diff options
Diffstat (limited to '_sfp')
| -rw-r--r-- | _sfp/bosa-tosa-rosa.md | 4 | ||||
| -rw-r--r-- | _sfp/ont-wo-mac.md | 26 | ||||
| -rw-r--r-- | _sfp/sfp-standard.md | 14 |
3 files changed, 22 insertions, 22 deletions
diff --git a/_sfp/bosa-tosa-rosa.md b/_sfp/bosa-tosa-rosa.md index a893170..2558603 100644 --- a/_sfp/bosa-tosa-rosa.md +++ b/_sfp/bosa-tosa-rosa.md @@ -5,8 +5,8 @@ nav_order: 3 layout: default --- -In optical-electrical conversion, special components called TOSA (Transmitter Optical Sub Assembly) and ROSA (Receiver Optical Sub Assembly) are used to convert the signal. -They are responsible for translating the optical signal into a corresponding electrical signal or vice versa, which inputs or outputs symbols corresponding to the optical values. These values, which we refer to as unprocessed or RAW values for simplicity, are not standard signals and must be converted into standard signals[^huawei]. +In optical-electrical conversions, special components called TOSA (Transmitter Optical Sub Assembly) and ROSA (Receiver Optical Sub Assembly) are used to convert the signal. +They are responsible for translating the optical signal into a corresponding electrical signal and viceversa, which inputs or outputs symbols corresponding to the optical values. These values, which we refer to as unprocessed or RAW values for simplicity, are not standard signals and must be converted into standard signals[^huawei]. TOSA and ROSA are essential components in the uni-directional transceivers which transmit on one fiber optic strand and receive on the other fiber optic strand. diff --git a/_sfp/ont-wo-mac.md b/_sfp/ont-wo-mac.md index b1e6faf..06c9c72 100644 --- a/_sfp/ont-wo-mac.md +++ b/_sfp/ont-wo-mac.md @@ -5,22 +5,22 @@ nav_order: 2 layout: default --- -PON technologies unlike Ethernet are not P2P but one-to-many with two device types: ONU (Optical Network Unit)/ONT (Optical Network Terminal) and OLT (Optical Line Terminal). Both devices can be manufactured using the SFP form factor[^tibit]. +PON technologies, unlike Ethernet, are not P2P but one-to-many with two device types: ONU (Optical Network Unit)/ONT (Optical Network Terminal) and OLT (Optical Line Terminal). Both devices can be manufactured using the SFP form factor[^tibit]. -The OLT provides an integrated access box for Passive Optical Networks. OLTs are typically chassis with one or more line cards inside, and on each line card there is one or more PON transceiver, usually in SFP form factor. Each line card is connected to a secondary switch that provides line card aggregation to the Ethernet uplinks. OLTs are often a mixture of Layer 2 and Layer 3 switching with traffic shaping on a per-customer, per-service basis[^tibit]. +The OLT provides an integrated access box for Passive Optical Networks. OLTs are typically chassis with one or more line cards inside, and on each line card there is one or more PON transceiver, usually in the SFP form factor. Each line card is connected to a secondary switch that provides line card aggregation to the Ethernet uplinks. OLTs are often a mixture of Layer 2 and Layer 3 switching with traffic shaping on a per-customer, per-service basis[^tibit]. -The communication within the SFP PON transceiver is neither MMI nor Ethernet, outside [SFP standards](/sfp-standard.md), but rather it is an *equivalent electrical symbols of optical transmission* (which is simply the input/output of the [BOSA](/bosa-tosa-rosa.md)) that for simplicity's sake we call **PON RAW communication** (also referred to as SFP w/o PON MAC). All the PON management part is left to the line card itself. Each equivalent electrical symbol of optical transmission is a separate dialect, distinct from other dialects. Furthermore, as one can easily guess, this communication is not standard and is not within the signalling standards ([^sfprate],[^sfprate2],[^sfpplusstandard]) but it is compliant with some portions of the MSA [^sfpstandard],[^sfpplusstandard],[^sfpplusmi]. This requires extreme compatibility between ONT and transreciver. This design choice is made for several reasons: -- *size*: the size of an OLT w/o PON MAC is very similar to that of an MMI or Ethernet transceiver, and the size of an OLT with the integrated PON MAC far exceeds that of the standard SFP form factor -- *dissipative heating capacity*: the dissipative heating capacity of an OLT with PON MAC is higher than a normal transceiver, such as a 1 or 10 Gbps Ethernet link. -- *duplication*: there is a double `MAC` → `MMI` conversion (`MMI` → `MAC` → `PHY` → `MAC` → `OLT CPU`) -- *repairability*: since lasers often have a shorter lifetime than other ICs, it is good to be able to change only the transceiver +The communication within the SFP PON transceiver is neither MII nor Ethernet, outside [SFP standards](/sfp-standard.md), but rather it is an *equivalent electrical symbols of optical transmission* (which is simply the input/output of the [BOSA](/bosa-tosa-rosa.md)) that for simplicity's sake we call **PON RAW communication** (also referred to as SFP w/o PON MAC). All the PON management part is left to the line card itself. Each equivalent electrical symbol of optical transmission is a separate dialect, distinct from other dialects. Furthermore, as one can easily guess, this communication is not standard and is not within the signalling standards ([^sfprate],[^sfprate2],[^sfpplusstandard]) but it is compliant with some portions of the MSA [^sfpstandard],[^sfpplusstandard],[^sfpplusmi]. This requires extreme compatibility between ONT and transceiver. This design choice is made for several reasons: +- *size*: the size of an OLT w/o PON MAC is very similar to that of an MII or Ethernet transceiver, and the size of an OLT with the integrated PON MAC far exceeds that of the standard SFP form factor; +- *dissipative heating capacity*: the dissipative heating capacity of an OLT with PON MAC is higher than a normal transceiver, such as a 1 or 10 Gbps Ethernet link; +- *duplication*: there is a double `MAC` → `MII` conversion (`MII` → `MAC` → `PHY` → `MAC` → `OLT CPU`); +- *repairability*: since lasers often have a shorter lifetime than other ICs, it is good to be able to only change the transceiver. -Despite this, there is a vendor that sells OLT SFP with PON MAC[^tibit]. The following pictures show an OLT SFP with PON MAC part and a transreciver without PON MAC. It is interesting to see that the latter is much longer and requires an additional heatsink. +Despite this, there is a vendor that sells OLT SFPs with PON MAC[^tibit]. The following pictures show an OLT SFP with PON MAC part and a transreciver without PON MAC. It is interesting to see that the latter is much longer and requires an additional heatsink. {% include image.html file="ont-wo-mac/tibit.png" alt="PON OLT with MAC" caption="PON OLT with MAC" %} {% include image.html file="ont-wo-mac/huawei.png" alt="PON transceiver for OLT w/o MAC" caption="PON transceiver for OLT w/o MAC" %} -Similarly, the same argument can be made for OLT SFPs, especially in 10E-PON and XGS-PON there are a lot of transreceivers w/o PON MAC and few ONTs with PON MAC. In this case the reasons are similar to the previous ones. It is also clear that ONTs w/o PON MAC require a PON MAC part within the end device that supports the relevant communication protocol. +Similarly, the same argument can be made for ONT SFPs, especially in 10E-PON and XGS-PON there are a lot of transceivers w/o PON MAC and few ONTs with PON MAC. In this case, the reasons are similar to the OLT SFPs'. It is also clear that ONTs w/o PON MAC require a PON MAC part within the end device that supports the relevant communication protocol. The following pictures show some operating diagrams of some ONT with PON MAC and ONT w/o PON MAC[^SFPP-XGS-ONU-MAC-ASC-I-C],[^SFPP-XGS-ONU-N1-I-C],[^MSOG22-xD6C-xxT1]. @@ -40,7 +40,7 @@ graph TD end; subgraph CS[Cage SFP] E --> |I2C| Controller; - C --> |MMI on Tx - Rx| MAC; + C --> |MII on Tx - Rx| MAC; Controller[Controller I2C] --> Switch; MAC --> Switch; end; @@ -65,11 +65,11 @@ graph TD end; ``` -# Why are there no ONT w/o MAC on Hack GPON? +# Why are there no ONTs w/o MAC on Hack GPON? -For utility reasons all SFPs w/o PON MAC are not illustrated on Hack GPON as they are not modifiable like ONT with MAC (they require two inter-compatible devices). +For usefulness reasons, all SFPs w/o PON MAC are not illustrated on Hack GPON as they are not modifiable like ONTs with MAC (they require two inter-compatible devices). -In particular, the SFP ONU of the AVM Fritz!Box 5530/5590 belongs in this category, and that the above-mentioned devices are not compatible with any other SFP using MMI/Ethernet/Fibre Channel, while for example the FreeBox or IliadBox supports both ONU w/o PON MAC and some SFP with MAC. +In particular, the SFP ONU of the AVM Fritz!Box 5530/5590 belongs in this category, and that the above-mentioned devices are not compatible with any other SFP using MII/Ethernet/Fibre Channel, while for example the FreeBox or IliadBox supports both ONU w/o PON MAC and some SFP with MAC. In general, these devices do not have enough customisation to allow the required parameters to be changed other than the GPON Serial Number and GPON Ploam Password. This means that in most scenarios these devices with ONT w/o MAC are not flexible enough to be used as a replacement for an ISP-provided ONT. diff --git a/_sfp/sfp-standard.md b/_sfp/sfp-standard.md index 6825ace..892f772 100644 --- a/_sfp/sfp-standard.md +++ b/_sfp/sfp-standard.md @@ -7,22 +7,22 @@ layout: default The organisation that developed SFPs (MSA SFP) has always been very cautious about defining a hardened list of admissible signals for SFPs, their first standard only providing pinout, form-factor and dissipative capacity specifications. It is up to the manufacturer to decide which communication to use in the Tx and Rx pins[^sfpstandard]. -After the SFP standard entered the market, in the early 2000s Ethernet and Fibre Channel, the MSA SFP also started to standardise signalling, starting with [^sfprate] and [^sfprate2] which define a list of admissible standard signalling limited to the capabilities of the current form factor SFP. +After the SFP standard entered the market, in the early 2000s with Ethernet and Fibre Channel, the MSA SFP also started standardising signalling, starting with [^sfprate] and [^sfprate2] which define a list of admissible standard signalling limited to the capabilities of the current form factor SFP. With the need to increase the heat dissipation characteristics of the modules (in order to increase speeds) and to allow some additions to the EEPROM, an additional standard, called SFP+[^sfpplusstandard],[^sfpplusmi],[^xenpak_xfp], was developed, which contains all the aforementioned improvements. The 16GFC, 20GFC signalling for Fibre Channel and the 10 Gbps and 2.5 signalling for Ethernet were also included in the updated [^sfprate] and [^sfprate2] standard. Some of these are also included in [^sfpplusstandard] locking the SFP+ standard to a tenth of signalling, all other signals should fall under the SFP standard[^sfpstandard], but they can use the extended SFP+ management interface[^sfpplusmi]. -The Ethernet signals are all very similar, but there are some differences between Base-X and MMI. The media-independent interface (MMI) was defined in the IEEE 802.11u standard, was originally defined as a standard interface to connect a Fast Ethernet MAC block (i.e. CPU, switch) to a PHY chip (i.e. twisted pair, fiber optic, etc.) in a standardised way. The main advantage is that the MMI can be used without redesigning or replacing the MAC hardware. Thus any MAC may be used with any PHY, independent of the network signal transmission media[^ethernet]. +The Ethernet signals are all very similar, but there are some differences between Base-X and MII. The media-independent interface (MII) was defined in the IEEE 802.11u standard. It was originally defined as a standard interface to connect a Fast Ethernet MAC block (i.e. CPU, switch) to a PHY chip (i.e. twisted pair, fiber optic, etc.) in a standardised way. The main advantage is that MII can be used without redesigning or replacing the MAC hardware. Thus any MAC may be used with any PHY, independent of the network signal transmission media[^ethernet]. The main difference is the physical media over which the frames are: -- *Base-X* is based on the Ethernet PHYsical Layer (level 1) and this standard uses the 8B/10B coding (or other encodings as specified in the EEPROM), and *MMI* is based on the Ethernet MAC Device (level 2, the device that actually makes and receives Ethernet frames)[^ethernet]. -- In *Base-X*, auto-negotiation is limited to flow-control (and duplex, which is not really used since it's always full-duplex), and in *MII*, auto-negotiation (AN) also allows the PHY to indicate to the MAC the post-PHY link speed. Even though the MAC-to-PHY SGMII link is always 1000Mbps, it supports 10, 100 and 1000Mbps past the PHY and the MAC need to know this to space out the bits properly (e. g. if the external link is 100Mbps, each bit on the SGMII link is sent 10 times)[^ethernet]. +- *Base-X* is based on the Ethernet PHYsical Layer (layer 1) and this standard uses the 8B/10B coding (or other encodings as specified in the EEPROM), and *MII* is based on the Ethernet MAC Device (layer 2, the device that actually makes and receives Ethernet frames)[^ethernet]. +- In *Base-X*, auto-negotiation is limited to flow-control (and duplex, which is not really used since it's always full-duplex), and in *MII*, auto-negotiation (AN) also allows the PHY to indicate to the MAC the post-PHY link speed. Even though the MAC-to-PHY SGMII link is always 1000Mbps, it supports 10, 100 and 1000Mbps past the PHY and the MAC needs to know this to space out the bits properly (e. g. if the external link is 100Mbps, each bit on the SGMII link is sent 10 times)[^ethernet]. -The MII can be used to connect a MAC to an external PHY using a pluggable connector, or directly to a PHY chip on the same PCB. In the first case it is also used in SFP connectors, for example to allow connections between two MAC blocks without passing through a PHY (i.e. passive DAC). -This technology, and in particular its evolutions such as RGMII[^rgmii], SGMII[^sgmii], QSGMII[^qsgmii], XGMII[^intel], USXGMII[^xilinx], is widely used as a communication bus over SFP, in addition to IEEE BaseX[^ethernet]. The 2.5 SGMII or HSGMII[^altium] and 10 SGMII or XSGMII[^aquantia] standards are specifics that increase the clock speed of the SGMII standard without redefining it. +MII can be used to connect a MAC to an external PHY using a pluggable connector, or directly to a PHY chip on the same PCB. In the first case it is also used in SFP connectors, for example to allow connections between two MAC blocks without passing through a PHY (i.e. passive DAC). +This technology, and in particular its evolutions such as RGMII[^rgmii], SGMII[^sgmii], QSGMII[^qsgmii], XGMII[^intel], USXGMII[^xilinx], is widely used as a communication bus over SFP, in addition to IEEE BaseX[^ethernet]. The 2.5G-SGMII or HSGMII[^altium] and 10G-SGMII or XSGMII[^aquantia] standards are specifics that increase the clock speed of the SGMII standard without redefining it. RGMII, SGMII, 1000BaseX standards allow a speed of 1 Gbps, 2.5BaseX and HSGMII standards of 2.5 Gbps, the XGMII, XSGMII, USXGMII and 10GBaseT of 10 Gbps. --- -[^xenpak_xfp]: With the advent of higher speeds MSA has developed several new interfaces, such as XENPAK, X2, XPAK, XFP, but the newest standard is the transceiver is called SFP+. Based on the same form factor as SFP, it is smaller than its predecessors and has lower power than XFP. SFP+ has become the most popular socket on 10GE systems because it shares a common physical form factor with legacy SFP modules, allowing higher port density than XFP and the reuse of existing designs for 24 or 48 ports in a 19-inch rack width blade. +[^xenpak_xfp]: With the advent of higher speeds MSA has developed several new interfaces, such as XENPAK, X2, XPAK, XFP, but the newest standard is the transceiver is called SFP+. Based on the same form factor as SFP, it is smaller than its predecessors and has lower power than XFP. SFP+ has become the most popular socket on 10GbE systems because it shares a common physical form factor with legacy SFP modules, allowing higher port density than XFP and the reuse of existing designs for 24 or 48 ports in a 19-inch rack width blade. [^sfpstandard]: *Specification for SFP (Small Formfactor Pluggable) Transceiver* INF-8074 [^sfprate]: *SFP Rate and Application Selection* SFF-8079 [^sfprate2]: *SFP (Small Formfactor Pluggable) Rate and Application Codes* SFF-8089 |