Fortran
array temporaries
can be implicitly created when
non-contiguous
array slices are passed to procedures.
This copy of data can impact performance.
Older compilers like GCC 8.x (including 8.5), which may be the only installed choice on some HPC systems, have occasional bugs related to array temporaries that can lead to incorrect results or crashes.
It is possible that disabling optimization can change the bug behavior, but this can make the program run vastly slower and it’s just for debugging.
Use
git bisect
or similar techniques to find the commit that introduced the bug.
If the bug is introduced by a new / changed function call using an actual array argument that indexes into a non-contiguous array, that is a hint that the problem may be related to array temporaries.
Enable compiler flags to help identify array temporaries in GFortran:
Here, A(:, 4:16:3) creates a non-contiguous array slice, leading to the creation of an array temporary when passed to myfun.
Manually allocate an contiguous auxiliary array variable and copy the data into it before passing it to the function:
real,allocatable::temp(:,:)allocate(temp(size(A,1),5))! Adjust size as needed
temp=A(:,4:16:3)! Copy data into contiguous array
x=myfun(temp)! Pass the contiguous array
deallocate(temp)!Cleanup
Obviously this process is tedious, so we only employ it when needed to workaround a compiler bug.
Many programming languages (including LaTeX) have directives or pragma–formatted comments that tells the compiler to do something specific.
LaTeX pragma: de factodirectives
are
compatible
with popular LaTeX editors including TeXstudio.
Specify LaTeX compiler and bibliography toolat the top of the main/root LaTeX document like:
% !TeX program = xelatex
% !BIB program = biber
LaTeX projects often use Biber for bibliography management.
The above directives tell TeXstudio to use XeLaTeX compiler and Biber bibliography tool when compiling the document.
Find the MAC address of a Bluetooth device in Windows from Windows Device Manager under Bluetooth, (select device in list), Properties, Details, Bluetooth Device Address.
List all Bluetooth devices and their MAC addresses from PowerShell:
The standard
git bisect
command help to quickly identify the specific commit that introduced a bug in a Git repository.
In projects where branches with lots of commits were previously merged, there may be known-bad commits with build failure unrelated to the current bisect search.
It’s possible to
skip
such commits during the bisect process, or
avoid
specific commits altogether.
For projects using Git Submodules, an additional command
git submodule update --init --recursive
at each step of the bisect process is required to update the submodules to the correct commit.
One can make a simple shell script to use with
git bisect run
to automate this process, or simply run the command manually after each bisect step.
TeXstudio and VS Code are among the IDEs that support remote collaborative code / document editing with real-time updates.
The program
Teamtype
is free, open-source, serverless, peer-to-peer, non-AI tech that enables no-cost real-time collaboration.
2 or more users connect via a Teamtype shared link code - no account or signup is used.
Git can be used to capture specific versions or milestones of the LaTeX document.
For general code editing and development with VS Code, Teamtype is also supported.
Overleaf has progressively added restrictions to its free plan, making TeXstudio (or VS Code)with Teamtype a good alternative for collaborative LaTeX editing.
To have essential security features with an on-premises instantiation of open-source
Overleaf Community Edition,
a paid Overleaf Pro plan is required.
Alternatives:
server-based collaborative team editing is possible with VS Code Live Share.
Crixet has an appealing free plan, watch to see how the features evolve vs. pricing
Papeeria free version is quite restrictive, only 1 private project
Scienhub free version lacks Git integration and has limited storage
Since 1972, standard VHF AM used 25 kHz channel spacing generally worldwide.
In 1999, European regulations began a mandatory phase-in of 8.33 kHz channel spacing for VHF AM aviation communications to increase the number of available channels.
The rest of the world generally remains on 25 kHz spacing for aviation.
One factor driving the 8.33 kHz spacing in Europe was the number of relatively close proximity airports and large number of small nations each having their own aviation communication needs.
Other countries like China and the United States also have regions with many airports and a lot of air traffic, but each have a unified regulatory framework managing all VHF airband communications.
Anyone who has used AM CB radio or AM broadcast radio would understand why 25 kHz channel spacing is reasonable to avoid adjacent channel interference with real-world AM transmitters and receivers in most reasonable proximities.
Even without overmodulated, splattering transmitters, an AM receiver can experience noticeable adjacent channel interference when within about 1 km of another AM transmitter operating on a channel just 10 kHz away.
If the CB radio transmitter is operating properly, 20 kHz spacing is adequate when not closer than about 1 km.
This observation can be loosely derived from a typical AM CB radio (say the Cobra 19 mini) receiver’s selectivity performance of say 6 dB at 7 kHz spacing and 60 dB at 10 kHz spacing.
Consider also the dynamic range of the receiver, akin to
Sherwood Engineering’s Receiver test data,
which doesn’t include CB radios.
Assume the CB radio transmitter ERP is 36 dBm (4 watts) and the receiver sensitivity for 20 dB SINAD is -113 dBm (0.5 µV).
At 1 km distance, free-space path loss at 27 MHz is about 61 dB.
This first order approximation neglects important factors like radiation pattern distortion from ground reflections.
Assuming 0 dB receiver antenna gain, received signal strength would be about -25 dBm, which can exceed the maximum dynamic range of the receiver front end, if we arbitrarily assume a receiver
dynamic range
of about 80 dB.
However, the typical
environmental noise floor
on 27 MHz - neglecting skip or other anthropogenic interference - may be ~ -110 dBm in the AM receiver bandwidth, so the dynamic range impacts of a legal CB radio transmitter 1 km away may not be noticeable.
An adjacent channel signal 10 kHz away (neglecting modulation bandwidth) would be attenuated by about 60 dB, resulting in a received adjacent channel signal strength of about -85 dBm.
The adjacent channel signal would be strong enough to cause noticeable interference on the desired channel, and unwantedly open the squelch if no desired signal is present.
With 20 kHz spacing, the adjacent channel signal would be attenuated by over 80 dB, approaching the general receiver’s dynamic range limit vs. the noise floor, and would be much less likely to cause noticeable interference, assuming the transmitter is clean.
The emissions mask in FCC part 95.979 requires better than -35 dBc in the band 8 kHz to 10 kHz from the carrier, and better than -59 dBc beyond 10 kHz from the carrier.
The radio transmitter will often do better than -59 dBc at 20 kHz from the carrier and increasingly better beyond that.
Switching to AM VHF airband radio, a typical installed receiver such as the Icom IC-A220 has receiver selectivity at the 25 kHz step of 6 dB at 3 kHz spacing and 60 dB at 22 kHz spacing, and at the 8.33 kHz step of 6 dB at 2.8 kHz spacing and 60 dB at 7.4 kHz spacing.
A typical handheld airband transceiver such as the Yaesu FTA-250L RX selectivity in the 25 kHz step of 6dB at 8 kHz spacing and 60 dB at 25 kHz spacing, and at the 8.33 kHz step of 6dB at 2.8 kHz spacing and 60 dB at 8.3 kHz spacing.
Thus we see how at airband, where the frequency plans would be made to slice up former 25 kHz step frequencies into distant frequency reuse at 8.33 kHz step, while avoiding close-by overload, the 8.33 kHz step is quite useful.
To put this back into CB radio terms, suppose CB radio is popular in a town.
The typical use I’ve seen is for local groups to be at least 20 kHz apart to avoid adjacent channel interference.
For example, CB channel 20 (27.205 MHz) is more popular as a “local” channel than channel 18 (27.175 MHz) to avoid adjacent channel interference to/from channel 19 (27.185 MHz).
Since CB radio use by definition is entirely ad hoc, users generally space themselves 2 channels (20 kHz) apart in real-world observations.
For airband VHF AM in Europe, channels of 8.33 kHz spacing are successfully used with proper frequency planning considering physical separation to avoid adjacent channel interference and overload.
Given the above, I think President should have make like 8
P-channels
and not have had “P4” as CB channel 18.
They should have instead done like P1 = CB1, P2 = CB3, P3 = CB12, P4 = CB14, P5 = CB22, P6 = CB23, P7 = CB27, P8 = CB29.
This avoids CB channels 6, 9, 16 (SSB use), 19, and 20 and the SSB channels 30-40.
However those are only hypothetical channel assignments, and President did not in fact do that.
Given the existing President P-channel assignments, I think President should consider adding P6-P9 like:
P6 = CB14, P7 = CB22, P8 = CB23, P9 = CB27.
Again that’s hypothetical.
First we need to get more P-channel or at least FM + CTCSS users on CB radio.
Zello
and
Motorola Wave
are two popular PTT (push to talk) instant group voice communications apps.
These apps are used by neighborhood watch groups, citizen rescue groups, and more.
Note that these apps require internet connectivity via cellular or WiFi.
To make group voice PTT apps more convenient, consider a speaker-mic with PTT button to hear much louder audio than with the cell phone.
The key advantage of speaker mics include loud audio output and not needing to fumble around with a phone or radio in the hand, perhaps clipping the speaker-mic to one’s outer garment.
The speaker is several times larger than the internal smartphone speaker, meaning the fidelity and sound level can be much improved.
Pryme Radio Products makes several models of Bluetooth
PrymeMAX PTT phone speaker mics.
Most Bluetooth Speaker-Mics are “open”, meaning they can work with any app suited to accept them.
However, some Bluetooth Speaker-Mic are “proprietary”, meaning they work only with their specific app.
The problem with proprietary Speaker-Mics is that if the manufacturer goes away, you lose service.
Consider a speaker-mic with rotary volume control as it allows changing volume with gloves or inside a pocket without looking at it.
Among the key specifications to consider are speaker output (should be at least 1 watt) and battery life.
Public safety and critical infrastructure may use MCPTT (Mission Critical Push to Talk) solutions over 5G networks, in contrast to the PTToC used for commercial and personal applications.
These work at a lower network layer than typical apps like Zello and Wave.
The Open Mobile Alliance help define specifications for release as ETSI standards that network carriers and OEMs (infrastructure and devices) can comply with.
MCPTT works at a lower network layer to provide QoS and prioritization with high reliability.
This may also includes “talkaround” via
5G sidelink,
where devices can communicate directly without other wireless infrastructure.
These capabilities are NOT directly present in Wave, Zello and similar apps freely available.
For Bluetooth speaker mics that connect to traditional walkie-talkies, note that some radio manufacturers have proprietary Bluetooth aspects that make generic Bluetooth Speaker-Mics incompatible.
In those cases, an external radio Bluetooth adapter like PrymeBlu may be needed to make the connection.
On macOS when using the default “AppleClang” compiler in a Fortran project where GFortran objects are linked with C / C++ AppleClang objects, the ld linker may emit warnings like:
ld: warning: could not create compact unwind for ...: register .. saved somewhere other than in frame
ld: warning: could not create compact unwind for ...: registers .. and .. not saved contiguously in frame
This is an actual issue because C++ exception handling will not completely work when this warning is emitted from C++ code coupled with Fortran code.
In general, using C++ exception handling within C++ code that is linked with Fortran code will work just fine, except when this warning is issued.
The solution is to use GNU GCC C++ compiler with GFortran instead of mixing AppleClang with GFortran.
Specifying environment variable:
LDFLAGS="$LDFLAGS -Wl,-no_compact_unwind"
removes the warning, but this also disables C++ exception handling so it is not recommended.
if(NOTDEFINEDok_compile)message(CHECK_START"Checking for C++ exception handling with Fortran")try_compile(ok_compilePROJECTexception_checkSOURCE_DIR ${CMAKE_CURRENT_LIST_DIR}/exception_checkCMAKE_FLAGS-DCMAKE_LINK_WARNING_AS_ERROR:BOOL=onOUTPUT_VARIABLE_out)# _out is used for CMake < 4.0
if(NOTok_compileOR_outMATCHES"ld: warning: could not create compact unwind for")message(CHECK_FAIL"C++ exception handling will not work reliably due to incompatible compilers:
C++ compiler ${CMAKE_CXX_COMPILER_ID} ${CMAKE_CXX_COMPILER_VERSION}
Fortran compiler ${CMAKE_Fortran_COMPILER_ID} ${CMAKE_Fortran_COMPILER_VERSION}" )else()message(CHECK_PASS"OK")endif()endif()
Inexpensive integrated circuits like the On Semiconductor SA572 can be used for compandored audio.
Compandored audio increases the perceived SNR by greatly reducing audio gain for signals below a threshold, and greatly increasing gain for signals above that threshold.
It could be of interest to amateur radio operators to have such a circuit for noisy HF SSB / AM / FM.
CB radios before digital NRC features used compandoring e.g. Cobra SoundTracker or Realistic “Audio Clarity Enhancer”.
Narrowband FM two-way radios (2 kHz maximum deviation) often use compandoring to improve the audio quality.
Note that compandoring has a downside – when communicating with non-compandored stations, the audio may sound unnatural.
Even communicating with another compandored station may sound odd if the settings are not matched well.
Compandoring has fallen a bit out of favor due to DSP-based noise reduction (NRC) generally doing a better job of improving apparent SNR with fewer audio artifacts vs. compandoring.
A compandor IC such as used in commercial two-way radio can improve the apparent audio SNR of analog voice communications, whether AM / FM / SSB.
The circuit thresholds and compresses transmitted audio, in a better fashion than the simpler clipping circuits sometimes used in ham radio.
The circuit also has the capacity to “expand” the receive audio when the transmitting station is transmitting “compressed” audio.
The apparent SNR increase can be on the order of 20 dB if the transmit and receiver compandoring settings are matched.
Note that this apparent increase is caused by increasing the average power of the transmission; peak power is still limited by the radio.
The compandor IC SA572NG is available from
Mouser
for $2.
It needs a few common resistors and capacitors external to the IC as well.
The 220 MHz ACSSB two-way radio system in the USA was a remarkable last gasp of analog technology trailing off as DMR, NXDN, P25, and other digital modes began to dominate the land-mobile radio landscape.
ACSSB squeezed into 5 kHz channel spacing in contrast to the 12.5 - 25 kHz common for two-way radio FM channels.
Digital two-way radio modes achieve 6.25 kHz effective channel spacing via TDMA (DMR, P25 Phase II) or FDMA (NXDN).
A vital contemporary use of 220 MHz band is for data telemetry (e.g. SCADA via GE Vernova / MDS radios) and
Positive Train Control (PTC).
Technical deep dive links:
220 MHz ACSSB system design manual fascinating with good LTR and conventional system design info applicable to common radio systems.
While many UHF intercoms (e.g. RTS BTR-800) use wideband (40 kHz deviation) FM, there is a design house that made ACSSB intercoms for nuclear power plants due to the uniquely severe multipath and RF environment
Political analysis on FCC spectrum auctions including 220 MHz band
The construction requirements for SMRs in the 220, 800, and 900MHz band across the 1990s and 2000s led to some interesting scenarios.
One that I saw in the Lansing, Michigan area in the late 1990s was an 800 MHz 5 channel LTR system that used 5 Realistic analog scanning receivers set to a single frequency each, with 5 EF Johnson 8600-series mobile radios set to talkaround on the repeater output frequency.
The scanners and mobile were each wired to an LTR controller panel to make a functioning trunked repeater system.
This actually worked, but was certainly not to the RF performance of a dedicated repeater design.
This was done after the SMR licensee lost 5 of their original 10 channel license due to failure to meet construction requirements.
An attempt to provide a 220 MHz
LTR-Net system
and data telemetry network across 80% of the continental USA and 2/3 of the population was made by Cornerstone SMR, who was one of the 220 MHz
auction 72 winners,
but adoption stalled and Cornerstone SMR was
denied a waiver
on their 220 MHz license construction requirements in 2018.
Cornerstone SMR aimed to raise about
$5 million in 2003-2004
from investors to build out the wide-area 220 MHz LTR-Net network.
Civil theft litigation
due to an employee stealing nearly $500k led to an appeal ending in 2015 (Cornerstone SMR, Inc. v. Bank of Am., N.A. , 163 So. 3d 565, 568 (Fla. 4th DCA 2015)) and set a precedent in Florida case law.
Cornerstone SMR’s idea to use their 220 MHz spectrum for a nationwide IoT network and proposed business
merger
to make innovative IoT hardware ultimately led to
litigation filed
in Delaware in Sept. 2025.
