Rectifier waveform troubleshooting for ADC data

Match the symptom to the first check

These symptoms suggest where to look. They do not prove a particular fault.

• Alternate humps appear to be missing: check the rectifier output, diode orientation and measurement reference. Also check whether samples are missing downstream.
• Peaks have flat tops: check for ADC input clipping before changing graph limits.
• Values are far outside the ADC code range: inspect byte decoding, signedness and packet alignment.
• The trace contains mostly zeros or one repeated value: check the input, then look for binary data being parsed as text or incomplete serial reads.
• The waveform has occasional spikes or sudden jumps: inspect framing, buffer overruns and dropped samples.
• The shape looks reasonable but its period is wrong: verify the sampling clock, missing-sample handling and horizontal-axis units.
• The shape and period are correct but the voltage is inaccurate: check reference voltage, divider ratio and calibration.

Work in signal order. Changing several settings together makes it harder to identify the cause.

Prove the voltage before opening the graph settings

First establish what circuit you are measuring. Do not connect a mains-referenced rectifier directly to an ADC or a computer interface. An isolated low-voltage supply does not make every possible ground connection harmless.

Confirm isolation and grounding from the circuit documentation before attaching test leads. An earth-referenced oscilloscope ground or a grounded computer connection can short part of a bridge rectifier if attached to the wrong node. Do not assume the bridge negative terminal and either AC terminal are interchangeable references. Use an appropriately rated differential or isolated measurement arrangement where required. Never disconnect protective earth to float equipment.

If the isolation or instrument ratings are uncertain, stop and arrange an electrical assessment.

With a suitable measurement setup, compare the signal at the rectifier output with the signal presented to the ADC input:

• An unfiltered full-wave rectifier produces successive positive humps, approaching zero around each AC zero crossing.
• A bridge with a reservoir capacitor typically produces a DC level with ripple rather than separate humps returning to zero.
• Half-wave operation leaves a much longer interval between conducting portions of the waveform.

For a periodic AC input, full-wave ripple repeats at twice the input frequency. A 50 Hz input therefore produces 100 Hz ripple; a 60 Hz input produces 120 Hz ripple.

Record the input frequency, rectified peak, minimum voltage and measurement reference. State which node each reading describes. A peak-to-peak value alone does not identify the peak voltage applied to the ADC.

Check the ADC input range and conditioning

Read the converter documentation for its permitted input range, reference voltage and absolute limits. The conversion range and the damage limits are different specifications. Staying below a damage limit does not establish that a measurement is valid.

For a unipolar ADC, verify that the signal remains within its permitted positive range and does not make prohibited negative excursions. Inspect the waveform at the converter pin, after any divider, filter or buffer. A correct signal at the bridge does not prove that the converter receives it correctly.

If a resistive divider is used, its ideal ratio is:

V_ADC = V_rectifier × R_bottom / (R_top + R_bottom)

Choose the ratio from the highest credible input voltage, component tolerances and reference variation, not just one measured peak. Check the converter's source-impedance and acquisition-time requirements as well. An excessively high divider resistance can prevent the sampling input from settling correctly.

An unsigned 8-bit result has 256 possible codes, from 0 through 255. Repeated top-code readings suggest clipping, but can also result from a decoding fault. Compare them with the voltage measured at the ADC pin before deciding which applies.

Verify sampling and the graph time base

Confirm the conversion rate actually used by the ADC. A configured rate, a serial read rate and a graph refresh rate are three different things.

At 62.5 ksample/s, equally spaced conversions are 16 microseconds apart because the sample interval is 1 / sample rate. That rate is ample for resolving the basic shape of 100 Hz or 120 Hz ripple, provided those samples reach the application intact. It does not by itself establish adequate bandwidth for switching edges or other fast components. Those require appropriate input filtering and sampling design.

For uniformly sampled data, assign time using:

t[n] = t_start + n / sample_rate

Use that interval in the waveform data or build an explicit time array. Do not label sample indices as milliseconds without applying the conversion.

Serial packets often arrive in bursts. Their arrival times are not the original conversion times. If samples can be lost, use documented sample counters or acquisition timestamps where available. Joining the remaining samples with an unchanged interval can conceal gaps and distort the apparent period.

Inspect the raw serial traffic before converting it

Capture the received bytes before any numeric conversion. Use both a hexadecimal view and, when useful, a text view. Compare the capture with the transmitting firmware's documented format.

A steady, known input makes this check easier. If safe for the circuit, disconnect the rectifier with power removed and substitute a suitable low-voltage source within the ADC's permitted range. Confirm its reference connection before powering the measurement chain.

Determine whether the stream contains:

• Raw binary samples.
• Decimal or hexadecimal text.
• Headers, lengths, channel identifiers or sample counters.
• Checksums or other packet checks.
• Line endings or another defined delimiter.

For example, hexadecimal bytes 32 30 30 0D 0A represent the text value 200 followed by a carriage return and line feed. They are not five independent voltage samples.

Check baud rate, data bits, parity and stop bits against the transmitter configuration. Apply termination-character handling only when it matches the protocol. In arbitrary binary data, a byte that resembles a line ending may be a valid sample.

Check throughput and decode complete records

Calculate whether the link can carry the configured acquisition rate. With eight data bits, no parity and one stop bit, each transmitted byte normally occupies ten serial bits, including its start bit.

If each 8-bit sample occupies one byte, 62,500 samples per second therefore require at least 625,000 serial bits per second before packet overhead. Text encoding requires more capacity because a single sample can occupy several characters. Also allow for headers and other protocol fields.

A faster converter does not necessarily produce a faster delivered stream. The transmitter may buffer, decimate or drop samples. Establish which behavior the firmware implements rather than inferring it from the graph.

Then make the decoding match the format:

• For documented one-byte unsigned samples, extract each payload byte as one value from 0 through 255.
• For text samples, retain incomplete text between reads and parse only complete records.
• For multibyte values, follow the documented byte order, signedness and sample width.
• For framed packets, validate the header, length and any supplied integrity check before accepting the payload.

One serial read is not necessarily one packet. A read may return half a record or several records together. Keep a receive buffer and consume complete records from it.

Do not reinterpret several independent sample bytes as one 32-bit integer. That changes both the values and the number of plotted points. Likewise, a decimal text parser is not a decoder for arbitrary binary bytes.

Apply voltage scaling only after the codes make sense

Once the raw codes follow a known input predictably, convert them to engineering units using the ADC's documented transfer function.

For an ideal unipolar 8-bit converter, the nominal code width is often expressed as:

LSB = V_reference / 256

Use the actual datasheet convention for endpoint handling and offset. If a divider precedes the ADC, refer the measured voltage back to the rectifier by applying the inverse divider ratio.

Check several known input levels across the usable range. A single point cannot distinguish an offset error from a gain error. Make sure the reference measurement is sufficiently accurate for the result you need.

Calibration belongs here, after the waveform shape, timing and decoding are correct. It cannot recover information lost through clipping, aliasing or dropped bytes.

Confirm the repair from input to saved data

Reconnect the rectifier only after the acquisition chain behaves correctly with the known input. Repeat the electrical checks and capture several complete ripple periods.

Confirm that:

• The ADC input remains within its permitted range.
• The plotted period matches the measured input and expected rectification frequency.
• Raw codes and scaled voltages agree at selected points.
• Sample counts and packet checks show no unexplained losses.
• Saving and reopening the capture preserves values, timing and units.

Keep the raw capture, decoded samples, acquisition settings and a simple wiring diagram together. That record separates a demonstrated correction from a graph that merely looks plausible.

If saved samples are correct but a production screen still shows older values, investigate factory screen update delay separately. Measurement accuracy and presentation freshness are different checks.

Questions