libera/#neo900/ Monday, 2018-06-18

houkimeShe also plays MMOs on it) It is some xiaomi mid-range stuff. 250$00:00
houkimeI mean even if it is not really often now, it quickly becomes ok to have a constant connection.00:01
Joerg-Neo900O just can tell my Cat S60 with a quite beefy 3000some mAh battery can do hotspot for maybe 4 hours, or less when much data transferred00:02
Joerg-Neo900>>The S60 comes loaded with a hefty 3,800mAh battery.<<
infobotsomebody said power was
houkimeThere should be some data rate that is tolerable for the battery. Like, datarate which increases your daily energy consumption only by 10% or so.00:17
houkimeone can calculate this average datarate and see how feasible is this slow-but-frequent network.00:18
houkimethe thing is that you can send packets different routs to manage the load00:19
houkimeand if we're speaking about phones, number of nodes and route alternatives can be large.00:19
Joerg-Neo900err route alternatives?00:20
houkime*alternative routes00:20
Joerg-Neo900I think the most important factor is if the WLAN is a client or an AccessPoint aka AdHoc mode00:20
houkimewhen you can connect through A and B but can through C and D00:20
Joerg-Neo900I don't see how this applies to a WLAN device00:21
Joerg-Neo900note how power saving mode in N900 WLAN makes the device WLAN deep-sleep more than 90% of time, only waking up for a short moment to check if the beacon from AP it's associated to has the "data pending" flag set00:23
Joerg-Neo900obviously in AdHoc and AP mode the device needs to send beacons itself instead00:24
Joerg-Neo900refer to extreme increase in power consumption when N900 WLAN is *NOT* associated to an AP and needs to run the receiver all the time00:25
Joerg-Neo900funny enough the complex and high bandwidth receiver eats more power than the transmitter00:26
Joerg-Neo900and this seems to be valid for all existing WLAN chipsets00:26
houkimeon phones you can probably do wakeup-calls via cell-related circuitry instead?00:29
Joerg-Neo900not when you want to do a peer2peer mesh00:31
Joerg-Neo900when you just want to route a call via AP to phone's WLAN and to e.g. a SIP client there, you don't need any fancy tricks, my N900 is capable of doing this for like 3 days in standby on WLAN medium PowerSavingsMode00:32
Joerg-Neo900prolly 5+ days when I shut down IRC00:33
Joerg-Neo9007 if I shut down cell as well00:34
Joerg-Neo900not keep an established callof course, just receive an imbound call any time, in standby00:36
houkimei want phones to be able to resend packets to other phones but without having to constantly run WLAN receiver.00:38
Joerg-Neo900dang, I got rsyslog-ng rules to route all my server remote syslogs to a separate file. It works for all messages, except for those which still show up in default syslog file:   2018-06-17T23:02:12+02:00 UbiqPKD kernel: last message repeated 12 times00:39
Joerg-Neo900how would you resend a WLAN data packet to a phone that doesn't constantly run WLAN receiver?00:40
Joerg-Neo900I mean, you could power up WLAN every 30s or somesuch, which would maybe cut the WLAN power consumption by half, but gives you a average 15s latency per hop00:41
houkimeVia making use of the phone already listening to his cell. Imitate a cell signal which onboard soft will decode as a wakeup for WLAN.00:41
Joerg-Neo900omotate? that's neither possible nor legal00:42
Joerg-Neo900and cell modem employs exactly same deepsleep methods, shutting down the RX for several seconds between very short listening intervals00:43
Joerg-Neo900the standby times cranked up from 10h to 5 days when that feature got introduced around 2000 minus a few years00:44
Joerg-Neo900and it's the reason why it takes up to 10s until your hear a call signal when you place a call to a mobile phone00:45
Joerg-Neo900so you don't gain anything much by switching from WLAN to GSM_ff00:46
Joerg-Neo900plus to imitate a BTS you as well need to constantly send "beacons" (not really beacons) on the cell's service channel00:47
Joerg-Neo900the only commonly available low-energy_standby medium-range RF protocol is BLE00:48
Joerg-Neo900I don't know how they do it but they manage to run their receivers pretty humble on power consumption00:50
houkimethay may use field-powered receivers actually.00:51
houkimeso that each session is initiated with a strong pulse00:51
Joerg-Neo900NFC does00:51
Joerg-Neo900BLE not00:51
Joerg-Neo900I raher think that pairing is an important part that allows the receiver to listen to very narrow band and well defined wake signal only00:52
houkimenfc does, but infc is short range. I mean one can make a receiver being awaken by a strong pulce on a given frequency that actually powers up a pin for a moment.00:53
Joerg-Neo900so sleep for 99ms, check 1ms for a well known pattern on a well defined frequency00:53
houkimenono, i mean if you have some tuned antenna you can basically power a wakeup latch for a moment with a wakeup pulse itself.00:54
Joerg-Neo900would require only 100ms of WAKE signaling from the sender, and only a 1% duty cycle of a simple receiver00:55
houkimeI mean  it can be literally sleeping intil brutally poked to action.00:55
Joerg-Neo900you can, but signal amplitude drops with almost cube exponent of distance00:55
Joerg-Neo900and BT only has max 100mW usually00:56
houkimecan store some power and release as a pulse00:56
Joerg-Neo900theey don't do this00:56
Joerg-Neo900it would fail in 3m distance even for a 5W pulse00:57
Joerg-Neo900and it would void any hope for getting a certification of the BT transmitter00:58
Joerg-Neo900but basically they could do something almost similar, using a pretty passive receiver design for the narrow band receiving00:59
Joerg-Neo900such passive receiver wouldn't have a sufficient bandwidth and sensitivity for any useful data transfer rate, but it can sense a known pattern like 9800Bd UART characters01:01
houkimeok, so we might need an open versionof this.01:01
Joerg-Neo900sou you could send the character "!" serially encoded with start and stop bit for a 100ms and receiver could sense that within 2ms listening01:02
Joerg-Neo900TX would send a 100 "!" and each takes 1ms, RX listens for 2ms and can receive one complete char/byte01:03
Joerg-Neo900if RX doesn't receive the "!" then it goes to sleep for 98ms again01:04
Joerg-Neo900for even better duty cycle, the RX checks for a carrier frequency and if none detected it can sleep again after as little as 0.01ms01:05
houkimeseems like latency will be a thing again though. Many hops, each is taking minimum 2 ms.01:07
houkimein a combined net of routers phones and doishlines might be ok though01:08
Joerg-Neo900minimum isn't even relevant here. average would be 50ms in this scheme01:09
Joerg-Neo900but yu can tune those parameters in a wide range, like 2 or 3 magnitudes01:10
Joerg-Neo900and actually a low bandwidth 'passive' receiver might be so humble you could run it 100% duty cycle01:11
Joerg-Neo900then you just have the latency from cranking up the high bandwidth RX after you receive a WAKE01:11
Joerg-Neo900and I think BT4.0 does something like this with WLAN actually01:12
Joerg-Neo900or was it BT5?01:12
Joerg-Neo900anyway wake signal on BT, real data communication on WLAN01:13
Joerg-Neo900but yes, mesh networks have gigh latencies, that's design immanent01:16
Joerg-Neo900you already run into this probem when using WLAN repeaters01:17
Joerg-Neo900each repeater adds 1 (at least) to the N in 1/N bandwidth of the network01:18
Joerg-Neo900so wuth 4 repeaters you get 1/5 of the genuine bandwidth01:19
Joerg-Neo900or 1/1001:19
Joerg-Neo900since every data package gets transferred no earlier than it been completely received, on a repeater01:20
Joerg-Neo900*starts getting transferred*01:21
houkimesoo... protocols need to be dumbed down/redone to support faster hopping?01:25
houkimelike "don't know what it is will just repeat it it it hits target?"01:26
houkimeso we have a packet and an antipacket "received" spreading through net01:26
Joerg-Neo900as long as you don't find a physical trick to receive and send data packages same time on same Radio access Technology, you're short of luck with speeding up stuff01:27
Joerg-Neo900some WLAN repeaters use 2.4->5GHz to avoid that problem01:28
Joerg-Neo900or vice versa01:28
Joerg-Neo900making a WLAN receive packets on 2.4 GHz while sending on 2.4GHz+20MHz usually fails for physical reasons of RX pre-amp getting saturated01:30
Joerg-Neo900even when you use physically separated antennas that are like 20cm apart from each other01:31
Joerg-Neo900so for all usual scenarios, it's TX and RX are mutually exclusive01:33
Joerg-Neo900in a mesh you also need to invest time into routing, unless you want to send a 'broacast' packet avalanche through the complete mesh01:45
Joerg-Neo900and you for sure don't want a packet to run in circles through the mesh ad infinitum01:46
houkimeavalanche+ lifetime counter?01:47
houkime-1 every hop01:47
houkimethough too much stress01:48
Joerg-Neo900possible. needs reading of at least the packet preamble with the counter, before you can start retransmitting the decremented counter packet01:48
Joerg-Neo900you don't get around the latency per hop01:48
Joerg-Neo900then you also probably don't want to relay/retransmit packets that have a RX error, so you need to receive the complete packet, do the checksum calculation and worst case ask sender to retransmit, before you start relaying01:52
houkimeAAAAAAAAAAAAAPacket, each time one letter gets reduced.01:55
houkimethen avalanche will produce shorter and shorter packets and no checks are needed01:56
Joerg-Neo900go simulate this, with each node havong 5 neighbours, and a node count > 50. You'll pull your hair01:57
Joerg-Neo900go send a second and a third packet from different origin nodes concurrently01:58
Joerg-Neo900the thing will explode right into your face01:58
Joerg-Neo900do you kknow chain letters?01:59
Joerg-Neo900or, simple example. Flu waves sweeping through a country02:00
Joerg-Neo900those are even harmless since you usually don'T catch same flu a second time02:00
Joerg-Neo900read about internet routing particularly of multicast packets02:02
houkimeWe can actually make use of WIFI being geografical. If it is not an AAAAAAAAAPacket but an ABCDEFGpacket with letters being assigned by your location in a deterministic algorithm.02:04
houkimeone hop still consumes one letter.02:05
houkimebut repeat is not following if letters don't match.02:05
houkimethough this will need special periodic pulces to update routing02:07
Joerg-Neo900protocols for mash networks exist. Refer for example B.A.T.M.A.N.02:07
houkimeok. will look into it later.02:13
houkimecan you explain a bit a purpose of C1122 and R1124? They are between agnd and agnd16:38
Joerg-Neo900off topic: please sign - read for more context.17:55
houkimelet's hope petition makes it through. Ok, now Microsoft-led Github will represent oss developers in debates over whether or not code uploads need to be checked for copyright infringements. What can possibly go wrong19:07
houkimeJoerg-Neo900: soo... what' with C1122 and R 1124? Even with star connection this doesn't seem to make much sense - therefore I son't really know what configuration is needed for this to work as intended.19:11
Joerg-Neo900will check19:12
Joerg-Neo900eeshow: error while loading shared libraries: cannot open shared object file: No such file or directory19:13
Joerg-Neo900make: *** [Makefile:56: view] Error 12719:13
Joerg-Neo900I see... C1122 R1124 are probably cruft for full symmetry of the differential pair line that is mic signal19:17
houkimeR1124 actually breaks geometrical symmetry.(19:23
houkimehm, or maybe not19:24
houkimeshouldn't C1108 also lead to agnd instead of gnd?19:27
houkimeseems like in terms of symmetry the biggest offence is... err... big video switch19:43
Joerg-Neo900yes, should be AGND19:44
Joerg-Neo900oops C1108, sorry need to check19:45
Joerg-Neo900yes, AGND of course19:45
Joerg-Neo900good catch :-)19:46
Joerg-Neo900actually could be both, but AGND makes more sense19:48
houkimeso, how is "differentiality" is maintained while mic+ goes through video switch and mic- doesn't?19:48
Joerg-Neo900try as good as it gets. It's basically a concept, not a recipe19:52
houkimehow at all a differential line can have gnd as - .19:53
houkimeI mean, they are supposed to be antiphase19:53
houkimebut good gnd is always in place19:53
houkimeit doesn't oscillate.19:53
Joerg-Neo900it's about minimizing current interference from digital to analog GND, and trying to keep the analog signal and GND currents as parallel and close to each other as possible. It's not a impedance matched differential line though19:55
Joerg-Neo900makes sense?19:56
houkimeso that current xcancel each other in farfield?19:57
Joerg-Neo900as much as possible19:57
Joerg-Neo900the magnetic fields19:58
houkimewell, will try.19:59
Joerg-Neo900think of it like faintly similar to a shielded coax cable19:59
Joerg-Neo900or rather: twisted pair19:59
Joerg-Neo900the better the "twisting" the smaller the susceptibility to any magnetic interference20:00

Generated by 2.17.0 by Marius Gedminas - find it at!