A GNSS receiver front end does not choose its reference oscillator by frequency. The frequency was fixed the day the baseband was picked. What is left to decide is what the board can hand the oscillator. A rail with a tolerance, a patch of copper under the shield can, and a receiver chip that either squares the clock itself or expects to steer it.
The Epson X1G0042010062 is a 32 MHz TCXO in a 3.2 mm × 2.5 mm × 0.9 mm body. On the sheet it carries the model name TG-5006CE-39H. It runs from 2.85 V to 3.15 V, holds ±0.5 × 10⁻⁶ over −30 °C to +85 °C, drifts ±1.0 × 10⁻⁶ in the first year, draws 1.5 mA maximum and starts within 2.0 ms. The output is a clipped sine wave.
Four of the six constraints are cleared without argument: frequency, board area, current and drift. The other two are where it goes marginal. The rail is a ±5 % window around 3.0 V, so a board that only offers 3.3 V puts this part above its maximum before anything else is considered. And there is no voltage-control pin, so a receiver that steers its reference with a DAC cannot use this part at all. Those two sentences decide most designs, and both are visible only after you read past the feature list.
The constraints that actually decide this
Six constraints survive contact with a real GNSS board, and only one is about accuracy at room temperature.
The first is the reference the baseband expects. A 32 MHz input is not programmable on this part. There is no divider and no select pin. If the receiver wants 26 MHz, the conversation ends here.
The second is the supply the board can actually provide. The recommended range is 2.85 V to 3.15 V, a 3.0 V nominal with ±5 % at both ends. It will not hang off a 1.8 V logic rail without a boost.
The third is the error budget the receiver vendor hands you, usually a parts-per-million allowance the reference may fill.
The fourth is production spread, how far the part sits from nominal after two reflow passes.
The fifth is current on a rail that never sleeps. A receiver that loses lock pays in reacquisition time, so the oscillator runs for the whole mission.
The sixth is the output format the receiver input expects. This part delivers a clipped sine wave, not a CMOS square.
| Constraint | Datasheet figure | Where |
|---|---|---|
| Reference frequency the baseband expects | 32 MHz, fixed | p. 2 |
| Rail the board can supply | 2.85 V to 3.15 V | p. 2 |
| Error budget contribution over temperature | ±0.5 × 10⁻⁶, −30 °C to +85 °C | p. 2 |
| Spread after two reflows | ±2.0 × 10⁻⁶ after two reflows | p. 2 |
| First-year drift | ±1.0 × 10⁻⁶ per year | p. 2 |
| Current on the always-on rail | 1.5 mA max | p. 3 |
| Time from power-up to a usable clock | 2.0 ms max | p. 3 |
| Board area under the shield can | 3.2 mm × 2.5 mm × 0.9 mm | p. 2 |
| Output the receiver input must accept | Clipped sine, 0.8 V to 1.5 V peak to peak | p. 3 |
| Output load the part is specified into | 10 kΩ // 10 pF | p. 2 |
Mapping each constraint to a datasheet number
Take the rail first, because it is the constraint that kills the most designs. The recommended operating conditions give Vcc as 2.85 V minimum, 3 V typical and 3.15 V maximum. A 3.3 V rail is above that ceiling, and decoupling does not fix a DC overshoot. It needs a 3.0 V rail that holds ±5 % across load steps and temperature. The frequency/voltage coefficient is ±0.2 × 10⁻⁶ for a Vcc change of ±5 %, so the supply does not have to be perfect, but it does have to stay inside the window.
The stability chain is where the arithmetic happens. Initial tolerance before reflow is ±1.0 × 10⁻⁶, and after two reflow cycles it becomes ±2.0 × 10⁻⁶, because the sheet folds the reflow shift into that second figure. Temperature adds ±0.5 × 10⁻⁶ across −30 °C to +85 °C. Load adds ±0.2 × 10⁻⁶ for a 10 kΩ // 10 pF load that varies by ±10 %. Supply adds ±0.2 × 10⁻⁶. Aging adds ±1.0 × 10⁻⁶ in the first year. Add the worst case and the part sits a few parts per million off nominal before the receiver sees it. That is normal for a plain TCXO, and it is why receivers that care either calibrate against the network or steer the reference.
Which brings back the missing pin. The pinout is four pads: N.C., GND, OUT and Vcc. A receiver that compensates its own reference by writing a DAC has nowhere to write. If your baseband datasheet has an AFC output and a pull-range requirement, this part is the wrong architecture, not the wrong grade. The TG-5006CJ parts in the same family carry the same four pads and the same 2.85 V to 3.15 V rail, so switching siblings does not buy you a control pin either.
The output format is the second architectural question. A clipped sine wave at 0.8 V to 1.5 V peak to peak into 10 kΩ // 10 pF needs a DC-cut capacitor of 0.01 µF at the OUT pin. That capacitor is not inside the package, and the sheet says so twice. On the receiver side, the reference input has to contain a squaring stage. Most GNSS front ends do. If yours expects a square clock, you are adding a comparator, and that comparator contributes jitter that no TCXO specification will cover.
ESD is the one dimension the sheet leaves open. There is no HBM or CDM number anywhere, only the instruction to use standard precautions through handling, assembly, test and shipment. Plan protection as if the part had no rating, because in this document it does not.
Current is the easy one. The maximum is 1.5 mA at 25 °C into the specified load, while the 38.4 MHz sibling X1G0041310011 is rated at 2.0 mA, so the number moves with frequency. On a receiver that runs continuously, that is a fixed line in the power budget, not a duty-cycled one.
Start-up is 2.0 ms maximum, defined as reaching 90 % of final amplitude. On a receiver that stays powered this does not matter; on a duty-cycled one, it sets the floor on recovery.
Where the part stops being the right answer
Three boundaries are worth writing down before layout release.
The first is the temperature range. −30 °C to +85 °C is the only operating range on the sheet, and storage is −40 °C to +90 °C. There is no −40 °C option and no +105 °C option. An outdoor enclosure in a cold climate will spend time below −30 °C, and a sealed box in sun will push the footprint above +85 °C. Neither is a fault of the part; both put it outside its specified window, where the drift number above no longer applies.
The second is phase noise. The electrical characteristics table prints harmonics at −8.0 dBc and then stops, with no L(f) row and no dBc/Hz figure at any offset. That is not the same as a noisy part, but a phase-noise mask from your receiver vendor cannot be closed with this document alone. The TG-5006CJ siblings do print the table: X1G0041310006 lists −85 dBc/Hz at 10 Hz offset through −148 dBc/Hz at 10 kHz and beyond, on the same 2.85 V to 3.15 V rail. Epson specifies the parameter when it wants to. Here it is absent, so plan a measurement.
The third is board area, and it goes the other way. At 3.2 mm × 2.5 mm this is the larger of the two bodies Epson builds to this family. The 2016 parts, at 2.0 mm × 1.6 mm × 0.73 mm, cover the same ground with 0.5 × 10⁻⁶ temperature stability, the same 1.5 mA ceiling and the same clipped sine output. X1G0041310009 does it at 19.2 MHz, X1G0041310010 at 26 MHz and X1G0041310011 at 38.4 MHz, all with ±1.0 × 10⁻⁶ initial tolerance. None of them is drop-in either: the land pattern changes with the body. So if the only reason for the larger body is that 32 MHz was on a list, check whether the smaller one carries the frequency your baseband needs.
Two of those 2016 siblings are a trap. X1G0041310013 and X1G0041310014 are also 16.367667 MHz and 16.368 MHz parts and look identical in the headline, but their frequency/temperature characteristic is ±2.0 × 10⁻⁶, four times wider than the ±0.5 × 10⁻⁶ above, and their initial tolerance is ±0.8 × 10⁻⁶. Same body, same rail, same current, a different stability class. That difference shows up in the row, not on the first page.
Bring-up checklist
- Confirm the rail is 3.0 V, not 3.3 V. Measure at the part, not the regulator output, and include the receiver's load step.
- Decode the ordering code before you release the BOM. The product number is X1G0042010062 followed by a two-digit suffix, and the sheet names 00 as the recommended code. It also lists 01, 11 and 14 against different packing conditions, so the suffix decides what arrives.
- Fit the bypass capacitor between Vcc and GND, 0.01 µF to 0.1 µF, as close to the part as the layout allows. The datasheet asks for it twice.
- Add the 0.01 µF DC-cut capacitor at the OUT pin. It is not in the package, and without it the DC level at the receiver input is whatever the oscillator sits at.
- Leave pin 1 unconnected. The sheet says not to connect the N.C. pin to any other lead, including to another N.C. pin.
- Keep the metal lid clear. It is connected to GND through pin 2, so anything conductive resting on the cap is a short to ground.
- Keep signal, supply and ground traces off the area under and opposite the part, including inner layers.
- Bring the supply up monotonically and do not reapply power until the bypass and bulk capacitors have discharged. The handling section calls this out as a condition for correct start-up.
- Verify peak-to-peak at the receiver input after the DC-cut capacitor, against the 0.8 V minimum. The oscillator pin tells you nothing about what the receiver sees.
- Time the first fix after a cold start against the 2.0 ms figure.
Frequently asked questions
Does the 3225 body save board area?
It is the larger of the two bodies in this family, so the honest answer is no. The 2016 parts are 2.0 mm × 1.6 mm × 0.73 mm and carry the same 0.5 × 10⁻⁶ temperature stability, the same 1.5 mA maximum and the same clipped sine output. The larger body is easier to place and inspect, and keeps more copper between the pads near an antenna feed. Pick it for assembly margin, not area.
Is −30 °C to +85 °C enough for an outdoor receiver?
That depends on the enclosure, and the number to trust is footprint temperature, not air temperature. A sealed box in sun runs well above ambient, and a mast-mounted unit in a cold climate spends real time below −30 °C. There is no wider grade in this family, so if either end is in your operating profile, this is the wrong choice rather than a marginal one.
Do I need external components around it?
Two, and both are specified. A bypass capacitor of 0.01 µF to 0.1 µF between Vcc and GND, mounted as close as the layout allows, and a 0.01 µF DC-cut capacitor at the OUT pin. The load the part is specified into is 10 kΩ // 10 pF, and the DC-cut capacitor is not inside the package. There are no load capacitors to tune, because the compensation is internal.
How do I order it for volume, and what about supply over a multi-year build?
The orderable number is X1G0042010062 plus the two-digit suffix, with 00 named as the recommended code on the packing page. The sheet lists 1000 pieces per reel, 2000 pieces per reel, any quantity per reel and a tape-cut vinyl bag as the conditions behind the four codes, so suffix and packing are decided together. Reels follow EIA-481 and IEC-60286. On longevity, 32 MHz sits inside a family already covering 16 MHz to 38.4 MHz, which is the useful signal on paper, but a lifecycle and PCN check still belongs before a multi-year build.
Sourcing note
The X1G0042010062 ships through Mandu with the full Epson specification sheet. Send us your rail voltage, your measured footprint temperature and your annual quantity, and we will confirm which suffix and packing code your order needs before you freeze the layout. If your receiver steers its reference from a DAC, say so in the first message, because this family has no control pin.