FCC ranting has been moved to its own thread. Please continue your happy thoughts there.
http://www.snbforums.com/threads/ranting-about-the-fcc.24323/
http://www.snbforums.com/threads/ranting-about-the-fcc.24323/
Does An AC Router Improve N Device Performance?
- Thread starter L&LD
- Start date
sfx2000
Part of the Furniture
Getting back to the thread and to recap...
1) Better RF front-ends - PA's are better, LNA's are better... more radio chains add additional performance - an AC class radio is pretty darn good for something that costs as little as it does.
2) Better MAC/Baseband Chipsets - an "AC class" chipset flat out needs more horsepower at the baseband level for 5GHz - most of them now have at least some embedded SRAM, and an Cortex-A9 equivalent core running the firmware - 2.4GHz benefits from this on many Router/AP's as it typically the same for both at the higher end...
The firmware in these chips - much better than before - they're smarter, faster, and generally better - things that used to be a challenge in the early days of 11n aren't a challenge these days - mixed-mode (11n with legacy) is better supported, narrow vs. wide channels, OBSS co-existence, rejection of interference, the list goes on...
3) 11ac itself - it's a cleaner spec for legacy, in other words, it's a better 11n (and 11g for that matter), at least for Wave1 functionality - remains to be seen where things land once some of the special sauce in Wave2 starts getting folded in..
So I'm not surprised - an 11ac AP can be a much better AP for 11n/11g STA's - and the results, and reports from other, tend to show that the article is on target...
1) Better RF front-ends - PA's are better, LNA's are better... more radio chains add additional performance - an AC class radio is pretty darn good for something that costs as little as it does.
2) Better MAC/Baseband Chipsets - an "AC class" chipset flat out needs more horsepower at the baseband level for 5GHz - most of them now have at least some embedded SRAM, and an Cortex-A9 equivalent core running the firmware - 2.4GHz benefits from this on many Router/AP's as it typically the same for both at the higher end...
The firmware in these chips - much better than before - they're smarter, faster, and generally better - things that used to be a challenge in the early days of 11n aren't a challenge these days - mixed-mode (11n with legacy) is better supported, narrow vs. wide channels, OBSS co-existence, rejection of interference, the list goes on...
3) 11ac itself - it's a cleaner spec for legacy, in other words, it's a better 11n (and 11g for that matter), at least for Wave1 functionality - remains to be seen where things land once some of the special sauce in Wave2 starts getting folded in..
So I'm not surprised - an 11ac AP can be a much better AP for 11n/11g STA's - and the results, and reports from other, tend to show that the article is on target...
VPonwireless
New Around Here
Thanks Tim for interesting article and data. A few observations came up when reading the results.
You said that 2.4 GHz baseline downlink throughput between Intel N 7260 (2x2, 802.11n) and Asus RT-N66U (3x3, 802.11n) was 28.5 Mbit/s. This test location D is somewhat distant and around RX level measured at – 60 dBm. You also said that path loss from AP to this point was 17 dB.
1. Path loss must be higher since AP puts out about +15 dBm and you measure about – 60 dBm.
=> So path loss would be around 75 dB or so. You may mean additional attenuation/path loss comparing measurements close by the AP and at remote location D?
2. The baseline throughput of 28.5 Mbit/s with 802.11n and 2x2 sounds low. I of course do not know your RF environment so estimating it is a bit difficult. However, your other tests seemed to achieve 25-30 Mbit/s throughput levels with 802.11g standard only, which sounds about right in clean RF/decent signal level.
=> The –n baseline numbers what you used for calculating the percentage gains in benefit of –ac may be too low and not be accurate. Better baseline values would have been be expected for normally operating 802.11n equipment with 2x2 MIMO. Something seems not right. Maybe –n was not enabled or not used for some reason at all during baseline tests?
3. You reported 321% improvement to 2.4 GHz downlink throughput when switching the Asus RT-N66U to Asus RT-AC68U and still using the 2x2 –n only client with HT20. This 321% increase on top of referenced 28.5 Mbit/s means you measured about 120 Mbit/s DL throughputs with 2x2 –n standard devices with HT20. Theoretical maximum for 802.11n MCS 15 (2x2, short guard interval, HT20) is 144.4 Mbit/s. You would have achieved 83% efficiency between end user measured and theoretical rate. This is very unlikely/not possible to be the case due to significant overhead in the 802.11 protocol. In addition, you were remote, not close to AP, which means rate downshifting and increased retries getting further away for max values.
=> Measured –ac AP throughput values seem not accurate. Are you sure –ac APs and Intel N 7260 device did not decide to use HT40 at 2.4 GHz? This could explain too high values, too close to theoretical maximums. Sometimes HT40 is enabled by default at 2.4 GHz band in consumer grade APs.
=> => I do not think the presented throughput improvement % values are accurate when comparing properly and comparably configured and operating –n and –ac systems.
Regards,
Veli-Pekka Ketonen
Twitter @VPonwireless
You said that 2.4 GHz baseline downlink throughput between Intel N 7260 (2x2, 802.11n) and Asus RT-N66U (3x3, 802.11n) was 28.5 Mbit/s. This test location D is somewhat distant and around RX level measured at – 60 dBm. You also said that path loss from AP to this point was 17 dB.
1. Path loss must be higher since AP puts out about +15 dBm and you measure about – 60 dBm.
=> So path loss would be around 75 dB or so. You may mean additional attenuation/path loss comparing measurements close by the AP and at remote location D?
2. The baseline throughput of 28.5 Mbit/s with 802.11n and 2x2 sounds low. I of course do not know your RF environment so estimating it is a bit difficult. However, your other tests seemed to achieve 25-30 Mbit/s throughput levels with 802.11g standard only, which sounds about right in clean RF/decent signal level.
=> The –n baseline numbers what you used for calculating the percentage gains in benefit of –ac may be too low and not be accurate. Better baseline values would have been be expected for normally operating 802.11n equipment with 2x2 MIMO. Something seems not right. Maybe –n was not enabled or not used for some reason at all during baseline tests?
3. You reported 321% improvement to 2.4 GHz downlink throughput when switching the Asus RT-N66U to Asus RT-AC68U and still using the 2x2 –n only client with HT20. This 321% increase on top of referenced 28.5 Mbit/s means you measured about 120 Mbit/s DL throughputs with 2x2 –n standard devices with HT20. Theoretical maximum for 802.11n MCS 15 (2x2, short guard interval, HT20) is 144.4 Mbit/s. You would have achieved 83% efficiency between end user measured and theoretical rate. This is very unlikely/not possible to be the case due to significant overhead in the 802.11 protocol. In addition, you were remote, not close to AP, which means rate downshifting and increased retries getting further away for max values.
=> Measured –ac AP throughput values seem not accurate. Are you sure –ac APs and Intel N 7260 device did not decide to use HT40 at 2.4 GHz? This could explain too high values, too close to theoretical maximums. Sometimes HT40 is enabled by default at 2.4 GHz band in consumer grade APs.
=> => I do not think the presented throughput improvement % values are accurate when comparing properly and comparably configured and operating –n and –ac systems.
Regards,
Veli-Pekka Ketonen
Twitter @VPonwireless
zerodegrekelvin
Regular Contributor
@VPonwireless,
I do not agree with your comments, Tim's results are very accurate, test methodology are also correct. I am a wireless engineer myself and Tim's result are very inline with what I tested myself, both in RF chamber and open air, I do see the improvement from 802.11n chipset to 802.11ac chipset for 3x3, I work with both QCA and Broadcom enterprise chipset.
Tim's inventory of wifi routers for the test are quite representative for his test. I myself recommend to my colleague to replace their 802.11n for our newer 802.11ac even if we do not have 802.11ac client because I do see huge improvement at our office for 802.11n clients connected to 802.11ac AP, the sustain high MCS data rate speak for itself.
But of course you can rerun Tim's test yourself to see if you do see discrepancy and publish your results.
Thanh
Hi VP,
Thanks for your comments. I did not measure RSSI during testing. The -60dBm figure you cite I think comes from this sentence:
I made sure all routers were forced to 20 MHz mode in 2.4 GHz.
Looking at my past 2.4 GHz open air results from the Charts with N300 and N600 class routers, measured values ranged from 5 to 55 Mbps. So the 28.5 Mbps I got as a reference is within reason.
No attempt was made to adjust router settings to optimize throughput. I was trying to see what a typical user would get from an "out of the box" experience.
Thanks for your comments. I did not measure RSSI during testing. The -60dBm figure you cite I think comes from this sentence:
The "path loss" figures I cited were the differences between RSSI readings using inSSIDer made with the test notebook next to the router and then in Location D. inSSIDer is not an accurate signal level measurement tool and even less so the lower the signal level. In hindsight, I should not have quoted those figures.
I made sure all routers were forced to 20 MHz mode in 2.4 GHz.
Looking at my past 2.4 GHz open air results from the Charts with N300 and N600 class routers, measured values ranged from 5 to 55 Mbps. So the 28.5 Mbps I got as a reference is within reason.
No attempt was made to adjust router settings to optimize throughput. I was trying to see what a typical user would get from an "out of the box" experience.
Last edited:
VPonwireless
New Around Here
Assuming you really get in practice the max possible end user throughput 120 Mbit/s with the AC-68 (2x2, HT20, 2.4 GHz), RF conditions must be excellent including very strong signal level. If you get only 28.5 Mbit/s results with RT-66U (2x2, HT20, 2.4 GHz) within those same excellent RF conditions, something is not right with this baseline setup. Asus RT-66U has received excellent reviews and performed well in many tests offering much higher throughputs. Your RT-66U data with Intel N 7260 client device seems to indicate opposite.
Unfortunately Wi-Fi equipment have all kinds of issues which may cause low performance, including driver problems with specific versions and interoperability issues between certain chip sets. Problems include also devices not following the made settings, like even though you set bandwidth to HT20 it does not yet mean that's the case. To be sure, you would need to monitor what really happens in the air interface, does the -ac AP really stick to HT20 continuously and make results comparable? Poorly operating client-AP pair may be the reason for the low baseline results which in turn drive the high % gains with new -ac APs. I do not think your 4.2x higher downlink throughput results represent something that can be generalized to be valid for typical users.
I have also AC68U AP at home. I completely agree that it's a great router (when running -n only, -ac disabled) and works clearly better than my other ones. The magnitude of difference is what I do not find feasible.
Unfortunately Wi-Fi equipment have all kinds of issues which may cause low performance, including driver problems with specific versions and interoperability issues between certain chip sets. Problems include also devices not following the made settings, like even though you set bandwidth to HT20 it does not yet mean that's the case. To be sure, you would need to monitor what really happens in the air interface, does the -ac AP really stick to HT20 continuously and make results comparable? Poorly operating client-AP pair may be the reason for the low baseline results which in turn drive the high % gains with new -ac APs. I do not think your 4.2x higher downlink throughput results represent something that can be generalized to be valid for typical users.
I have also AC68U AP at home. I completely agree that it's a great router (when running -n only, -ac disabled) and works clearly better than my other ones. The magnitude of difference is what I do not find feasible.
sfx2000
Part of the Furniture
@ VPonwireless - fair comments - this, I think was a quick test, and while perhaps not as instrumented as those of us that eat WiFi cheerios for breakfast - the tests were sound, and the methodology repeatable and consistent.
Tim Higgins does not have all the Lab gear that developers might have, where we can get very deep into the PHY/Radios, and the ability to peek/poke bits in the WiFi chipset - but at that level, we're not talking consumer level activities.
RSSI in his tests are what's reported by the NIC driver, which as we all know, isn't the most reliable of constructs.. the same card when presented with VHT20/40 parameters can seriously go off the deep-end, depending on driver versions - don't get me wrong, the Intel 7260 is a fine card, but sometimes it burps odd info to the WDM layer in the Windows stack.
I stand behind Tim's article, as I've also done the free-space casual testing, as well as on the bench and actually measuring things out with a reference Agilent box, across multiple NIC vendors (Intel, Broadcom, QC-Atheros, RealTek, etc...) - and my observations, while test methodology was different, concur with his findings.
You seem like a really smart guy, and we welcome your participation in our little corner of the web.
sfx
Tim Higgins does not have all the Lab gear that developers might have, where we can get very deep into the PHY/Radios, and the ability to peek/poke bits in the WiFi chipset - but at that level, we're not talking consumer level activities.
RSSI in his tests are what's reported by the NIC driver, which as we all know, isn't the most reliable of constructs.. the same card when presented with VHT20/40 parameters can seriously go off the deep-end, depending on driver versions - don't get me wrong, the Intel 7260 is a fine card, but sometimes it burps odd info to the WDM layer in the Windows stack.
I stand behind Tim's article, as I've also done the free-space casual testing, as well as on the bench and actually measuring things out with a reference Agilent box, across multiple NIC vendors (Intel, Broadcom, QC-Atheros, RealTek, etc...) - and my observations, while test methodology was different, concur with his findings.
You seem like a really smart guy, and we welcome your participation in our little corner of the web.
sfx
sfx2000
Part of the Furniture
And while my results don't reflect "huge" improvements, the results are measurable and repeatable - and there is a reasonable amount of improvement, both close-in, and at the edge of coverage.
It's the papa-john's of wireless - better MAC SW + better Radios/Baseband = better Wifi
sfx
It's the papa-john's of wireless - better MAC SW + better Radios/Baseband = better Wifi
sfx
VPonwireless
New Around Here
Sfx, thiggins and zerodegrekelvin, appreciate your responses and welcome. Please, do not get me wrong. I do like very much the tests and great reports presented by snb. I look forward to see new tests and reports! I find the snb forum topics very interesting.
Since I do enterprise/operator Wi-Fi/wireless performance for living, I have encountered all kinds of surprises and findings. I try to share them regularly at different forums, like Wireless LAN Pro's conferences (WLPC), IEEE802.11HEW/ax standardization meetings and our company blog.
Since I do enterprise/operator Wi-Fi/wireless performance for living, I have encountered all kinds of surprises and findings. I try to share them regularly at different forums, like Wireless LAN Pro's conferences (WLPC), IEEE802.11HEW/ax standardization meetings and our company blog.
sfx2000
Part of the Furniture
I think most of the folks you mentioned are in the same camp as you - we're doing WiFi as a living, and that you've joined our community is a very good thing - please stick around... We don't see you as someone who is attacking the group, but as someone who can contribute in a constructive manner.
You'll find folks with a lot of practical experience here... and great opportunities to share insight.
I've done the IEEE 802 rounds, so I get the challenges there.
sfx
if you have any questions/concerns - feel free to pm me here...
sfx2000
Part of the Furniture
FWIW - you're taking an 11ac NIC and putting it into 11n - should see ok performance there in 5GHz - actually since the code running in the chipsets, it should also be a good 11n compared to earlier chipsets..
My concerns are putting VHT20/40 into 2.4Ghz... there's currently not a clean way to do this, and no clear consensus upon vendors, as this goes well beyond what is published in 802.11-2012, and WiFi Alliance isn't taking the initiative here..
the RT-AC68U is a favorite for many on SNB - I have issues with the board support package and how this resolves into different routing use-cases, but as an AP, it's a decent device...
VPonwireless
New Around Here
____________________________
At home, I run AC68 in AP mode. Initially with AC68 (with default config running -ac rates) I had problems using Skype from the next room with iPhone 6 at 5 GHz. Voice quality was very poor. Then I updated the FW to 3.0.0.4.376_3626 and problems disappeared. To my surprise, FW upgrade also disabled -ac completely. Ever since, the AP has been running also voice nicely, but its not any more using -ac at all.
With -n and even more with -ac, AP rate control has significant room for improvement. It's very typical with APs that rate control tries to use very high rates and then packets fail. Rate control continuously uses too high rates and this results as very high retransmission rates. This in turn quickly becomes a problem for real time applications like voice. In addition, overall network capacity drops since air utilization goes up due to high retry rates. It's not uncommon to have AP retry rates at 20-40% range or even higher. Clients seem to handle rates better than AP and typically retry rates seem to be about half of the rates that APs use. My experience at home where voice worked better without -ac rates, seems to correlate with this also.
Having -ac rates (VHT20/40) supported at 2.4 GHz has gained some support with 802.11ax standardization as one item and for example Cisco has been pushing that. But this takes long time. In the mean time, we have proprietary implementations. Not sure really what is the value unless rate control algorithms also improve.
sfx2000
Part of the Furniture
Hmmm... well, with your AC68, losing 11ac mode isn't good - have you tried a hard reset on the unit?
http://www.snbforums.com/threads/faq-nvram-and-factory-default-reset.22822/
Also, iDevices seem like like having 11h mode enabled - I think for the factory firmware, this is turned off, but one can jump into the shell and query/set those nvItems...
Tend to agree with you on Rate Adaptation - spec says what to do, but they don't define the thresholds/watermarks - that's left to implementation...
I've been following 11ax, but with a proposed final spec around 2019, my guess is that it's going to be a while...
The VHT20/40 in 2.4GHz - while non-standard, starting to see that vendors are using general consensus to get everyone on the same page as a defacto standard at least. But I've seen drivers where they crash as they run out of memory... I generally recommend going mixed b/g/n in 2.4Ghz, and disabling the VHT modes...
http://www.snbforums.com/threads/faq-nvram-and-factory-default-reset.22822/
Also, iDevices seem like like having 11h mode enabled - I think for the factory firmware, this is turned off, but one can jump into the shell and query/set those nvItems...
nvram set wl0_reg_mode=h
nvram set wl1_reg_mode=h
nvram commit
reboot
nvram set wl1_reg_mode=h
nvram commit
reboot
Tend to agree with you on Rate Adaptation - spec says what to do, but they don't define the thresholds/watermarks - that's left to implementation...
I've been following 11ax, but with a proposed final spec around 2019, my guess is that it's going to be a while...
The VHT20/40 in 2.4GHz - while non-standard, starting to see that vendors are using general consensus to get everyone on the same page as a defacto standard at least. But I've seen drivers where they crash as they run out of memory... I generally recommend going mixed b/g/n in 2.4Ghz, and disabling the VHT modes...
VPonwireless
New Around Here
Thanks. I have not tried reset. -ac is disabled even in the settings since the SW upgrade. Could try to activate it there first. On the other hand, now it works better for voice/Skype than before when -ac was enabled...
Standardization would need to include true performance requirements for equipment.
Standardization would need to include true performance requirements for equipment.
azazel1024
Very Senior Member
The best way to do it is simply getting higher gain antennas. A more powerful radio/amp can often lead to bad things(tm). On top of that, you need a more powerful radio on both the basestation and client to get truely meaningful gains. So basically bridges only is where you might see true benefits.
sfx2000, did I miss something? 40MHz/VHT mode in 2.4GHz is standards compliant for 802.11n. It is listed there in the standard. However, the wifi alliance requires (wifi alliance is a trade group, they don't control the IEEE 802.11 specs) products to back down to 20MHz in 2.4GHz if they detect an overlapping 2.4GHz network within a certain RSSI.
That doesn't mean that 2.4GHz 40MHz is often a good idea, but it IS standards compliant. It is one of the 802.11n-2009 optional extensions. About the only company I know of that does not enable this option is Apple. Absolutely everyone else allows this, and technically Apple implements it in a fashion too, as they set the fat channel intolerant flag (if the Apple product was completely 40MHz incompatible, it wouldn't even set this flag), so when a 40MHz enabled basestation communicates with an Apple client is backs off to 20MHz during Tx and Rx to the Apple client before resuming 40MHz operation.
What I'd love to see, but I am pretty sure I won't, is a big wad of the 3.6GHz spectrum added to the ISM spectrum for unlicensed usage. A nice 160MHz chunk would be awesome. Granted, incumbents, but I can dream. IMHO, 2.4GHz is often too penetrative in urban and suburban environments, but 3.6GHz would be a nice compromise between being penetrative and non-penetrative. Probably still wouldn't be great for dense urban environments, but it would be a lot better than 2.4GHz and 5.2-5.9GHz both in a lot of ways. Likely will never happen. The best I can probably hope for is the 5GHz DFS channels to have some restrictions eased on them and/or incumbents shifted some to free up a few more channels.
sfx2000, did I miss something? 40MHz/VHT mode in 2.4GHz is standards compliant for 802.11n. It is listed there in the standard. However, the wifi alliance requires (wifi alliance is a trade group, they don't control the IEEE 802.11 specs) products to back down to 20MHz in 2.4GHz if they detect an overlapping 2.4GHz network within a certain RSSI.
That doesn't mean that 2.4GHz 40MHz is often a good idea, but it IS standards compliant. It is one of the 802.11n-2009 optional extensions. About the only company I know of that does not enable this option is Apple. Absolutely everyone else allows this, and technically Apple implements it in a fashion too, as they set the fat channel intolerant flag (if the Apple product was completely 40MHz incompatible, it wouldn't even set this flag), so when a 40MHz enabled basestation communicates with an Apple client is backs off to 20MHz during Tx and Rx to the Apple client before resuming 40MHz operation.
What I'd love to see, but I am pretty sure I won't, is a big wad of the 3.6GHz spectrum added to the ISM spectrum for unlicensed usage. A nice 160MHz chunk would be awesome. Granted, incumbents, but I can dream. IMHO, 2.4GHz is often too penetrative in urban and suburban environments, but 3.6GHz would be a nice compromise between being penetrative and non-penetrative. Probably still wouldn't be great for dense urban environments, but it would be a lot better than 2.4GHz and 5.2-5.9GHz both in a lot of ways. Likely will never happen. The best I can probably hope for is the 5GHz DFS channels to have some restrictions eased on them and/or incumbents shifted some to free up a few more channels.
sfx2000
Part of the Furniture
Higher gain antennas - generally don't help for the most part - this is a myth propagated by vendors that can sell a 1 dollar part for 20 dollars... but people buy them, and belief bias make folks think they work.
Higher gain antennas increase both Signal and Noise on the channel at the AP they're installed on - so basic a near net-zero gain.
VHT20/VHT40 is a backport of 11ac, it's not included in 11n - I think your confusing HT with VHT...
VHT is 11ac - and 11ac was designed specifically to exclude the 2.4GHz ISM band in the IEEE 802.11ac working group.
Now we're seeing chipset vendors deploy VHT modes in 2.4 in any case - I suppose the only good news in that is that while initially, they did different approaches (BRCM, for example, advertised TurboQAM as a vender specific attribute, whereas Marvell, being second, actually put VHT stanza's into their beacon frames) - seems like Marvell's approach was perhaps a better one, and VHT is slowly becoming a defacto standard if not outright formal - VHT is 11ac in 2.4GHz for all intents and purposes, but nobody will call it that...
WideChannels are in 2.4GHz 11n because if they weren't included, there was a camp that would have walked out of TG-n and we probably wouldn't have a spec at all... it was a compromise to move things forward... I know, I was there in the meeting when that compromise was presented - I was working on 802.16m at the time, but a lot of us were following what was going in over in the TG-n group.
Apple has been one of the few that insist on not using Wide Channels in 2.4GHz - and for valid reasons - think Bluetooth which Apple is heavily into integrating into all of the products.
I like their approach to be honest..
Up until the Airport Extreme AC, Apple used to set the Fat Channel Intolerant bit on their AP's, so other vendor AP's, when, and if they properly do an OBSS scan, should actually disable the secondary channel, at least for a while...
The Wide Channel debate has been going on for over ten years now - started with Atheros and their SuperG bonding...
When looking at WideChannels in 2.4- few things...
1) Wide Channels really don't double actual transfer rates, it's more along the line of 30 percent at best
2) A loss of about 3dB in most client/AP link budgets with WideChannels - in 2.4GHz, this is huge, as 3dB is half power..
3) WideChannels in most environments are impacted by overlapping other AP's - the signal there will be non-linear across the sub-carriers
4) Interference with other users/technologies using the ISM band - WiFi isn't the only player
So when you take these things into consideration, I see very little benefit of running wide channels in 2.4GHz, and a lot of reasons why not to.
VPonwireless
New Around Here
azazel1024
Very Senior Member
Bah, yeah, I was thinking HT40, not VHT40. It should really be standardized at some point, if for no other reason than there are at least times it can help (and VHT20 as an extension of that reasoning).
In response to your numbers
1) I'd imagine this depends heavily on the SINR. In my low interference semi-rural environment, the >dozen different APs, extenders and routers I have tested, I think I average about 105% increase in performance, roughly in line with the modulation rate difference. I have seen a few where it is lower, but generally even then you are talking an 80-90% gain in performance. Example, my Archer C8 at close range jumps from 14.2MB/sec to 28.5MB/sec. even at fairly extreme range it jumps from 2.5MB/sec to 3.5MB/sec. It generally takes the absolute edge of reception for 20MHz to perform better than 40MHz. Exception being, a lot of interfernence
2) True, but again see where that is only an issue in high interference environments and/or extreme range
3) True for high interference environments. Many people don't operate in such environments. Not allowing something that doesn't impact a large number of people is silly. I am more than agreeable to better and mandatory neighboring SSID detection to bump down link rates to 20MHz
4) Trying to test the crap out of possible interference sources, 20MHz seems to introduce as much interference as 40MHz (testing both bluetooth and 2.4GHz dongles), as in either I see interference or I don't, and the magnitude of the interference appears roughly similar to both the wifi and the other 2.4GHz user. As for received interference, testing things like microwave ovens, 2.4GHz baby monitors (I have a 2.4GHz set and a 900MHz set) and 2.4GHz phones (I have an old one, switched to DEC only recently), 40MHz still performs much better than 20MHz when receiving interference from those sources and the magnitude of the received interference appears roughly similar (ex. when setting my laptop directly under my microwave, starting a Rx or Tx file transfer and then turning the microwave on, I see roughly a 20% reduction in performance with 40MHz channels, with 20MHz I see roughly the same).
I see every reason to run 40MHz channels and little reason not to, unless you are in a crowded environment or you need absolutely maximum range. In my own TH, comparing 20MHz to 40MHz with about a dozen nearby networks, 40MHz performed roughly 70% better on the same floor with the router (at the time I only had a 2.4GHz router and client), but about 25% worse than 20MHz if you were in the basement or upstairs. So I choose 20MHz, both to be a better neighbor, but also becuase performance not on the same floor was SO bad, I needed to get every last bit (same floor to my laptop I might be able to get 9-10MB/sec on 2.4GHz 20MHz and about 16-17MB/sec on 40MHz, but a floor up or down it was at best 2-3MB/sec and often as little as 300-500KB/sec.)
For antennas, again, sure, in a high interference situation, that can be true. In a relatively low one, higher gain antennas can help out. I've done quite a bit of testing, granted only with 3 different routers, but alternating between 3, 5 and 7dBi omnis and 2 different manufacturers' antennas. There are diffinitely times where a lower gain can perform better than a higher gain, but in most cases I see about a 15-20% gain each time you step up antenna gain by 2dBi at medium and long range. It isn't across the board and it is mostly 5GHz gains, 2.4GHz seems to be more like 5-10%. 802.11ac deffinitely seems to benefit a LOT from higher gain antennas, probably because of the SINR requirements for 256QAM to work. With my 11ac routers, on 5GHz, I see nice 15-20% gains even at short range (example, my Archer C8, line-of-sight and 10ft from the router going from 3dBi antennas to 5dBi results in an increase from 55MB/sec receive to 61MB/sec receive and 40MB/sec transmit to 46MB/sec transmit).
In response to your numbers
1) I'd imagine this depends heavily on the SINR. In my low interference semi-rural environment, the >dozen different APs, extenders and routers I have tested, I think I average about 105% increase in performance, roughly in line with the modulation rate difference. I have seen a few where it is lower, but generally even then you are talking an 80-90% gain in performance. Example, my Archer C8 at close range jumps from 14.2MB/sec to 28.5MB/sec. even at fairly extreme range it jumps from 2.5MB/sec to 3.5MB/sec. It generally takes the absolute edge of reception for 20MHz to perform better than 40MHz. Exception being, a lot of interfernence
2) True, but again see where that is only an issue in high interference environments and/or extreme range
3) True for high interference environments. Many people don't operate in such environments. Not allowing something that doesn't impact a large number of people is silly. I am more than agreeable to better and mandatory neighboring SSID detection to bump down link rates to 20MHz
4) Trying to test the crap out of possible interference sources, 20MHz seems to introduce as much interference as 40MHz (testing both bluetooth and 2.4GHz dongles), as in either I see interference or I don't, and the magnitude of the interference appears roughly similar to both the wifi and the other 2.4GHz user. As for received interference, testing things like microwave ovens, 2.4GHz baby monitors (I have a 2.4GHz set and a 900MHz set) and 2.4GHz phones (I have an old one, switched to DEC only recently), 40MHz still performs much better than 20MHz when receiving interference from those sources and the magnitude of the received interference appears roughly similar (ex. when setting my laptop directly under my microwave, starting a Rx or Tx file transfer and then turning the microwave on, I see roughly a 20% reduction in performance with 40MHz channels, with 20MHz I see roughly the same).
I see every reason to run 40MHz channels and little reason not to, unless you are in a crowded environment or you need absolutely maximum range. In my own TH, comparing 20MHz to 40MHz with about a dozen nearby networks, 40MHz performed roughly 70% better on the same floor with the router (at the time I only had a 2.4GHz router and client), but about 25% worse than 20MHz if you were in the basement or upstairs. So I choose 20MHz, both to be a better neighbor, but also becuase performance not on the same floor was SO bad, I needed to get every last bit (same floor to my laptop I might be able to get 9-10MB/sec on 2.4GHz 20MHz and about 16-17MB/sec on 40MHz, but a floor up or down it was at best 2-3MB/sec and often as little as 300-500KB/sec.)
For antennas, again, sure, in a high interference situation, that can be true. In a relatively low one, higher gain antennas can help out. I've done quite a bit of testing, granted only with 3 different routers, but alternating between 3, 5 and 7dBi omnis and 2 different manufacturers' antennas. There are diffinitely times where a lower gain can perform better than a higher gain, but in most cases I see about a 15-20% gain each time you step up antenna gain by 2dBi at medium and long range. It isn't across the board and it is mostly 5GHz gains, 2.4GHz seems to be more like 5-10%. 802.11ac deffinitely seems to benefit a LOT from higher gain antennas, probably because of the SINR requirements for 256QAM to work. With my 11ac routers, on 5GHz, I see nice 15-20% gains even at short range (example, my Archer C8, line-of-sight and 10ft from the router going from 3dBi antennas to 5dBi results in an increase from 55MB/sec receive to 61MB/sec receive and 40MB/sec transmit to 46MB/sec transmit).
sfx2000
Part of the Furniture
Well, perhaps in your case - WiFi networks, like I've said before are like snowflakes - each one is unique... you're probably an edge case where you're not fighting 20 other overlapping AP's...
In any event...
WiFI is inherently reverse link limited - Access Points generally have more gain and Tx power than the clients - putting big antennas or turning up Tx power compound the issue - that's why we have design link budgets...
Location helps much more than big antennas, more power... put the AP's where the people are - most of the issues regarding throughput and connectivity here are mainly due to poor locations of either the AP's or the clients.
And I will absolutely, and respectfully, disagree with you on 40MHz channels in 2.4GHz... might work for you in an edge case, but in general best practices, stick with narrow channels... many times it'll get better range, and similar if not better performance.
I see very little need to even consider enabling VHT20/40 on the Access Points, as clients that support it are extremely rare, and no need to introduce interoperability issues in the 2.4GHz band.
In any event...
WiFI is inherently reverse link limited - Access Points generally have more gain and Tx power than the clients - putting big antennas or turning up Tx power compound the issue - that's why we have design link budgets...
Location helps much more than big antennas, more power... put the AP's where the people are - most of the issues regarding throughput and connectivity here are mainly due to poor locations of either the AP's or the clients.
And I will absolutely, and respectfully, disagree with you on 40MHz channels in 2.4GHz... might work for you in an edge case, but in general best practices, stick with narrow channels... many times it'll get better range, and similar if not better performance.
I see very little need to even consider enabling VHT20/40 on the Access Points, as clients that support it are extremely rare, and no need to introduce interoperability issues in the 2.4GHz band.
