General:
Impact of Notching and Tone-Mapping on the Frame

_______________________________________________

Question:
The influence of the notching and tone-mapping mechanisms on the frame are not explicitly described. To which part of the frames are these mechanisms applied?

Answer:
Notching is used for preamble, FCH and data. The notched tones are not used.
Tone-mapping only applies to data. The tonemap allows the receiver to adaptively inform the transmitter which tones it should use to send data and which tones it should use to send dummy data bits that the receiver will ignore.

Frame Control Header: Frame Length Field ([PHY] §5.5)

_______________________________________________

Question:
The Frame Length field needs clarification.
The specification states: "The frame length bit field gives the number of symbols in the frame based on the formula: Number of Symbols = FL*4"
The exact definition of "Number of Symbols" is not given. What does it correspond to?

Answer:
The number of symbols is defined as the total number of OFDM symbols for the data we transmit.

Frame Control Header: CRC5 ([PHY] §5.5)

_______________________________________________

Question:
The CRC5 is not completely defined. In particular, the following elements are missing:

- Initialization of the shift register.
- Computation of the final result.
- CRC-5 field endianness.
- Number of data bits for CRC-5 computation (34 bits looks wrong).

Answer:
The shift register is initialized with "0". The final computation of the crc5 is as crc5 = crc5 XOR 0x1F; it is complemented to get the final result. If the CRC5 is computed as "10011", the FCCS field will be filled as "0011-1". The number of data bits for CRC5 computation is 28 bits (34 bits is the FCH length, including PDC-DT plus ConvZeros).

Frame Control Header: Use of the Scrambler ([PHY] §5.5)

_______________________________________________

Question:
In the specification, the FCH is not scrambled. Therefore, it doesn't give the FCH the pseudo-random distribution needed for good transmission. Has it been forgotten in the specification?

Answer:
FCH is also scrambled using the same algorithm.

Convolutional Encoder ([PHY] §5.7.2)

_______________________________________________

Question:
The order of the X and Y bits in the output binary stream is not defined. What is the order?

Answer:
The output of the encoder is XY: X goes first.

Repetition coding: Encoding ([PHY] §5.7.3 and 5.7.4)

_______________________________________________

Question:
The repetition process is not completely defined. In particular, the operation to perform the encoding process is not described. How is the repetition done?

Answer:
The repetition is done bit-by-bit. If the input bit stream is 0123, the output should be 0000111122223333 (assuming repetition 4).

Interleaver: Padding ([PHY] §5.8)

_______________________________________________

Question:
At the interleaver input, if needed, the data must be padded to completely fill the interleaving matrix. This padding is not described. Which data should be used?

Answer:
The padding uses "0" to fill the matrix.

Interleaver: Usage for Multiple Bits per Symbol Modulations ([PHY] §5.8)

_______________________________________________

Question:
There is a dependency between the interleaving matrix and the modulation used. How does the interleaver work when using a modulation with several bits per symbol?

Answer:
In this example of how to perform the multiple bits per symbol modulation, assume 2 bits per symbol and 3 subcarriers. The input binary stream is "123456789". Bits are being read "123456789" into the interleaver matrix to be interleaved (? is the padding bit). The interleaver matrix is shown below (the interleaving has not been done to simplify the example):

1

2

3

4

5

6

7

8

9

?

?

?

After completing interleaving, bits are being read as 12, 34, 56, and so on, to be placed into the tones.

DBPSK/DQPSK Mapping ([PHY] §5.9)

_______________________________________________

Question:
The mapping uses differential modulation, meaning that each symbol uses the previous one as a phase reference. So the first symbol needs a fixed reference. What is the reference for the first symbol?

Answer:
The first symbol of FCH uses the phase of the last preamble as reference. The first symbol of data uses the phase of the last FCH symbol as reference.

Tone-Mapping ([PHY] §5.13 and 5.5)

_______________________________________________

Question:
The TM field (used in the FCH) is a 9-bit field and represents the used carriers. Each bit makes it possible to enable/disable 6 carriers. G3-PLC can use up to 36 carriers, between 35 kHz and 90 kHz. The mapping is not explained. Two interpretations are possible:

- The 9-bit field represents all the 54 carriers existing in the CENELEC A band. Given that, as the G3-PLC doesn't use the carriers below 35 kHz, the first 3 bits of the tone-map are unused (fixed to 0).
- The 9 bits are mapped to the 36 G3 carriers, and each bit controls 4 carriers and not 6.
Which one is the right one?

Answer:
The first 6 bits (LSB) in the 9-bit field are used and each of the 6 bits represents one sub-band (6 carriers). So, the total is 36 carriers. The last 3 bits in the 9-bit field are default to 0 (not used).

Segment Control

_______________________________________________

Question:
In the [MAC] specification, a "Segment Control" header is added before the normal [802.15.4] MAC header (MHR).
The content of this header is not completely defined, several fields are only named, without unit or content description, and the integration in the normal MAC header processing is not defined. It impacts in particular the security, as the MAX header is protected by a MIC.
Does the MAC integrity cover the segment control to protect it from modification?
The fields that are not completely defined are SC and SL and relate to frame segmentation. This will be treated in Q013.

Answer:
Segment Control is included in MHR as indicated in Table 2.1 on page 21 of the G3 specification and is protected by MAC security.

MAC Segmentation

_______________________________________________

Question:
The frame segmentation performed by the MAC layer is not defined at all.
The segmentation process description must include:

- Description of the segmentation fields used (LSF, SC and SL) in content, unit and limit values.
- Description of the fixed and segmented part of the frame.
- Example of an exchange of segmented frames.
- Definition of the timing and time-out needed for the segmentation process.
- Definition of minimum buffer and number of parallel segmentation process allowed.
- Information that will be used to discriminate segmented frames from different devices.
- Retry mechanisms must be described for "per segment" and "per frame" retry (if present).
- With the current version of the specification, it is not possible to implement an inter-operable MAC segmentation.

Answer:
Definition of the fields:

- Last Segment Flag (LSF): Boolean, value as defined in [MAC], set to indicate the last segment in a transmission.
- Segment Count (SC): Integer, 6 bits. Set to zero for the first fragment and incremented for each following fragment.
- Segment Length (SL): Integer, 10 bits. The length of the MAC frame, in bytes.

If segmentation is needed to transmit a MAC frame, the segments will be created as follows:

- Fixed part: The MAC header (MHR, from Segment Control to Auxiliary Security Header) and FCS are present in each segment.
- Variable part: The MAC payload is segmented in the data block. Each data block is processed as a normal MAC payload for CRC or security.

Segments are created by the concatenation of the fixed part and one data block.
To simplify the reassembly process and allow out-of-order transmission and retransmission, every data block, except the last one, have the same size.
The size used is the maximum input data that can fit in a PHY frame (variable value, depends on modulation and carriers used) with the MHR and FCS size removed.
The Sequence Number, Destination Address and Source Address fields act together as an identifier for the transmission, in order to separate segments sent from different devices. Retransmitted packets are distinguished using the same Sequence Number.
The normal acknowledge mechanism is used for retransmission of lost segments. The number of retry is macMaxFrameRetries.
A mechanism to acknowledge the complete MAC frame transmission is not needed as every segment has been acknowledged.
There is no specific timeout for de-segmentation defined in the standard. It's up to the implementation. However, an arbitrary multiply of aEIFS can be used for the timeout of each segment as default.
The minimum requirement for a device is to be able to perform one de-segmentation at a time.

Acknowledgement Mechanism

_______________________________________________

Question:
The Acknowledgement mechanism is not completely defined.
In an acknowledge frame, the PHY fields PDC and TM are used to host a value that links the data frame to the acknowledge frame. This value is defined as the MAC CRC16 present in the FCS field.

Questions:

- The byte order to store the CRC16 in those two fields is not defined. What is the used endianness?
- The TM field is 9 bits long so 1 bit has to be unused. Which one is it — first, last? What is its value?

Answer:
The MAC CRC16 is packed in the PHY header as follows: For a CRC16 value of 0xABCD, the PDC field value is 0xCD and the TM[0:7] is 0xAB, TM[8] is forced to value 0.

Tone-Mapping Configuration

_______________________________________________

Question:
In the §5.13 of [PHY], the algorithm used to choose the optimal PHY parameters is not described. This may be needed to obtain a homogeneous PLC network.
There is a need to describe how the following parameters are chosen:

- Modulation
- Disabled carriers
- Gain for each 10 kHz band

If these are considered implementation details, at least a general algorithm must be defined and key values must be outlined:

- Minimal SNR limit for each modulation.
- Minimal SNR limit for carriers deactivation.
- Limits for TXGAIN and TXCOEF to avoid breaking the CENELEC A band usage specification.

Note that the figure 12 is incorrect, as the tone-map response is sent inside a MAC command frame (as defined in [MAC] §2.3.1).

Answer:
The choice of modulation mode and disabled carriers are decided by the channel estimation algorithm which is implemented at the receiver. There are many ways to achieve that. The specific choice is out of the scope of the G3 specification.
Second, the limits for TXGAIN and TXCOEF to avoid breaking the CENELEC A band usage specification should be computed off-line based on CENELEC A band specification EN50065 to make sure the transmitter power will not exceed the spectrum regulation.
Third, thank you for noticing this. Yes, Figure 12 is not right. It will be removed from the PHY specification and we will also add a link to the [MAC] specification §2.3.1 on this point. We will let you know once these changes are made.

Minimal Number of Usable Carriers

_______________________________________________

Question:
The minimal number of active carriers is not defined. Below which number of available carriers do we consider a link as unusable?

Answer:
This is decided by the receiver implementation and this value should be programmable. The specific value should be determined based on the optimization of system performance. So, the specific value definition is out of the scope of the G3 specification.

TXCOEF

_______________________________________________

Question:
Several questions regarding TXCOEF amplification:

- The TXCOEF parameter covers a part of the CENELEC A band that is not used by the G3 specification (between 10 and 30 kHz). Do we need to force a value of 0 for those coefficients or do we just ignore them in the tone-map response?
- The CENELEC A band is parted in 9 sub-bands for the tone-map selective carrier disabling, but in 8 sub-bands for the TXCOEFF feature. Is this an error in the specification?
- Does the requirement for a flat spectrum emission from §6.6 of [MAC] apply when TXCOEFF feature is used?

Answer:
The TXCOEF parameter covers a part of the band that is not used by the G3 specification (between 10 and 30 kHz). We can just set the value to be 0 for those parts.
Second, please note that CENELEC A band is partitioned into 6 sub-bands and those 6 sub-bands locate between 36 and 90 kHz. However, TXCOEFF covers bands from 10-90 kHz. These two are different.
Third, when TXCOEFF is used, we still need to make sure the transmitter emission spectrum is below the spectrum mask defined by regulation. However, the transmitter output spectrum may not be flat.

Phase detection mechanism

_______________________________________________

Question:
Regarding the AC phase detection mechanism, what is the frequency that the 8-bit counter is actuially counting?

Answer:
The 8-bit counter is counting from 0 to 255 in one period of the mains. So, the counter ticks are 20/255 ms for 50 Hz main and 16/255 ms for 60 Hz.

General byte ordering

_______________________________________________

Question:
What is the byte ordering, from the data input to the physical symbol?

Answer:
The order is Most Significant Byte first.

Notching spectral density

_______________________________________________

Question:
In the chapter §6.2 "TRANSMIT SPECTRUM MASK" of the PHY specification, figure 15 describes the spectrum when a notch is inserted to cohabit with S-FSK PLC. The following paragraph explains that the notched frequencies must have a spectral density 25 dB below the active carriers. But the figure presents a reduction of 50 dB instead. What is correct?

Answer:
The notched frequencies must have a spectral density 25 dB below the active carriers (minimum requirement). The figure just shows an illustration of the concept with a deeper notch. The figure will be modified to clarify that the requirement is 25 dB, but a deeper notch can be used.