Usb 3.0 device could be slowing down your wifi on you Mac.

[srinivasvaradaraj]

Folks,

Short version:
Your usb 3.0 device could be slowing down your wifi on you Mac. Read on ..

Long version:

I figured I would write this up to help save time for others. For the past weekend I was driving myself crazy trying to figure out the reason. This isn’t a technical explaination but an observation and possibly a resolution.

So here is my setup:
Late 2013 macpro (cylinder edition), OSX 10.10.2 .
Asus RT-AC68R.
The router and the computer are on the opposite ends of the same room, approx 6-7 feet apart.
There is only one device that (this computer) connects on this AP (AC via 3x3).

For the most part the mac and the router are stable with great TX and RX speeds. However, last weekend, I noticed that my mac would not negotiate beyond 866.7Mbps (2x2). I had recenty updated the asus firmware from 3.0.0.4.374.5656 to 3.0.0.4.378.4376 (latest at the time of this write up). As always I had reset and rebuilt the router after the firmware updates. So, naturally given this was the most recent change, I assumed that this was the cause for the reduced negotation and reduced throughput.

So, to troubleshoot and fix this I downgraded the router several times, tried alternate firmwares (Merlins) with different versions but none solved the issue. I then tried all the normal wifi troubleshooting, changing channels, setting the router to auto .. etc etc, none solved the issue.

Finally, it occurred to me that problem may actually be on the Mac side rather than the router itself. To confirm this, following several threads on mac focussed forums, I tried to downgrade the driver from 7.1 to 7.0 for the computer (.kext modification). However, this didn’t work. I did all the normal mac stuff Cmd+Opt+P+R.. but none of those worked either.

I then enabled wifi logging on the mac, Option+click on wifi. I studied the logs only find out the when wifi was toggled there was a reference to USB 3.0 devices. This give me the clue that something with USB 3.0 was potentially causing issues. I read several papers about the interplay between USB 3.0 and 2.4Gz

https://support.apple.com/en-us/HT201163(last one)
http://www.usb.org/developers/whitepapers/327216.pdf

I knew that unshielded usb 3.0 interfered with 2.4GHz but didn’t know about possible 5.0GHz. To verify if this was the cause, I disconnected all USB3.0 devices from the mac and sure enough I was back to 1300mbps TX and RX. I was able to reliably re-create the problem by disconnecting and reconnecting the USB 3.0 external HDD.

Anyway, I am thinking that the issue may not be direct interference from the device/cable but is a compensation at the driver level. The engineers may have chosen to drop the wifi card to a 2x2 configuration for both 2.4GHz and 5GHz to stablize wifi in the presence of USB 3.0 device.

Here are some throughput speeds between same two hosts on my network.

Without USB 3.0 device plugged in:
[root@clusterpc2 ~]# iperf -c 10.0.1.146 -t 30 -r
------------------------------------------------------------
Server listening on TCP port 5001
TCP window size: 85.3 KByte (default)
------------------------------------------------------------
------------------------------------------------------------
Client connecting to 10.0.1.146, TCP port 5001
TCP window size: 65.7 KByte (default)
------------------------------------------------------------
[ 5] local 10.0.1.76 port 39302 connected with 10.0.1.146 port 5001
[ ID] Interval Transfer Bandwidth
[ 5] 0.0-30.0 sec 2.27 GBytes 650 Mbits/sec
[ 4] local 10.0.1.76 port 5001 connected with 10.0.1.146 port 50058
[ 4] 0.0-30.0 sec 2.77 GBytes 793 Mbits/sec

With USB 3.0 device plugged in:
[root@clusterpc2 ~]# iperf -c 10.0.1.146 -t 30 -r
------------------------------------------------------------
Server listening on TCP port 5001
TCP window size: 85.3 KByte (default)
------------------------------------------------------------
------------------------------------------------------------
Client connecting to 10.0.1.146, TCP port 5001
TCP window size: 65.7 KByte (default)
------------------------------------------------------------
[ 4] local 10.0.1.76 port 39303 connected with 10.0.1.146 port 5001
[ ID] Interval Transfer Bandwidth
[ 4] 0.0-30.0 sec 1.52 GBytes 436 Mbits/sec
[ 5] local 10.0.1.76 port 5001 connected with 10.0.1.146 port 51855
[ 5] 0.0-30.0 sec 1.99 GBytes 569 Mbits/sec

Conlusion:
Given the state that Apple devs have chosen to reduce wifi speeds in the presence of USB 3.0 devices, and we know this is the cause. Its sad that we can’t have both on the Mac platform given these combinations.

 

[azazel1024]

There is no interference on 5GHz from USB3. Zero, totally different frequency realm (its in the 2.4GHz area). Now there might be a driver/OS bug going on with the Mac, but it has nothing to do with direct or indirect interference on 5GHz. To properly compensate for interferance on 2.4GHz, you would NOT want to drop spatial streams, you'd want to drop frequency width to 20MHz, which all Macs do anyway (they are not 40MHz 2.4GHz capable).

Also, on the Mac side, if the USB3 ports are properly positioned and shielded and you aren't running the USB3 cable right up by the wifi antennas (which are in the screen IIRC), there should be no effective 2.4GHz interference either.

My HP Envy 4t with an Intel 7260ac in it shows no USB3 interference with a USB3 thumb drive as well as external HDD connected on 2.4GHz. Performance with no USB3 devices plugged in

25.2MB/sec on 2.4GHz.

w/ USB3 devices plugged in, 25.2MB/sec 2.4GHz.

At long range from my router no USB3 devices plugged in, 2.7MB/sec 2.4GHz

with USB3 devices plugged in, 2.7MB/sec

The USB3 devices were active at the time (transfering files between the thumb drive and the USB attached HDD).

The distance between the USB3 ports and the wifi antennas is around 10 inches, give or take a bit. The ports are also fully enclosed.

Most routers with properly designed ports also show little to no USB3 interference. Properly grounded and shielded USB3 ports generate little interference, the same with cables and end devices. The amount of noise generated by USB3 is relatively little, so it needs to be VERY close to a wifi antenna unless there is very poor design (even then, needs to be pretty close). The difference between a poor design and a good one is on the order of 30dB of noise difference, which is a LOT. Same with distance, Intel's white paper mostly looked at USB3 ports placed within an inch or two or wifi receivers (IE things like a USB Wifi adapter in the port next to a USB3 hard drive, both ports unshielded).

Well shielded + proper distance from wifi antennas can easily put USB3 interference below the typical 2.4GHz noise floor. Put right up next to the wifi receiver with unshielded ports and you can easily swamp something like a bluetooth reciever (which is extremely low power) or 2.4GHz performance at medium to long ranges.

 

[srinivasvaradaraj]

There is no interference on 5GHz from USB3. Zero, totally different frequency realm (its in the 2.4GHz area). Now there might be a driver/OS bug going on with the Mac, but it has nothing to do with direct or indirect interference on 5GHz. To properly compensate for interferance on 2.4GHz, you would NOT want to drop spatial streams, you'd want to drop frequency width to 20MHz, which all Macs do anyway (they are not 40MHz 2.4GHz capable).

Also, on the Mac side, if the USB3 ports are properly positioned and shielded and you aren't running the USB3 cable right up by the wifi antennas (which are in the screen IIRC), there should be no effective 2.4GHz interference either.

My HP Envy 4t with an Intel 7260ac in it shows no USB3 interference with a USB3 thumb drive as well as external HDD connected on 2.4GHz. Performance with no USB3 devices plugged in

25.2MB/sec on 2.4GHz.

w/ USB3 devices plugged in, 25.2MB/sec 2.4GHz.

At long range from my router no USB3 devices plugged in, 2.7MB/sec 2.4GHz

with USB3 devices plugged in, 2.7MB/sec

The USB3 devices were active at the time (transfering files between the thumb drive and the USB attached HDD).

The distance between the USB3 ports and the wifi antennas is around 10 inches, give or take a bit. The ports are also fully enclosed.

Most routers with properly designed ports also show little to no USB3 interference. Properly grounded and shielded USB3 ports generate little interference, the same with cables and end devices. The amount of noise generated by USB3 is relatively little, so it needs to be VERY close to a wifi antenna unless there is very poor design (even then, needs to be pretty close). The difference between a poor design and a good one is on the order of 30dB of noise difference, which is a LOT. Same with distance, Intel's white paper mostly looked at USB3 ports placed within an inch or two or wifi receivers (IE things like a USB Wifi adapter in the port next to a USB3 hard drive, both ports unshielded).

Well shielded + proper distance from wifi antennas can easily put USB3 interference below the typical 2.4GHz noise floor. Put right up next to the wifi receiver with unshielded ports and you can easily swamp something like a bluetooth reciever (which is extremely low power) or 2.4GHz performance at medium to long ranges.

Agreed with everything you said, the thing that bothered me was that explicit drop even with the 5 GHz by the driver when a USB 3.0 device was plugged in. In my case, I could replicate the problem when I plugged a USB 3.0 flashdrive both directly at the computer and via a USB hub. Good points !!

Anyway, I fully intend to report/complain my observations to apple support. If anything will be done or not IDK.

 

[stevech]

USB3 is high speed digital data - square waves. That generates a lot of harmonics and can go well above 5GHz.

I'd think that this would not be an issue if you use good quality cables with good shielding, with ferrite bead or ring filters, and avoid plastic USB hubs,

 

[azazel1024]

USB3 is high speed digital data - square waves. That generates a lot of harmonics and can go well above 5GHz.

I'd think that this would not be an issue if you use good quality cables with good shielding, with ferrite bead or ring filters, and avoid plastic USB hubs,

It can, but look at Intel's white paper. Unless that was in a faraday cage, the results they found show effectively no noise generated in the 4.4-6GHz range from USB3. It is showing almost a flat line from 4.4-6GHz at ~-92dBm/100kHz power level with a TINY spike right at 5GHz up to about -85dBm/100kHz.

This is compared to the 0-~3GHz range which shows -60 to -40dBm/100kHz and roughly -60dBm/100kHz right at 2.4-2.5GHz. About 2,000x the generated noise in the relevant 2.4-2.5GHz band as there is in the 5.2-5.9GHz range (which, again, I'd bet a lot that the noise shown in 4.4-6GHz is background noise).

The Intel test for the white paper placed a USB3 hard drive in a cage with a spectrum analyzer "close" (without a measurement of distance given) to the USB3 hard drive and it showed in the 2.4-2.5GHz range an increase in the noise floor of roughly 20dB peaking at 2.43GHz and dropping to roughly 15dB on either side of this (which coincidently means if you are having USB3 interference issues, means you might want to try the higher 2.4GHz channels).

I'll granted not all USB3 devices are exactly equal, but they all have to communicate on roughly the same set of frequencies over the cable, otherwise the USB hosts can't hear each other.

When looking at the connector on a notebook in the 2.4-2.5GHz range, the noise floor is roughly 25dB higher with a USB3 device connected. With proper shielding on the connector it is only raised by 15dB (IE 10dB of mitigation).

I'll grant they didn't look at 5GHz at all for the white paper, but it certainly looks like the impact on the UNI-I-III bands would be negligible to nothing, even probably if you had unshielded USB3 ports and devices and placed then next to a USB Wifi dongle. On 2.4GHz, it might be pretty bad in a case like that. In a case where you have the wifi antennas several inches from the USB3 port and USB3 device, the impact even on 2.4GHz isn't likely to be huge, but probably will be noticable at longer range (where an increased noise floor of even a few dB might impact things).

 

[sfx2000]

USB3 is high speed digital data - square waves. That generates a lot of harmonics and can go well above 5GHz.

I'd think that this would not be an issue if you use good quality cables with good shielding, with ferrite bead or ring filters, and avoid plastic USB hubs,

Interference is why Apple disabled USB3 on the Airport Extreme and Time Capsules - the chipset is fully capable, but the potential impact to the AP with poorly shielded cables/cases was enough not to implement USB 3 on their AP's...

FWIW - the USB3 interference issue has nothing to do with Apple, has everything to do with the signalling rates - Intel has a great whitepaper on it - link below:

http://www.intel.com/content/www/us...al-bus/usb3-frequency-interference-paper.html

sfx

 

[stevech]

and what of eSATA?

 

[azazel1024]

and what of eSATA?

I can't find SATA3/6Gbps information, but it looks like SATA2 uses 1.5GHz carrier with SATA1 being lower frequency than that.

So it is at least possible some of SATA3/6Gbps might operate near enough to the 2.4GHz realm, but I am thinking probably not.

As for Apple, it isn't a bad design choice for their APs to not use USB3. However, proper design would prevent a whole lot of interference. Just looking at Intel's white paper, just enclosing the USB3 port properly drops interference by a good 10dB with the USB3 port being right next to a spectrum analyzer. So if they properly positioned the antennas, enclosed the USB3 port and probably shielded the traces/USB controller on the board, they'd probably generate negligble interference. Granted the USB3 end device could generate interference too, but you'd have to locate it close to the antennas.

According to Intel's charts, you aren't looking at a whole lot of interference. It messes with something like a mouse dongle big time, but you are also talking about locating a source of interference within an inch or so of the receiver there. Just simple shielding even that close showed no impact on the receiver. Now do that plus locate the port all of 3-4 inches from the antennas and now lets talk about Wifi being a good 15-20dB stronger than most mouse dongles/bluetooth...the only excuse for really needing to drop USB3 is because you have a poor design (otherwise how are all of these other manufacturers now managing USB3 in their AP/routers and not having 2.4GHz issues. Only issues I saw were in some of the first designs with USB3 ports where they had a bad design/improper testing).

 

[stevech]

I can't find SATA3/6Gbps information, but it looks like SATA2 uses 1.5GHz carrier with SATA1 being lower frequency than that.

So it is at least possible some of SATA3/6Gbps might operate near enough to the 2.4GHz realm, but I am thinking probably not.

As for Apple, it isn't a bad design choice for their APs to not use USB3. However, proper design would prevent a whole lot of interference. Just looking at Intel's white paper, just enclosing the USB3 port properly drops interference by a good 10dB with the USB3 port being right next to a spectrum analyzer. So if they properly positioned the antennas, enclosed the USB3 port and probably shielded the traces/USB controller on the board, they'd probably generate negligble interference. Granted the USB3 end device could generate interference too, but you'd have to locate it close to the antennas.

According to Intel's charts, you aren't looking at a whole lot of interference. It messes with something like a mouse dongle big time, but you are also talking about locating a source of interference within an inch or so of the receiver there. Just simple shielding even that close showed no impact on the receiver. Now do that plus locate the port all of 3-4 inches from the antennas and now lets talk about Wifi being a good 15-20dB stronger than most mouse dongles/bluetooth...the only excuse for really needing to drop USB3 is because you have a poor design (otherwise how are all of these other manufacturers now managing USB3 in their AP/routers and not having 2.4GHz issues. Only issues I saw were in some of the first designs with USB3 ports where they had a bad design/improper testing).

Remember, these digital signals are square waves. The transition time from 1 to 0 and vice-versa is on the order of a nanosecond. A square wave consists of the sum of many harmonically related signals. If the signal was a pure sine wave- virtually no harmonics generated.
Here's a simple to understand explanation.

 

[sfx2000]

and what of eSATA?

eSATA - should be fine - use good quality cables, goes without saying, but that's general guidance for any external connection.

 

[sfx2000]

As for Apple, it isn't a bad design choice for their APs to not use USB3. However, proper design would prevent a whole lot of interference. Just looking at Intel's white paper, just enclosing the USB3 port properly drops interference by a good 10dB with the USB3 port being right next to a spectrum analyzer. So if they properly positioned the antennas, enclosed the USB3 port and probably shielded the traces/USB controller on the board, they'd probably generate negligble interference. Granted the USB3 end device could generate interference too, but you'd have to locate it close to the antennas.

I don't think anyone can fault a tier 1 OEM like Apple for bad design choices...

I think one of the challenges with putting USB3 on a router -- and you hit it right on the head -- is that the end-point is mere inches away from the 2.4GHz radio, and interference here, well, that would impact every node attached to the Router/AP, as the local radio is the one being impacted.

It's not just the shielding of the ports and controllers, but having high-quality shielded cables and external devices...

For me, it's not that big of a deal, as I don't hook up external drives/devices to my routers in any event, that is why I have a dedicated NAS box...

But it brings to mind - I have an Asus USB-56AC 802.11ac USB dongle - it supports USB3, and I've never really compared performance across USB2 and USB3 in 2.4GHz (it's set up to run in 5GHz for obvious reasons.)

 

[Razor512]

When operating in USB 2 mode, USB 3 ports do not make use of the extra spectrum. This is why WiFi adapters such as the netgear A6210 will restart into USB 2 mode when you attempt to connect to a 2.4GHz AP.

Some routers will also run their USB 3 ports in USB 2 mode when the 2.4GHz radio is in use. other that do not drop the rate, will use a RF shielding can over the USB 3 port and the traces.

 

[sfx2000]

When operating in USB 2 mode, USB 3 ports do not make use of the extra spectrum. This is why WiFi adapters such as the netgear A6210 will restart into USB 2 mode when you attempt to connect to a 2.4GHz AP.

Some routers will also run their USB 3 ports in USB 2 mode when the 2.4GHz radio is in use. other that do not drop the rate, will use a RF shielding can over the USB 3 port and the traces.

Shielding of the port - in many of the AP designs I've seen, it's a bit of hand applied kapton tape, and it shields the socket only... and no cans around the controller - in any event, once the signal leaves the socket, it's up to the cable and its shielding...

 

[azazel1024]

True, but just looking at Intel's testing, just basic shielding of the port itself is good for ~10dB reduction in noise...which considering that for 2.4GHz wifi, USB3 just doesn't produce overwhelming interference in general (you'd get more interference having a microwave oven running 10-20ft away from your client/router).

You can get terrible interference but it takes a nice convergence of poor router design, bad cable routing and bad USB3 device shielding. Which of course means it DOES happen, but vaguely intelligent design of everything just doesn't produce overwhelming noise. It may be a detriment if you are attempting to connect at extreme range, but most people are going to have higher inteference from local 2.4GHz sources that are co-channel than what USB3 will produce (in a GOOD design).

One thing I do wonder is if USB3.1 happens to have shifted away from the range of frequencies that'll generally produce 2.4GHz interference. Since AFAIK it is the same pin/wire count for the ports and cables, I'd assume they are getting to twice the bandwidth through higher frequency/wider frequency. Also hopefully they thought through some of the issues with Wifi/bluetooth and USB3 and intentionally designed around it.