I've now got two wifi connections on my Pi5 running Bookworm.
Following the latest update a day or two ago both come up
automatically on reboot connecting to the same access point
with the same frequency and channel. Using ifconfig it's
possible to turn them on or off individually.

The internal wifi shows a ping time of 5-10 ms, the USB
external Archer T3U is generally under 2 ms, perhaps because
it reports a stronger (~95% vs 75%) signal strength.

Using both interfaces together seems to result in much worse
performance than either one alone, which I didn't expect at all.
Ping times reach tens of seconds under load, for example. The
effect relliably appears a few seconds after applying the load
using the chromium browser.

I'm a little surprised that two interfaces are worse than one.

Has anybody else seen this behavior?

Thanks for reading,

bob prohaska

--- SoupGate-Win32 v1.05
* Origin: Agency HUB, Dunedin - New Zealand | Fido<>Usenet Gateway (3:770/3)

On 2024-11-24, <bp@www.zefox.net> <bp@www.zefox.net> wrote:

I'm a little surprised that two interfaces are worse than one.

Do you have separate IP addresses for the two interfaces?

cu
Michael
--
Some people have no respect of age unless it is bottled.

--- SoupGate-Win32 v1.05
* Origin: Agency HUB, Dunedin - New Zealand | Fido<>Usenet Gateway (3:770/3)

Lawrence D'Oliveiro <ldo@nz.invalid> wrote:

On Sun, 24 Nov 2024 18:50:55 -0000 (UTC), bp wrote:

I've now got two wifi connections on my Pi5 running Bookworm. Following
the latest update a day or two ago both come up automatically on reboot
connecting to the same access point with the same frequency and channel.

Using both interfaces together seems to result in much worse performance
than either one alone, which I didn't expect at all.

If they are on the same frequency/channel, it’s no surprise they’re interfering with each other. Can you set them on different, non-
overlapping frequencies/channels?

I wondered about that, but the initial settings were "automatic" and
the connection software didn't try to put them on different channels.
Presently the internal wifi remains "automatic" and the USB wifi is
set to channel 5. At this point I'll be interested mostly in whether
the setting is honored (tried it once before, still got channel 2).

There's still something funny about the two channels when used
separately. The internal wifi shows ~5 ms pings to the access point
with no other traffic but seconds of delay when watching Youtube
videos. The external wifi alone shows ~2 ms delays when idle and
only about 4 ms when watching YouTube. The internal wifi bogs down
very badly under load, in a way that is new compared the initial
Bookworm installation.

On the plus side, at least the internal wifi can connect now. For
quite some time it didn't connect at all. I'm using the same hardware,
only software updates have been applied.

Thanks for writing,

bob prohaska

--- SoupGate-Win32 v1.05
* Origin: Agency HUB, Dunedin - New Zealand | Fido<>Usenet Gateway (3:770/3)

On Sun, 24 Nov 2024 18:50:55 -0000 (UTC), bp wrote:

I've now got two wifi connections on my Pi5 running Bookworm. Following
the latest update a day or two ago both come up automatically on reboot connecting to the same access point with the same frequency and channel.

Using both interfaces together seems to result in much worse performance
than either one alone, which I didn't expect at all.

If they are on the same frequency/channel, it’s no surprise they’re interfering with each other. Can you set them on different, non-
overlapping frequencies/channels?

--- SoupGate-Win32 v1.05
* Origin: Agency HUB, Dunedin - New Zealand | Fido<>Usenet Gateway (3:770/3)

Michael Schwingen <news-1513678000@discworld.dascon.de> wrote:

On 2024-11-24, <bp@www.zefox.net> <bp@www.zefox.net> wrote:

I'm a little surprised that two interfaces are worse than one.

Do you have separate IP addresses for the two interfaces?

Yes. From the GUI controls there doesn't appear to be any other choise.

cu
Michael

--- SoupGate-Win32 v1.05
* Origin: Agency HUB, Dunedin - New Zealand | Fido<>Usenet Gateway (3:770/3)

Knute Johnson <knute2024@585ranch.com> wrote:

I'm curious, what is the point? Do you want it to go faster, use a
wire. Are you going to connect them to different access points? Do you
have two different networks? You can plug in multiple USB NICs as well.

The point is to first understand why wifi connectivity went from usable
to unusable. Then, if possible, to fix it.

The first thought was interference from other access points nearby.
No consistent evidence has been found, but I'm still looking.

The second thought was a faulty upgrade to Bookworm (this is on a Pi5)
There's a sliver of support for that idea, since performance changed
following some upgrades.

Third, a problem with the access point. That seems unlikely since other
hosts connect successfully and reliably.

Right now using a usb-wifi adapter seems to fix the problem. To me that suggests a problem with the internal WiFi hardware or the software that
drives it. The Raspberry Pi forums have some accounts of poor wifi behavior
on Pi5s, but not many and no unifying features are obvious.

Thanks for writing,


bob prohaska

--- SoupGate-Win32 v1.05
* Origin: Agency HUB, Dunedin - New Zealand | Fido<>Usenet Gateway (3:770/3)

On 11/24/24 12:50, bp@www.zefox.net wrote:

I've now got two wifi connections on my Pi5 running Bookworm.
Following the latest update a day or two ago both come up
automatically on reboot connecting to the same access point
with the same frequency and channel. Using ifconfig it's
possible to turn them on or off individually.

The internal wifi shows a ping time of 5-10 ms, the USB
external Archer T3U is generally under 2 ms, perhaps because
it reports a stronger (~95% vs 75%) signal strength.

Using both interfaces together seems to result in much worse
performance than either one alone, which I didn't expect at all.
Ping times reach tens of seconds under load, for example. The
effect relliably appears a few seconds after applying the load
using the chromium browser.

I'm a little surprised that two interfaces are worse than one.

Has anybody else seen this behavior?

Thanks for reading,

bob prohaska

I'm curious, what is the point? Do you want it to go faster, use a
wire. Are you going to connect them to different access points? Do you
have two different networks? You can plug in multiple USB NICs as well.

--

Knute Johnson

--- SoupGate-Win32 v1.05
* Origin: Agency HUB, Dunedin - New Zealand | Fido<>Usenet Gateway (3:770/3)

On 24/11/2024 23:01, bp@www.zefox.net wrote:

Knute Johnson <knute2024@585ranch.com> wrote:

I'm curious, what is the point? Do you want it to go faster, use a
wire. Are you going to connect them to different access points? Do you
have two different networks? You can plug in multiple USB NICs as well.

The point is to first understand why wifi connectivity went from usable
to unusable. Then, if possible, to fix it.

The first thought was interference from other access points nearby.
No consistent evidence has been found, but I'm still looking.

The second thought was a faulty upgrade to Bookworm (this is on a Pi5) There's a sliver of support for that idea, since performance changed following some upgrades.

Third, a problem with the access point. That seems unlikely since other
hosts connect successfully and reliably.

Right now using a usb-wifi adapter seems to fix the problem. To me that suggests a problem with the internal WiFi hardware or the software that drives it. The Raspberry Pi forums have some accounts of poor wifi behavior on Pi5s, but not many and no unifying features are obvious.

What case do you use - is it metal?
If so, I would suggest disabling the internal WiFi, and using the
stronger signal you get from the USB adapter

--
Chris

--- SoupGate-Win32 v1.05
* Origin: Agency HUB, Dunedin - New Zealand | Fido<>Usenet Gateway (3:770/3)

Chris Townley <news@cct-net.co.uk> wrote:

What case do you use - is it metal?

Not metal. It's clear plastic. The internal wifi worked just fine,
displaying signal quality in the 75-80% range when first set up.

There's some chance I misunderstand the meaning of "signal quality".
It appears to be a simple log measure of received power. Maybe there's
more to it. The man pages for wavemon don't indicate.

If so, I would suggest disabling the internal WiFi, and using the
stronger signal you get from the USB adapter

In effect, that's what I've done. I'd still like to understand why
it was necessary.

Thanks for writing,

bob prohaska

--- SoupGate-Win32 v1.05
* Origin: Agency HUB, Dunedin - New Zealand | Fido<>Usenet Gateway (3:770/3)

On Sun, 24 Nov 2024 20:58:42 -0000 (UTC), bp wrote:

There's still something funny about the two channels when used
separately. The internal wifi shows ~5 ms pings to the access point
with no other traffic but seconds of delay when watching Youtube
videos. The external wifi alone shows ~2 ms delays when idle and
only about 4 ms when watching YouTube.

Are you sure the unused interface is turned completely off -- no radio transmissions at all?

--- SoupGate-Win32 v1.05
* Origin: Agency HUB, Dunedin - New Zealand | Fido<>Usenet Gateway (3:770/3)

Lawrence D'Oliveiro <ldo@nz.invalid> wrote:

On Sun, 24 Nov 2024 20:58:42 -0000 (UTC), bp wrote:

There's still something funny about the two channels when used
separately. The internal wifi shows ~5 ms pings to the access point
with no other traffic but seconds of delay when watching Youtube
videos. The external wifi alone shows ~2 ms delays when idle and
only about 4 ms when watching YouTube.

Are you sure the unused interface is turned completely off -- no radio transmissions at all?

No, the only measure I have is wavemon running on the Pi5. It would
be nice to ask the access point what it can see, but that feature is
not offered, AFAIK.

When ifconfig takes the unused interface down wavemon stops reporting
for it. At the moment there's a mismatch between what wavemon reports
and what the GUI stataus indicator of the Pi5 reports. The latter states
that both internal and external wifi interfaces are present and active,
but ifconfig shows only the external wifi interface active with an IP
number. The GUI indicator is also greyed out, as if historical.

Between what ifconfig reports and wavemon displays I tend to think
the unused interface is inactive. However, I'm not certain.

Thanks for writing!

bob prohaska

--- SoupGate-Win32 v1.05
* Origin: Agency HUB, Dunedin - New Zealand | Fido<>Usenet Gateway (3:770/3)