A 14-bit oscilloscope digitizes each sample into 1 of 16,384 levels (2 to the power of 14), 64 times the resolution of an 8-bit scope's 256 levels. That is the job of a digital oscilloscope's converter: turn the analog signal into discrete levels while preserving the waveform details that matter. Waveform details smaller than one level may disappear into the record, which is why bit count helps determine what the instrument can resolve.

The math turns bit depth into an ideal step size quickly. Divide the full-scale range by the level count. For an ideal converter at a 5 V full scale, an 8-bit converter steps in about 19.5 mV increments, while a 14-bit converter steps in about 0.31 mV increments. A 2 mV ripple can sit inside a single ideal 8-bit step; at 14 bits, it spans about six levels. Treat this as an illustrative quantization calculation, not the smallest signal a given oscilloscope can resolve: real measurement capability also depends on vertical scale and offset, front-end noise, ENOB, bandwidth, and the probe setup, which the sections below cover.

Nominal bits set the ceiling, but ENOB and the noise floor show what the instrument can actually deliver. Noise and distortion reduce real-world resolution, which is why effective number of bits matters.

ENOB is a measured quantity, not just a nominal bit count. IEEE Std 1057-2017 covers terminology and test methods for digitizing waveform recorders, including sine-wave-fit methods, while IEEE Std 1241-2023 provides common terminology and test methods for evaluating analog-to-digital converters. When a datasheet quotes ENOB, the follow-up question that matters is: at which frequency was it measured?

The front end affects the resolution before the ADC sees the signal. A low-noise amplifier chain helps keep the delivered bits closer to the nominal bit count. Keysight's published support data makes the point with system-level numbers: general-purpose 8-bit InfiniiVision models measure system ENOB around 6.9, while the HD3 series measures system ENOB above 10.4 (Keysight support KB on oscilloscope ENOB). Pair that with the HD3's about 50 µVRMS noise floor (at 2 mV/div, 50 Ω, 1 GHz), and the 14 bits stay useful in practice, not just on the spec sheet.

Read any high-resolution oscilloscope datasheet with this checklist:

• ENOB at your working bandwidth, not just the headline value
• Noise floor at the V/div settings you actually use
• Native ADC bits versus software-enhanced mode bits
• Sample rate while the full resolution is active
• Input sensitivity range for your smallest signals

One more habit worth keeping: check how sampling rate is specified alongside resolution, because the 2 specs trade against each other in converter design.

Precision numbers you can verify beat adjectives: see the 14-bit InfiniiVision HD3 (HD304MSO) offer and request a quote.

Buy the bits your signals need: millivolt work often needs 14 bits, while logic timing usually does not. Here is a working map. Treat the values as application-dependent guidance rather than hard thresholds: the resolution a measurement needs also depends on vertical range and offset, bandwidth, front-end noise, probe attenuation, and the measurement uncertainty you need.

Two buying triggers cover many real cases. The first trigger is small signals riding large offsets. Measuring millivolt ripple on a 48 V or 400 V bus means the signal of interest occupies only a sliver of the input range, and a high-resolution front end leaves more levels for that detail. The second trigger is deep dynamic range in one capture. PSRR sweeps and regulation artifacts may sit 40 to 80 dB below the carrier or rail.

Power and EV examples show why this matters. Analog Devices' signal-chain power guidance (Perez and Pasaquian, Analog Dialogue, March 2021) puts it plainly: every high-speed signal chain has a maximum allowable rail ripple, established before the power tree is designed. Verifying millivolt-class limits on a 1 V rail is measurement work where an 8-bit front end may not provide enough resolution. In EV work, the signals of interest get even smaller relative to the rail: millivolt-level ripple and transients on battery-management supply rails need to be seen and correlated with bus activity, which is where high vertical resolution and a low noise floor pay off.

If your work sits in the top half of the table, you may be able to keep the budget lower and skip the premium tier. For automotive-leaning work, the automotive oscilloscopes glossary frames where scopes fit in vehicle labs.

Not sure which row you live in? Browse the certified used inventory and request a quote with your measurement list.

Many modern scopes offer resolution-enhancement modes, and they can be useful once you understand the trade-offs. Hi-res, or boxcar, filtering averages adjacent samples inside one acquisition. It adds apparent bits but costs bandwidth. Averaging combines repeated acquisitions. It adds apparent bits but requires a repetitive, stable signal. Neither mode restores a single-shot transient that the ADC did not resolve.

The trade-off is well documented in converter research. R. H. Walden's IEEE survey observed resolution falling roughly 1 bit for every doubling of sample rate, and Boris Murmann's public ADC Performance Survey (1997-2026) tracks the same frontier today. Hi-res and averaging modes approach that limit from the other side: they add apparent bits by giving up bandwidth or requiring repetition. In the HD3, a native 14-bit ADC at 3.2 GSa/s is a deliberate architecture choice. It preserves native bit depthon each acquisition across its specified instrument bandwidth, instead of relying on filtering or repeated captures to add apparent resolution. Its 15- and 16-bit high-resolution settings, by contrast, use bandwidth limiting. Other 14-bit architectures make their own trade-offs, so check the bandwidth at which a datasheet specifies full native resolution.

Resolution-enhancement modes can be the right answer when the signal is slow, repetitive, or bandwidth-tolerant and the scope is already on the bench. The averaging glossary entry covers the math. When the problem is a one-time transient with millivolt detail, native hardware resolution matters most.

Seven checks should shape the purchase: ENOB, noise floor, rate, memory, channels, software, documentation.

• ENOB at your bandwidth: ask for the measured value and test frequency.
• Noise floor at your V/div: check µVRMS at the sensitive settings you will actually use.
• Sample rate at full resolution: confirm the instrument maintains its native bit depth at the sample rate you need.
• Memory depth: long captures at full rate need enough points. The HD3 includes 20 Mpts standard memory, with 50 or 100 Mpts options.
• Channels: 4 analog plus 16 digital channels can cover rail-plus-bus correlation without a second instrument.
• Software licenses: confirm what is included versus added, including power analysis (HD300PWRA), automotive decode (HD300AUTA), aerospace decode (HD300AERA), embedded protocols (HD300EMBA), or the HD300BDLA bundle.
• Documentation: calibration certificate and warranty options on the specific unit; one avoidable failure costs more than the instrument gap (78% of customers lose more than $25k a day when equipment fails). Quality-system-driven labs: ask for Keysight Premium Used units.

On certified used listings, configuration is part of the purchase decision. Can't compare units side by side? Request the configuration sheet and calibration documentation for the exact unit before you commit. Our team can confirm installed options against your requirements so the instrument that arrives matches the plan. The broader used oscilloscope buying guide and the 15-features checklist cover the general-purpose criteria this list extends.

Fit by buyer: if you defend purchases in design reviews, the HD3 gives you one defensible instrument with documented calibration and warranty options. If you run a research lab, check current education and research discount terms, and keep the documentation ready for grant or audit requirements.

Before you order, run the pre-order check:

• Bandwidth license: HD304MSO bandwidth is set by license: 200 MHz standard, with 350 MHz, 500 MHz, and 1 GHz upgrade options. Ask which bandwidth the specific unit carries and what the upgrade path costs before you decide the tier. Request a quote on the unit, and our team can confirm the installed configuration.
• Memory option: confirm whether the unit includes 20 Mpts standard memory or 50/100 Mpts options for long full-rate captures.
• Software: confirm which HD300-series licenses are installed versus quoted as add-ons.
• WaveGen: confirm whether the 100 MHz WaveGen through the HD3WAVEGEN license option is enabled on the specific unit, especially if frequency response analysis matters to your workflow.
• Documentation: confirm that the calibration certificate is current and warranty options are selected. Accuracy and precision claims depend on that paperwork.

Match the tier to your signals: see the InfiniiVision HD3 (HD304MSO) offer and request a quote.
 

The purchase price is only one part of the ownership math. As of October 2026, certified used direct from Keysight runs 40 to 90% below new list price, plus 5% for buying online and 15% for education and research customers. Certified used units can include calibration documentation and up to 5 years of warranty options. Keysight Premium Used instruments pass a final 101-point quality check, and KeysightCare support can back the bench.

Downtime belongs in the comparison. Keysight reports that 78% of customers lose more than $25,000 a day when equipment fails, which is why warranty options and support belong in the cost discussion, not outside it. If usage is short-term, weigh the buy vs rent comparison. When procurement wants as-new paperwork, the Keysight Premium Used line may be the better path. Research labs can check the education and research offers page for current discount terms.

Fourteen bits can make millivolt-level problems measurable, so you can act on them. Three bench cases show where the extra resolution matters.

Power integrity. A 3.3 V rail with 2 mV of switching ripple can look flat on an 8-bit scope because the ripple sits inside a single 19.5 mV step at a 5 V full-scale setting. At 14 bits, the same ripple spans about six levels, so its shape, frequency, and load dependence are easier to measure. PSRR verification pushes further. Measuring 60 dB of rejection means resolving an output residual 1,000 times smaller than the injected disturbance, which is where dynamic range, not just bit depth, decides the measurement.

Automotive and EV. Millivolt ripple riding a 48 V rail or a 400 V bus is a small signal on a large offset. Some battery cell monitors, such as the Texas Instruments BQ79718-Q1, specify ±1 mV cell-voltage accuracy, but that is an absolute DC accuracy figure, a different characteristic from oscilloscope vertical resolution. The scope's job is to show the millivolt-level ripple and transients on those rails and correlate rail behavior with bus activity, which is where the HD3's high resolution and low noise floor are directly relevant. On the HD304MSO, HD300AUTA decodes the serial communication protocols (CAN/CAN FD/CAN XL/LIN/SENT) on the same 4+16 channels, so the rail event and the bus reaction can land in one capture.
Aerospace and defense. DO-160 Section 16 power-input testing includes fast transients and longer interruptions, run as separate test procedures. Fourteen bits across a deep memory record of up to 100 Mpts supports both captures, and HD300AERA decodes ARINC 429 and MIL-STD-1553 beside the rails. For MIL-STD-461 pre-screening, the FFT plus the HD3's low noise floor can help surface wideband conducted-emission candidates before accredited lab testing.

For the measurement mechanics behind these cases, related guides go deeper: the bode plot oscilloscope guide covers loop and PSRR sweeps, the DO-160 pre-compliance guide covers Section 16/17 screening, and how an oscilloscope works covers the fundamentals.

Why Do High-Resolution Measurements Still Go Wrong?
Most failed high-resolution measurements spend the bits in the setup, not the instrument. The most common causes:

• A 10:1 passive probe attenuates the signal toward the noise floor. For power-rail work, PDN measurement practice often favors terminated coax connections over standard probes (Istvan Novak's PDN probing guidance)
• V/div set too coarse: the signal uses only a small part of the ADC range, which wastes available levels
• Bandwidth-limit filters left off, which digitizes noise the measurement does not need
• Ground loops and long ground leads adding artifacts that the ADC records as part of the signal
• Outdated calibration making microvolt readings harder to defend

The fixes cost little: right-size the probe (the refurbished oscilloscope probes guide covers the options), fill the screen with the signal, engage the 20 MHz limit when the measurement allows, and keep the paperwork current per the oscilloscope calibration guide and the certificate of calibration explainer.
Building the full bench? See the used MSOX4154A listing for the value tier and request a quote.

Three takeaways belong in the purchase memo. Judge instruments by ENOB and noise floor, not just the bit count on the box. Let the application map guide your tier, because 8 bits still earns its keep on logic and general debug. Use the certified used path to put native 14-bit resolution on the bench well below list price, with up to 5 years of warranty options.

Browse the certified used inventory and request a quote.

Match the tier to your highest frequency and your modulation needs; Keysight's current signal generators run from 9 kHz to 54 GHz natively, with frequency extenders reaching past 100 GHz. Every row below is verified against the official Keysight product pages.

The certified used instrument we highlight is the PSG E8267D vector signal generator (100 kHz to 44 GHz, up to 80 MHz modulation bandwidth): the microwave vector source for radar, electronic warfare, and satellite test benches, with good availability certified used now, including Keysight Premium Used units. On the RF tier, note an important lifecycle fact: Keysight's product page announces that the MXG N5182B is to be discontinued, naming the N5186A as its successor. That makes certified used the practical way to put MXG-class phase noise (-146 dBc/Hz typical at 1 GHz, 20 kHz offset, with Option UNX; check installed options per listing) on a bench at a fraction of a current-generation price; confirm calibration support for your planning horizon when you request a quote. For high-volume 6 GHz work on a tighter budget, the EXG pair (N5171B analog, N5172B vector) covers manufacturing test in a rack-friendly form factor, and the vector signal generators category page shows the wider certified-used lineup.

Signal sources can age well when maintained: performance depends heavily on the synthesizer design, and periodic calibration helps keep the instrument accurate over time. That makes the used-equipment question mostly a provenance question.

Buying certified used equipment direct from Keysight removes most of the usual used-market risk. Keysight Premium Used units are refurbished at the factory that builds new instruments, inspected against a 101-point checklist, fully calibrated, and pass the same final test as a new product. Keysight Used units ship in working condition with calibration as listed, at deeper savings. Both tiers offer up to 5 years of warranty options. Calibration is one detail to insist on for any signal source, because a generator with drifted level accuracy or degraded spectral purity fails quietly: your measurements move, and nothing on the front panel says why. Confirm the calibration status and installed options per listing before you order.

The buying decision comes down to four questions. What is the highest frequency your test plan needs? Do you need analog purity or vector modulation? What phase noise can your most sensitive measurement tolerate? For arbitrary waveforms, how much usable bandwidth and play time does the sample-rate math give you? Answer those questions, and the model table above should narrow the choice quickly: Trueform for the bench, X-Series for RF, PSG or VXG for microwave.

Certified used from Keysight brings these instrument families within reach of tighter budgets, with warranty options and Keysight calibration behind them. Browse the certified used signal generators and request a quote for the configuration that fits.