**4.1. Applicability of low-resolution ADCs (single user case)**

In [14] we have shown that in the 2-PPM case and under certain conditions, low-resolution ADCs can almost achieve the full resolution *E*b/*N*0-performance. One important condition is the number of pulse repetitions *N*s within one modulated symbol, which should not be too small. For the one bit ADC case, *N*<sup>s</sup> = 8 and *N*<sup>s</sup> = 20 just correspond to quantization penalties of 2 dB and 1.5 dB, respectively, cf. Fig. 11(a). If the resolution is increased from 1 bit to 2 or 4 bits, the penalty may decrease, but only if the input level of the ADC is well controlled by an additional gain-control circuit. In [14] we have also proven that a 1 bit ADC with its inherent clipping characteristic offers a superior interference rejection capability.

Pulse repetition *Ns*

**4.2. Multiple access codes for time hopping PPM**

codes as suggested for analog receiver implementations.

**Table 1.** *d*int/*d*ref at 1% packet error ratio for SOP tests

**4.3. Performance of simultaneously operating piconets**

(a)

random-codes are assumed. BER is 10−3.

1-b

desirable.

Quantization

 Loss (dB) 2-PPM non-coherent detection in multipath, = 32 ns, =1 *T N* int *<sup>u</sup>* BER = 10-3

simulated analytical

**Figure 11.** Power penalty due to 1-bit quantization (a) and multi-user interference (b), where

Loss due to MUI (dB)

Fig. 11(b) shows the power penalty due to MUI for a network with 11 users, where perfect transmit power control and a full resolution ADC was assumed. Random codes were applied. For a given processing gain *N*<sup>s</sup> · *N*h, the penalty depends strongly on the ratio of the parameters *N*<sup>s</sup> (number of non-zero pulses) and *N*<sup>h</sup> (number of hopping positions). Since *N*s determines the ADC quantization induced penalty, too, we conclude that *N*s on the order of 8 leads to a good trade-off between the quantization loss and the MUI penalty. In [14] we have shown that this rule does not only apply to random codes but also to optical orthogonal

A test geometry of simultaneously operating piconets (SOP) is shown in Fig. 12, where a single co-channel interference is considered. The reference distance *d*ref (desired piconet 1) is chosen such that the power at the receiver is 6 dB above the receiver sensitivity threshold. The interfering transmitter (uncoordinated piconet 2) operates at the same power as the transmitter of piconet 1, but at a distance *d*int to the reference receiver. We have considered the IEEE 802.15.4a channel model 3 (indoor LOS) and the channel model 4 (indoor non-LOS) [9], where random TH codes with *N*<sup>s</sup> = 10 and *N*<sup>h</sup> = 8 are applied altogether with forward error correction (Reed Solomon code with a rate of 0.87). The results shown in Table 1 prove clearly that digital receiver implementations outperform analog ones. Furthermore, 1-bit ADCs are

> Channels Analog DCMF (full) DCMF (1 bit) CM3 1.53 0.64 0.32 CM4 1.05 0.71 0.45

*Lm=* 64 *L =m* 80 *L =m* 100

simulated

*Lm=* 64

Pulse repetition *Ns*

(b)

2-PPM non-coherent detection in multipath, = 32 ns, = 11 *T N* int *<sup>u</sup>*

*L =m* 80

*L =m* 100

Non-Coherent UWB Communications 119

**Figure 9.** Block diagram of the DCMF-based non-coherent receiver shown in the complex baseband for (a) energy detection and (b) differential detection. (c) A NBI suppression scheme using a soft-limiter.

**Figure 10.** *E*b/*N*<sup>0</sup> performance improvement of a fully digital 2-PPM non-coherent receiver compared to its analog counterpart for the multipath single-user case. The integration window extents over *T*int = 32 ns. A full-resolution ADC is assumed.

We have also investigated the applicability of Sigma-Delta ADCs, especially if *M*-ary Walsh modulation is used, cf. Section 4.4. The results show that the full resolution performance can be obtained for an oversampling rate of 4. Since the power consumption of an ADC depends linearly on the sampling rate, but exponentially on the resolution [8], Sigma-Delta ADCs can thus be considered as attractive candidates.

**Figure 11.** Power penalty due to 1-bit quantization (a) and multi-user interference (b), where random-codes are assumed. BER is 10−3.

### **4.2. Multiple access codes for time hopping PPM**

10 Will-be-set-by-IN-TECH

digital

*τB*

(c)

2-PPM non-coherent detection in multipath, *T*int = 32 ns

<sup>100</sup> <sup>101</sup> <sup>102</sup> <sup>0</sup>

Number of pulses *N*s per bit

**Figure 10.** *E*b/*N*<sup>0</sup> performance improvement of a fully digital 2-PPM non-coherent receiver compared to its analog counterpart for the multipath single-user case. The integration window extents over *T*int = 32

We have also investigated the applicability of Sigma-Delta ADCs, especially if *M*-ary Walsh modulation is used, cf. Section 4.4. The results show that the full resolution performance can be obtained for an oversampling rate of 4. Since the power consumption of an ADC depends linearly on the sampling rate, but exponentially on the resolution [8], Sigma-Delta ADCs can

**Figure 9.** Block diagram of the DCMF-based non-coherent receiver shown in the complex baseband for (a) energy detection and (b) differential detection. (c) A NBI suppression scheme using a soft-limiter.

analog

(a)


()∗

(b)

Re{·} <sup>∑</sup>*T*int*<sup>B</sup>*


ADC DCMF

*g*R(*t*)

1

*E*b/*N*0

ns. A full-resolution ADC is assumed.

thus be considered as attractive candidates.

2

3

improvement

4

5

 in dB at

*p*b = 10

−3

6

7

8

9

Fig. 11(b) shows the power penalty due to MUI for a network with 11 users, where perfect transmit power control and a full resolution ADC was assumed. Random codes were applied. For a given processing gain *N*<sup>s</sup> · *N*h, the penalty depends strongly on the ratio of the parameters *N*<sup>s</sup> (number of non-zero pulses) and *N*<sup>h</sup> (number of hopping positions). Since *N*s determines the ADC quantization induced penalty, too, we conclude that *N*s on the order of 8 leads to a good trade-off between the quantization loss and the MUI penalty. In [14] we have shown that this rule does not only apply to random codes but also to optical orthogonal codes as suggested for analog receiver implementations.

### **4.3. Performance of simultaneously operating piconets**

A test geometry of simultaneously operating piconets (SOP) is shown in Fig. 12, where a single co-channel interference is considered. The reference distance *d*ref (desired piconet 1) is chosen such that the power at the receiver is 6 dB above the receiver sensitivity threshold. The interfering transmitter (uncoordinated piconet 2) operates at the same power as the transmitter of piconet 1, but at a distance *d*int to the reference receiver. We have considered the IEEE 802.15.4a channel model 3 (indoor LOS) and the channel model 4 (indoor non-LOS) [9], where random TH codes with *N*<sup>s</sup> = 10 and *N*<sup>h</sup> = 8 are applied altogether with forward error correction (Reed Solomon code with a rate of 0.87). The results shown in Table 1 prove clearly that digital receiver implementations outperform analog ones. Furthermore, 1-bit ADCs are desirable.


**Table 1.** *d*int/*d*ref at 1% packet error ratio for SOP tests

**4.5. Advanced narrowband interference suppression schemes**

interference, since the DCMF has a frequency dependent transfer function.

power.

**5. Conclusions**

oversampling rate of 4.

Nuan Song, Mike Wolf and Martin Haardt *Ilmenau University of Technology, Germany*

below 10.6 GHz, amended ECC/DEC/(06)04.

**Author details**

**6. References**

98-153.

In [13] we have presented a new NBI-mitigation technique, which is shown in Fig. 9c). It is based on a soft limiter, where the soft limiter itself was originally proposed to suppress impulse interference [2]. The thresholds of the soft limiters are adjusted according to NBI

Non-Coherent UWB Communications 121

We have shown that the proposed receiver can effectively mitigate the NBI, if the threshold factor and the input level of the subsequent ADC are chosen appropriately. Furthermore, the performance improves if the ADC resolution is increased. In the presence of the OFDM interference, the proposed scheme could also be used, but it is required to adjust the threshold dynamically. It should be noted that the performance also depends on the frequency of the

We have derived concepts for energy efficient impulse radio UWB systems with a low transceiver complexity. These concepts are especially suitable for wireless sensor networks operating at low data rates. The *E*b/*N*0-performance of non-coherently detected 2-PPM and DPSK is very similar. It differs by 3 dB in favour of DPSK. However, if the multipath combining should take place in the analog domain, i.e., by means a simple integrate and dump filter, the difficulty to realize analog broadband delays makes it almost impossible to use differential detection and thus DPSK. On the contrary, digital receiver implementations enable advanced modulation schemes and offer superior interference rejection capabilities. With low-resolution ADCs, only a small quantization loss is observed. Compared to the full-resolution case, a one-bit receiver shows a higher MUI suppressing capability. Sigma-Delta ADCs can be considered as attractive candidates for the analog to digital conversion. Our results show that the full resolution performance can be obtained for an

[1] 05-58, F. [2005]. Petition for Waiver of the Part 15 UWB Regulations Filed by the

[2] Beaulieu, N. C. & Hu, B. [2008]. Soft-Limiting Receiver Structures for Time-Hopping UWB in Multiple Access Interference, *IEEE Transactions on Vehicular Technology* 57. [3] Commission, F. C. [2002]. First Report and Order: Revision of Part 15 of the Commission's Rules Regarding Ultra-Wideband Transmission Systems, ET Docker

[4] ECC [2007]. ECC Decision of 24 March 2006 amended 6 July 2007 at Constanta on the harmonised conditions for devices using Ultra-Wideband (UWB) technology in bands

Multi-band OFDM Alliance Special Interest Group, ET Docket 04-352.

Uncoordinated Piconet 2

**Figure 12.** The SOP test geometry with a single co-channel interference

### **4.4. Power efficient Walsh-modulation**

In [17], we have proposed two advanced (low data rate) transmission schemes based on *M*-ary Walsh-modulation, namely repeated Walsh (R-Walsh) and spread Walsh (S-Walsh). For both schemes, the fast Walsh Hadamard transformation can be used to efficiently implement the demodulator. Whereas the more implementation friendly R-Walsh transmission is favorable for data rates of up to 180 kbps (*M* = 8 or *M* = 16), S-Walsh transmission with *M* ≥ 32 is an option for higher data rates.

In [15], we have compared R-Walsh transmission with *M*-PPM. It has been shown that R-Walsh works well with a 1-bit quantization. Fig. 13 shows that in the case of Walsh modulation, the quantization loss (compared to the full resolution case) is only about 1.5 dB — independently of *M* and *N*s. However, if a strong near-far effect is present, *M*-PPM outperforms R-Walsh with respect to the MA-performance.

**Figure 13.** Quantization loss due to a one-bit ADC versus the number of repetitions *N*<sup>s</sup> at *p*<sup>b</sup> = 10−<sup>3</sup> for *M*-Walsh and *M*-PPM in multipath channels (analytical estimation).
