In all honesty… I stopped using Mikrotik mobile solutions, incuding wireless in 2015. I had to connect more then 200 locations in EU via 3G using Sierra Wireless and Novatel modems. The speed was not a problem as I didn’t expect much, but making a stable connection even in cities was undoable plus all of our 411A routerboards had bloated capacitiors which we could only throw into the garbage alongside with modems.
This was my story until yesterday. A friend got me a new LHGGR. I live in the hills and currently use Xiaomi Poco XT G4 phone for my internet. Using USB cable between Poco and computer I am able to get 100mbit/30mbit. The poco is stationed inside a house.
Using LHGGR that i stationed outside I get 3,5mbit/30mbit and that’s it. I was turning the antenna for 3-4 hours, checking cells… cant even watch Netflix anymore. For the joke I also used a very USED Tenda which was also able to get atleast 30 mbit DL/20 UL. Mind you, you just put a SIM card in that Tenda and it works…
Do you have CA working (if you have a device supporting it)? MIMO? CQI? The parameters you’ve provided are not enough. Please do
/interface/lte/monitor 0 once
and show us everything from primary-band and belog. And do it while having a running download, otherwise some values might not be reliable.
I live in the hills
I was turning the antenna for 3-4 hours, checking cells
That does not really describe your situation. Is this a kind of location were you only have a single BTS available, a few km away? Are there multiple BTS to choose from, but all are far away? Are any close to you? Maybe directional antenna is not the best option.
and currently use Xiaomi Poco XT G4
This phone has 5G as far as I can see, which none of LHGs can do. Do you receive 5G signal on the phone? Any chance the mobile provider throttles the traffic when they detect internet sharing? You can force outbound ttl to always be 64 to pretend that you only have one device. You can override outgoing TTL on Mikrotik routers by firewall mangle rules.
Which bands does your mobile provider cover in your location? You could run
/interface/lte/cell-monitor 0 max-age=120
for every BTS you can point the antenna at to check that.
Yeah I know it’s not enough. The board name is LHGGR, it’s a new one with a giga port. I can’t do anymore tests on this device. Yesterday this dish got destroyed. Too much rain and it came in LAN port. I consider it my fault as I didn’t “duck tape” the “port hatch”. The modem can’t be seen by the LHGGR anymore. But still, I mean you put a phone on the table and share the wifi and you get 100mbit+ DL link. With “pro” equipment you need to change 100000 settings and it still doesn’t work well. The last speed was 15/30 mbit/s. Thank god I was able to agree with the seller to return this. He told me it was the first device sold in Slovenia… I’m going back to POCO. The phone is in LTE mode, not 5G, but it supports 5G as most of newer phones do. Thank you for reply.
With “pro” equipment you need to change 100000 settings and it still doesn’t work well
However that is true in general for Mikrotik’s functions like routing, firewalling or WiFi, and that’s where Mikrotik shows its strength, I can’t really agree on that for LTE. The amount of settings is absolutely minimal, you can disable bands and lock cells and that’s it. With LTE what is offered to you, like carrier aggregation, modulation and encoding, traffic throttling all depend on operator’s will and environment conditions. User Equipment can’t control those.
The board name is LHGGR, it’s a new one with a giga port.
So the one with LTE Category 6 modem. Then I’d say the problem might rather be in bad choice of the device. This device does not support all bands used in Slovenia and I see that not all operators have been allocated the same channel width on all bands. Maybe LHG LTE 18 would do better in this situation.
I expect your mobile phone, since it has 5G, does at least cat 18 or 20 on LTE.
Yes, I understand but still… I need a fairly good internet connection to be able to work from home. The current situation is that the phone is doing a much better job sharing an internet connection than LHGGR in MY CASE. Like I said I dont need so many functions, I just need a stable working internet connection and being able to use the mobile phone at the same time. Also my download was always so bad compared to upload. My provider has a LTE speed limit set at 150/50. I was able at some point to get 49,7 mbit/s UL but the DL was below 1 mbit… in bet scenario I was able to lock to a cell 230 and get 15 mbit DL max and cca 20 mbit UL. And again POCO is set in LTE mode only as 5G is not working in these parts. Using the same card the performance of POCO is cca 100/30 which is what I would need from LHGGR and would be satisfied. I don’t event bother how to position the phone when it’s sharing the connection… I just talked to a reseller and he is going to lend me a SXT unit as I dont need over 100 mbit DL. I’m going to pick it up later today and test it over the weekend.
The problem with MT’s offerings in LTE/5G group is that included (and supported in general) modems tend to be a generation or two old. Which means they will almost always perform worse (or even much worse) than any contemporary smart phone. Because they don’t support carrier aggregation to nearly the same level as modern phones do.
This, unfortunately, holds true even in areas with weak mobile network signal as antennae tend to have mediocre (or even poor) gain in lower frequency bands (which are used by mobile operators to cover those remote areas).
Things get even worse as in those remote areas cells may get even more congested during peak hours (because their capacity is low and yet there are users who want to use service). And this is where user’s own equipment may perform worse than (comparable) MNO’s equipment … in Slovenia there’s at least one MNO which offers “mobile broadband” and does utilize QoS to prioritize certain classes of service for those customers (IPTV in particular). And obviously those QoS settings don’t get applied to customers who don’t subscribe to those “prefered” services (those come strictly together with MNO-provided CPEs).
i did that route around 4 years ago after my dsl died.
got an sxt. the connection was decent (50-60mbit) but it had constant, many many times per day micro-stutterings (like a second or two) when no data would pass. tried everything, changing ros version, modem fw, bands… had that problem for months.
in the end got rid of the sxt and got myself a huawei b818 with external log periodic antennas connected in the attic (10m cable). my connection didn’t drop once since then, and speeds were rarely below 200mbit.
make yourself a favor, get a set of MIMO outdoor antennas and hook them up to whatever modern indoor modem you can find that supports CA that your provider supports (and which are available on your remote area). in my country, B818 is best buy if we’re talking about 4G.
for 5G i don’t know, i got fiber and got rid of all 4G equipment.
LTE CAT6 only does 2CA … and there’s always possibility that device either doesn’t support frequency band MNO uses at all … or that it doesn’t support band combination which would give best results. Modern devices tend to support really wide rande of bands and many CA combinations (3CA as well, CAT6 doesn’t).
I was talking about QoS parameters on LTE side, which is definitely not something user settable.
Neither do I. Personally, I’d follow advice by @igorr29 … get yourself a mediocre (but modern!) LTE modem with antenna ports. Then attach a pair of Iskra P-56 antennae (the page I linked is about a bundle of two antennae which is what you need) and point them in direction with best performance (try to avoid directions with lower RSRQ … that’s where value is more negative). These antennae are surely ugly as a sin, but are very decent performers in the whole wide band supported (unlike MT’s offerings which shine in high frequency bands but suck big time in lower bands).
It says for this exact device: LTE Category 6 (300Mbit/s Downlink, 50Mbit/s Uplink)
LTE CAT6 should suffice getting 100mbit DL
That’s not how LTE works. Let’s go through 2 examples of real world scenario, not marketing:
General information:
MIMO can provide bandwidth multiplied up to the amount of antennas if the eNB also can support the same amount of antennas
Carrier Aggregation can provide bandwidth multiplied up to the amount of carriers
64QAM modulation can provide bandwidth of up to 37.5Mb/s per antenna per 10MHz of carrier width
QAM256 modulation can provide bandwidth of up to 48.95Mb/s per antenna per 10MHz of carrier width.
Example 1, device is LTE Category 6 with 2x2 MIMO, 2xCA, 64QAM. Your operator offers 10MHz- and 20MHz-wide carriers. The maximum possible bandwidth is 300Mb/s. Let’s add some real-life conditions:
The supported bands of the device don’t cover all bands used by the operator (in general or at your location, or what the operator is willing to serve you due to some network conditions). There is only 1 band you get connected to. CA benefit is gone. You’re down to 150Mb/s
The operator put you on a carrier 10MHz wide and does not offer another carrier on this band (so no in-band CA). You are down to 75Mb/s.
MIMO works more or less OK, like 150%. You are down to 56.25Mb/s
You get a more or less OK link regarding SINR. Modulation and symbol rate are 50% efficient. You are down to 28.125Mb.s
There are other users sharing the eNB with you, it’s evening and everyone is watching Netflix. Your device is granted 10% of Resource Blocks. Your throughput is down to 2.8Mb/s
Example 2, device is LTE Category 18 with 4x4 MIMO, 5xCA, 256QAM. Your operator offers 10MHz- and 20MHz-wide carriers. The maximum bandwidth is a bit tricky to tell, because with CA not all carriers can operate with 4x4 MIMO. The specification of the LTE Cat 18 modem used by Mikrotik is 1.2Gb/s. So fast! Who needs that, eh?
Let’s add some real-life conditions:
Your device can cover all bands from the operator. Your operator in your location offers 4 bands. 2 of those bands offer 10MHz- and 20MHz-wide carriers. 2 other offer only 10MHz-wide carriers.
The operator assigns you 1x 10Mhz + 2x 20MHz carriers. In fact its 10+20 on one band and 20 on another band, but that does not matter.
The operator does not support 4x4 MIMO at your location, only 2x2.
Maximum bandwidth is thus 489.5Mb/s
MIMO works as in the previous example, at 150%. You are down to 367.125Mb/s
The signal quality is somehow similar. Let’s say you get 50% of useful bandwidth just like with the previous example, but now it’s QAM256. You’re down to 183.56Mb/s.
Again you compete with other Netflix watchers and you’re getting only 10% of RBs allocated. But now you are at 18Mb/s.
This information, if you’re still willing to stick with Mikrotik for an outdoor device, leaves you with only 2 options: LHG LTE 18 (directional high-gain antenna with 2x2 MIMO) and ATL LTE18 (directional, not-so-high gain antenna with 4x4 MIMO). And if you don’t need outdoor, there are Mikrotik Chateau also with Cat 18 or with 5G.
Buying an LTE modem / route without first understanding what is available from you provider at your location will result in sub-optimal experience. And this applies to smartphones too, the very cheap phones also can’t do multiple CA either. However I understand this might be harder to spot, because people often buy phones with contracts, so they hardly ever have to pay a real price of few hundred, or a thousand (iPhone) € for one, and will not buy those cheap 100€ phones. And the Mikrotik LHGG LTE6 you’ve mentioned in the original post costs half of the Poco XT G4 smartphone. So yes, this smartphone definitely has a higher LTE category modem. It has all the right to work better if there’s good enough signal that does not really need that directional antenna of Mikrotik LHG series to be received.
General rule of thumb is: sub-2GHz carriers are never 4x4, they are mostly 2x2 (if not 1x1) … with possible exception of 1.8GHz band (B3) in Europe on modernized towers. The above-2GHz carriers may be 4x4 or not, depends on age (and legacy) of cell tower installations. Newer are more likely 4x4, older (which were first equipped for 3G or older) are usually 2x2.
And regarding calculation of achievable throughputs:
Usually it doesn’t work that bad … active users have certain requirements, e.g. when watching netflix they might try to get 5Mbps (and failing to get that, they are downswitched to lower quality/bitrate streams). So if cell capacity can’t support all those users, then they indeed get a portion of resources each. However, if the cell capacity is ample, then all those netflix users will use those 5Mbps … perhaps there will be more netflix users (in case of congested cell tower some will just give up) … but there will still be capacity left. So when running a speedtest, one would get much more than the minimum share in such case.
But this is, as you showed, very hard to predict because other users in the same cell coverage are bastartds and don’t want to let you to get best service
I work occasionally from a remote location where only mobile connection is possible and a single tower (with multiple cells, but facing different directions, so I can’t switch) serves a pretty big area and from my valley that single tower is the only one I can connect to. At night I can get 180Mb/s. During the day 60-100Mb/s. When holidays come and people visit their summer houses, available bandwidth drops to 10Mb/s for a few days on LHG LTE 18.
I agree this is not usual, but I wanted to give an edge case (however real) example how capabilities of a higher category LTE modem, even though seem unnecessary, as having 150 or 300Mb/s seems often enough for everyday use, make a huge difference under high network load or when CA is impossible for particular device.
I think those Cat 4 and Cat 6 Mikrotik devices have very few uses. They would work good only on condition that the whole neighbourhood gets proper copper or fiber Internet access and only very few people use LTE. Like a small town with everybody having DSL and just a few houses far away on LTE. But is it always like that? My personal case is very different: hardly anybody has a landline there.
Maybe those devices are useful for some automation, IOT and such. But if somebody needs Internet at home, they should go with Cat 18 devices even if gigabit speed does not seem necessary.
e.g. when watching netflix they might try to get 5Mbps
…
in case of congested cell tower some will just give up
For streaming, hopefully they will. But they can still browse web pages at even lower speeds and not give up Or their phones just sync, upload photos, fetch the latest podcasts and so on. There is always some traffic, even if users don’t purposefully generate it. Might be small but it adds up.
Well, tabling the RF stuff… i.e. If you want to optimize RF, it at least a dozen things to research/measure/do/tweak. And no doubt you’re better off with a high-category modem vs a lower one – but trying quantify speeds linearly does not give enough credit to 3GPP specs that try to ensure fairness, which goes well beyond just modem’s category. Anyway…
Where you can also see “underperforming LTE speeds” is at the IP / Layer3… Once a modem connects, it has an MTU. That MTU is very often lower than standard internet/ethernet 1500 MTU. i.e. tower backhauls are sometimes tunneled over internet, and also underlying SS7 network employs more tunneling. So one quick thing to try with the problematic LHGGR, is set the MTU on /interface/lte to match the [connected] LTE network’s MTU. If you search the logs after startup for “mtu”, you might find a message that says:
lte1 mbim: network advertises lower mtu XXXX
So you can use that number as the MTU for the /interface/lte. And also adding a corresponding change-mss mangle rule on out to lte with clamp-to-mtu in /ip/firewall/mangle. While PMTUD should work to detect the MTU, it does not always work…so if your having speed problem, it’s likely worth it to statically set MTU on the LTE interface.
Shouldn’t MSS clamping be done by operator’s TCP Proxy (they are often running on mobile networks, at least for APNs without inbound access)? But you’re right, it’s probably a good idea to check it.
Hard to know. Kinda my point… there is even more ways for things to go wrong even if RF were dialed in…
The idea to make sure MTU is right is any fragmentation would persist all the way to the carrier’s Layer3 IP gateway. Perhaps it can be “corrected” there… but better if TCP packets remained whole from the start, for a lot of reasons.
And, importantly, the LTE network employ retransmission between “tower” and modem. But this comes at the cost of latency. And often TCP congestion control mechanisms, interpret the add latency as “congestion” and that will cause TCP to slow down sending – well before it even hits the modem. While modern OSes, generally use newer TCP congestion control algos that don’t weigh latency as much as older ones. And modern browsers try QUIC for web pages, which better than TCP over LTE. But fundamentally TCP just isn’t ideal over LTE. And in NO case does a TCP packet getting fragmented help.
On top of this, some short speedtest does not let TCP congestion control to ramp up (nor the tower allocating more CA/etc based on higher traffic needs)…all making numbers look worse.
Maybe not the reason for the OP’s slowness… Just the signal does not look terrible & 4Mbs vs 100Mbs seems beyond modem category differences… kinda why I’m not sure the RF is at issue here…
I very much doubt that reduced MTU (if it really is) is result of backbone topology. S1 interfaces (between nodeB and MME or SGW) are definitely tunnelled if they (tres)pass outside MNO’s own backhaul network. So even if there are layers of overhead to cater for, the MTU perceived by mobile station will not correlate to it (much). Also air interface on mobile networks has “native” MTU much lower than 1500 bytes, so LLC does fragmentation/defragmentation all the time and this doesn’t affect data service (i.e. IP) MTU. BTW, L2 in mobile networks goes between mobile station/modem and PGW, everything between them is transparrent.
SS7 has nothing to do with data in mobile technologies beyond 3G, it’s used for signalling (between mobile station and core network … also between different core network boxes) in circuit-switched networks (read: voice calls and short messaging in 2G and 3G, also CS data in 2G and 3G … the one which is true dial-up and gets charged per minute).
Fair enough, “SS7” is technically incorrect, just old and lump all “carrier stuff” as SS7. I’m just not convinced it’s a device/antenna selection issue in this case. e.g. SINR is 15db.
Only point was the LTE carrier can send a requested MTU to the modem. And on Mikrotik this does NOT automatically get used – it’s just logged. Why the carrier wants a lower MTU is above my pay grade, but seems wise to use it if it being reported lower in Mikrotik’s log. We just don’t know that’s the case here…
And at least in US, both major carrier use a lower MTU…now this is an academic point for example only…since the LHGGR nor ATL will NOT work in US:
Or their phones just sync, upload photos, fetch the latest podcasts and so on. There is always some traffic, even if users don’t purposefully generate it. Might be small but it adds up.