Usually, a 3G mobile phone manufacturer faces two kinds of issues: the ones related to 3GPP standard compliance and the ones related to UE performances. The first category is ruled by objectivity since it refers to functional part of the device, while the second one is closer to subjective evaluation. Usually, for the second group of issues, a comparative testing session should be performed before claiming Device Under Test is poorer then the Reference. Otherwise, the fact that an UE doesn’t reach its maximum performances, as described by its release version, in a certain testing environment doesn’t mean necessarily that the UE has a problem.
One of the most reported performance issue for an UE is for sure the Throughput rate problem, either in DL or in UL or in both directions. Each UE belongs to a certain WCDMA release or HSPA category. Based on this, the maximum R99 or HSPA throughput is defined. Of course, in ideal testing conditions (no interference, perfect radio link quality, etc) this maximum rate should be normally achieved. But in real life testing environment, the degradation of the maximum throughput rate is explicable, without indicating an issue on UE side. So the main question in this case is the following: how can we separate the throughput rate degradation because radio link conditions from the one linked with a potential UE problem (the processor not able to handle all the data blocks, the buffers too small, the RF part introducing interference on DL because continuous Tx transfer on UL, etc)?
This paper will try to describe the steps which are executed at Physical Layer side in case a Throughput rate issue has to be analyzed.
In the case of R99 connection, for a release 6 UE for example, the maximum throughput on DL and UL can be 384 kbps. This is reached if the minimum TTI of 10 ms is set by the network for the TrCH transporting the maximum number of blocks, i.e. 12, each block having the size 336 bits, 16 bots per block representing the CRC. In this case the theoretical throughput is computed as 12x(336-16)x100.
The first thing to be special-list in such a scenario is of course the physical connection configuration sent by the network. If all the above mentioned parameters are not on the indicated values, the maximum throughput can’t be reached. If these parameters are ok, then the next step would be to check and count the rate of CRCs=1, meaning that the data blocks were not correctly decoded on UE low layer side. If the amount of CRCs=1 subtracted from the theoretical maximum throughput leads to a value equal to the reported throughput, then the only thing that remains to be confirmed is the fact that CRC rate is explained by the radio link quality. If yes, the analysis is ended, if not, RF performances on UE side should be verified by specific tests.
If the obtained value after CRC subtraction still doesn’t reach the reported throughput, it means that not all the data blocks correctly received are carrying payload data. This is another discussion, more on L23 side: based on the transfer type (UDP or TCP), acknowledge is expected or not on the opposite direction (for example on UL for the data sent on DL) – this will reduce the throughput rate, but in this article we will neglect this aspect, since it has a systematic and computable character. On L23 side also headers are to be removed before obtaining the final throughput rate. In this case longer data packets, with same fixed header, would ensure the maximum possible throughput.
Figure 1 depicts the above mentioned steps.
In the case of HSDPA transfer, so on DL direction, for a category 8 UE for example, the maximum achievable throughput is 7.2 Mbps. The first step to analyze a potential HSDPA throughput issue is to check the network allocation on HS-SCCH channel. There are 5 sub-frames, each of 2 ms long on each frame. On HS-SCCH channel the network informs the UE with 2 slots before if a sub-frame is allocated to the UE or not. On those allocated sub-frames the important parameters are the following: number of used codes, block size, used modulation. Based on these parameters and the network allocation, the theoretical throughput rate can be computed: maximum block size, as specified in 3GPP 25.214, is 14411 bits for the selected UE category when 16 QAM modulation and 10 HS-PDSCH are used, while the maximum allocation is 5 sub-frames per each 10 ms frame. In these conditions the theoretical maximum throughput is 14411x5x100. So if the maximum throughput is not reached, the combination of the parameters is verified and the new theoretical throughput is computed. Based on the UE reported CQI, the transport block size and the modulation are verified against the 3GPP table. If the values are not matching, meaning that the network is using smaller block size and lower modulation (QPSK), then it is purely a network decision and the lower throughput rate (or at least part of it) is explained by this network behavior. Then the network allocation is also investigated. Based on the observed allocation and used parameters, a real throughput is computed. If this throughput is still bigger than the reported one, the next step would be to count the number of CRCs=1 and to compute the error rate.
A very important question that appears in the context described above is the following: was the UE reported CQI as good as possible? In the scenario when the network really follows the UE recommendation as uses the appropriate parameters (block size, modulation, etc), a lower throughput versus a reference UE may be obtained if the DUT is reporting a lower CQI. This leads again to RF performances check, because the CQI is dependent of the radio link quality indicators.
And finally, another cause for a lower throughput rate may be the lost ACKs on UL, i.e. the NodeB is not able to correctly receive the ACKs sent by UE on UL. This will lead to unwanted retransmissions on DL, those blocks being discarded by the UE low layer since they were already sent to MAChs.
Figure 2 describes the above mentioned steps.
In conclusion, there are some clear check points during R99 or HSDPA throughput rate issue analysis. Usually, most of the problems can be identified performing the described steps for the investigation procedure. Still, there might exist also particular reasons for a lower throughput rate: sample issue (damaged hardware, bad RF calibration tables, local interferences produced by different tools/ signal generators. Also, when it comes to network allocation, lower allocation on DUT versus the reference may be explained by other UE reported capabilities which are not directly linked with data transfer but which may be taken into consideration by the network, SIM type, etc.
So as one observes, many aspects are to be taken into consideration when analyzing such throughput problems, but following the presented processes, solution can be found in short time.