Pendulum oscillator production testclock module testmulti-channel frequency counterCNT-104S
Oscillator & Clock Module Production Test
Parallel multi-channel counting, gap-free statistics and data automation: measure frequency, jitter and start-up behavior of every oscillator in a single fixture cycle.

In short
In oscillator and clock module production test the bottleneck is usually takt time, not accuracy. Every part needs its frequency accuracy, start-up time and jitter measured, and rotating several single-channel counters and aligning the data afterwards holds up neither the takt nor the confidence in the result. The Pendulum CNT-104S is the core of this setup: four channels measure in parallel and gap-free, all four signals share one timebase, and 20 million results per second are reduced to statistics inside the instrument. Where budget is tight or two channels are enough, the CNT-102 fits. The CNT-91/91R with TimeView software plots warm-up drift and PLL lock as frequency against time. Before committing, a short rental lets you prove the line takt first.
The real bottleneck on the line: takt time and data alignment
Outgoing inspection of oscillators, TCXO/OCXO modules and clock buffers has to answer at least three questions per part: is the frequency error within spec, how long after power-up does it settle, and how large is the period jitter once settled. Each measurement is easy on its own. The hard part is doing all three within the production takt, with the results from four outputs (or four parts) landing on the same time axis.
When several single-channel counters take turns, each one's trigger instant and timebase drift is an independent error source, and aligning the data afterwards only folds the instrument-to-instrument differences into the yield decision. What the line needs is one instrument that watches several signals at once, with every reading sharing the same timebase by construction.
Four parallel, gap-free channels: the CNT-104S architecture

The four input channels of the Pendulum CNT-104S record in parallel with no gaps, with better than 7 ps time resolution, 13 digits per second of frequency resolution, 20 million measurements per second, and a 50 ns per-channel sample interval in block measurement mode. One fixture cycle can capture the frequency sequences of four parts (or one four-output module) at the same time, and the mean, standard deviation and extremes are computed inside the instrument, so millions of samples never need to move to a PC first.
The standard channels cover up to 400 MHz. For GHz-class clocks or RF oscillators, add the Channel C RF input with a 3 GHz or 10 GHz hardware base; the 10 GHz version can be upgraded by software to 24 GHz. The timebase is a TCXO or OCXO option. The accuracy ceiling of a production pass/fail decision is set by the timebase, so if the plant already has a 10 MHz reference, lock to it externally.
When two channels are enough: the CNT-102 trade-off
Not every line needs four parallel inputs. The CNT-102 keeps the same generation's measurement architecture with two independent parallel channels, 14 ps time resolution, the same 13 digits per second of frequency resolution, an internal measurement rate of 1 million per second and 170,000 per second in block mode. It suits part-by-part testing of single-output oscillators, or using the second channel on a reference signal for phase or delay comparison.
Confirm the channel count is really sufficient before you choose: moving from two channels to four later means a new instrument, not an upgrade. If the same station has to test a four-output clock distributor, choosing the CNT-104S from the start saves trouble.
Seeing how frequency changes with time: TimeView and the CNT-91

Production decisions look at numbers; development and failure analysis look at waveforms. TimeView software with a CNT-90, CNT-90XL, CNT-91 or CNT-91R turns the counter into a modulation domain analyzer that plots frequency, period or phase against time. Oscillator warm-up drift, PLL lock overshoot and frequency modulation from supply noise, none of which are visible in a list of numbers, appear directly as curves. Free-running capture runs at up to 250,000 samples per second, and repetitive sampling reaches an equivalent 100 MSa/s.
The CNT-91 has 35 ps single-shot time resolution and internal memory for up to 3.5 million samples, so jitter and time-interval distribution statistics complete on the instrument. The 91R version has a built-in rubidium timebase for sites without an external reference.
How to introduce it: rent to prove the takt, then choose the model
The benefit of a multi-channel counter only shows under real line takt. Evencat offers short- and long-term rental across the Pendulum line, so you can run a lot through a CNT-104S or CNT-102 during pilot production, confirm station throughput, data export format and the SCPI automation flow, and then decide whether to buy or extend the rental. As the authorized distributor in Taiwan we can speak directly to calibration documents and lead time.
Instruments for this job




Further reading
- Frequency Counter Selection Guide: Resolution, Channels and Time Base Digits/s, single-shot resolution, channel count, microwave input and time base: five specifications that decide which counter you need
- Time Interval and Jitter Measurement: TIE, Period Jitter and the Resolution Floor Jitter is a clock edge that does not arrive at the instant it should. To measure it, first understand how a counter measures time
- Laboratory Time and Frequency References: GNSS-disciplined Rubidium and Standalone Sources One 10 MHz and one 1 pps make every instrument in the laboratory tell the same time, and here is what to do when there is no satellite signal
Image credits: Pendulum Instruments (Manufacturer material, used with distributor authorization)
Need an instrument set for this job? Tell us the quantity to measure, the uncertainty you are aiming for and the conditions on site.
