[IEEE 2013 IEEE 78th Vehicular Technology Conference (VTC Fall) - Las Vegas, NV, USA (2013.09.2-2013.09.5)] 2013 IEEE 78th Vehicular Technology Conference (VTC Fall) - eICIC Performance Calculations Based on Idle Mode Measurements in Two Heterogeneous Networks

eICIC performance calculations based on idle mode measurements in two heterogeneous networks Jürgen Beyer, Ole Klein Deutsche Telekom, Germany Abstract — Introducing low power pico eNodeBs in dense urban macro networks is a promising way to meet the future mobile radio traffic demand. Many simulations clearly show the benefit of such heterogeneous networks (HetNet) for capacity and performance in particular if enhanced inter-cell interference coordination (eICIC) is applied. Due to the lack of hardware supporting eICIC, the corresponding measurement results from urban HetNets are somewhat rare. Thus, to estimate the eICIC performance under real radio conditions we use idle mode scanner measurements to compute the downlink throughput. We consider measurements in our LTE HetNet trial and in a part of our commercial UMTS HetNet. Results derived from 11 outdoor pico cells in quite different environments and HetNet structures are presented. On average, eICIC together with advanced UE receivers expand the pico cell range about 70%. UEs close to the pico cell border in the expanded pico area will get significantly higher throughput compared to the non-eICIC case. Keywords — Heterogeneous LTE network, measurements, eICIC performance computations I. INTRODUCTION According to most forecasts smartphones and tablet computers will squeeze out the 'classical' voice-only mobile phones in the next few years. Both devices are designed for mobile internet access which essentially contribute to the expected enormous increase in mobile broadband data traffic demand [1]. In addition some operators offer mobile access as substitution of fixed services as DSL even in urban areas which will further increase the mobile data traffic. The challenge to mobile network operators is to offer the mobile radio network capacity which is required to fulfil their customer's demands. Since the LTE point-to-point transmission capacity is reaching the theoretical limits [2] only a few options are left to further increase the network capacity. One option is to introduce low-power eNodeBs with antennas well below the surrounding roof tops. In heterogeneous networks (HetNets) such pico eNodeBs are running under over-laying macro cells. To achieve the desired spectral efficiency pico and macro cells have to run on the same frequency (=co-channel deploy- ment) which, however, leads to a small cell size [5] [6]. Introducing eICIC in 3GPP specification Rel. 10 overcomes this restriction. It enables cell range expansion (RE) by adding a bias to the pico cell Reference Signal Received Power (RSRP) together with resource partitioning between macro and pico in the time domain. RE – which is also possible before Release 10 [7] - enables to capture more traffic for more efficient pico cell operation and an improved load balance between macro and pico cells [2] [8]. Furthermore, RE enables a shift of the pico cell border to the optimal uplink handover point [2] [8] giving – due to lower interference – higher uplink capacity of the pico cells [7] [8]. In the downlink UEs associated with the pico site (pUE) in the expanded pico cell area suffer from high macro cell interference as confirmed by a trial [7]. In eICIC the time domain resource partitioning reduces this negative impact of the macro cell. In a part of the subframes the macro cell only transmits common control signals, but no user data. In those almost blank subframes (ABS) pUEs have much lower macro cell interference. Simulations show that even pure RE increases the HetNet capacity [2] [9] which, however, depends on the spatial user distribution [3] [6] and the location of the pico relatively to the macro site [6]. But with ABS the HetNet capacity is further increased [2] [9]. The pico cell range in the ABS is now mainly limited by the interference of the macro common control channels which could be further reduced by advanced UE receivers with interference cancellation [2] [4] [9]. In order to get measurement based eICIC performance results in addition to simulations we apply idle mode scanner measurements in two HetNets where all picos are deployed outdoor. In this paper we present the results of this investigation. The measurement areas and set-up are intro- duced in Section II. The computation approach is briefly described in Section III and in Section IV we present results covering the cell range for different bias, the eICIC performance and for the case of none-mute macro during the ABS. Finally, assuming a certain building wall penetration loss we present indoor performance results achieved with the outdoor picos. II. MEASUREMENT AREA AND TRIAL SET-UP The performance computations described in this paper are based on scanner measurements. Apart from other things, a scanner measures simultaneously the received power from several cells per measurement point. In UMTS cells the primary common pilot channel (pCPICH) power is measured and in LTE cells the Reference Signal Received Power (RSRP). Only those values - together with the corresponding cell identifier - are required for our computations. The measurements in Darmstadt - a medium sized city with 150,000 inhabitants - were performed in our LTE HetNet trial area which was built-up together with Huawei [5]. The 978-1-4673-6187-3/13/$31.00 ©2013 IEEE trial network is located in urban and dense urban environment in vicinity to the city centre. The goal of the HetNet trial was to study the performance of individual pico cells in different environments which is also done in this paper. An overview about the HetNet is given in Figure 1. The HetNet consists of 1 macro site and 8 pico sites. The macro eNodeB antennas are located on top of a building well above the surrounding roof tops. The pico site antennas are mainly mounted on walls of buildings at heights of between 2 and 8m. The measurements in Berlin were performed in a part of our commercial UMTS HetNet close to Brandenburg Gate. The HetNet is shown in Figure 2. No macro site exists in the surrounding park resulting in a quite large distance between pico and macro sites. The macro close to Pico B1 has reduced TX power. The picos have been deployed in order to handle the enormous volume of mobile voice and data traffic during specific events which take place in that area. The antennas of the picos are mounted on street lambs and traffic lights. Due to range limitations of the UMTS scanner system the measured values of only 3 picos are suitable for the eICIC investigation. The macro antennas have a gain of more than 15dBi whereas the pico site antennas have a gain well below 5dBi. The carrier frequency is 2.6GHz in Darmstadt and 2.1GHz in Berlin. The transmission power (TX power) per macro sector and TX path is 20W and the pico TX power is 1W per TX path. When we refer to a specific pico site in this paper we simply call it 'Pico Dx' (Pico D4, Pico B1 etc) where the capital letter represents the city ('D' for Darmstadt, 'B' for Berlin). III. THROUGHPUT CALCULATION BASICS In our measurements the scanner detects the received power of up to 8 cells per measurement point which is sufficient for a precise downlink signal to interference plus noise power ratio (SINR) computation. Our interest is to analyse the pico-macro interaction of individual pico cells. To achieve this in the computation we 1) regard only one pico (index jRP) per computation run and 2) eliminate the pico-pico interference from the denominator of the SINR formula. Assuming LTE scanner measurements we get ∑ ≠ ∈ + = cellN ji Mi thiG jG NP P SINR ,4 ,4 η RPjjMj =∨∈ (1) with P4G Received power in LTE cells (=RSRP) M Class of macro cell index j Index of the downlink best serving cell i Index of interfering cells jRP Index of the regarded pico cell Ncell Number of cells at the measurement point η Resource utilisation of interfering cells Nth Thermal noise power. To apply the measurements in the UMTS HetNet in Berlin for LTE SINR computations (1) has to be modified. Assuming that the pCPICH TX power is defined by ΔpCPICH and that in LTE the eNodeB TX power is equally shared between the sub- carriers we get the relationship between the power the scanner receives in UMTS (P3G) and in LTE (P4G) pCPICHscGG NPP Δ⋅⋅= 43 (2) ΔpCPICH Part of the TX power allocated for pCPICH Nsc Number of sub-carriers Applying this power relationship in (1) we compute the SINR for LTE from UMTS scanner measurements. Two ABS subframe types are possible [12]: Normal subframe with 8 reference symbols and MBSFN subframe with 2 reference symbols in the 1st slot (MBSFN: Multimedia broadcast single frequency network). We assume normal ABS subframes without interference cancellation. We compute the throughput based on the SINR to throughput mapping as used for radio network planning. The MIMO performance is predicted according to the approach in [10] which is also implemented in our radio network planning tool PegaPlan. For the performance computation we regard a single UE which gets certain cell resources as given in Table I. D1 D5 D3 D4 D6 D2 D7 D8 200m200m Pico ID Macro Figure 1: The Darmstadt LTE HetNet with vector street and 3D building data as background. The circle segments represent the eNodeB antennas. The network consists of 1 macro site (red antenna symbols) and 8 pico sites (blue antenna symbols). Measurements are performed along all streets in this area. Darmstadt HetNet: Pico IDs with a 'D'. B2 B1 B3 Street 1 Park Park Street 2 500m500m Figure 2: The part of the Berlin UMTS HetNet regarded in this paper. Only those macro sites (red symbols) are plotted which are best server or up to 6dB below the best server. Only the labelled picos are considered. The measu- rement routes are along streets 1 and 2. Berlin HetNet: Pico IDs with 'B'. TABLE I. ASSUMPTIONS FOR THE THROUGHPUT CALCULATION Bandwidth 20 MHz MIMO order 2 Resource utilisation of interfering cells 70% ABS pattern 1 of 8 subframes ABS subframe type Normal (not MBSFN) Numb. mute macros in ABS 1 Cell resources per UE In ABS: 100%; otherwise 12.5% (1/8) IV. CELL RANGE AND PERFORMANCE RESULTS A. Cell radius To determine from measurements how much eICIC expands the cell size is not a simple task. After trying some approaches we found that the cell radius is the most relevant measure for the pico cell size, although there are clear arguments against that [5]. Nevertheless, we were looking for a quite fair cell radius definition and apply finally a semi- numerical procedure. In 10 degree circle segments in the horizontal plane we collect those measurement points which are assigned to a pico cell and compute the 80th percentile of the distance from the pico site to measurement point (=radius). Afterwards we generate plots with radius versus the horizontal angle. Taking into account several constraints we manually compute the average of up to four radius peaks which have to be well separated in the horizontal plane. This average is regarded as the cell radius. In the case of eICIC a measurement point is assigned to a pico cell if the received powers P fulfil the criteria. macropico PbiasP ≥+ (dB) (3) A bias of 9dB is the upper limit according to the 3GPP specification Rel. 10. In particular the cell-specific reference signal (CRS) the macro further transmits in the ABS generates high interference to pico UEs which avoids a higher bias [4]. With advanced UE receivers applying CRS interference cancellation (IC) a bias of 18dB might be feasible [4]. Please note that IC might even be required to achieve the 9dB bias. Figure 3 shows the cell radius derived for those three bias values. Without bias, the radius of most pico cells in Darmstadt is around 100m – which corresponds well with the radius obtained in [5]. Pico D1 has a clear lower radius since it is located very close to the macro site and, moreover, the macro has LOS to almost the complete pico cell area. Both together give the quite small cell radius of 45m. The Picos D2, D7 and D8 have a cell radius of more than 120m. They are far away (>350m) from the macro and LOS to the macro exists only in a very small portion of the pico cells. The pico cell radius in the Berlin HetNet is clearly larger. One reason is the larger distance between macro and pico sites whereby no strong interfering macro cell exists. Furthermore, LOS to the serving pico NodeB antenna exists for all measurement points in the Berlin HetNet. For such conditions similar large pico cell ranges are also observed in the Darmstadt HetNet where we regard them as exceptions and consider them with a lower weight in the cell radius computations. Thus, the cell ranges obtained in the Berlin HetNet could be regarded as typical for the combination of LOS and weak macro interference. The impact of the bias on the cell range is quite different for the various pico sites. The largest relative increase of the radius (from 45m to 105m) is obtained for Pico D1 - the pico with the single, but very strong interfering macro. A further large increase of the radius is given for Pico D2 - the only pico site which is not located at a street crossing. Without bias the pico cell area is almost limited to the street where the pico is located. With bias the Pico D2 cell area expands significantly in the none-line of sight (NLOS) region leading to the clear increase of the cell radius. Overall, in the Darmstadt HetNet a bias of 9dB shifts the cell radius from around 110m to 145m on average. Such a large relative expansion is not obtained for the Berlin picos where 9dB bias increases the cell range by around 18% at maximum. 18dB bias affects the cell range of the various picos also quite differently. In particular for the Picos D4 and D6 the 0 100 200 300 400 D1 D2 D3 D4 D6 D7 D8 B1 B2 B3 18 dB 9 dB 0 dB Pico ID R ad iu s ( m ) Figure 3: The pico cell radius obtained in both HetNets for different bias values. Bias=0dB means no eICIC; 9dB is the limit in 3GPP Rel. 10; 18dB is the limit expected to achieve with advanced UE receivers and CRS-IC. Darmstadt HetNet: Pico IDs with 'D'; Berlin HetNet: Pico IDs with 'B'. Pico D5 is not considered. 0 2 4 6 8 10 0 5 10 15 D1 D3 D7 B3 Pmacro - Ppico (dB) Th ro ug hp ut (M b/ s) Figure 4: The downlink throughput in ABS over the difference of the received power from macro and pico sites (=ΔPmp). The graphs represent groups of picos as given in Table II. The assumptions for the throughput computation are summarised in Table I. higher bias results in a significantly larger cell range which we could not explain by environmental effects. On average, a cell radius of 185m is obtained in the Darmstadt HetNet. In the Berlin HetNet the higher bias of 18dB does not further clearly increase the cell range. Finally it should be mentioned that we obtain a relevant NLOS pico cell coverage only in the specific situation of Pico D2. Thus, the values for the co-channel pico cell range given in this paper are valid for LOS only. B. ABS performance In the eICIC performance computations we assume that every 8th subframe is an ABS – which is the lowest number currently suggested in 3GPP [8]. We regard the downlink throughput in ABS (simply called ABS throughput below) in dependency of the difference ΔPmp = Pmacro – Ppico (dB) (4) of the received powers P and compute the piecewise average of the ABS throughput in 1dB ΔPmp-intervals. Plotting the graphs of the piecewise average together we realise 4 groups of picos with very similar graphs. For better over- view we show only one graph per pico group in the figures. The groups are given in Table II and the comments in Table II show that this pico classification could also be well explained. TABLE II. OVERVIEW ABOUT THE PICO GROUPS SHOWN IN FIGURES 4-7 AND THE CORRESPONDING PICO IDS REPRESENTED BY THE GROUPS. Picos in plots Associated Pico Comments D1 D1 Very close to macro site, macro has LOS to complete pico cell. D3 D2, D3, D4, D6 Only 1 interfering macro cell D7 D5, D7, D8 Pico cells are in the inter-sector region of the macro site B3 B1, B2, B3 Many interfering macros Figure 4 shows the downlink ABS throughput. We assume that only the macro with the highest downlink power is involved in the eICIC process and interferes in the ABS with its common control signals only. The other macros are interfering in the ABS corresponding with their resource utilisation (see Table I). The highest ABS throughput is obtained for Pico D3 (and the other picos in that group; this fact is not further mentioned below) which has only one interfering macro whose interference is significantly reduced in the ABS. Furthermore, LOS exists to most measurement points resulting – together with low interference in ABS - in a good MIMO performance [10]. Pico D1 has also only one interfering macro, which, however, generates even in the ABS so much interference that at small ΔPmp the performance is much worse than for D1. For larger ΔPmp the performance of D1 and D3 is comparable. The Picos D7 and B3 have more than one interfering macro and muting only the strongest macro in ABS does not give such high ABS throughputs as for the other picos. Although the conditions in Darmstadt and Berlin HetNet are very different the graphs for D7 and B3 are quite similar. This is also obtained in all other ABS throughput plots and, for better overview again, we consider only Pico D7 in the further ABS throughput plots. Finally it should be mentioned that assuming 10% cancellation error [11] interference cancellation clearly increases the ABS throughput of Picos D1 & D3 but only slightly affects the performance of Picos D7 & B3 (several interfering macros). To estimate the performance benefit of eICIC also requires consideration of the throughput of a macro UE in the expanded pico cell area (=tpmacro). For easy comparison the macro UE gets also 1/8 of the cell resources. Let's further introduce tpABS as the pico UE ABS throughput and compare the macro throughput in Figure 5 with the pico ABS throughput in Figure 4. For the Picos D1 and D3 tpmacro ≅ tpABS is given for ΔPmp = 7dB and with further increasing ΔPmp tpABS becomes much lower than tpmacro. Please note the unrealistic large slope of tpmacro for D1 and D3 due to non-existing further interfering macros. For the Picos D7 and B3 tpmacro ≅ tpABS is achieved around ΔPmp = 5dB and with increasing ΔPmp the throughput difference increases significantly. Hence, in all considered scenarios the throughput is clearly reduced when a UE is shifted from macro to the pico ABS for ΔPmp >> 7dB. Therefore, to achieve a HetNet capacity gain for large ΔPmp requires several pico cells per macro sharing the resource that 0 2 4 6 8 10 0 5 10 15 D1 D3 D7 B3 Pmacro - Ppico (dB) Th ro ug hp ut (M b/ s) Figure 5: The throughput of a UE associated to a macro cell in the expanded pico cell area (=tpmacro). The UE gets 1/8 of the macro cell resources which corresponds with the cell recourses assumed in ABS. 0 2 4 6 8 10 0 5 10 15 D1 D3 D7; B3 Pmacro - Ppico (dB) Th ro ug hp ut (M b/ s) Figure 6: ABS throughput if the macros transmit user data with 10dB lower TX power during the ABS. Dashed: Mute macro (= Figure 4). the macro is keeping free for eICIC. High time dynamic ABS allocation might further improve the HetNet capacity [2]. C. ABS performance in the case of none-mute macro The HetNet capacity could be further increased if the macro isn't muted in the ABS but transmits user data with clearly reduced TX power. Assuming 10dB lower macro power for user data the ABS throughput for such a scenario is shown in Figure 6 together with the mute macro case (dashed lines). For Picos D1 & D3 – both have single interfering macro - the non-mute macro decreases the ABS throughput by less than 8%. For the Pico D7 (many interfering macros) the ABS throughput is only slightly reduced. Thus, we might conclude that as long as the macro TX power for user data in the ABS is sufficiently reduced the none-mute macro has only minor impact on the pico throughput. D. Indoor conditions Adding the building wall penetration loss in (1) enables estimation of the eICIC performance for indoor conditions achieved with outdoor picos. Assuming a building loss of 20dB we obtain almost the same cell ranges as given in Figure 3 for outdoor. Please note that – according to a previous discussion - LOS between pico site and building is required to get those pico cell ranges for indoor. For such scenarios we obtained similar pico cell ranges from LTE indoor performance measurements [5]. The ABS throughput computed for indoor conditions is given in Figure 7. The assumed building loss of 20dB significantly reduces the ABS throughput only for the picos D7 & B3 at small ΔPmp. This might be caused by the large cell range of more than 100m (partly much more) those picos have for 0dB bias. However, this has to be further investigated. For ΔPmp>7dB the ABS throughput for outdoor and for indoor deviates only slightly. Please note that for indoor conditions and ΔPmp >12dB more measurement points are not covered by the pico and, thus, they are not considered in the piecewise average. But even including them the indoor ABS throughput is only slightly lower than in Figure 7. V. CONCLUSION In this paper we present results of eICIC performance calculations based on idle mode scanner measurements in HetNets in Berlin and Darmstadt, Germany. We apply a cell radius as the measure for the cell range expansion due to the power bias. Without bias the cell radius is 110m on average, 9dB bias expands the cell radius to 145m and 18dB bias to 185m. For LOS in long wide streets together with quite low macro interference we obtain a cell range of around 300m without bias and 350m with bias. We further compute the indoor coverage from outdoor picos assuming 20dB building wall loss and we get similar cell ranges as for outdoor. Regarding the downlink performance in almost blank subframes (ABS) the 11 picos considered could be classified into three groups: 1) Picos with one very strong interfering macro, 2) picos with one interfering macro and 3) picos with several interfering macros. The ABS performance of those groups differ a lot for small ΔPmp (= Pmacro – Ppico; P=received power). The lowest performance is obtained for group 3, in particular for ΔPmp >5dB. In all groups a UE shifted from a macro to a pico ABS at ΔPmp >>7dB will get a much lower downlink throughput. For indoor conditions the ABS throughput is significantly reduced only for group 3. We further compute the ABS performance for the case the macro transmits in the ABS user data but with clearly reduced power. Assuming a 10dB lower macro TX power in the ABS has no significant negative impact on the ABS throughput of picos in group 3 which is the most likely scenario. REFERENCES [1] Cisco VNI Forecast 'Cisco Visual Networking Index: Global Mobile Data Traffic Forecast Update, 2011 – 2016', Cisco Public Information, February 2012. [2] M. Vajapeyam et al, 'Downlink FTP performance of heterogeneous networks for LTE-Advanced', Proc. IEEE. Int. Workshop on Hetero- geneous Networks, Kyoto, June 2011. [3] K. Balachandran et al, 'Cell Selection with Downlink Resource Partitioning in Heterogeneous Networks', Proc. IEEE. Int. Workshop on Heterogeneous Networks, Kyoto, June 2011. [4] A. Damnjanovic et al, 'UE's role in LTE Advanced Heterogeneous Net- works', IEEE Communication Magazine, pp 164 – 178, Feb. 2012 [5] J. Beyer et al, 'Performance measurement results obtained in a hetero- geneous LTE field trial network', IEEE Vehic. Tech. Conf., Dresden, June 2013. [6] S. Landström et al, 'Deployment Aspects of LTE Pico Nodes', Proc. IEEE. Int. Workshop on Heterogeneous Networks, Kyoto, June 2011 [7] P. Ökvist, A. Simonsson, 'LTE HetNet Trial – Range Expansion including Micro/Pico Indoor Coverage Survey', IEEE Vehic. Tech. Conf., Quebec, Sep. 2012. [8] I. Güvenc et al, 'Range expansion and inter-cell interference coordi- nation for picocell networks', IEEE Vehic. Tech. Conf., San Francisco, Sep. 2011. [9] K. I. Pedersen et al, 'eICIC Functionality and Performance for LTE HetNet Co-Channel Deployments', IEEE Vehic. Tech. Conf., Sep. 2012, Quebec [10] J. Beyer et al, 'A measurement based approach to predict the MIMO throughput of the LTE downlink in RF planning tools', IEEE Vehic. Tech. Conf., San Francisco, Sep. 2011. [11] B. Soret et al, 'CRS Interference Cancellation in Heterogeneous Networks for LTE-Advanced Downlink', Proc. of Int. Workshop on Small Cell Wireless Networks 2012, pp. 6797 – 6801 [12] Stefania Sesia et al, 'LTE. The UMTS Long Term Evolution. From Theory to Practice', 2nd edition, John Wiley, 2011 0 2 4 6 8 10 0 5 10 15 D1 D3 D7; B3 Pmacro - Ppico (dB) Th ro ug hp ut (M b/ s) Figure 7: ABS throughput for indoor conditions assuming a building penetration loss of 20dB (solid lines). Dashed lines: Outdoor (=Figure 4). /ColorImageDict > /JPEG2000ColorACSImageDict > /JPEG2000ColorImageDict > /AntiAliasGrayImages false /CropGrayImages true /GrayImageMinResolution 200 /GrayImageMinResolutionPolicy /OK /DownsampleGrayImages true /GrayImageDownsampleType /Bicubic /GrayImageResolution 300 /GrayImageDepth -1 /GrayImageMinDownsampleDepth 2 /GrayImageDownsampleThreshold 2.00333 /EncodeGrayImages true /GrayImageFilter /DCTEncode /AutoFilterGrayImages true /GrayImageAutoFilterStrategy /JPEG /GrayACSImageDict > /GrayImageDict > /JPEG2000GrayACSImageDict > /JPEG2000GrayImageDict > /AntiAliasMonoImages false /CropMonoImages true /MonoImageMinResolution 400 /MonoImageMinResolutionPolicy /OK /DownsampleMonoImages true /MonoImageDownsampleType /Bicubic /MonoImageResolution 600 /MonoImageDepth -1 /MonoImageDownsampleThreshold 1.00167 /EncodeMonoImages true /MonoImageFilter /CCITTFaxEncode /MonoImageDict > /AllowPSXObjects false /CheckCompliance [ /None ] /PDFX1aCheck false /PDFX3Check false /PDFXCompliantPDFOnly false /PDFXNoTrimBoxError true /PDFXTrimBoxToMediaBoxOffset [ 0.00000 0.00000 0.00000 0.00000 ] /PDFXSetBleedBoxToMediaBox true /PDFXBleedBoxToTrimBoxOffset [ 0.00000 0.00000 0.00000 0.00000 ] /PDFXOutputIntentProfile (None) /PDFXOutputConditionIdentifier () /PDFXOutputCondition () /PDFXRegistryName () /PDFXTrapped /False /CreateJDFFile false /Description > /Namespace [ (Adobe) (Common) (1.0) ] /OtherNamespaces [ > /FormElements false /GenerateStructure false /IncludeBookmarks false /IncludeHyperlinks false /IncludeInteractive false /IncludeLayers false /IncludeProfiles true /MultimediaHandling /UseObjectSettings /Namespace [ (Adobe) (CreativeSuite) (2.0) ] /PDFXOutputIntentProfileSelector /NA /PreserveEditing false /UntaggedCMYKHandling /UseDocumentProfile /UntaggedRGBHandling /UseDocumentProfile /UseDocumentBleed false >> ] >> setdistillerparams > setpagedevice