Reference
How far SDI goes, and who actually decides
No SMPTE standard states a maximum cable length for SDI. What they state is how much signal loss a receiver has to cope with — and the distance is that budget divided by the cable's attenuation at one-half the clock frequency.
This is why you can find two entirely reputable distance charts that disagree about the same cable by a factor of two. Belden 1694A at 3G-SDI is published as 78 metres and as 155 metres, and both are right: one assumes a receiver that copes with 20 dB of loss, the other one that copes with 40. The cable did not change. The assumption about the equaliser at the far end did.
So this page does not reprint a distance chart. It gives you the two inputs — published attenuation, and the loss budget — and computes the answer for both budgets, so the number comes with the assumption attached. The computation is checked at build time against three figures published in the wild, including both of the contradictory ones above.
The formats
Note the rates. "12G-SDI" runs at 11.88 Gb/s and "6G-SDI" at 5.94 — the names are rounded up. Every one of these also has a variant divided by 1.001, for the 59.94-family frame rates, and equipment is expected to handle both.
| Format | Standard | Year | Serial rate | Half clock | Carries |
|---|---|---|---|---|---|
| SD-SDI | SMPTE ST 259 | 1997 | 0.270 Gb/s | 135 MHz | 480i59.94, 576i50 |
| HD-SDI | SMPTE ST 292 | 1998 | 1.49 Gb/s | 742.5 MHz | 720p, 1080i |
| 3G-SDI | SMPTE ST 424 | 2006 | 2.97 Gb/s | 1485 MHz | 1080p50/60 |
| 6G-SDI | SMPTE ST 2081 | 2015 | 5.94 Gb/s | 2970 MHz | 2160p30, 1080p120 |
| 12G-SDI | SMPTE ST 2082 | 2015 | 11.88 Gb/s | 5940 MHz | 2160p60 |
| 24G-SDI | SMPTE ST 2083 | 2020 | 23.76 Gb/s | 11880 MHz | 2160p120, 4320p30 |
- SD-SDI — The 270 Mb/s variant is the one anybody means. ST 259 also defines 143 and 177 Mb/s composite rates that have effectively vanished.
- HD-SDI — The rate is 1.485 Gb/s, or 1.485/1.001 for the 59.94-family frame rates. Both are "HD-SDI" and equipment must handle either.
- 3G-SDI — Two mappings, and they are not interchangeable: Level A carries 1080p directly, Level B carries two multiplexed HD streams. A device that speaks only one will show nothing on the other.
- 6G-SDI — Named for a rate it does not have. 5.94 Gb/s, rounded up in the marketing to 6.
- 12G-SDI — The rate is 11.88 Gb/s. This is where a single coax stops being comfortable and the receiver equaliser starts deciding your infrastructure for you.
- 24G-SDI — Specified, and rare. Most cable datasheets do not publish an attenuation figure at 11.88 GHz, which is its own kind of answer.
Why half the clock frequency
SDI is scrambled NRZI. The scrambling guarantees transitions so the receiver can recover a clock without a separate one, and the consequence is that the signal'shighest fundamental frequency is half the bit rate — the alternating-bit pattern, one transition per two bits. Everything above that is harmonics.
So the loss that matters is the loss at that frequency, and that is the point every cable datasheet and every SMPTE receiver specification is written against. A 12G link is an eye diagram at 5.94 GHz, and coax attenuation goes as roughly the square root of frequency, which is why doubling the rate does not halve the distance — it costs rather less than that.
The failure mode is a cliff, not a fade. An SDI link is error-free, then briefly sparkly, then gone, over a very short additional distance. There is no gradual softening to warn you, which is exactly why designing to the last metre of a published chart is a bad trade for the few pounds of better cable it saves.
How far, on which cable
Each cell gives the reach at a 20 dB budget and at 40 dB. The lower figure is what a cautious design assumes; the upper is what ST 2082-1 calls typical for a modern receiver. Your actual receiver is somewhere between them and its datasheet is the only thing that will tell you where.
| cable ↓ format → | SD-SDI | HD-SDI | 3G-SDI | 6G-SDI | 12G-SDI | 24G-SDI |
|---|---|---|---|---|---|---|
| 1855AMini RG-59, 23 AWG | 159–318m3.84 dB/100ft at 135 MHz | 65–131m9.31 dB/100ft at 750 MHz | 46–91m13.38 dB/100ft at 1500 MHz | 31–62m19.65 dB/100ft at 3000 MHz | 21–42m28.97 dB/100ft at 6000 MHz | not publishedno attenuation figure at this frequency |
| 1694ARG-6, 18 AWG | 272–544m2.24 dB/100ft at 135 MHz | 113–226m5.4 dB/100ft at 750 MHz | 78–156m7.8 dB/100ft at 1500 MHz | 53–106m11.5 dB/100ft at 3000 MHz | 34–69m17.7 dB/100ft at 6000 MHz | not publishedno attenuation figure at this frequency |
| 4694RRG-6, 18 AWG, 4K UHD | 290–581m2.1 dB/100ft at 135 MHz | 120–239m5.1 dB/100ft at 750 MHz | 84–167m7.3 dB/100ft at 1500 MHz | 58–115m10.6 dB/100ft at 3000 MHz | 39–79m15.5 dB/100ft at 6000 MHz | 27–53m23 dB/100ft at 12000 MHz |
Attenuation is taken at the nearest published datasheet frequency at or above the half-clock point — 750 MHz for a 742.5 MHz signal, 1500 for 1485, 3000 for 2970, 6000 for 5940. That rounds against the result, so these figures are marginally pessimistic, which is the correct direction to be wrong in.
SD-SDI is the one format conventionally budgeted differently. The figure long associated with ST 259 is 30 dB rather than 20 or 40, which lands between the two columns — around 408 m on 1694A. The 20 dB column is correspondingly over-cautious for standard definition, and the 40 dB column optimistic. Everything above SD is read straight off the band.
Three published figures this model reproduces
The build fails if any of these stops matching, which is what makes the claim that the charts differ only in their budget checkable rather than merely assertable:
- 1694A at 3G-SDI, 20 dB — 78 m. Matches the 78 m widely quoted for 1694A at 3G.
- 1694A at 3G-SDI, 40 dB — 156 m. Matches Belden's own 155 m for 1694A at 3G.
- 4694R at 12G-SDI, 40 dB — 79 m. Matches the 257 ft (78 m) published for 4694R at 12G.
The first two are the same cable and the same format, quoted at half and at twice the distance by different sources. Neither is wrong; they answer different questions, and only one of them says which question it answered.
Work it out for your own link
The budget is the input, not a constant — that is the whole point. Set it to what your receiver's datasheet claims, or sweep it to see how much of your run depends on that one number.
This calculator needs JavaScript. The table above covers the same cables at 20 and 40 dB.
Cable only. Connectors, bulkheads and patch panels all add loss and none of them are in this figure — nor is any knowledge of the receiver you actually own. Attenuation is taken at the nearest published frequency at or above the half-clock point, so the answer leans slightly pessimistic.
What the standard actually says about receivers
SMPTE ST 2082-1:2023 §6.2, in full: "Receivers operating with input cable losses in the range of up to 40 dB at one-half the clock frequency are typical; however, receivers designed to work with greater or lesser signal attenuation are acceptable."
That is not a requirement. It is a description of the market, with an explicit statement that conforming equipment may be better or worse. A receiver that manages 25 dB is compliant; so is one that manages 45. Neither has to say so on the front panel.
Belden's own note on their distance table makes the practical consequence explicit: "reputable manufacturers aren't designing all their 12 GHz receivers to reach the maximum loss level, which could result in distance reductions of up to 50%." A 79 metre figure for 4694R at 12G is therefore a 40 metre figure on some real equipment, and nothing in the chart tells you which you have.
The things that reach site
- Pathological signals are real and they are worst-case by design.Certain video content scrambles into long runs with heavy DC content — the SDI "checkfield" or pathological test pattern exercises exactly this. A link that passes colour bars all day can fail on a particular frame of real content, because bars are a deliberately non-stressing signal. ST 2082-1 says so in as many words in its jitter note. If a link is marginal, test it with a pathological pattern, not with bars.
- Level A and Level B 3G are not the same signal. Both are 2.97 Gb/s and both are "3G-SDI". Level A carries 1080p as one stream; Level B multiplexes two HD streams. Equipment that speaks only one shows black on the other, and the fault looks like a dead cable.
- Reclockers and equalisers are not the same thing either. An equaliser compensates cable loss; a reclocker regenerates the timing. A distribution amplifier that does both resets your entire loss and jitter budget and is the cheapest way to double a run. One that only buffers does not.
- A 12G link will often carry 3G down cable that is wrong for 12G, which makes commissioning misleading. Test at the format you intend to run, at the length you intend to run it, with the equipment you intend to use.
Sources
- 12G-SDI electrical specification, receiver loss budget and jitter — SMPTE ST 2082-1:2023, published openly by SMPTE and read directly. §6.2 is quoted verbatim above; the 11.88 Gb/s rate and the 75 Ω generator and receiver impedances are from §6.1 and §6.2.
- Cable attenuation — Belden technical datasheets, read directly: 1855A, 1694A and 4694R. Every figure in the reach table is computed from these.
- The loss-budget method, and the 50% receiver caveat — Belden, "Going the Distance with Serial Digital Coax", which states that their table "is based on attenuation or signal loss at one-half the signal frequency as defined by SMPTE", that the values are "for cables only", and that receiver design can cost up to half the distance.
- Format rates and standards numbers — SMPTE ST 259, ST 292, ST 424, ST 2081, ST 2082 and ST 2083. Only ST 2082-1 was read directly; the others are paywalled and their rates are cross-checked between secondary sources.
- The 20 dB budget is widely associated with ST 292 and ST 424 receivers and is used here as the conservative case. It has not been verified against those standards, which are paywalled — unlike the 40 dB figure, which is quoted from ST 2082-1 §6.2 above.
Assembled 7 September 2026 with AI assistance. The reach figures arecomputed, not measured: cable attenuation from published datasheets, divided into a loss budget, with no allowance for connectors, bulkheads or patch panels, and no knowledge of the receiver you actually own. They are a design starting point and an explanation of why published charts disagree — not a promise about a specific link. Test the real path at the real format before you trust a number on this page.
Companion pages: Ethernet cabling and XLR, AES3 and DMX.