WO1986001952A1

WO1986001952A1 - A method and apparatus for limiting the amplitude of a frequency modulated signal

Info

Publication number: WO1986001952A1
Application number: PCT/GB1985/000437
Authority: WO
Inventors: David Michael Windram; Richard Morcom
Original assignee: Independent Broadcasting Authority
Priority date: 1984-09-21
Filing date: 1985-09-20
Publication date: 1986-03-27
Also published as: GB8423939D0; EP0194307A1

Abstract

Amplitude limiting of a frequency modulated signal, in particular a sampled signal such as a multiplexed analogue component television signal, in order to reduce the bandwidth taken up by the signal. A number of problems arise when applying limiting to a signal, in particular limiting especially a sampled signal can produce aliasing effects due to harmonics. Also if any filtering is applied to the signal after limiting much of the effect of the limiting is removed. The present invention overcomes these problems by applying differing degrees of limiting to the signal depending upon the frequency being limited at a particular time. In one embodiment the signal is filtered prior to the application of limitation, so that more of some frequencies are allowed to pass through to the limiter than others. A post-filter is applied to the signal after limitation to correct for the effects of the pre-filter.

Description

- 1 -

A Method and Apparatus for Limiting the Amplitude of a Frequency Modulated Signal

The present invention relates to a method for limiting the amplitude of a frequency modulated signal, in particular an F.M signal in sampled form, and more particularly to signals in a multiplexed analogue component television signal. The invention also relates to apparatus for carrying out the method.

The colour difference and luminance components of the MAC signal used for DBS are specified in EBU Document SPB 28*_. In this document, the amplitude range of the colour difference signal is defined as 1.0V p-p (corresponding to 77% saturation).

If we define the red, green and blue signals in the studio to be in the range 0 to 1 volt, then the luminance and colour difference signal amplitudes for MAC, if coded linearly are given using vector and matrix notations, by:

volts.

However, in order to provide a chrominance band¬ width equal to that of the luminance in the compressed signal, it has been decided to limit the maximum chrominance amplitude to - 0.5^V (i.e. 1 volt peak to peak) to match the amplitude of the luminance signal.

The ideal process of limiting would result in any signal which exceeds the permitted -0. ^ range being reduced to the range +0.5 . A number of problems immediately arise when one considers that the source signal can be expected to be the Digital Studio standard with colour difference signals covering the full range, and with the U and V signals sampled at 6.75MHz. I the process of limiting produces harmonics some of which are removed by the band limiting filters on the MAC Coder output. This process of filtering also removes much of the effect of limiting. II the process of limiting before pre-emphasis has the effect of reducing the amplitude of the low frequencies which are themselves attenuated by the pre-emphasis network as well as the high frequencies, The present invention provides a method of limiting the amplitude of a frequency modulated signal which covers a range of frequencies , comprising limiting the amplitude of the signal by amounts which change over the frequency range covered. The present invention also provides apparatus for carrying out the method. Preferably, amplitude limitation of the signal involves a first step of applying a frequency filter to the signal, said filter having such a response as to allow the passage of more of some frequencies than of others, and a second step of amplitude limiting the filtered signal, whereby the signal is amplitude limited by an amount which varies in accordance with its frequency.

After limiting the signal is preferably corrected to compensate for the effects of the first filter. Such correction can be carried out by applying a second frequency filter to the limited signal, the second filter having a frequency response which is complementary to that of the first filter.

In the case of a sampled signal, e.g. MAC, the signal is preferably first sampled at higher than the normal rate (e.g. 13.5MHz instead of 6.75MHz in the case of MAC) before amplitude limitation is applied. This has the advantage that any aliasing effects are much reduced.

Features and advantages of the present invention will become apparent from the following description of embodiments thereof relating to the MAC chrominance signal given by way of example with reference to the accompanying drawings, in which: -

Figure la shows the effect of standard amplitude limiting on an analogue signal;

Figure lb shows the effect of standard amplitude limiting on a sampled signal;

Figure 2 shows a block diagram of apparatus used to evaluate the effects of limiting on various sampled signals;

Figure 3 shows a multiburst signal sampled at 6.75MHz without limiting;

Figure *_ shows the signal of Figure 3 but subjected to limiting; Figure 5 shows a signal similar to that of

Figure *4 but which has been sampled at twice the frequency;

Figure 6 shows the waveform of Figure 5 after being subjected to pre-emphasis;

Figure 7a shows a block diagram of one apparatus according to the present invention;

Figure 7b shows a block diagram of another apparatus according to the present invention;

Figure 8 shows the waveform of a multiburst signal after pre- and post-filtering but without limiting; Figure 9 shows the frequency responses of the - k -

pre- and post-filters used to produce the waveform shown in Figure 8;

Figure 10 shows the waveform which results from the apparatus shown in Figure 7 using filters whose frequency responses are shown in Figure 9;

Figure 11 is a waveform similar to that shown in Figure 10 but with a 0.65V pk-pk offset by 0.325V;

Figure 12 shows a waveform similar to that shown in Figure 10 but using different pre- and post-filters; Figure 13 shows a waveform similar to that shown in Figure 12 but with no limiting and

Figure 14 shows a waveform similar to that shown in Figure 11 but with the pre- and post-filters used for Figure 12.

The problems associated with the limiting of a sampled signal are shown in Figure 1. Figure la shows the spectrum of a 2MHz analogue chrominance signal, before compression, which has been limited in amplitude. The limiting is assumed to be symmetrical and the signal is assumed to be centred on OV. The harmonics fall outside the 3MHz chrominance bandwidth.

The effect of limiting a sampled e.g.. digital signal which is sampled at 6.75MHz is shown in Figure lb. The effect of the sampling is to produce in-band aliases, so that for example the 2MHz signal produces an alias due to the third harmonic at 0.75MHz.

Computer simulation has been used to determine the importance of this effect. The system used in the simulation is shown in Figure 2.

A sampled waveform generator which samples at a sampling frequency f s is indicated by reference numeral

1, a ^Mhardⁿ amplitude limiter by reference numeral 2, and a low pass filter by reference numeral 3» Figure shows as an example a 1.3^V p-p amplitude multiburst signal (4 μsec each of 0.5 - 1.0,. 1.5, 2.0 and 2.5MHz) which has been generated with a sampling frequency of 6.75MHz, and low pass filtered by convolution with an interpolation function which gives the channel a relatively sharp cut-off at 3«375MHz. The five bursts can be seen to be accurately reproduced by the system.

Figure 4 shows the same signal with the same sampling frequency of 6.75MHz, but with a hard limiter of range -0.5V introduced before the output low pass filter as in Figure 2. The 0.5MHz signal is satisfactorily limited, but signals of 1MHz and above show a peak to peak amplitude of greater than one volt and also significant alias components. It is clear from this plot that direct limiting of the digital studio signal before MAC encoding is not a satisfactory process.

Limiting of the analogue signal does not produce aliases. However, we have found that the sampling frequency can be raised to the point where aliases are unimportant. Raising the sampling frequency to 13«5MHz eliminates aliases due to the third harmonic, and with the relatively low level of limiting in use here, the effect of the aliases of the fifth harmonic may well not be signi icant. This is illustrated in Figure 5« In this, the

1.3V p-p multiburst signal is sampled at 13«5MHz before limiting and low pass filtering (the low-pass filter has the same frequency response in all examples given in this specification) . The level of alias can be seen to be reduced to a much lower level although not eliminated.

The major effect which remains is that the higher frequencies show peak to peak amplitudes of up to 1.2V.

If the sampling frequency is raised further, then this latter effect remains, which is a disadvantage. Limiting of the analogue colour component signals would appear to be a possible alternative. However, limiting without filtering at the analogue inputs to the ADCs for the MAC Codec produces the effects described above. Likewise limiting at the coder output prior to low pass filtering produces the same effect.

The process of both limiting and filtering is a possibility at the coder input. Similarly, limiting at the coder output, post low pass filter is possible, although in the latter case, the modulator video band- width is likely to be such as to re-filter the limited signal.

The process of limiting and filtering produces waveforms very similar to those of Figure 5_. with the high frequencies rising in amplitude to up to 1.2V p-p due to removal of the harmonics.

The process of limiting and filtering in the analogue domain can therefore also be considered to be unsatisfactory.

Figure 6 shows the result of the use of a pre- emphasis network El on a signal which has been sampled at 13.5MHz, limited and then filtered. This is equivalent to the application of pre-emphasis to an analogue limited and filtered signal.

As expected, the pre-emphasis further lifts the high frequencies and in particular the signals above 2MHz. It is in order to accommodate the frequencies above 2MHz in the chrominance signal that limiting has been intro¬ duced, and this plot indicates that the critical high frequencies can reach a very high amplitude due to the problems of limiting.

It is necessary therefore to consider processes which can provide better limiting of the chrominance signal - without excessive complexity.

The discussion above suggests that the use of a sampling frequency higher than 6.75MHz in the limiter provides the necessary reduction in level of aliases in the limiter for the relatively small levels of limiting required here. The proposals described here are based on digital implementation using 13«5MHz sampling of chrominance in the uncompressed domain. The conversion from the 6.75MHz studio standard and conversion back to 6.75MHz are conventional interpolation, filtering and downsampling processes which are not considered further here. Choice of the sampling frequency which reduces the effects of aliasing for any signal in general can be made by suitable experimentation or computer simulation with the arrangement of Figure 2 i.e. by gradually increasing the sampling frequency f until any aliasing effects are reduced to a satisfactory level.

Figures 7a and 7b illustrate two possible methods according to the present* invention by which the limiting process can be carried out. Both methods have been investigated, but for the relatively small amount of limiting required (-—-*^*1.3J 1 maximum), method (b) has been found to be relatively simple and to provide satisfactory results. This is the method used here to illustrate the possibilities of improvements. Effects such as rounding errors etc., are not considered here, but may dictate that in practice, the more complicated network of Figure 7(a) should be used.

Figure 7^a shows a circuit for limiting an F.M signal comprising an input A, a high pass filter 4, an amplitude limiter 5 and an output E. The input signal is applied to a mixer C as well as the high pass filter 4. The signal from the high pass filter 4 is applied to a negative mixer B as well the amplitude limiter 5_* The signal from the amplitude limiter 5 is applied to the other input to the negative mixer B. The output from the negative mixer B is applied to the other input to the mixer C. Correction terms for high frequencies at high amplitudes can be applied at the output of the negative mixer B. With this circuit the signal is limited in dependence on its frequency. Suitable circuit character- istics (e.g. filter value) can be determined by using this circuit as the "hard" limiter 3 of Figure 2, and experimenting with values, or using computer simulation.

Figure 7b shows a generally simpler circuit comprising two filters F and F_ and an amplitude limiter 7- The input is applied to the filter F- , the output of which is applied to the limiter 7- the out¬ put of which is applied to the filter F„, The output of the filter F„ is a suitably limited signal.

In Figure 7b, the transversal filters for F- and F- should ideally be complementary. In practice this is very difficult to achieve exactly, but the errors can be reduced to negligible proportions.

Figure 8 illustrates as an example the result for a 1.3V pp amplitude multiburst signal using a -1,0, 20,0,-1 pre-filter and 1,0,20,0,1 post-filter preferably ^{a 0 , ~}f ^{, 0 ,}Tl^{, 0 ,} Tc ^{, 0} P^re-^{filter and a} 0 >J2 ^{, 0 ,} T2 ' ⁰ 'T2 ' ⁰ post-filter (both filters normalised to unity gain at

DC) but with limiting. The response of these is shown in Figure 9« The pre-filter rises to approximately + 1.5dB at 3-4 MHz and approximately + IdB at 5 MHz from OdB at zero frequency. The post-filter falls to approximately - 1.5dB at 3_*4 MHz and approximately - IdB at 5 MHz from OdB at zero frequncy. Although the product of the responses is not exactly unity, the results are sufficiently close (that of a -1/398,0,0,0,400/398,0,0,0 -1/398 filter). Figure 10 shows as an example the result with the same pre- and post-filters, but with a limiter set to a maximum of IV p-p between the filters. The result of this process can be considered to give satisfactory limiting for the case of a multiburst signal. - q^~ -

The pre-filter F- has the effect of allowing through to the limiter more of some frequencies than others, which means that more amplitude limitation is applied to some frequencies than others. The post- filter F„ corrects for the effects of the pre-filter F- . By applying more amplitude limitation to some frequencies than others

- . σ -

satisfactory amplitude limitations can be achieved without the disadvantageous effects described above.

Satisfactory performance of the limiter for a signal away from the centre of the channel is an essential requirement. Such a test signal is a multiburst of amplitude O.65V pk-pk centred on +0.325V. This signal has been processed through the limiter above, and the results are shown in Figure 11. This suggests that a filter/limiter/filter combination can provide a satisfactory method of limiting the amplitude of the colour difference signal as generated in a digital studio.

In general, for any signal which it is required to amplitude limit using this process, suitable filter values can be chosen by experiment or computer simulation using a circuit such as Figure 2 with the hard limiter being replaced by the filter/limiter/filter combination of Figure 7b (or by the circuit of Figure 7a). -The filter values can be varied until a satisfactory amplitude limited output is obtained. If it is required to use this process to amplitude limit an analogue F.M signal, the sampled waveform generator 1 of Figure 2 would be dispensed with when determining filter values. In the case of the MAC chrominance signal sampled at 13.5MHz the filters with the above characteristics have been found to produce suitable results but it will be appreciated the filters with similar response would also produce good results.

The process shown in Figures 7a and 7b are essentially one of 'dynamic limiting¹ , in which the amount of amplitude limitation applied varies according to the frequency being limited. It is interesting to consider the logical extension of this in which only the high frequencies are limited in amplitude.

Using the same multiburst test signal as before, with 1.3V amplitude, but this time with a -1,0,8_*0,-1 pre- filter and 1,0,8,0,1 post filter (which filters have - I I -

similar responses to that shown in Figure 9) and with limiting to 1.3V p-p, we get the result shown in Figure

1 8 1 12 for a 0,-*r-,0,^■**^**,0, - ^O pre-filter and a

0,— ,0,— ,0,— ,0 post-filter. The same signal pre- and post-filtered, but without limiting is shown in Figure 13»

Again, satisfactory performance is obtained for a multiburst sweep signal centred away from the centre of the channel, as shown in Figure 14. Here, the low frequencies are now altered, and the high frequencies are slightly reduced in amplitude. The small amount of limiting of the high frequencies is adequate, since the high frequency amplitude is well below IV p-p.

This type of limiting provides the possibility of faithful reproduction of bright colours in large areas of a picture, while providing limiting in areas of high frequency colour which could otherwise produce interference into- adjacent channels.

For MAC receivers which use a digital implemen¬ tation, the design of the input ADC and particularly the signal range is very important. The signal range is affected by both the chrominance and luminance signals.

The luminance signal occupies a nominal range of 1.0 volts. However, such a signal also has a significant noise component. For a carrier to noise ratio of lldB (just above threshold), the unweighted noise at the ADC input is at a level of about -26dB (depending on the exact channel filter response) . Assuming such noise to have a Gaussian distribution, then the noise can be expected to have peak values approximately 3 x rms values. This means that at lldB C/N ratio, the luminance signal covers a total range of about 1 + 2X.15 = 1.3 volts. In order to avoid noise clipping and rectification on black and white, the ADC input range for luminance should be at least 1.3 volts. The chrominance signal as defined in EBU Document SPB 284 also occupies 1.0V range so that with noise, an ADC range of 1.3 Volts as required by the luminance is totally adequate.

For signals outside the 1.3V range, as could occur on occasional noise spikes, correct behaviour of the ADC with limiting, and not overflow, is essential to avoid for example black dots in white areas. If such an ADC range and behaviour is used, then the possibility exists for the use of the full I.3V range of chrominance for low frequencies in future extended definition trans- missions, with these peaks being handled correctly at high C/N ratios by conventional receivers , but that at low C/N ratios, the noise spikes of the chrominance signals only would be limited by the ADC.

It will be appreciated that the method and apparatus of the present invention is not only useful with MAC chrominance signals, but could be used when limiting other sampled signals or even analogue F.M signals.

Claims

- ' 3 -Claims :

1. A method of limiting the amplitude of a frequency modulated signal which covers a range of frequencies, comprising limiting the amplitude of the signal by an amount which is arranged to vary over the frequency range covered.

2. A method in accordance with claim 1, comprising the first step of applying a frequency filter to the signal, said filter having such a response as to allow the passage of more of some frequencies than others, and a second step of amplitude limiting the filtered signal, whereby the signal is amplitude limited by an amount which varies in accordance with its frequency, in accordance with the frequency response of the filter.

3. A method in accordance with claim 2, comprising the further step of applying correction to the limited signal to compensate for the action of said filter.

4. A method in accordance with claim 3. wherein said correction is applied by passing the limited signal through a further filter, the further filter having a frequency response substantially complementary to said filter.

5. A method in accordance with any preceding claim, wherein the signal to be amplitude limited is a sampled signal.

6. A method in accordance with claim 5- wherein the signal is a MAC chrominance signal sampled at higher than the normal sampling rate.

7. A method in accordance with claim 6, wherein the sampling frequency is 13.5MHz.

8. Apparatus for limiting the amplitude of a frequency modulated signal, comprising a first filter arranged to receive the signal and to allow more of some frequencies to pass than others, and an amplitude limiter arranged to receive and limit the filtered signal, whereby the signal is amplitude limited by an amount which varies in accordance with its frequency, in accordance with the frequency response of the filter.

9. Apparatus in accordance with claim 8, comprising circuit means having an input arranged to receive the signal to be limited and an output arranged to output the amplitude limited signal, said first filter being connected between said input and said amplitude limiter, and a second filter connected between said amplitude limiter and the output.

10. Apparatus in accordance with claim 9. wherein said second filter has a frequency response which is complementary to that of the first filter.

11. Apparatus in accordance with any of claims 8 to 10, wherein said signal is a MAC chrominance signal sampled at 13«5MHz and said first filter has a frequency response rising to approximately + 1.5dB at 3.4MHz and approximately + IdB at 5MHz from OdB at zero frequency.

12. Apparatus in accordance with claim 9 or 10, wherein said signal is a MAC chrominance signal sampled at 13.5MHz and said second filter has a frequency response falling to approximately - 1.5dB at 3.4MHz and approximately - IdB at 5MHz from OdB at zero frequency. - \≤ -

13. Apparatus in accordance with claim 8, comprising circuit means having an input arranged to receive the signal to be limited, an output arranged to output the limited signal, said first filter being arranged to receive the signal from the input and to output the filtered signal to the limiter and to one input of a negative mixer, the output from the limiter being arranged to be applied to the other input of the negative mixer, and a positive mixer having one input being arranged to receive the signal to be limited from the input and another input arranged to receive an output signal from the negative mixer, the output from the positive mixer being the output of the circuit means.

14. Apparatus in accordance with claim 13_. wherein there is provided means for applying a correction term to the output from the negative mixer.