GB2401016A - Apparatus for rapidly outputting a waveform template - Google Patents

Abstract

A 256-sample representation of a short-duration Ultra Wideband (UWB) pulse is generated faster than would be possible via a Digital Analogue Convertor (DAC) by rippling a pulse (xn) several times along a delay bank (10). The output template yn comprises the summed output from 16 mixers (9), each fed by one input from the delay bank 10 and one from a bank of 16 memories (12) via a DAC (11). The first memory contains taps b0,16,32...240, the second b1,17,33...241 etc. After the first pass yields taps 0-15, a counter (13) is incremented, its value C fed into the memories and the pulse rippled though the loop once more, further yielding taps 16-31. After completing a full sequence of 16 loops yielding 256 values, Y is output and the next impulse X arrives. Reconstructing the pulse via an equivalent Finite Impulse Response (FIR) filter would require some 256 mixers and DAC's, ie. 16 times more hardware.

Description

APPARATUS FOR RAPIDLY OUTPUTTING A WAVEFORM TEMPLATE

This invention relates apparatus for rapidly outputting a waveform template particularly for impulse based radio systems such as ultra widcband radio systems.

s Impulse-based radio systems can operate by conveying information by virtue of variability of successive pulses in a signal. It includes ultra wideband (UWB) which is a based on the transmission of discrete very narrow pulses, typically of sub nanosecond duration. Typically the duty cycle of transmission of such pulses is in the order of 1%. If transmitted on a regular basis the signal would generate a line 0 spectrum. The lines of the spectrum could interfere with the operation of any other radio systems whose RF bandwidth encompassed the frequency of one of theses lines. In order to mitigate this effect, a time hopping code is commonly introduced.

This provides a pseudo random variation in the times at which the pulses are transmitted within the nominal available period. This results in many more spectral lines at closer spacing and with lower power. The line spacing is then related to the length of the time hopping code. The use of time hopping codes allows several users to co-exist in the same bandwidth; the receiver will expect a pulse at a particular time and only listen for a pulse in a narrow window of time. If a different time hopping code is devoted to each link then interference between two links will take place only when the time of the received pulse from the interfering transmitter coincides with the time of the received pulse from the wanted transmitter. If synchronized orthogonal codes are used then this will never happen. In a practical, real world, environment this is impossible to achieve and codes with good cross correlation properties over a wide range of time offsets are used.

The pulses may be modulated in several different ways. Typically pulse position modulation is used wherein the time of transmission of the pulses is varied from the expected time of arrival. The pulse is transmitted either with a slight delay or advance and so the receiver will determine this and will register a "0" or "l", depending on a late or early pulse; hence digital information can be conveyed. The reception problem therefore becomes one of determining the precise timing of the received pulse, or more specifically determining whether a received pulse arrived earlier or later than the nominal timing as specified by the time hopping code.

In practical UWB communications links, the pulse radiated from the transmitter not only has a direct path to the receiver; it also reaches the receiver via reflections off ob jects in the environment. This is known as multi-path interference.

With UWB systems, this multi-path interference affects the received pulse by spreading the pulse shape (typically a 0.5ns Gaussian monocycle), into a series of pulses whose average amplitude decreases over a relatively long time span (e.g. 1 OOns). This means that the energy of the transmitted pulse is spread over time and lo therefore the informational content of the signal is also spread out over time. A typical pulse in a real system can be represented by that shown in Figure 1. As shown it extends over a finite period and comprises a number of ripples. In known systems the received waveform is correlated with a single Gaussian monocycle to enable the detection as to whether the pulse is slightly late or early. In practice therefore this correlation is over a short portion of the pulse and the correlation essentially looks for the first zero-crossing point i.e. the first cycle as shown by arrow A. The problem with this method is that the extra ripples contain energy and therefore information and such correlation loses about 12 dB in signal to noise ratio.

To recover this informational content the inventor has determined that it is advantageous to measure not only the direct path clement of the received pulse, but also all of the multi-path reflected elements i.e. over the entire waveform or a substantial portion thereof. In other words it would be advantageous to correlate the received pulse with a template pulse that had a shape that corresponded to a pulse that had travelled through the current multi-path reflecting environment.

In accordance with the present invention, an apparatus for rapidly outputting a waveform template, the waveform being stored as a set of discrete values, comprises a) a numberp of memory units, each memory unit storing a number q of the discrete values and where =pq; b) a number of DAC's each connected to its corresponding memory unit; c) a number p of time delays arranged in series; d) a number p of mixers, each mixer having an input from one of the DAC's and an input from an individual connection adjacent a time delay; e) an integrator having an input from each mixer and whose output is representative of the waveform template to be output; and f) means to send a pulse to ripple q times along the time delay bank such that in operation, the analogue value representing successive discrete values are output one at a time from each memory via the DAC's to the integrator, and after s each ripple or loop this operation is repeated until all discrete values have been output and wherein no adjacent (successive) discrete values are stored on the same memory unit.

Preferably, the apparatus includes a counter which determines the amount of the repeats, this counter having an input into the memories.

l o Preferably, the method of storing a signal shape template comprises storing a number of discrete and consecutive values ma ={m, m2, m3 mn}, on a (ring of) numberp of memory units such that each memory stores a number q of the discrete values, the values being mz, where i are separated byp (mz+p, mz2p,mz+3p, mz+qp) where z represents the memory unit number.

Preferably, a series of discrete values representing a waveform in consecutive order are output by a method comprising outputting a single value from each of a set of memories in tum, via a DAC, from the first to the last memory, and then repeating this until all of the discrete values have been output consecutively.

Preferably, switching between the memories is perfommed by sending a pulse to ripple through a series of time delays, allowing the output of one of the discrete values from one memory unit at a time.

Preferably, the output of the discrete unit from each memory unit is performed by the output from each memory being connected via its corresponding DAC to a mixer, each mixer being connected to different points in the time delay 2s array.

In accordance with a second aspect of the present invention, a receiver comprises an antenna to receive the signal containing pulses, a scanning receiver connected to the antenna adapted to take values of successive input pulses to provide a representative pulse, and means to output the representative pulse to a correlator to be correlated with an incoming pulse in synchronization, wherein the means to output the representative pulse comprises apparatus as according to the first aspect.

Preferably, the receiver further comprises means to feed the representative pulse output to the correlator and also to the scanning receiver in order to reduce non- l ineariti es.

Preferably, reduction of non-linearities is by means of pre-distortion.

The invention will now be described in more detail by means of example only and with reference to the following figures of which.

Figure 2 shows a simplified block diagram of one embodiment of the invention.

Figure 3 shows a diagrammatic representation of how incoming pulses can 0 be sampled to provide the template pulse.

Figure 4 shows an example of one implementation of the pulse template generator component of figure 2.

Figure S shows an example of a further preferred implementation of the pulse generator component of figure 2.

Example 1 Basic configuration of overall system Fig. 2 shows the basic components of one example of a system according to an embodiment of the invention. The system comprises an antenna l which is input via switch 2 to a scanning receiver 3. The scanning receiver operates as described above and effectively scans successive pulses so as to provide a pulse template. A necessary input to the scanning receiver is from a time hopping sequence generator 4 which tells the scanning receiver when to expect the next pulse. The scanning receiver generates the pulse shape (template) and updates it on a regular basis. The template or pulse representation is then fed into a pulse template output generator 5 which in the example is a Finite Impulse Response Filter (FIR) which stores the template as a series of coefficients or alternatively termed "tap values". The job of the FIR is to output at the correct time the template pulse so as to be correlated with an incoming pulse. Hence the FIR has an input from the time hopping sequence generator. Operation of the FIR will be described in more detail below. The system s also includes a correlator 6 which is shown in general by a dotted line and includes a mixer 7 and an integrator 8. This correlator functions to correlate the received pulse and pulse template; the latter of which is timed to be output for correlation in order to ascertain whether the incoming pulse is slightly delayed or advanced. In other words in normal operation with the switch in position "b", the pulse from the antenna arrives at the mixer either slightly before or slightly after the time that the pulse from the FIR filter gets to the mixer. This time difference will depend upon whether a O or a I is being transmitted. The time hopping code sequence generator acts to zero the integrator ready for each pulse. This integrator integrates the output l o of the mixer over time. This is then thresholded to give a decision output of a I or a 0.

The switch has 2 positions "a" and "b". When the switch is in position "b", the scanning receiver samples pulses from the antenna and the resultant waveform shape is then stored and passed onto the FIR as a set of tap values, as described above. For calibration purposes, the switch can be put in position "a". The pulses produced by the FIR are then fed back into the scanning receiver and the actual output of the FIR compared with the previously stored waveform shape sampled from the antenna. By comparing these two waveform shapes, the FIR tap values can be modified until the two waveforms match. This feedback system can compensate for non-linearity in the FIR filter electronics and in the delay times between each stage of the FIR.

Components of the system will now be explained in more detail below.

Generation of Pulse Template To capture the shape of a received pulse, a scanning receiver is used. The scanning receiver uses a time hopping code sequence generator to supply the variable T which is the time between one pulse and the next successive pulse. The pulse is sampled at a time increments of T+6, where is a small increment, i.e. a small fraction of the pulse duration. T is of course variable where a time hopping code is used. Fast but repetitive waveform shapes are captured by sampling the pulse amplitude at successive pulses, but at steadily increasing offsets. This slowly builds up an image of the pulse shape, and is similar to techniques common to many designs of oscilloscope. Fig. 3 represents 5 successive pulses. Samples of values are taken at five points a, b, c, d, and e and these are used to build the representative pulse. This overcomes the problem of sampling the same signal pulse at 5 time instants which is practically impossible due to the short duration of the pulse.

Pulse Constructor Having obtained the average shape of the received pulses, the next task performed by the system of Fig. 2 is to reconstruct that pulse at a time indicated by a time hopping code sequcoce generator. It is extremely difficult to reproduce representative pulse quickly especially at a precisely defined time for the purpose of l o correlating with an incoming pulse. Various methods of achieving this with varying effectiveness are now discussed below: a) Distal to Aralocue Converter Where the values representing the pulse shape (template) are stored in memory it is not practical to output them on a digital to analogue converter (DAC) as the pulse is so short in duration that a DAC would not be quick enough.

b) FIR This may be implemented using also e.g. a finite impulse response filter FIR.

The taps of the FIR can be initially taken directly from the output of the scanning receiver as they are already in the required form. Figure 4 shows a representation of a FIR filter. This comprises a number of mixers q input to which are a series of coefficients (be, bl, b2...) representing the pulse template i.e. values of amplitude with time. The mixers also have an input from one of a series of time delays 10.

These delays are extremely short and can be implemented by flip flops. They are cascaded to form a bank. In operation the input xn is a pulse which ripples along the time delay bank. In this way the output value y is successively bo' be, b2, b3, be,, b5 factored by the input along the x line.

The equations defining this filter are as follows.

In =bo X,? +6, Xn_i +b2 X,,-2 +. +bq x'' y q Yn=bJ xnJ J=o It can be seen from the above equations that where xn consists of a unit impulse at l n=0 and is zero thereafter and the FIR taps ho to hq are the samples from the scanning receiver, then the output Yn will be a reconstruction of the required template pulse waveform. The delays are implemented as discrete digital elements and the propagated impulse is mixed (i.e. multiplied) by the analogue tap values.

The disadvantage of the above such conventional FIR filter implementation for this application is the large number of taps required. Each filter tap requires its 0 own digital to analogue converter and mixer, adding to the cost and power usage to the design.

For a typical UWB implementation that uses a 0.5ns impulse and ringing of up to 128ns, this would require a FIR filter with 256 mixers and digital to analogue converters.

c) Novel pulse template generator.

To remedy the above problems, a novel design has been developed by the inventor. This enables practical implementation of the invention whereby generation of the analogue representation of the pulse signal is possible which overcomes the problems of using a conventional FIR circuit.

Fig. 5 shows an example of a system according to an example of the novel filter design to reconstruct the pulse signal, where the pulse signal comprises 288 discrete values varying with time. It has some similarities to the conventional FIR of Fig. 2, but there are only 16 taps comprising a series of 16 time delays and mixers.

2s To each mixer is an input from a DAC, 11, of which there are 16, each DAC having an input from a memory, l 2.

In operation, as a pulse ripples down the set of time delays, as input into a counters so that at only one of the points lo, It, I2, I3, I4. is there is a voltage present, such that only one at a time the value b from each memory (via the DAC) is output onto the integrrator and thus the output therefrom, yn. On the first pass therefore, the first address contents of each memory are thus output to line Yn in consecutive order.

This process is repeated such that the contents of the second address of each memory are thus output to line y,, in consecutive order, etc. At the end of the time delay bank the pulse inputs to a 4 bit counter, 13. This counts the number of times a I pulse has ripple down the time delay bank. This value C is also fed into the each of the memories. This procedure will now be described in more detail below.

The reset procedure for this circuit would be to preload the counter with O and to clock each of the 16 memory chips so that their first address contents are; presented at their outputs, then to increment the counter to 1. Each of the 16 memory lo chips has 16 address locations that are filled with the filter tap values. MemO i contains taps 0,16,32...240, Meml contains taps 1,17,33...241 and so on up to MemlS which contains taps 15,31,47...255.

The impulse now arrives at X and proceeds to travel through the discrete I (non-clocked) delays. The impulse at lo is multiplied by be the output of D/AO and I fed to Y via the summer. The D/AO output voltage be is set by the binary value from the first memory's first address contents. When the impulse travels through the delay to It it then becomes multiplied by be and fed to Y via the summer.

The impulse at 11 also triggers MemO to output the memory contents whose address is C to D/AO. At this point C is 1. This process continues until the impulse reaches l l 5. After this delay, the impulse is fed into the counter. It is also fed back to the beginning of the FIR filter. By the time that the impulse has travelled through all 16 delays and back to the beginning, the first memory and D/A have had time to set their output voltage. The counter ensures that the impulse travels around the loop 16 times only. After completing a full sequence of 16 loops, the memory contents and loop counter are automatically set to the correct values for the next impulse to arrive at X. This system arrangement conveys 256 tap values to the output Y in response to a single input impulse from the time hopping code sequence generator whilst only I using 16 delays, mixers, D/As and memories. Due to the looping structure, the response times required of each individual D/A and memory are divided by 16 relative to that required by a system that feed all the taps to a single stage of D/A and memory. The looping structure also uses 16 tines less hardware than a conventional FIR filter implementation. These characteristics act to reduce the cost and power consumption of the circuitry.

It can be seen from this application specific example that different division ratios could be chosen to trade off number of hardware elements versus speed of I operation of each of those taps. If slower hardware elements are desired, then more elements must be used and the size of the loop counter correspondingly reduced.

Conversely, if less hardware is required, then fewer but faster elements may be used and the loop counter increased so that the impulse travels through each of the; elements more times.

in The principle advantages of the invention are a large proportion of the i received radio energy is used to determine the informational content of the system.

This confers increased signal processing gam and therefore system efficiency, unlike existing systems. Compared to a fixed filter system, these techniques permit adaptive I matching to the changing multi-path reflection environment. This process maintains I system efficiency even if the communications link is mobile, unlike other solutions.

The core technology of the cyclic-FIR filter circuit permits the precise time synchronised generation of fast, arbitrary time varying waveform shapes without requiring the ultra high-speed electronics that would be required for a conventional FIR or single A/D implementation. This technology is not restricted to the given ultra wideband receiver application.

Claims (9)

1. An apparatus for rapidly outputting a waveform template, the waveform being stored as a set of discrete values comprising: l a) a number p of memory units, each memory unit storing a number q of the discrete values and where =pq; b) a number of DAC's each connected to its corresponding memory unit; c) a number p of time delays arranged in series; ; d) a number p of mixers, each mixer having an input from one of the DAC's and lo an input from an individual connection adjacent a time delay; i e) an integrator having an input from each mixer and whose output is representative of the waveform template to be output; and, f) means to send a pulse to ripple q times along the time delay bank such that in l operation, the analogue value representing successive discrete values are output l one at a time from each memory via the DAC's to the integrator, and after each ripple or loop this operation is repeated until all discrete values have been output and wherein no adjacent (successive) discrete values are stored on the same memory unit.

2. An apparatus according to claim 1, including a counter which determines the amount of the repeats, this counter having an input into the memories.

3. Apparatus according to claim 1, wherein the method of storing a signal shape template comprises storing a number of discrete and consecutive values ma = {ma, m2, m3 mn}, on a (ring of) numberp of memory units such that each memory stores a number q of the discrete values, the values being mz, where i are separated byp (mz+p, mzi2p,mzsp, . m+qp) where z represents the memory unit number.

4. Apparatus according to claim l, wherein a series of discrete values representing a waveform m consecutive order are output by a method comprising outputting a single value from each of a set of memories in turn, via a DAC, from the first to the last memory, and then repeating this until all of the discrete values have been output consecutively.

5. Apparatus according to claim 4, wherein switching between the memories is performed by sending a pulse to ripple through a series of time delays, allowing the output of one of the discrete values from one memory unit at a time.

6. Apparatus according to claim S. wherein the output of the discrete unit from each memory unit is performed by the output from each memory being connected l o via its the corresponding DAC to a mixer, each mixer being connected to different points in the time delay array.

7. A receiver comprising an antenna to receive the signal containing pulses, a scanning receiver connected to the antenna adapted to take values of successive input pulses to provided a representative pulse, and means to output the representative pulse to a correlator to be correlated with an incoming pulse in synchronization, wherein the means to output the representative pulse comprises apparatus according to any preceding claim.

8. A receiver according to claim 7, further comprising means to feed the representative pulse output to the correlator and also to the scanning receiver in order to reduce non-linearities.

9. A receiver as claimed in claim 8, wherein the reduction of nonlinearities is by means of pre-distortion.