Regarding cell locking: I already explained (perhaps not in this thread) how mobility in mobile networks works. In short: when device is connected to certain cell, cell sends a set of “neighbours” to measure. By “neighbours” I mean handover candidates, identified by certain properties (those depend on technology, on LTE that’s eARFCN (frequency) and PCI (numerical identifier)). Then device does the measurements, it can take a while because measurements on non-serving eARFCN can be only done during idle periods (and cell does schedule those if needed so device throughput can be reduced during that time). When measurements are done (can be as they’re made, it doesn’t have to be complete measurements), device reports them back to the serving cell.
This is when all those “cell lock”, “band lock” and “band affinity” come into the play. Device can simply omit some measurements from reports (or they might even not be done, depends on implementation of device RF chipset). So if device is set to “cell lock”, then it won’t report any of “neighbours” as being detected. In case of “band lock”, device will report “neighbours” from same eARFCN and with “band affinity” device will report only “neighbours” from the preferred bands. Device can try to prioritize certain neighbours by reporting them sooner or it can report their signal strength and quality (RSRP and RSRQ) better than they actually are (but this kind of innovative reporting can be detrimental later).
The “neighbour” reporting in LTE is only started when device measures serving cell below certain thresholds, these thresholds are set per cell by MNO. After device reports “neighbours”, network decides how to handle the situation. Most often it orders handover to a better cell. Which is it? Could be it’s not the cell with best signal, MNO might configure some affinity (to “move” devices to bands with higher capacity even if signal strength of those at device is slightly lower … strategy is entirely up to MNO and depends on network equipment functionality, some vendors offer more than others).
So: if device doesn’t report certain “neighbour”, then it won’t be ordered to handover to that neighbour.
Indeed some MNOs are a bit “innovative” when trying to distribute devices between different cells. If your MNO is one of those, then my condolences.
It’s worth to mention that there is different behaviour when device is idle (i.e. registered to netwoork, but no data transfer is going on) and when device is active (i.e. during data transfers). Different MNOs play slightly different games here, but most often they are not really pushing devices to cells with lower signal strength when devices are idle. When devices start to transfer data, they can be hand-over to higher-capacity cell pretty quickly.
Another thing is carrier aggregation. Again candidates for component carriers have to be measured by device before resources are scheduled for certain device on those carriers. When device has some of locks configured, it may or may not allow for CA on omitted bands/cells.
Re. question #3: it’s been a while since I quit the telco job and I don’t need mobile networks for broadband access on any of “my” locations. So I really don’t have any usable opinion here. My ex employer uses some Cisco routers with LTE modems for business customers (we are one, we use such setup on one of our remote locations to provide backup for primary link which is optical line), but that doesn’t mean anything, it’s a Cisco shop for business customers. And I’ve no idea what kind of LTE modem is built in (I guess I could ask though). And they use the ugly Iskra antennae, they are really good and perform up to promises from the product catalogue. My guess is that the setup comes a bit pricier than ATL though.
Re. speeds on low bands: there’s a definition that 4G speeds should exceed 1Gbps and on LTE this only happens with full CA performing at 100%. If CA is not at it’s maximum (and often it’s not), then max speeds are more around 100Mbps (if you’re lucky) or lower, so 3.9G (3.5G was used for HSDPA which could do around 15Mbps in downlink). As others already explained, lower bands offer larger coverage but capacity is lower, so yes, this means lower throughputs (even if signal strength is excellent) … often shared between higher number of users making actual user throughput even lower. So indeed one has to adjust expectations in these scenarios. Stable throughput of around 5-10Mbps may be maximum achievable. However, sometimes CA (for downlink) is possible even if higher frequency cell (which has higher capacity) is not suitable as stand-alone cell (e.g. because Tx power of device is not high enough to get uplink going) and in such case fownlink throughput can be quite decent, uplink won’t shine though because it’ll use the low-capacity low-frequency cell.
… Any thought on the idea of “mucking” with band selection?