A typical component enquiry reads: two hundred thousand infrared emitters, two hundred thousand photodiode receivers, quote both. The two lines are treated as independent purchases, and each is optimized on its own — the brightest emitter available, the most sensitive receiver available. The assembled unit then behaves worse than either datasheet suggests, because the two devices were never specified against each other.
Jiangmen City Yiyang Infrared Semi-conductor Co., Ltd (Yiyang Optoelectronics YY) has produced infrared emitters, photodiode receivers and matched emitter–receiver pairs in Jiangmen, Guangdong for over 13 years, at 600K pieces per day, exporting to more than 40 countries. A large share of the customization work that comes through the factory is not a single-device request at all. It is a request to make one device fit another: adjust the power split between the two chips in an emitter, widen or narrow the emission angle, change the active area of the receiver, or shift a wavelength bin so that the two curves overlap where they should. The individual figures are covered in the companion articles on emitter specification and photodiode receiver specification; this article is about what changes when the two devices are specified against each other.
What follows covers the three variables that bind a pair together, what each one constrains, and how to decide which side of the pair to change.
The Pair Is the Component, Not the Two Parts
An optical link has a budget. The emitter puts a certain optical power into a certain solid angle; the geometry spreads that power over a distance; the receiver collects it over its active area and converts it to current. Every term in that chain is a property of one device, but the result is a property of the combination. Two examples make the point.
| Situation | What a component-by-component approach produces | What pair-level specification produces |
|---|---|---|
| Emitter changed to a higher-output bin | Receiver amplifier saturates at short range; AGC hunts | Output bin chosen against the receiver's linear range, not maximized |
| Receiver changed to a larger active area for more signal | Higher dark current and capacitance, slower response, more ambient pickup | Active area and emitter angle sized so the beam fills the area without overspill |
The pair is also a mechanical unit. In a reflective sensor or a probe, the emitter and receiver sit at a fixed baseline and a fixed angle. That geometry, not the datasheet, decides how much of the emitted cone lands on the active area. A pair specified with the geometry included can be delivered as a matched set; a pair specified as two separate line items cannot. Yiyang Infrared supplies dual-wavelength DIP emitters, two-pin SMD patch emitters and matching receivers, and can quote them as a set.
1. Red to Infrared Power Ratio
In a pulse oximeter, the emitter package holds two chips: a visible red at roughly 660nm and a near-infrared at roughly 905nm or 940nm. The receiver measures the transmitted light at both wavelengths and the ratio between the two signals carries the information. That ratio only works if both channels arrive at the receiver in a usable range.
The two chips are not equally efficient. A red AlGaInP chip and a near-infrared AlGaAs chip produce different radiant flux for the same forward current, and they age at different rates. A design that drives both at 20mA and does nothing else will find one channel much stronger than the other. The power ratio is the customization that corrects for this at the source.
What the ratio actually sets
- Dynamic range usage. If the red channel is five times stronger than the infrared, the analog front end must either attenuate red or amplify infrared, and the weaker channel ends up with fewer effective bits. Matching the delivered power at the factory recovers that range.
- Drive symmetry. When the two channels are close in output, one drive current can serve both, which simplifies the LED driver circuit and removes a per-unit calibration.
- Ratio stability over life. The two chip materials degrade at different rates under the same stress. Starting from a deliberately offset ratio can leave room for the drift.
How the ratio is specified
The ratio should be stated as a target with a tolerance, at a defined forward current and temperature, and with a defined measurement geometry. "Red:infrared = 1:1.2 ±15% at 20mA per chip, measured on our fixture" is a specification. "Balanced output" is not. The 660nm chip typically runs at a forward voltage of 2.0–2.4V and the 905nm chip at 1.3–1.5V, so the two channels do not share a common drive voltage either — the ratio and the drive circuit have to be designed against each other. The 660/905 dual-wavelength two-pin transmitter is the standard version of this package.
| Ratio item | Adjustable on the emitter side | Adjustable on the receiver side |
|---|---|---|
| Relative output of the two chips | Chip selection and binning; drive current per chip; internal reflector geometry | Not directly — the receiver sees what arrives |
| Channel-to-channel balance at the amplifier | Yes — the source of the correction | Gain per channel in the AFE, but that is a board change |
| Ratio tolerance across a batch | Yes — set by chip sorting and test limits | Receives the consequence |
The practical conclusion is that the ratio belongs on the emitter order. It is the only place it can be corrected without touching the PCB. Where the ratio change also requires a different lead form, that is a separate drawing item — see lead length, pitch and epoxy customization — and the ordering and packaging guide covers how the pair is supplied and scheduled.
2. Emission Angle Against Active Area
Emission angle and active area are two ends of the same calculation. The emitter radiates into a cone; the receiver collects over an area at some distance. Whether the cone underfills, fills or overfills that area determines both the signal level and the sensitivity to misalignment.
| Condition | Signal | Alignment tolerance | Ambient pickup |
|---|---|---|---|
| Cone much narrower than the active area (underfill) | Strong on axis, but loses power that misses the area | Poor — small misalignment loses a large fraction | Low |
| Cone matches the active area (fill) | Good, and stable | Moderate to good | Moderate |
| Cone much wider than the active area (overfill) | Weaker, most power lands outside the area | Good — the spot moves within the area | Higher, because the wide cone also collects more stray light |
The half-angle convention matters here. A viewing angle quoted as 2θ½ describes the full cone at half intensity, so a "20°" emitter has a half-angle of 10° from the axis. At a separation of 8mm, a 10° half-angle illuminates a spot roughly 2.8mm across; at 3mm separation the same emitter covers about 1.1mm. A receiver with a 1.2mm active area is well matched at 3mm and badly underfilled at 8mm — the same two parts, a different distance, a different result.
This is why both sides are customizable in the same conversation. A narrower emission angle concentrates power for a long, narrow path; a larger active area collects more of a wide cone but costs dark current and speed. The right adjustment depends on the distance the design has to work over, and on whether the mechanical alignment is tight or loose. A 660/905 two-pin patch emitter is normally paired with a receiver whose active area is chosen for the probe's geometry, not for maximum sensitivity. The SMD patch receiver is the form used in disposable and reusable probes; its active area is specified against the probe geometry, not maximized.
Active area as a customization
The photosensitive chip size can be specified on a photodiode receiver. Three consequences follow from making it larger:
- More signal per unit irradiance — the collected current scales with area.
- More dark current — leakage also scales with area, which raises the noise floor and matters most at low light levels.
- More capacitance, slower response — a larger junction is a larger capacitor, and the RC limit moves down.
For a pulse oximeter, the signal is strong and the modulation rate is the pulse rate, so a moderate area is a reasonable trade. For a beam-break sensor with a fast edge requirement, a small area with a narrow emitter cone is the better pairing. The customization question is therefore not "how large can it be" but "what area matches this emitter at this distance with this response requirement".
3. Wavelength Bin Alignment
An emitter's output is a band with a peak; a receiver's response is a curve. The two overlap over a region, and the useful signal comes from that overlap. Sorting each device to a bin does not automatically align the two bins, which is the point most often missed.
A receiver with a spectral response from 840nm to 1100nm responds across the whole near-infrared range, so for a 940nm emitter the alignment looks trivial. It is not, for two reasons. First, the receiver's responsivity is not flat across that band — it peaks somewhere in the middle and falls off toward both ends, so where the emitter sits on that curve changes the delivered current per watt. Second, the emitter's peak shifts with temperature and current, typically by a few tenths of a nanometre per degree. A receiver chosen because it covers 940nm may be operating on the shoulder of its response curve rather than near the peak, and the temperature drift then moves the system further down the shoulder.
| Pairing item | Why alignment matters | How it is specified |
|---|---|---|
| Emitter peak versus receiver peak responsivity | Delivered current per watt depends on where the emitter sits on the response curve | Peak wavelength of each, with bin tolerance |
| Bin width | A wide bin admits parts whose peak differs by enough to change the ratio | Bin limits, e.g. ±5nm or ±10nm |
| Temperature drift | Peak moves with junction temperature; a narrow-band system can walk off the band | Operating temperature range plus the drift figure |
| Filter passband (black epoxy) | A filtered receiver passes a band, not the whole spectrum; the emitter must sit inside it | Passband limits, not just a nominal wavelength |
For a dual-wavelength pair the alignment problem doubles, because two peaks must both land usefully on one response curve, and the ratio between them must stay inside tolerance. Sorting the emitter into tighter bins is the usual answer, and it is a factory operation rather than a design change.
4. Black Filtered Epoxy: Where Filtering Enters the Pair
The black epoxy front end on a receiver is an optical filter integrated into the package. It passes the working band and blocks visible light. In a pulse oximeter probe the receiver sits against skin under ambient light, and an unfiltered photodiode would convert that ambient illumination into a large DC offset that the amplifier must then reject.
Filtering is a pair-level decision because it changes the effective sensitivity: a black filtered receiver delivers less current than a clear one for the same incident infrared, and the emitter's output must be sized against the filtered figure rather than the unfiltered one. The common failure is specifying a receiver from a clear-body datasheet and then ordering the black version for the field, which quietly removes a share of the link budget. The bridge-type receiver head is an example of the black filtered form, and the 5mm black receiver tube is the through-hole equivalent.
The same logic applies to emitters with a blue infrared-transmitting body: the tint suppresses part of the visible band, which is useful where the emitter's faint red glow is undesirable, but it also changes the delivered output. Every body colour option on either side of the pair is an optical element, and the pair budget has to be calculated with the body colour applied.
5. Matched Pairs as a Delivered Set
Once the ratio, the angle and the bins are settled, the remaining question is consistency. A pair that is correct on paper but drifts unit to unit during production is not a matched pair; it is two components with a shared specification.
Consistency comes from three factory operations, and it is worth asking which of them apply to a quoted pair:
- Chip sorting before packaging. Emitter chips sorted into output and wavelength bins; receiver chips sorted for responsivity and dark current. Without this, the final test rejects rather than the incoming sort.
- Test at the delivered condition. Photocurrent, dark current and spectral response measured on finished parts, not on wafers — packaging shifts the optical result.
- Sampling from the same production run. Pairs drawn from one batch share chip lot, epoxy lot and process conditions, which is what makes the ratio hold across a shipment.
Yiyang Infrared runs final inspection for photocurrent, dark current and spectral response on finished components in its own packaging workshop, and can supply matching emitting and receiving components as a complete set on request. For a probe programme, ordering the pair as a set is usually cheaper than reconciling two independently binned shipments. The 660nm/905nm optical pair guide covers how the pair and the SMT process interact on the probe board.
Which Side of the Pair Should Be Adjusted
Customization requests are frequently aimed at the wrong device. A decision sequence that avoids the round trip:
| Symptom | Likely cause | Adjust on |
|---|---|---|
| One channel much weaker than the other | Chip efficiency mismatch, not a receiver problem | Emitter — power ratio |
| Signal adequate but varies unit to unit | Bin spread, or alignment tolerance in the assembly | Both — tighter bins plus a wider angle or larger area |
| Amplifier saturates at short range | Too much delivered power at the near end | Emitter — lower bin, or narrower drive range |
| Response too slow for the modulation rate | Receiver active area too large | Receiver — smaller area, with a narrower emitter angle to compensate |
| Poor performance under bright ambient light | Unfiltered receiver body | Receiver — black filtered epoxy, then re-check the emitter budget |
| Works at 20°C, degrades when warm | Wavelength drift off the filter or response shoulder | Emitter — tighter bin; receiver — wider passband |
What to Send for a Pair Quotation
- Application and separation distance — the geometry that sets how much of the cone reaches the area.
- Wavelengths required, and whether they are fixed by a standard or by the measurement principle.
- Target power ratio with tolerance, and the drive current it is defined at.
- Receiver active area preference, or the response time that constrains it.
- Body colour for each device — clear, blue-transmitting or black filtered — and whether filtering is required by the environment.
- Assembly method — through-hole or SMT — since it fixes the package family on both sides.
- Annual quantity and whether a matched set is wanted, because set supply changes how the lots are drawn.
- Existing design figures if the pair is a replacement, so the new parts can be aimed at the same operating point.
A Pair Specified Together Saves a Calibration
Power ratio, emission angle against active area, and wavelength bin alignment are three ways of saying the same thing: the emitter and the receiver form one optical system, and its behaviour is decided at the point where the two are specified. Correcting a mismatch afterwards means a per-unit calibration, an attenuator, or a board change — all more expensive than a binning instruction on the purchase order.
Yiyang Infrared customizes red and infrared power ratio, emitting angle, package size and special wavelength options on the emitter side, and photosensitive area, spectral response range and body colour on the receiver side, with matched emitter–receiver sets available as a complete pair. Send the geometry, the required ratio and the ambient conditions, and the pair can be specified against each other rather than in parallel.
Related reading: the role of optoelectronic components in intelligent sensing, the supplier evaluation checklist, and the product knowledge archive. For company information see about Yiyang Infrared; for purchasing questions, the FAQ and contact page.
