They seem to learn fast, also they have a lot of good engineers. We should expect some kind of response to Stuxnet and I guess we have established by Stuxnet that electronic warfare is OK for countries to do against each other.
It is going to be much harder to stomach the day some Air-force guy is taken out by a drone attach in Virginia with a missile to his car as he is delivering his children to Kindergarten.
According to Snowden, NSA use vulnerabilities in both Huawei, Cisco and other manufacturers gear to spy on traffic but if the vulnerabilities in Cisco, Juniper and others are planted there by NSA they might suspect that other parties have bigger difficulties spying on Cisco gear.
Most likely though it is more a question on wanting to favour American industry like when NSA did industrial espionage against Brazilian, German or other countries companies and share data with US companies.
I think you should save your money even though it is a noble idea.
Sources of randomness are not that hard to come by if you are on a physical machine and without having read how the linux kernel is doing it I assume it is rather bullet proof. If you only return hashed versions of your pool of randomness and add to the pool at every call it will become impossible to predict what the next output will be without knowing the content of the pool and the random elements you add to your pool.
If you at each call to the random() function add samples from the sound card, hash the pool, add free disk space, hash the pool, add cpu temperature, hash the pool, add content of memory location pointed to by lowest 32 bits of the pool, hash the pool. If you use a secure block cipher as hash function it should be peachy.
LTE is not CDMA. LTE is OFDM which is FDMA/TDMA in combination, very much like GSM. GSM in its simplest form have communication between handset and base-station on only one carrier while LTE use multiple sub-carriers but it is nothing like the CDMA mess of IS95/WCDMA.
I would be a little concerned that the silicon mono-layer would grow a natural oxide very fast and thus consume the silicon?
The solution in a HEMT transistor is cool in this respect. It is using an un-doped IV-V semiconductor next to a highly doped layer and excess carriers will form a two-dimensional electron gas at the interface. The carriers will move along the surface of the un-doped semi-conducter that since it is un-doped have better mobility and fewer defects than doped material. It must be something along this property they try to re-create with a silicon mono-layer.
If that is the case, it must be specified how a SIM card request this blocking from the phone. Otherwise this is not likely to work between different manufacturers of phones and SIM cards. If there is a specified way of doing this it must be within the GSM protocol to do so.
Alternatively this is a behavior specified by certain network operators who buy phones and SIM cards in bulk and mandate an in-official spec extension from both the SIM card and the phone manufacturer.
In the latter case I think the problem is with the operator. You cannot blame Nokia, Motorola, Samsung, Apple etc., from making business with AT&T, Vodafone, Hutchinson and the like. If an extra feature is a requirement for selling to these operators in the first place what are you to do? The customer is always right and in the subsidized markets the customer is the operator and not the punter using the phone.
Although the GSM specifications say that a phone should pop up a warning when it connects to a station that does not have encryption, SIM cards disable that setting so that alerts are not displayed. Even though the GSM spec requires it, this is a deliberate choice on the cell phone makers, Paget said."
I am not sure I understand the above text. If it is the SIM card disabling the setting, why is this then labeled a deliberate choice by the cell phone makers?
Also I have seen at least on numerous Nokia mobile phones that an icon in the display notify you at least in some instances when encryption is disabled. (This happen quite frequently in e.g. China).
Could you elaborate on your claim that a reciever with ~0dB noise figure can detect a signal
15 times smaller than a reciever with ~1dB noise figure? I would expect around 1dB improvement
in performance.
You are talking about the noise figure of the reciever and that will differ between handsets
as will the reciever algorithms re-creating the transmitted information from the decoded signal.
For a given standard one implementation might require 3dB s/n to reach a given BER but
other implementations might need 2dB or 4dB depending on the amount of signal processing
and the quality of the algorithms used. For a given sensitivity you can then design your receiver
with higher noise figure if your algorithms are better or you can live with lower baseband
performance if you design the analogue front-end with lower noise.
You write: "the minimum detectable signal is often defined as the point where the signal power equals the input referred noise power.". This will be the case for a system and an algorithmic implementation where you can detect a signal with 0dB s/n. For systems with coding gain you will typically be able to detect signals with negative s/n and in WCDMA mode a phone is likely to be able to decode a signal with -20dB s/n.
It might be possible to maintain a "gcc-light" compiler written fully in C and then have the gcc build scripts completing this boot-strap compiler first. The gcc-light do not need to be fast or effective since it will only be used for boot-strapping. It might even be possible to make it as a pre-processor converting c++ into C.
In the RoHS directive there is an exception for
high-lead solder used in flip-chip style packages.
TFA mention a mixed scenario where you mix
high-lead and eutectic solder. Not sure if that is
excempt and also not sure that this combination
have been used by nVidia.
Why dont you buy one of the lowest end Nokia phones?
They all come with a speaker for
hands-free and can be set to silent or discrete
mode and you can choose how the phone should
alert you for incoming text messages.
There is no thing in CDMA that will prevent
you from transmitting as much power as you like
unless this is built into the system by the
designer. Doing this is a design choice
but has nothing to do with whether you are making
a CDMA, TDMA, FD or any combination of these system.
In CDMA2000 and WCDMA the systems are made
with a power control loop similar to what is
implemented in GSM. The base station will ask
the handset to increase or decrease transmitted power depending on the current S/N estimate.
This enables the used frequency/code to be
re-usable in a closer neighbouring cell.
This will mean that the base-station will
receive all handsets with the same power that
can be as low as possible.
One problem with CDMA is that is is difficult
for the base-station to send different power levels to each handset since modulation noise
from a power-full signal will distort a less
power-full one. This is simple in TDMA-based
systems and widely used in e.g. GSM and allow
for fast re-use of frequencies.
The big problem in CDMA-systems is that they
are not made for packet data. The way the
current systems have implemented the power
control loop mean that the transmitter have to re-main open for the control channel even if
there is only little data to transmit.
The ISS is in low-earth orbit. So low that it requires propulsion to stay in orbit due to atmospheric drag. If you put more stuff up there you need more fuel to keep this in orbit so it is not for free to "park it close by". It is not going to stay there unless it is receiving a frequent boost from the friendly ATV, Progress or Shuttle.
Re-usable makes sense only if it is cheaper and uses fewer resources than disposable. I have yet to hear the case for re-usable toilet paper. Why should re-usable space-ships make sense?
Limiting your in-put bandwidth to 20-40MHz is a big help against blocking signals.
The most power-full blockers for a base-station will either be out-of-band blockers
like TV broad-casting, radar or other applications or it will be handsets from
competing cellular providers that operate outside the providers own licensed band.
It would make sense for an operator to order base-stations with a duplex filter only
wide enough to cover the frequencies for which the operator have a license.
All modern handsets have power control so the base-station will order handsets to
reduce transmitting power so the s/n is not bigger than necessary. This will enable
faster frequency re-use and more effective spectrum useage. An added bonus is that
this reduce the dynamic range for the base-station and consequently makes it easier
to implement an A/D converter that can handle multiple channels.
Since a handset will move around and can roam between multiple operators the
requirements for a handset is bigger. You need the full dynamic range since yout
own operator may have weaker signals than other carriers in your present cell and
you still need to decode yuor own channel.
Adjacent Channel Power requirements can be handled by the on-chip low-pass filters. In
CDMA2000 you have a channel bandwidth of around 1200kHz so you implement channel select
low-pass filters with 600kHz bandwidth which is very do-able on-chip. These filters are
analogue (they have to be before the A/D).
You can get around the channel select SAW filter by using direct conversion.
In a modern handset (not software radio) you would use direct conversion meaning that
you would create two sets of mixing products between your incoming signal and two copies
of an LO running at the carrier frequency 90 degrees out of phase with each other.
The desired signal would be represented by the resulting complex signal from the two
mixers (real and imaginary part) and the output from the mixers will be low pass filtered
to remove blocking signals from the carrier frequency.
Implementing a low pass filter on-chip is so much easier than implementing an on-chip
band-pass filter for channel selection at an IF.
You are right about the 18 bits for the dynamic range. The three bits for having
enough s/n of the wanted signal is still needed though so it would be 21 bit.
You are also right that this is slightly academic since no-one would implement a
handset in this way.
I am also rather sure that even the base-station in the story is using a band-select
filter in front of the LNA to limit the signal to in-band blockers.
Nyquist apply to the signal you want to process. If you want to do your filtering in the
digital domain you will first have to capture both your wanted signal and the signals you
seek to suppress. Otherwise you have your selectivity in your hard-ware and not in your
software.
The function of "ADC with lots of front-end bandwidth" is what the hardware in the
RF front-end is doing in a traditional radio system.
In GSM you must be able to detect your own signal at say -108dBm while you have a blocker
at 0dBm. Every 3dB is a doubling in power, 0dBm is 1mW so -108dBm is 1/(2^36)=1.45e-11 mW.
If you need 3 bit s/n in your wanted signal to decode it you will need at least 3+36 bits
to represent the samples.
As mentioned in the parent post you can do sub-sampling (If the carrier is at 1000MHz
you can sample at a fraction like 100MHz) as long as your signal have a bandwidth smaller than
half the sampling frequency. The fly in the ointment is that your blocker in GSM can be from
DC to 12.7GHz and a blocker at any multiplum of the sampling frequency will interfere with
the desired signal. If you are allowed 5 exceptions to the blocker requirement you will need
to sample with at least 12.7GHz/5 meaning >2GHz sampling frequency.
While this is possible it is not the simplest or the lowest power consuming way of doing
it.
This is why in any practical radio receiver you will have band filtering before your A/D
converter and this filter is NOT implemented in SW.
??? I do not understand this Exchange thing and how this could be a problem.
I were working for close to 10 years at a company who migrated to exchange for
reasons I still do not understand. This was not a problem for me for the exchange
server allow people to connect using IMAP4 which is no problem for most Email clients.
The only issue I ran into was that the windows users had limited mailbox-sizes
and could not send mail when they had more than 100MB in their in-box. Since the check
for some reason is made at the client-side I had no limitation and could grow my inbox
and let the "Your mail-box is over the size limit" Emails be eaten by the spam filter.
For many years I had mostly ignored CDMA systems and worked primarily on TDMA systems like GSM+GPRS+EGPRS, IS-136, PDC, DECT, PHS and PDC. I were of the impresseion that CDMA systems like
IS-95, CDMA2000 and WCDMA R99+HSDPA were overhyped but I assumed they deep down had some merit despite the hype.
How dissapointing to figure out that the guys who worked on these standards had largely missed out on 10 years of development in GSM so now we are stuck with something that is a solution to yesterdays problems rather than a step forward.
WCDMA is focussing so much on circuit switched data. It seems like the people making it was much more inspired by SONET and similar technologies and not at all by Ethernet, EGPRS or other packet switched technologies. The WCDMA system seems to be made for video calls and nothing but that. What an idiotic non-killer app. If video telephony was such a good idea why do we not do it from our networked PCs?.
In WCDMA and in HSDPA you need to maintain a circuit-switched 15kb/s full-duplex control channel whenever you are non-idle on the network. This means that for a chat-session, joy-stick motions or coffee-pot statistics from a networked appliance you occupy what corresponds to two voice-sessions of guaranteed quality-of-service bandwidth before you start transmitting any data.
I do not doubt that WCDMA will mature, but right now the system seem so much more primitive than EGPRS and some serious work is needed to make it compatible with packet switced operation.
Slashdot Mirror
The Lenovo Scrollpoint have been my solution the last 8-10 years. Very good product - I bought an extra for the office.
They seem to learn fast, also they have a lot of good engineers. We should expect some kind of response to Stuxnet and I guess we have established by Stuxnet that electronic warfare is OK for countries to do against each other.
It is going to be much harder to stomach the day some Air-force guy is taken out by a drone attach in Virginia with a missile to his car as he is delivering his children to Kindergarten.
According to Snowden, NSA use vulnerabilities in both Huawei, Cisco and other manufacturers gear to spy on traffic but if the vulnerabilities in Cisco, Juniper and others are planted there by NSA they might suspect that other parties have bigger difficulties spying on Cisco gear. Most likely though it is more a question on wanting to favour American industry like when NSA did industrial espionage against Brazilian, German or other countries companies and share data with US companies.
I think you should save your money even though it is a noble idea. Sources of randomness are not that hard to come by if you are on a physical machine and without having read how the linux kernel is doing it I assume it is rather bullet proof. If you only return hashed versions of your pool of randomness and add to the pool at every call it will become impossible to predict what the next output will be without knowing the content of the pool and the random elements you add to your pool. If you at each call to the random() function add samples from the sound card, hash the pool, add free disk space, hash the pool, add cpu temperature, hash the pool, add content of memory location pointed to by lowest 32 bits of the pool, hash the pool. If you use a secure block cipher as hash function it should be peachy.
LTE is not CDMA. LTE is OFDM which is FDMA/TDMA in combination, very much like GSM. GSM in its simplest form have communication between handset and base-station on only one carrier while LTE use multiple sub-carriers but it is nothing like the CDMA mess of IS95/WCDMA.
I would be a little concerned that the silicon mono-layer would grow a natural oxide very fast and thus consume the silicon?
The solution in a HEMT transistor is cool in this respect. It is using an un-doped IV-V semiconductor next to a highly doped layer and excess carriers will form a two-dimensional electron gas at the interface. The carriers will move along the surface of the un-doped semi-conducter that since it is un-doped have better mobility and fewer defects than doped material. It must be something along this property they try to re-create with a silicon mono-layer.
If that is the case, it must be specified how a SIM card request this blocking from the phone. Otherwise this is not likely to work between different manufacturers of phones and SIM cards. If there is a specified way of doing this it must be within the GSM protocol to do so.
Alternatively this is a behavior specified by certain network operators who buy phones and SIM cards in bulk and mandate an in-official spec extension from both the SIM card and the phone manufacturer.
In the latter case I think the problem is with the operator. You cannot blame Nokia, Motorola, Samsung, Apple etc., from making business with AT&T, Vodafone, Hutchinson and the like. If an extra feature is a requirement for selling to these operators in the first place what are you to do? The customer is always right and in the subsidized markets the customer is the operator and not the punter using the phone.
I am not sure I understand the above text. If it is the SIM card disabling the setting, why is this then labeled a deliberate choice by the cell phone makers?
Also I have seen at least on numerous Nokia mobile phones that an icon in the display notify you at least in some instances when encryption is disabled. (This happen quite frequently in e.g. China).
Could you elaborate on your claim that a reciever with ~0dB noise figure can detect a signal 15 times smaller than a reciever with ~1dB noise figure? I would expect around 1dB improvement in performance.
You are talking about the noise figure of the reciever and that will differ between handsets as will the reciever algorithms re-creating the transmitted information from the decoded signal.
For a given standard one implementation might require 3dB s/n to reach a given BER but other implementations might need 2dB or 4dB depending on the amount of signal processing and the quality of the algorithms used. For a given sensitivity you can then design your receiver with higher noise figure if your algorithms are better or you can live with lower baseband performance if you design the analogue front-end with lower noise.
You write: "the minimum detectable signal is often defined as the point where the signal power equals the input referred noise power.". This will be the case for a system and an algorithmic implementation where you can detect a signal with 0dB s/n. For systems with coding gain you will typically be able to detect signals with negative s/n and in WCDMA mode a phone is likely to be able to decode a signal with -20dB s/n.
The noise floor is around -174dBm/sqrt(Hz) depending on temperature.
This will be the same for all phones
I get your point.
It might be possible to maintain a "gcc-light" compiler written fully in C and then have the gcc build scripts completing this boot-strap compiler first. The gcc-light do not need to be fast or effective since it will only be used for boot-strapping. It might even be possible to make it as a pre-processor converting c++ into C.
Since there are platforms for which C++ compilers exist you can compile the compiler on one of these and then cross-compile for the target platform.
This is also how you boot-strap a C-compiler on a platform it is not implemented for initially.
In the RoHS directive there is an exception for high-lead solder used in flip-chip style packages.
TFA mention a mixed scenario where you mix high-lead and eutectic solder. Not sure if that is excempt and also not sure that this combination have been used by nVidia.
Why dont you buy one of the lowest end Nokia phones?
They all come with a speaker for hands-free and can be set to silent or discrete mode and you can choose how the phone should alert you for incoming text messages.
Best thing is they are cheap.
CDMA? Not sure I am with you here?
There is no thing in CDMA that will prevent you from transmitting as much power as you like unless this is built into the system by the designer. Doing this is a design choice but has nothing to do with whether you are making a CDMA, TDMA, FD or any combination of these system.
In CDMA2000 and WCDMA the systems are made with a power control loop similar to what is implemented in GSM. The base station will ask the handset to increase or decrease transmitted power depending on the current S/N estimate. This enables the used frequency/code to be re-usable in a closer neighbouring cell.
This will mean that the base-station will receive all handsets with the same power that can be as low as possible.
One problem with CDMA is that is is difficult for the base-station to send different power levels to each handset since modulation noise from a power-full signal will distort a less power-full one. This is simple in TDMA-based systems and widely used in e.g. GSM and allow for fast re-use of frequencies.
The big problem in CDMA-systems is that they are not made for packet data. The way the current systems have implemented the power control loop mean that the transmitter have to re-main open for the control channel even if there is only little data to transmit.
The ISS is in low-earth orbit. So low that it requires propulsion to stay in orbit due to atmospheric drag. If you put more stuff up there you need more fuel to keep this in orbit so it is not for free to "park it close by". It is not going to stay there unless it is receiving a frequent boost from the friendly ATV, Progress or Shuttle.
Re-usable makes sense only if it is cheaper and uses fewer resources than disposable. I have yet to hear the case for re-usable toilet paper. Why should re-usable space-ships make sense?
Limiting your in-put bandwidth to 20-40MHz is a big help against blocking signals. The most power-full blockers for a base-station will either be out-of-band blockers like TV broad-casting, radar or other applications or it will be handsets from competing cellular providers that operate outside the providers own licensed band.
It would make sense for an operator to order base-stations with a duplex filter only wide enough to cover the frequencies for which the operator have a license.
All modern handsets have power control so the base-station will order handsets to reduce transmitting power so the s/n is not bigger than necessary. This will enable faster frequency re-use and more effective spectrum useage. An added bonus is that this reduce the dynamic range for the base-station and consequently makes it easier to implement an A/D converter that can handle multiple channels.
Since a handset will move around and can roam between multiple operators the requirements for a handset is bigger. You need the full dynamic range since yout own operator may have weaker signals than other carriers in your present cell and you still need to decode yuor own channel.
Adjacent Channel Power requirements can be handled by the on-chip low-pass filters. In CDMA2000 you have a channel bandwidth of around 1200kHz so you implement channel select low-pass filters with 600kHz bandwidth which is very do-able on-chip. These filters are analogue (they have to be before the A/D).
You can get around the channel select SAW filter by using direct conversion.
In a modern handset (not software radio) you would use direct conversion meaning that you would create two sets of mixing products between your incoming signal and two copies of an LO running at the carrier frequency 90 degrees out of phase with each other.
The desired signal would be represented by the resulting complex signal from the two mixers (real and imaginary part) and the output from the mixers will be low pass filtered to remove blocking signals from the carrier frequency.
Implementing a low pass filter on-chip is so much easier than implementing an on-chip band-pass filter for channel selection at an IF.
You are right about the 18 bits for the dynamic range. The three bits for having enough s/n of the wanted signal is still needed though so it would be 21 bit.
You are also right that this is slightly academic since no-one would implement a handset in this way.
I am also rather sure that even the base-station in the story is using a band-select filter in front of the LNA to limit the signal to in-band blockers.
Why do you have two outputs from the IC?
Nyquist apply to the signal you want to process. If you want to do your filtering in the digital domain you will first have to capture both your wanted signal and the signals you seek to suppress. Otherwise you have your selectivity in your hard-ware and not in your software.
The function of "ADC with lots of front-end bandwidth" is what the hardware in the RF front-end is doing in a traditional radio system.
In GSM you must be able to detect your own signal at say -108dBm while you have a blocker at 0dBm. Every 3dB is a doubling in power, 0dBm is 1mW so -108dBm is 1/(2^36)=1.45e-11 mW. If you need 3 bit s/n in your wanted signal to decode it you will need at least 3+36 bits to represent the samples.
As mentioned in the parent post you can do sub-sampling (If the carrier is at 1000MHz you can sample at a fraction like 100MHz) as long as your signal have a bandwidth smaller than half the sampling frequency. The fly in the ointment is that your blocker in GSM can be from DC to 12.7GHz and a blocker at any multiplum of the sampling frequency will interfere with the desired signal. If you are allowed 5 exceptions to the blocker requirement you will need to sample with at least 12.7GHz/5 meaning >2GHz sampling frequency.
While this is possible it is not the simplest or the lowest power consuming way of doing it.
This is why in any practical radio receiver you will have band filtering before your A/D converter and this filter is NOT implemented in SW.
??? I do not understand this Exchange thing and how this could be a problem.
I were working for close to 10 years at a company who migrated to exchange for reasons I still do not understand. This was not a problem for me for the exchange server allow people to connect using IMAP4 which is no problem for most Email clients.
The only issue I ran into was that the windows users had limited mailbox-sizes and could not send mail when they had more than 100MB in their in-box. Since the check for some reason is made at the client-side I had no limitation and could grow my inbox and let the "Your mail-box is over the size limit" Emails be eaten by the spam filter.
For many years I had mostly ignored CDMA systems and worked primarily on TDMA systems like GSM+GPRS+EGPRS, IS-136, PDC, DECT, PHS and PDC. I were of the impresseion that CDMA systems like IS-95, CDMA2000 and WCDMA R99+HSDPA were overhyped but I assumed they deep down had some merit despite the hype.
How dissapointing to figure out that the guys who worked on these standards had largely missed out on 10 years of development in GSM so now we are stuck with something that is a solution to yesterdays problems rather than a step forward.
WCDMA is focussing so much on circuit switched data. It seems like the people making it was much more inspired by SONET and similar technologies and not at all by Ethernet, EGPRS or other packet switched technologies. The WCDMA system seems to be made for video calls and nothing but that. What an idiotic non-killer app. If video telephony was such a good idea why do we not do it from our networked PCs?.
In WCDMA and in HSDPA you need to maintain a circuit-switched 15kb/s full-duplex control channel whenever you are non-idle on the network. This means that for a chat-session, joy-stick motions or coffee-pot statistics from a networked appliance you occupy what corresponds to two voice-sessions of guaranteed quality-of-service bandwidth before you start transmitting any data.
I do not doubt that WCDMA will mature, but right now the system seem so much more primitive than EGPRS and some serious work is needed to make it compatible with packet switced operation.