although users can measure the characteristics of the ADC and front-end components combined together, it is difficult to analyze the characteristics of individual components separately. Therefore, many methods must be used to evaluate the quality of the oscilloscope's front-end circuit. Oscilloscope manufacturers usually use noise measurement and ENOB to evaluate the design quality of the oscilloscope's front-end and ADC.

However, when selecting an oscilloscope, one should not only evaluate ENOB or noise level, but also consider the overall performance of the oscilloscope. In particular, evaluating the oscilloscope noise level under different vertical settings and offset values ​​will help further confirm the quality of the oscilloscope measurement and determine whether the oscilloscope front-end and ADC converter design are stable enough.

Oscilloscope noise will increase unnecessary jitter and reduce design margin, and generally the higher the bandwidth of the oscilloscope, the more internal noise it will generate. This is because the oscilloscope will receive accumulated noise at high frequencies; while oscilloscopes with lower bandwidths can filter out these noises because of their lower frequency drop. Therefore, the most direct way to evaluate oscilloscope noise is to not receive anything on the input channel and view the root mean square (RMS) value of the voltage by changing the vertical sensitivity and offset value.

Mastering ENOB measurement factors: Signal source amplitude/frequency must be taken seriously

In fact, ENOB is measured by sweeping a sine wave with a fixed amplitude, and then capturing the voltage measurement results for evaluation and analysis. The analysis methods can be divided into time domain and frequency domain methods. The time domain measurement method is to calculate the ENOB value, that is, the noise, by subtracting the most ideal waveform presented by the voltage versus time relationship from the captured time domain data.

The noise may come from the front end of the oscilloscope, such as nonlinear phase and amplitude changes at different frequencies, or from the interleaving distortion of the ADC. When using the frequency domain method to calculate ENOB, the power associated with the main signal must be subtracted from the power of the entire broadband. In this way, the results obtained by the time domain and frequency domain methods can be consistent.

In addition, if you want to perform ENOB measurements or analyze the ENOB specifications provided by oscilloscope manufacturers, you must consider the following points. First, the spectral purity of the signal source used when measuring ENOB will affect the measurement results. Therefore, the signal source and the filter used in conjunction should ensure that the source ENOB is greater than the ENOB of the oscilloscope.

Secondly, the measured ENOB value is related to the amplitude ratio of whether the signal source fills the full screen of the oscilloscope. Using different signal sources that fill 75% or 90% of the full screen of the oscilloscope for measurement will result in different ENOB values. In view of this, the Joint Data Center (JDEC) standard of the International Electronics Interconnect Packaging Association recommends using 90% of the full screen as the amplitude for confirming the ENOB value.

Obviously, the ENOB value is closely related to the quality of the oscilloscope measurement. Therefore, when comparing or testing any effective bit specification, it is important to consider the impact of the amplitude and frequency of the signal to be measured on ENOB.

Universal ENOB Standard Helps to Identify the Pros and Cons of ADCs

It is worth noting that most of the ADCs in current oscilloscopes adopt pipelined or flash architectures, and the Institute of Electrical and Electronics Engineers (IEEE) has also specifically defined a method to use ENOB to evaluate the quality of ADCs to help users identify the advantages and disadvantages of various ADCs.

Among them, catheter ADC uses second-order or multi-order conversion to achieve a higher sampling rate. For example, the 20GSa/s ADC equipped with Agilent's 90000A series oscilloscope (Figure 1) is composed of eighty 256MSa/s sampling rate ADCs, which can provide a higher sampling rate.

Figure 1. ENOB graph of Agilent Infiniium 9000 Series oscilloscopes. ENOB values ​​vary by frequency, and each oscilloscope has a different ENOB graph. This ENOB graph shows the ENOB of the entire oscilloscope system, not just the ENOB of the 8-bit ADC.

As for the flash ADC, it has a set of comparators that can sample the input signal in parallel, and each comparator corresponds to a decoded voltage range. By feeding the signal to the logic circuit, this set of comparators will generate a code for each voltage range. However, each ADC technology has its limitations. For example, the linearity error of the flash ADC is higher, and the catheter ADC usually has more interleaving errors.

However, contrary to common sense, some oscilloscopes can provide more accurate measurements when not using the fastest sampling rate. This is because when using the fastest sampling rate, the oscilloscope may produce additional interleaving distortion and increase high-frequency noise.

Therefore, oscilloscope manufacturers conduct internal evaluations of the individual ADCs they use, and also evaluate the overall ENOB of the oscilloscope system. The resulting system ENOB will be lower than the ENOB of the individual ADCs. Since the ADC is a built-in component of the oscilloscope and cannot be used alone, only evaluating the ENOB of the overall system is meaningful.

To measure ADC quality, offset/phase/frequency distortion must be taken into account

In order to maintain stable ADC performance, another important application of ENOB is to measure the quality of the oscilloscope ADC. If the oscilloscope has a good ENOB, its timing error, frequency spike (usually caused by interleaving distortion) and low broadband noise will be very small. If the product application is mainly based on sine waves, the most suitable oscilloscope can be selected based on ENOB.

Generally speaking, users will not use up the 8-bit resolution of the oscilloscope ADC at the same time. In order to fully utilize the 8-bit vertical measurement range, users must zoom in on the waveform in order to use the entire vertical measurement range, but this will increase the difficulty of observing the signal and may cause the risk of ADC saturation and produce adverse effects. Not only that, front-end noise, harmonic distortion, and interleaving distortion will also reduce the effectiveness of the oscilloscope ADC. Therefore, 90% of the vertical range must be used to measure the signal. At this time, the user should reduce the oscilloscope 8-bit converter to 7.2 bits (90%×8 bits).

However, ENOB is a method to measure the quality of ADC and front-end components, but it ignores several other parameters, including ENOB does not consider factors such as offset, phase inconsistency and frequency response distortion. As shown in Figure 2, the measurement results of an input signal on two different oscilloscopes, although the two oscilloscopes have the same ENOB value, it can be clearly seen that the waveform displayed by one oscilloscope is closer to the actual input signal.

Figure 2. Scope 1 and Scope 2 have the same ENOB, but Scope 2 cannot truly represent the input signal due to factors such as offset and phase distortion errors.

This phenomenon means that ENOB does not take into account the offset error that may be introduced by the oscilloscope, because if two oscilloscopes have the same ENOB, they may display the same waveform but have different absolute voltage offsets. A better evaluation indicator can be obtained by adjusting the offset and measuring the noise, or evaluating the DC gain specification.

Ideally, all oscilloscopes have flat phase and frequency response, and the same roll-off characteristics, so that customers can select the right oscilloscope. However, this is not the case in reality. Oscilloscope manufacturers' product specifications usually do not provide phase and frequency response graphs, and ENOB does not take into account factors such as frequency response flatness or phase inconsistency.

In addition, if only oscilloscopes with higher ENOB are considered, the displayed input signal may not be more accurate. Since the frequency response and phase inconsistency of different oscilloscope models are different, for example, when two oscilloscopes with the same 6GHz bandwidth are used to observe a 2.1GHz sine wave, different waveforms may be obtained. One oscilloscope may have a slower bandwidth reduction and only a slight phase correction, resulting in a poor result; the other oscilloscope may have a frequency response with a peak value of more than 6GHz before the frequency is reduced, and uses an algorithm that can effectively correct the phase, and performs relatively better.

Tips for improving ENOB by using filter/capture mode

Given the importance of ENOB to oscilloscopes, the most effective way to increase ENOB is not only to purchase an oscilloscope with a higher ENOB from the beginning, but also to ask the oscilloscope manufacturer to provide the overall ENOB value of each oscilloscope model.

In addition, since most high-end oscilloscopes provide user-selectable bandwidth limiting filters, turning on the filter can limit the oscilloscope bandwidth and suppress high-frequency content such as interleaving errors and noise, thereby improving the ENOB value.

In addition, the oscilloscope can also use averaging or high-resolution acquisition modes to further measure repetitive signals to reduce broadband noise, and using these modes can also effectively perform more accurate measurements.

ENOB/Noise Level Double Check Oscilloscope Selection

In fact, the user's consideration of ENOB mainly depends on the items they want to measure and whether ENOB will affect the measurement results. Therefore, when selecting an oscilloscope, both the ENOB diagram and the noise level measurement results must be considered.

For example, if the signals to be measured are primarily basic sine waves, such as in some defense applications, ENOB may be an excellent evaluation criterion. Ask the oscilloscope manufacturer to provide an ENOB graph for the oscilloscope model you are interested in. However, remember that ENOB values ​​change with frequency, so you need to know the ENOB value of the selected model over the full rated frequency range.

However, if at certain frequencies, high-speed serial digital signals have harmonic components, they may pass through the measurement system without being affected by the reduction of effective bits. In this case, the oscilloscope noise level can be a more effective measure of accuracy.