|Jetson TK1 FFT Performance
||[Oct. 30th, 2014|06:59 pm]
Just ran a test on the Jetson TK1 cuFFT with a zero-copy host to device memory mapping as described here: http://arrayfire.com/zero-copy-on-tegra-k1/ . This probably not fully optimized as the device should be capable of 300 GFLOPs, but still seems pretty fast|
[FFT] is starting on Nvidia Kepler GPU/Jetson TK1...
Transforming signal cufftExecC2C
GPU code executed 10000 FFTs, bin size: 32768 in 1241.31 ms
1 FFT time: 124.131 us. Throughput 263.98 MSPS / 19.7985 GFLOPs
CUDA Device Query (Runtime API) version (CUDART static linking)
Detected 1 CUDA Capable device(s)
Device 0: "GK20A"
CUDA Driver Version / Runtime Version 6.0 / 6.0
CUDA Capability Major/Minor version number: 3.2
Total amount of global memory: 1746 MBytes (1831051264 bytes)
MapSMtoCores for SM 3.2 is undefined. Default to use 192 Cores/SM
MapSMtoCores for SM 3.2 is undefined. Default to use 192 Cores/SM
( 1) Multiprocessors x (192) CUDA Cores/MP: 192 CUDA Cores
GPU Clock rate: 852 MHz (0.85 GHz)
Memory Clock rate: 924 Mhz
Memory Bus Width: 64-bit
L2 Cache Size: 131072 bytes
Max Texture Dimension Size (x,y,z) 1D=(65536), 2D=(65536,65536), 3D=(4096,4096,4096)
Max Layered Texture Size (dim) x layers 1D=(16384) x 2048, 2D=(16384,16384) x 2048
Total amount of constant memory: 65536 bytes
Total amount of shared memory per block: 49152 bytes
Total number of registers available per block: 32768
Warp size: 32
Maximum number of threads per multiprocessor: 2048
Maximum number of threads per block: 1024
Maximum sizes of each dimension of a block: 1024 x 1024 x 64
Maximum sizes of each dimension of a grid: 2147483647 x 65535 x 65535
Maximum memory pitch: 2147483647 bytes
Texture alignment: 512 bytes
Concurrent copy and kernel execution: Yes with 1 copy engine(s)
Run time limit on kernels: No
Integrated GPU sharing Host Memory: Yes
Support host page-locked memory mapping: Yes
Alignment requirement for Surfaces: Yes
Device has ECC support: Disabled
Device supports Unified Addressing (UVA): Yes
Device PCI Bus ID / PCI location ID: 0 / 0
< Default (multiple host threads can use ::cudaSetDevice() with device simultaneously) >
deviceQuery, CUDA Driver = CUDART, CUDA Driver Version = 6.0, CUDA Runtime Version = 6.0, NumDevs = 1, Device0 = GK20A
|Walmart $20 toaster Oven temperature profiling
||[May. 20th, 2013|10:41 pm]
After ruining about 20 bucks worth of resistors and caps on the backside of the Hermes board I decided to plot my "reflow" oven temperature profile. I used a K thermocouple to measure the temp as close to the PCB as possible. The solder melted at 172° (10 degrees "below" the spec) which means the probe "lagged" behind the actual solder paste temperature. However plotting the temp over time probably gives a good indication if the process is at least within the ballpark of the correct reflow profile.
Per First Principles of Solder Reflow by EFD, Inc
I decided not to worry about exactly matching the solder profile and just get it "close enough for government work"
The reflow oven is a $20 1 KW toaster from walmart. It has a heating element at the top and bottom. In the "Toast" mode both heaters are on and the overall rate is too fast, resulting in solder paste splatter and SMD components jumping off the board! With 50% power (Bake) only the bottom element is on, providing 500 Watts of heat. Simply turning on the oven, without fancy controllers, provided a pretty good approximation of the recommended reflow profile:
- turn the heater on
- wait until peak temperature (or about 30-40 seconds after the solder paste becomes liquid)
- turn off the heater and open the oven door.
Here is the recommended starting profile for the 4300 Chip Quik w/lead Water based No-Clean solder Paste, from SMT Intl, LLC
And this is a generic Walmart 1KW oven at half power:
The toaster heats up a little slower than optimal in the preheat and reflow zones, but keeps a nice <1C° ramp up time during the "soak". I hope this is good enough for home brewing. :)
|Hermes HPSDR do it yourself. Part II
||[May. 19th, 2013|09:03 pm]
Getting ready to place top side components. All parts are sorted in sets of ten bags according to the BOM line number:
|Hermes HPSDR do it yourself. Part I
||[May. 16th, 2013|09:47 pm]
This is my first serious SMD project, utilizing 0.5 mm pitch QFN and QFP parts, stencils, solder paste, reflow (aka toaster) oven. After a small disaster I was able to make some progress and reflowed the back side of the PCB.
Here is the HERMES board compared to the Afedri SDR receiver:
A sack of 541 SMD parts for Hermes - from Digikey and Mouser
Soldering station - 5X magnified view from the EOS 40D live view
EOS Camera Movie Recordhttps://sourceforge.net/projects/eos-movrec/
Solder paste (leaded!) applied via stencil. Thanks to Rob Frohne for producing the stencils for "community" use. Applying the paste takes only a few minutes.
Back side SMD components installed:
Reflow success in a $20 toaster owen in BAKE mode. This was after the first attempt failed - I had the temperature raise too fast by turning on both top and bottom toaster oven heaters. This cause the SMD components to pop off the board like popcorn. I caught this early enough to stop the process before the solder paste melted, washed the board, and started over. In BAKE mode only the bottom heater is one. Total bake time about 6-7 minutes. The solder melted when my K thermocouple registered 165 C. (datasheet specifies 183C as melting point) . There is probably some inertia in the themperature sensor. I waited until the paste melted, then another 45 sec and opened the oven door. The resistors lined up perfectly under by the solder surface tension, but some capacitors and large coils are still a bit crooked. I'm happy with the result so far.
Front side of the PCB is next, which has a much more difficult component placement.
|SB-200 QSK mod using W7RY boards off ebay - relay timing
||[Mar. 27th, 2013|12:56 am]
I've finally completed the SB-200 QSK modification using the board designed by James Colville, W7RY, available on ebay. Here is some additional info about the modification as well as some interesting relay timing traces.
This board requires the following changes to the original SB-200:
- removal of the entire ALC circuit
- removal of the SWR metering bridge and circuits
- complete rewiring of the metering switch. The Rel Power and SWR positions now turn on the STANDBY mode. The Rel Power pot is disconnected.
- direct grounding of the 572b grids
- un-grounding of the R12 on the power supply board and wiring it to the (new) grid/plate metering circuit
After completing the mod I'm getting 750-800 watts on all bands, and a whopping 900 watts on 10m, with about 60W drive. This is with a new pair Chinese 572b and 125V mains voltage (see previous post on adjusting the Vf). The semi break-in operation is very smooth, almost as quiet as the Tx switching on the TS-590 radio.
There are three relays controlled by this QSK board: RF input reed relay, RF output vacuum relay (e.g. Jennings RJ1A), and a solid state (optocoupler) cathode bias relay K3. Pushing PTT will make the QSK board energize all three of them at the same time. The order of switching is influenced only by the transfer speed of each individual relay.
The actual contact closure will occur in the following order:
W8JI or NE7X recommend a different sequence for a fail-safe operation. It needs to be 3,1,2, i.e. the cathodes should be grounded last, which will turn ON the amplifier.
- input reed relay (spec 0.6 ms, measured ~0.4 ms) - Trace 2
- SS relay (spec 1ms, measured: ~0.5 ms) - Trace 1, and Trace 4
- vacuum relay (spec: 6-8 ms RJ1A, measured >1ms) - Trace 3
Because of this, slightly different, order the W7RY's board requires to be careful that the exciter is OFF until all three relays are done switching. If there is any RF present on the input before or at the same time with the PTT signal, it will make the output relay "hot switch". The one I got (jennings RJ1A) is not rated for hot switching, costs $60, and not easy to swap!
Most of the modern rigs will have a at least 10ms lead time for PTT, which should be safe for this board. But I'd rather have the correct timing on the T/R relays in the amp.
Trace 1. RF Vacuum relay and Cathode bias. Bias is removed ahead of the RF out relay:
Trace 2: PTT and RF Input relay 432us:
Trace 3. PTT and RF Out relay 1024us:
Trace 4. PTT and Cathode bias solid state relay 536us:
Note: I've attempted to correct the timing problem by doing a mod to the mod (hi). I'm currently testing and may post some info later.
|SB-200 adjusting 572b filament voltage (Корректировка напряжения накала в SB-200)
||[Mar. 19th, 2013|11:59 pm]
With a typical 125V mains voltage in my home, the initial filament reading was at 7V. This is outside the specs for 572b tubes, which require 6.3 +/- 0.3V.
Like many other HAMs I've added two short pieces of Teflon insulated wire (МГТФ
) to provide about 0.4V drop. This puts the tubes at the high end of allowable range (see W8JI's recommendations
), and gives enough room for mains to sag during peak loads, without dropping the filament voltage below the spec.
The teflon wires become slightly warm, and you can see the resistance dropping (Vf going up) after cooling them off by blowing air.
A true RMS multimeter is required for this adjustment. A cheap Chinese meter was off by over 0.5V (8%) !!!
|N1MM contest audio recording
||[Dec. 5th, 2012|11:30 pm]
A new plugin for N1MM qso recording. http://hamradiomap.com/qsorder/ This function was at one time built-in but disappeared from the Logger in 2012 because of some technical reasons...
From the yahoo group message:
""The option to record QSOs within the program has been removed. Ever
since it was introduced, the internal option has been problematical and
buggy. Windows 7 made it more so, with issues involving the enumeration
of sound cards and recurrent bugs that were difficult to sort out.
We recommend that users adopt one of the free or low-cost audio
recording programs out there that do MP3 recording (for smaller file
sizes) and also embed current clock time in the file, to enable you to
search the file by time in order to hear a particular QSO. We know that
RecAll and RecAll Pro have this capability, but there are no doubt a
variety of other ones out there, including freeware."
qsorder plugin restores this functionality.
||most recent entries