EUROCAE_Short Delays in Climax-V01

The European Organisation for Civil Aviation EquipmentWorking Group 67: “Voice over IP for ATM”

Working Paper

Short time differences in Climax, a theoretical studyVer 00.01

2006-04-01

By

Alf Nilsson (WG67 SG 1)(SAAB Communication – Sweden)

Short time differences in Climax Page 1 of 10


1. ABSTRACT

GeneralThere exist a requirement that for Climax the difference in one-way delay should be between 4 ms and 20 ms. The upper limit represents the value above which echo effects will become disturbing. The lower limit represents the value below which fading effects will become disturbing. This study tries to explain and visualize the fading mechanism and the impact on the frequency response in relation to the frequency range.

The study presented in this paper is theoretical and based on calculations in a lap-top under ideal circumstances. It is not a substitute for a practical test on radios, where also other possible practical effects may be presented.

ConclusionsThe effects shown in this study, appears to be dramatic and are certainly of importance in the TDM case with fixed delays. However, this study does not look into a practical VoIP system. For a practical VoIP-system the delay may vary and how this affects a real situation is not investigated here, but should be studied.

For short difference delay, shorter than 0.1 ms, the spectrum is very little affected in the 300 – 3400 Hz range. For 0.25 ms delay difference, there is one dip at 2000 Hz. From ca 1300 Hz to 2700 Hz the level is less than -6 dB relative to the maximum value. This indicates that the frequency region responsible for the intelligibility of the speech is affected.

For higher difference delays there are more dips, but they are narrower. And when the difference delay is more than 4 ms the number of dips and the frequency information lost is said to be acceptable, according to statements given during SG-1 discussions.



2. SCOPE

This working paper presents a theoretical study of the effect on the spectrum when a copy of a signal is added to itself after a short delay under ideal circumstances. It is not a substitute for a practical test on radios, where also other possible practical effects may be presented.

This study tries to explain and visualize the fading mechanism and the impact on the frequency response in relation to the frequency range for difference in time delay of 0 - 4 ms.

3. BACKGROUND

Two speech signals is added when a controller is listening to two receivers at the same frequency. The disturbing effects of echo and fading can be remedied by the use of Best Signal Selection (BSS), where one of the signals only (the best) is selected for listening.

The pilot does not have this possibility because he is already listening to one receiver. Hence, when in an offset-carrier (Climax) environment his headset signal will be the result from the reception from two or more transmitters transmitting the same speech signal. Therefore it is critical that the time difference between the received signals is within 4 - 20 ms. The propagation in the air is insignificant here as this is much shorter than the time difference in the ground segment.

For TDM-systems, with fixed delays for one and the same connection, this has been solved by delay lines adjusted to make sure the delays between different transmitters in the Climax-chain is kept within the above 4 - 20 ms. The upper limit represents the value above which echo effects will become disturbing. The lower limit represents the value below which fading effects will become disturbing.

4. RESULTS

4.1. Calculations

Fig 1 Adding the same signal delayed Δt

A fixed delay can be converted to phase for a frequency. The higher the frequency, the shorter the Period T, hence more radians for the same Δt. This can be expressed as:

Time period T = 1/f ¨

One period T represents 2π radians. The delay is therefore equivalent with an angle of

φ = Δt/T*2π = Δt*2πf



Let the signals be represented by two vectors A and B, and the resultant signal be vector C.

Fig 2 Adding the same signal delayed Δt

The magnitude (length) of C according to Pythagoras theorem will be:

4.2. Diagrams equal level

The results for different delays are presented below. The first diagrams are for the same magnitude of A and B which will give very deep minima. The next diagrams are where the B-vector is reduced by 3 dB which give less deep minima.

All calculations and diagrams assume that the wiring is such that the signals are in phase for no delay. The distance between minima will be the same also when the branches has been changed, but shifted a half section so the minima now will be where the maxima was.

As there are two signals added in phase for no delay, the summation signal will be 6 dB higher than one original signal.



Addition of two copies of the same signal, with delay

-30,0

-25,0

-20,0

-15,0

-10,0

-5,0

0,0

5,0

10,0

0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000

Frequency, Hz

Su

mm

atio

n le

vel,

dB

0.1 ms

0.25 ms

Fig 3 Summation of the same signal delayed 0,1 ms and 0,25 ms

Note that more than 6 dB is lost between 1300 Hz and 2700 Hz for 0,25 ms delay. The frequencies which are most important for the intelligibility of speech are affected.


-30,0

-25,0

-20,0

-15,0

-10,0

-5,0

0,0

5,0

10,0

0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000

Frequency, Hz

Su

mm

atio

n le

vel,

dB

0.5 ms

1 ms

Fig 4 Summation of the same signal delayed 0,5 ms and 1 ms



Important portions of the speech frequencies are still affected when the delay is 0,5 and 1 ms respectively.


-30,0

-25,0

-20,0

-15,0

-10,0

-5,0

0,0

5,0

10,0

0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000

Frequency, Hz

Su

mm

atio

n le

vel,

dB

0.25 ms

4 ms

Fig 4 Summation of the same signal delayed 4 ms compared to 0,25 ms

For 4 ms delay less consecutive parts of the frequency spectrum are lost. The critical bandwidth of the ear is 100 Hz up to about 350 Hz and then increases. For mid-frequencies the critical bandwidth is about a third octave. At 1000 Hz, the critical bandwidth is about 160 Hz. The 4 ms time difference results in 250 Hz between the notches. It has been stated during the meetings in EUROCAE WG67 SG-1 that 4 ms time difference is the minimum difference allowed for Climax operation.

If this has to do with the critical bandwidth of the ear has not been studied, it is just noted that the 4 ms time difference gives energy within all critical bands above around 1000 Hz. It is out of the scope of this paper to try to find a detailed explanation of why 4 ms is considered an acceptable value.



4.3. Diagrams 3 dB level difference

If there is a level difference, the bottoms of the notches will not be so deep. And the maximum level will be less than 6 dB. In this study the impact of a level difference of 3 dB has been studied.


-30,0

-25,0

-20,0

-15,0

-10,0

-5,0

0,0

5,0

10,0

0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000

Frequency, Hz

Su

mm

atio

n le

vel,

dB

0 dB diff

-3 dB diff

Fig 5 Summation of the same signal, level difference -3 dB and delayed 0,25 ms

This is shown only for the delay 0,25 ms. The effect is the same for all other delays. The bottom of the peak will now become about -11 dB and the top decreases with one dB from +6 dB to slightly less than +5 dB. The width of the range where the loss is more than 6 dB compared to the maximum value is about the same.

4.4. Adding more signals than two

This can happen where there is Climax operation of more than two stations. Due to the slightly different time differences, it is probable that the notches in practice will not be so deep, and the summation of three signals would be more flat. And it is also likely that there would be more difference in level from three stations than from two stations which also would flatten the spectrum somewhat. Two signals could in theory give the same level in the receiver and therefore cancel out. Signals from three stations have not been investigated in this study.

4.5. Latency in a radio network

Ref[1] describes measurements of the round-trip latency in already established connections for 142 radio channels in Sweden. Note that this is not VoIP, but TDM-connections. It must also be noted that Climax operation is not used in Sweden today, and therefore the measured values is as is and no delay-lines for compensation has been used.



Assuming the round-trip latency is twice the one-way latency; the given figures can be cut in half to get the one-way latency. The following table states measured one-way delays and the difference in delays for some frequencies used by the Swedish CAA. If they would be used in Climax-operation many of them would fall in the 0 – 4 ms range, hence need compensating delay lines. This report covers only lines within Sweden and does not cover lines elsewhere in Europe, but it seems likely that compensating delay lines are needed for many, if not all, TDM lines for Climax operation.

The lines in ref[1] are used in most cases for best signal selection (BSS) operation. The method where the channels were mixed during the BSS evaluation time (first 500 ms), was abandoned in favour of the method where the first received frequency was used during the BSS evaluation time. Switching to a better frequency took place if the BSS-function found that feasible. The reason for change of method was the fading effects described in this paper.

Station Frequencyround-trip

msone-way

ms

rel shortest

ms

rel next shortest

delayms

rel third shortest

delayms

rel fourth shortest

delayms

Bjurholm 7,9 3,95 0,0Måttsund 12,0 6,00 2,1 0,0Älvsbyn 16,4 8,20 4,3 2,2 0,0Vännäs 17,2 8,60 4,7 2,6 0,4Sundsvall rr 0,6 0,30 0,0Kramfors 3,8 1,90 1,6Arvidsjaur 13,4 6,70 0,0Gällivare 16,1 8,05 1,4 0,0Måttsund 18,1 9,05 2,4 1,0 0,0Storuman 18,7 9,35 2,7 1,3 0,3Sundsvall rr 2,0 1,00 0,0Kramfors 3,8 1,90 0,9 0,0Östersund 6,2 3,10 2,1 1,2Sundsvall rr main 2,0 1,00 0,0Sundsvall rr st-by 2,8 1,40 0,4 0,0Kramfors 5,2 2,60 1,6 1,2 0,0Gävle 7,1 3,55 2,6 2,2 1,0 0,0Umeå 16,5 8,25 7,3 6,9 5,7 4,7

Delay difference

125,60

129,55

131,05

132,15

135,02

Table 1 Example of difference in line delays, Sundsvall ATCC



Station Frequencyround-trip

msone-way

ms

rel shortest

ms

rel next shortest

delayms

rel third shortest

delayms

rel fourth shortest

delayms

Rydholm/Arn TWR 2,6 1,30 0,0Gävle 11,5 5,75 4,5 0,0Borlänge 18,9 9,45 8,2 3,7Rydholm/Arn TWR 2,5 1,25 0,0Eskilstuna 7,8 3,90 2,7 0,0Solvalla/Glia 11,5 5,75 4,5 1,9Malmen 10,1 5,05 0,0Solvalla/Glia 11,6 5,80 0,8 0,0Krokek 15,9 7,95 2,9 2,2Bromma 8,9 4,45 0,0Solvalla/Glia 11,6 5,80 1,4Granhult/Gustavsborg 2,4 1,20 0,0Rydholm/Arn TWR 2,6 1,30 0,1Rydholm/Karlslund 2,6 1,30 0,0Solvalla/Glia 11,6 5,80 4,5Solbacken/Gustavsborg 2,3 1,15 0,0Rydholm/Arn TWR 2,5 1,25 0,1Solbacken/Gustavsborg 2,2 1,10 0,0Rydholm/Arn TWR 2,4 1,20 0,1Solbacken/Gustavsborg 2,3 1,15 0,0Rydholm/Arn TWR 2,5 1,25 0,1Granhult/Karlslund 2,4 1,20 0,0Rydholm/Arn TWR 2,5 1,25 0,1Granhult/Karlslund 2,4 1,20 0,0Rydholm/Arn TWR 2,6 1,30 0,1 0,0Eskilstuna 8,9 4,45 3,3 3,2Rydholm/Arn TWR 2,6 1,30 0,0Malmen 10,1 5,05 3,8 0,0Gotland N 14,8 7,40 6,1 2,4 0,0Krokek 17,2 8,60 7,3 3,6 1,2 0,0Visby 21,6 10,80 9,5 5,8 3,4 2,2Granhult/Karlslund 2,4 1,20 0,0Rydholm/Arn TWR 2,5 1,25 0,1 0,0Berga 11,5 5,75 4,6 4,5Rydholm/Arn TWR 2,3 1,15 0,0Uppsala 6,6 3,30 2,2Rydholm/Arn TWR 2,6 1,30 0,0Filipstad 31,7 15,85 14,6Granhult/Gustavsborg 2,3 1,15 0,0Berga 9,7 4,85 3,7 0,0Gotland N 11,2 5,60 4,5 0,8 0,0Visby 21,9 10,95 9,8 6,1 5,4Rydholm/Arn TWR 2,6 1,30 0,0Malmen 10,1 5,05 3,8 0,0Gotland N 11,4 5,70 4,4 0,7 0,0Krokek 19,1 9,55 8,3 4,5 3,9Rydholm/Arn TWR 2,6 1,30 0,0Berga 9,6 4,80 3,5 0,0Krokek 15,8 7,90 6,6 3,1Solbacken/Gustavsborg 2,3 1,15 0,0Uppsala 7,1 3,55 2,4Solbacken/Arlanda TWR 2,2 1,10 0,0Gävle 10,9 5,45 4,4 0,0Borlänge 19 9,50 8,4 4,1 0,0Filipstad 26,5 13,25 12,2 7,8 3,8

133,70

134,20

135,20

130,40

131,13

132,48

133,45

Delay difference

118,20

118,28

120,15

120,50

126,65

118,40

119,40

119,63

123,75

129,18

129,45

129,95

124,10

Table 2 Example of difference in line delays, Stockholm ATCC



5. CONCLUSION

There is a lower limit of 4 ms time difference for Climax operation today according to statements in SG-1. This lower limit has been set a long time ago, and the time difference is controlled by adjustable delay-lines. This study has shown what happens to the spectrum, theoretically, when adding the same signal with less time difference than 4 ms.

It is out of the scope of this document to explain why 4 ms has been considered acceptable in the past.

It is probable that differences less than 4 ms, and especially differences below 1 ms has a detrimental effect on the spectrum.

It is also clear that with TDM-systems with fixed delay for an established connection, this may be a problem. It is not clear whether this is a real problem or not when it comes to VoIP. In VoIP-systems the delay will vary, and it is therefore less probable that the time difference will be in the range of 0 - 4 ms for a prolonged time. On the other hand, it is not clear how it is perceived when the time difference varies and passes the range 0 - 4 ms, which may be dependant on the implementation. This is for further study.

Round-trip latencies measured in Sweden (TDM), makes it likely that compensating delay-lines are necessary to keep the delay difference out of the range 0 – 4 ms in a TDM network, should they be used for Climax operation.

6. REFERENCES

1. “Latency in the radio network of the Swedish CAA”, Ver 01.01, Alf Nilsson, SAAB Communication, 2006-01-30


Documents

EUROCAE_Short Delays in Climax-V01