Continue reading on SmallNetBuilder.
Introduction To The TR-398 Wi-Fi Performance Test
- Thread starter thiggins
- Start date
Continue reading on SmallNetBuilder.
sfx2000
Part of the Furniture
Thanks, but I link to the spec two times in the What Is TR-398? section.
CrystalLattice
Regular Contributor
Right now a $35 piece outperforms most of the $200 wifi 5 pieces, so why in the world would anyone spend more than $35? Or $200?
What are you referring to?
Please note I've only shown results from four of the 11 tests (10 mandatory) in TR-398. So you can't draw any bottom line conclusions from the data.
sfx2000
Part of the Furniture
sfx2000
Part of the Furniture
Interesting test results in the article...
Do note that there have been performance and stability issues with the ath10k driver in both FOSS and QSDK versions for Wave 1 and Wave 2 for Peregrine - QCA9882 which is what the Linksys LAPAC1200 uses for the 5GHz side.
The QCA9557 (Scorpion) is old-school MIPS74k based for B/G/N, and is a solid performer across the board - and the test results show this.
The IPQ40xx, typically IPQ4018/IPQ4019 - is what many of the Mesh consumer devices use for 11ac for 2-stream A/B/G/N/AC - and there, the QSDK board support package is going to offer a lot of benefits, so finding a device built on QSDK is a better reference.
I suggest looking at something based on QSDK on IPQ40xx, which is also QSDK based (if licensed), where Wave 2 (and Wave 1) performance is better as an AC1200 class device.
(edit - merged post)
Anyways - I would suggest running the same test sequence across...
1) RT-AC68U - classic Wave 1 Broadcom design (or R7000) - Broadcom nailed early Wave1 AC1900 class and they're old but still great performers
2) WRT1900acV2/WRt1900acs - Wave 1 with Marvell's Wave 1 solution, which can be a strong player, SW not withstanding, but the Closed Source drivers are quote good properly configured (Linksys defaults are good for 5GHz out of the box, the 2.4GHz needs to be b/g/n mixed, and there, they're very good)
3) Netgear R7800 or Synology AC2600 - good QCA platform for Wave 2 on IPQ80xx
Reason for recommending most of these specific platforms is they all have SMA connections suitable for cabling up... and a good cross section for the community..
4) Trying to find a good replacement for IPQ40xx - gl.iNet's B1300 is a great choice for QSDK, but it's internal antennas only (as most of the fielded devices are), but it's a standalone router that is very close to Qualcomm's QSDK with all the bells and whistles as an AC1200 class Wave 2 device - it makes few excuses there.
Do note that there have been performance and stability issues with the ath10k driver in both FOSS and QSDK versions for Wave 1 and Wave 2 for Peregrine - QCA9882 which is what the Linksys LAPAC1200 uses for the 5GHz side.
The QCA9557 (Scorpion) is old-school MIPS74k based for B/G/N, and is a solid performer across the board - and the test results show this.
The IPQ40xx, typically IPQ4018/IPQ4019 - is what many of the Mesh consumer devices use for 11ac for 2-stream A/B/G/N/AC - and there, the QSDK board support package is going to offer a lot of benefits, so finding a device built on QSDK is a better reference.
I suggest looking at something based on QSDK on IPQ40xx, which is also QSDK based (if licensed), where Wave 2 (and Wave 1) performance is better as an AC1200 class device.
(edit - merged post)
Anyways - I would suggest running the same test sequence across...
1) RT-AC68U - classic Wave 1 Broadcom design (or R7000) - Broadcom nailed early Wave1 AC1900 class and they're old but still great performers
2) WRT1900acV2/WRt1900acs - Wave 1 with Marvell's Wave 1 solution, which can be a strong player, SW not withstanding, but the Closed Source drivers are quote good properly configured (Linksys defaults are good for 5GHz out of the box, the 2.4GHz needs to be b/g/n mixed, and there, they're very good)
3) Netgear R7800 or Synology AC2600 - good QCA platform for Wave 2 on IPQ80xx
Reason for recommending most of these specific platforms is they all have SMA connections suitable for cabling up... and a good cross section for the community..
4) Trying to find a good replacement for IPQ40xx - gl.iNet's B1300 is a great choice for QSDK, but it's internal antennas only (as most of the fielded devices are), but it's a standalone router that is very close to Qualcomm's QSDK with all the bells and whistles as an AC1200 class Wave 2 device - it makes few excuses there.
Last edited:
Tim,
I wish you the best on this change of direction.
I followed SNB for years. It was the best source of information for those of us who wanted to get them most out of our home network but didn't have the time to become network engineers.
For the past couple of years I have not spent much time on the site because it did not seem particularly relevant to me. My network issues were never fixed by the latest or greatest technology or the fastest connection. My network issues stemmed from several clients all competing for air time and access to the upstream internet link.
The ability to gracefully handle voip calls and a netflix stream or two while someone downloads a file is more important than raw AP to client speed.
Looking forward to your upcoming articles and reviews.
I wish you the best on this change of direction.
I followed SNB for years. It was the best source of information for those of us who wanted to get them most out of our home network but didn't have the time to become network engineers.
For the past couple of years I have not spent much time on the site because it did not seem particularly relevant to me. My network issues were never fixed by the latest or greatest technology or the fastest connection. My network issues stemmed from several clients all competing for air time and access to the upstream internet link.
The ability to gracefully handle voip calls and a netflix stream or two while someone downloads a file is more important than raw AP to client speed.
Looking forward to your upcoming articles and reviews.
L&LD
Part of the Furniture
You've missed a lot then!
Take a look at what amtm, (Diversion, pixelserv-tls, Entware), and the many scripts it supports can do for you.
The most relevant script doesn't even need the above (as of today), FreshJR QOS script should tame your network to work as you want it to.
https://www.snbforums.com/threads/r...ements-custom-rules-and-inner-workings.36836/
Even with a default installation, the improvements are dramatic. And FreshJR is working on a new version that will bring even more enhancements and capabilities soon.
L&LD
Part of the Furniture
And not a single word about radiation, that word is apparently taboo in the business
While I can see the potential need to have this aspect covered, what would be providing this information do for anyone of us? Even if we opted to not have WiFi on our premises, we are still subject to it everywhere else (and depending on circumstances, even within our homes).
And not a single word about radiation, that word is apparently taboo in the business
That part is already evaluated by regulatory bodies such as the FCC. You can see all the complete test data in FCC filings. Here's an example:
https://fccid.io/MSQ-RTHU00/RF-Exposure-Info/RF-Exposure-4071845
pege63
Very Senior Member
Yes but whoes data is that owned by whom, ordered by whom, is this is an independent measurement?
Or is it done by the industry itself on their terms, done by them self?
Those are done by specialized labs, who must provide complete test data to the manufacturer when they file with the FCC. So yes, they are independent, and falsifying test results could lead them in serious legal troubles. They are more trustworthy than what any typical reviewer might provide because of their legal liability.
Also, TR-398 is about wifi performance. EM radiations are unrelated to performance.
sinshiva
Very Senior Member
i'm right there with you. ax/6 has restored some faith that it all isn't just a race to more bandwidth for me, though
pege63
Very Senior Member
Does this only applies the United States where you can be sued for this or for the whole world itself?
But still it only describes the measurement of the Power Density and nothing about the Electric Field and Magnetic Field, why are these fields empty?
Last edited:
Ben Greear
New Around Here
Hello,
I have been working on implementing TR-398. I found that it is difficult to determine the proper attenuation since RSSI is never correlated with the attenuation or theoretical path loss specified in the document. I notice that you had some issues with the short/medium/far attenuation configuration as well. We did a 2-meter OTA distance test between a Netgear R7800 and one of our test systems, with /a/b/g station (to force high TX power and single stream) and round that RSSI on the receiving station was -24 on 2.4Ghz and -32 on 5Ghz. If we treat this as 'zero attenuation', then we can mostly pass the rate-vs-range and rx-sensitivity tests if we decrease over-all attenuation by 3 db as a fudge factor.
The receiver-sensitivity test requires attenuation-from-2m-baseline of 56 for 2.4Ghz MCS0 traffic. Adding to the baseline we recorded, that yields -80 RSSI as the cutoff point. I am very interested to know what RSSI you see at some of these steps (for instance the short/medium/long range test) and/or how you plan to calibrate this test. If we can all
use similar logic to calibrate, it should help anyone trying to compare results.
I have been working on implementing TR-398. I found that it is difficult to determine the proper attenuation since RSSI is never correlated with the attenuation or theoretical path loss specified in the document. I notice that you had some issues with the short/medium/far attenuation configuration as well. We did a 2-meter OTA distance test between a Netgear R7800 and one of our test systems, with /a/b/g station (to force high TX power and single stream) and round that RSSI on the receiving station was -24 on 2.4Ghz and -32 on 5Ghz. If we treat this as 'zero attenuation', then we can mostly pass the rate-vs-range and rx-sensitivity tests if we decrease over-all attenuation by 3 db as a fudge factor.
The receiver-sensitivity test requires attenuation-from-2m-baseline of 56 for 2.4Ghz MCS0 traffic. Adding to the baseline we recorded, that yields -80 RSSI as the cutoff point. I am very interested to know what RSSI you see at some of these steps (for instance the short/medium/long range test) and/or how you plan to calibrate this test. If we can all
use similar logic to calibrate, it should help anyone trying to compare results.
Hi Ben,
You're correct that translating the "2 meter distance" that is often referenced in TR-398 into actual attenuator settings is a challenge. You can't just set your attenuators to open air path loss values from the equation given in TR-398 5.2.2 paragraph C (P_L (dB)=20 Log10 (f_GHz ) + 20 Log10 (d_meters ) + 32.45), i.e. 46 dB for 2.4 GHz, 53 dB for 5 GHz, because there are other path losses due to cabling and antennas used.
In my setup, I have the additional concern of not exceeding the recommended maximum input level of -30 dBm for the octoScope PAL devices that I use as STAs. What I've found is that my setup generally requires that I add 18 dB of attenuation to any 2.4 GHz attenuation values stated in the TR-398 test specs and 0 dB for 5 GHz.
In the end, attenuation offsets are going to be specific to testbed configurations. So for whatever method you use, once you determine the values, you must use them for all devices you test to get valid relative performance comparisons. That's the approach I've always taken in all my testing.
Note that when establishing your values, use a four-stream DUT, since the higher number of streams will increase link gain. IMO, the R7800 is a good choice, since it's a four-stream AC router that's been around for awhile.
If you'd like to discuss this further, start a private conversation (click on my avatar) and shoot me your email.
You're correct that translating the "2 meter distance" that is often referenced in TR-398 into actual attenuator settings is a challenge. You can't just set your attenuators to open air path loss values from the equation given in TR-398 5.2.2 paragraph C (P_L (dB)=20 Log10 (f_GHz ) + 20 Log10 (d_meters ) + 32.45), i.e. 46 dB for 2.4 GHz, 53 dB for 5 GHz, because there are other path losses due to cabling and antennas used.
In my setup, I have the additional concern of not exceeding the recommended maximum input level of -30 dBm for the octoScope PAL devices that I use as STAs. What I've found is that my setup generally requires that I add 18 dB of attenuation to any 2.4 GHz attenuation values stated in the TR-398 test specs and 0 dB for 5 GHz.
In the end, attenuation offsets are going to be specific to testbed configurations. So for whatever method you use, once you determine the values, you must use them for all devices you test to get valid relative performance comparisons. That's the approach I've always taken in all my testing.
Note that when establishing your values, use a four-stream DUT, since the higher number of streams will increase link gain. IMO, the R7800 is a good choice, since it's a four-stream AC router that's been around for awhile.
If you'd like to discuss this further, start a private conversation (click on my avatar) and shoot me your email.
Similar threads
