Mittwoch, 31. Oktober 2018

DIV8RT development updates

As mentioned in a previous Post about my Receiver design, I have ordered a large amount of components to assemble a lot more units. Assembling this many boards takes a lot of time. If you don't organize your workplace and every process that it's going to need to build the device prior to the assembly it will almost certainly take a lot longer. I always sort and position most of the tools and parts used during assembly with about an arm length on my desk. There is nothing is more inconvenient and annoying than having to search for stuff while dealing with fragile electronics.






There are some more changes since the last post. First of all the receivers enclosure has changed to a 3d printed one as the old one was too fragile during the extensive testing that I have done. Also the overall volume has been reduced to shorten printing times and make it more useful in small airplanes.


The next change was more an addition. The antennas are now secured with some special electronic grade silicone. This prevents the antennas from being ripped of the connector or PCB. I got the silicone from Aliexpress for about 2Euro, so no large investment for a lot more security.

And finally I changed the pinout of the auxiliary Telemetry Port to match FrSky without having to build a special cable. This was as easy as routing the pins to the matching locations on the second revision of my PCB-Layout.

This project has already been one of the most time consuming things that I have ever done in my rare spare time and that is also why it takes so long to do updates or even get new features made. However I did a lot of software work to get the vario finally running the way it should. The most time went into a bug in the Processor itself. The barometric sensor had every now and then jumping values in the order of a couple kilometers. After many tests I decided to look at the data on the I2C lanes... Sometimes the processor decides to transfer an extra byte or even a zero byte. After quite some time someone pointed me to the errata sheet from ST, where this bug was already listed with a possible workaround. The workaround is to use I2C only with DMA except for single byte transfers... For me, who has never really used DMA before, there was a steep learning curve. But after stepping into many common pitfalls and general misunderstandings I got it working. Now the sensor was responding as expected.

Another really wanted feature was a automatic gain control of the amplifier stage. The problem without these few lines of code where missing packets and even complete signal losses on short distances (about 2-3m) due to overloads in the receivers input stage. Disabling the receivers amplifier during bind and high signal periods in combination with some window comparison did the trick. Now the antennas can almost touch each other without overloading. 


A huge software project for the future will be adding D16 LBT and a custom bootloader to update the firmware via the transmitter and therefore building a custom .frk file converting script. At the time I'm writing this, the first part of the d16 development is almost done. I'm already able to bind and receive channel values. I now need to look into failsave values, telemetry and lbt functionality. Last mentioned has to my knowledge not been reverse engineered entirely so some nice stuff to investigate and learn. So far the D16 protocol implementation has been a nice journey, because midelic and jry on RCGroups.com have done most of the hard work and I only have to implement it into my existing software.

Mittwoch, 16. Mai 2018

Saito radial engine repair

As almost every RC fan I'm fascinated by model engines. Most of them are however quite expensive to buy new or even used.

By the end of 2017 I got hold of a very cheap but broken Saito radial engine. With the experience that I have built over the years of flying gas and nitro engines I was very convinced that I would be able to get also this engine back up and running. I only did a few rough calculations if for the given price it was even possible to repair without paying more than I would have for a good running used one...
Immediately after the parcel arrived I wrote a list with all the broken and missing parts. The engine has already been disassembled by the previous owner, but he did a wonderful job packaging and marking all the parts. That made my job a lot easier...





The damage was quite substantial as the previous owner had an unplanned 'landing' at the airfield. With broken Cylinder, Rocker arm brackets and ripped screws there was a lot to be done. Luckily I was able to drill some of the broken screws out and save some more parts that I would have otherwise had to buy new. Also the crankshaft and the matching bearings where in good condition. Here just a few impressions. There where a lot more parts damaged but I think they illustrate it the best...




The whole cleaning and inspection took 2 days in total and afterwards I placed an order for the missing parts. Unfortunately the shop I ordered them from had to order them from Japan (Saito is a Japan company) so the delivery could have taken over 3 Month. Again having luck the parcel arrived by the middle of February. I also could have ordered the parts at a local German hobby shop but they would have cost more than twice as much. 




The day the parts arrived I started sorting and assembling of the parts. Unfortunately one of the cylinders had a large leak at the exhaust valve. I decided to try to grind it in using some abrasive cleaning solvent. However it did not work at all. Maybe if I would have ground it a lot longer than I already did. After that I bought a proper valve grinding kit for a few Euros. This grinding paste was way more on the coarse side and made fast work of the leak. Another night of valve grinding and screwing parts together later I had 2 cylinders in place with decent compression. The third cylinder got assembled in the following morning. The engine got to me with a Keleo customs exhaust. It was in really bad shape too. Having no exhaust but a ready to run engine I took a flip without a muffler.
It fired up at the third try and ran pretty smooth afterwards. 






With that result I was really happy having invested in the broken parts only the muffler being an open issue. Two weeks later I tried to get the muffler back into shape. Propane torch, huge clamps and a vise later I finally got it, so that it would fit the engine again. It's certainly not a thing of beauty anymore but the sound will not be affected by that. And as it will be screwed under a hood it won't be seen too.




Hopefully the repairs will last a long time and I will find a suitable airframe in the future.




Mittwoch, 22. November 2017

DIV8RTV Updates and lab measurements

In the last couple weeks I tweaked the Software a lot and also the receiver now has an enclousure. Also the MS5611 (high precision Barometer chip) arrived and the Software work can finally start to get also this feature reliable.
The enclousure is made from thick plastic film. First printed and then plotted to the right dimensions. The holes for the bind Button and the Leds have also been plotted out. Also the edges where cut half way through, to make bending straight edges easyier. The case is then folded into shape and secured with some superglue. (Oh and sorry FrSky for using your Label! I use these just for private use, so that should not be a huge problem I think (?).)
DIV8RTV with enclosure
Software wise has a lot changed from the last update. The receiver has now SmartHub enabled and also has programmable failsave. I still have to work a lot on the vario feature as the time for the heavy calculations is really limited. So far I've made 5 boards and have flown all of them without any issues up to 400m (RSSI still up to 75), limited by the size of the test plane. For almost all tests I used a HyperBipe from Hobbyking which has a wingspan of about 900mm. Nitro planes have been tested too, so vibrations are also tolerable. The total fight time of the receivers should have reached about 10 hours. As I said no issues so far.

I'll now order more parts to be able to produce in a small (for me possible to hand assemble) batch of about 50 receivers. Soldering many of these boards also requires some technics to follow while assembling. Otherwise it'll take a lot longer and will be more harmfull to yourself. Because of that I'll describe some details about my assembling process in the future. Facing the use of stencils and general soldering tips that might be useful for someone with not so much experience.
Fresh soldered and flashed receivers
The hardware has prooven to be working with the current layout. The only thing that will change from the first prototype is a small track to control the amplifier gain. 
I noticed some swamping issues when the receiver is very close to the transmitter (about 2m radius) which is really annoing, because the transmitter is spitting out telemetry lost warnings all the time. With the then implemented software controlable gain this issue should be fixed in the next revision by disabling the power Amplifier when the RSSI is reading over 100.

I attached a few impressions of the receiver in action. YouTube is a little slow with the upload so expect more to come(Already a little old from summer 2017)

The Measurements have been made! Last week I met with one of the Assistants of the local Institute for integrated Circuit design at my University. He kindly gave me the opportunity to use one of their Lab-Places to do my measurements. Huge thanks again for that!
The equipment that I used for the tests was a Rohde & Schwarz FSV. Also a HP benchtop Power supply and a IR-Camera FLIR T430sc. So, quite nice stuff to play with!


I started by getting used to the Spectrum Analyzer... This took longer than I thought initially but I've never used a instrument like this before. After I was somewhat knowing what I was doing. I connected a antenna to the input Port and switched my FrSky Taranis on. The result was about 100mW as expected by the specification, of cause depending on the distance between the FSV and the Transmitter.
Here some snapshots of the Transmitter idleing and in Range-Test mode. I had to change the attenuation in order to get the best resolution (and not to overload the FSV).
Transmitter power X9D | 30dB Att.
Transmitter power X9D RANGE MODE | 10dB Att.
I now connected the first receiver that I brought to the Lab. I brought many receivers to the Lab in order to get a more consistent result. Also I built 2 receivers with "fake" CC2500 and SE2431L from the chinese market. But more on that later...
The first receiver now got it's output power mesuared in peak hold mode (because that was the only mode I was aware of to use on FHSS (Frequency Hopping) systems. Should have informed myself a lot more than I did, but hell yeah... To get a uni-directional measurement I added a directional coupler and of cause connected the receiver to a propper uFL to SMA adaptor.
The receiver performed as good or better than what I could have ever imagined. Almost exactly 100mW! The insertion loss of the Balun network in the RF-Section that I used may be the reson for the slight attenuation, or even mesuarement errors that I have made by compensating the directional coupler...


After that I connected one of the boards with the fake chips to the FSV. And got really suprised... The output Power is at best 1000th of the non fake chip! Also there really is not a consistent Signal comming out of these...

Chinese ICs perform really bad!
I plan on taking a look inside one of these chips to see how they managed to copy the ICs. Also I may build another receiver with only the SE2431L beeing from china. To sort out only one IC not working properly. Even no I would never use these ICs in a real application, I really was quite intereseted in testing these.
After doing many similar measurements on all of the other receivers, with the same results, I started playing around with the FSV. Here a shot of how channel hopping looks like (The Amplitude is not relevant, because I used averaging to make it a more smooth movement):

Also I took a few shots with a IR-Camera (FLIR T430sc). The max Temperature is not really accurate in the way we used it! These shots where just for fun...


That's it for now! I hope you enjoyed this update.

Dienstag, 25. Juli 2017

DIV8RTV - A compact, cheap but powerful FrSky-Receiver

I finally got around to write something about the project I was working on for about 5 Month till now.
At first I'll describe something about it's features and after that I will go more into the technical aspects of the device and how I made the device itself.




Features (at the current hardware and software revision)

It's a Diversity 8 channel Receiver that features Telemetry and a high precision Variometer. With high precision Vario I mean <=10cm (delta) altitude recognition, wich is head to head with the current top of the line versions by some name Brands like Futaba. Also the Receiver has a small form factor (only 45mm x 25mm) furthermore it's relatively easy to assemble and use. Just set your Transmitter as per normal to D8 Protocol and hold the Bind button on the Receiver while switching it on. Then start the Bind sequence on the TX. After a few seconds the green Led on the Receiver will turn off. That's the binding done. The Telemetry is currently returning the current supply Voltage on the Servo Connectors, the current altitude and the current RSSI value. You can input everything from about 3.3 to 9 volts DC. The builtin protection against reverse polarity protects the Receiver from stupid mistakes, as they often happen on the field... The antennas are the same design as used by the official FrSky receivers. So if you cut one by accident you can just buy a new one, clip it on and continue flying. I also considered to enable the so called "SmartPort" feature that some official Receivers feature. That way you can connect any telemetry Item that you want. But it's not jet implemented. Additionally it features "Full-Range" capabilities by using a low noise amplifier. The practical Distance should be around 1.5 to 3km (for Telemetry a little less I think). The Receiver is now almost ready for some extensive testing in the Lab and on the field. 
Many thanks at this point to the Institute for integrated Analog and RF Circuit Design (IAS) at the RWTH-Aachen for making it possible for me to use some of their Rhode & Schwarz testing Equipment in their Laboratory. (Further Posts coming on this topic for sure!)

Design

The design was pretty straight forward, as the Protocol and Hardware used by FrSky have been discussed in many forums. Huge thanks to all of these people that have successfully reverse engineered most of the parts. Also huge thanks to fishpepper, who has made his implementation open Source. I based my Software around his one, as it is well written and easy to adjust for new hardware. 

But to get started, I've read a lot for the most part of this project in order to build up some knowledge about RF-Circuit design and the Protocol and Hardware used by FrSky. The design is a 4-Layer PCB with the inner Layers dedicated to VCC and GND. That makes routing of RF-Signals a lot easier. Especially if you don't have massive amount of knowledge and practice. But the board could be routed on a 2 Layer Board with some effort. The Design consists of a STM32F1 32Bit ARM Cortex M3 MCU and a CC2500 RF IC by Texas Instruments. Additionally I implemented a SE2431L to have a LNA (low noise amp) and PA (Power amp) on board to gain the Range. The other Components are more or less for convenience. A onboard EEprom is used to keep the current Bind and frequency offset correction information. An MS5611 is used to get the current altitude (quite expensive by the way). It get's reset to relative 0 when switched on. And some bypassing caps and filters are added as necessary. As already mentioned I used the same IPEX connectors for the antennas as FrSky does. The project is not 100% done yet, so I'll post some updates in the future and maybe make the Hard and Software OpenSource.
If you have any questions about this project don't hesitate to ask!



Donnerstag, 27. April 2017

Reverse engineering a TD025 compressor wheel shape

This is a post that I've written quite a while ago. I never released it, because I thought, that it's to boring. So keep that in mind while reading...


Some month ago I bought a compressor wheel for a TD025 turbocharger, which is a very common part used in cheaper cars, for some upcomming experiments with gas turbines. I plan to calculate and build up a gas turbine from parts that are easy to access. As the very popular DIY turbine KJ66 uses parts from an older turbocharger (which are really expensive these days), I initially came to the idea of trying to optimise the BOM (Bill of material) by using more standard parts. I plan to base my project around the existing KJ66 and MW54 as these are heavily optimised since the 1990s. Of cause changing a critical part of the design does lead to a lot of issues. The profile of the compressor wheel I plan to use is not that easy to find on the web and even after days of reasearch I wasn't pleased with the results. But because the shape is extreamly critical I thought about ways to specify the shape with simple methods.
The first method that I tried was mounting the wheel to a shaft and chucking it in my drill press to carefully mill some foam to the perfect shape. This worked out okay but not as well as I was hoping it would. Even the fine styrofoam that I used was still not fine enough to highlite the profile well enough for relaiable specification. In addition the small runout of the drillpress over shaped the profile. So further tinkering was required. The next option that came to my mind was a optical method by positioning a camera perpendicular to the compressors shaft and also in the center of view to minimize any optical distortion. That way I only have to take several images from different rotation positions of the compressor and overlay them to get the shape. So I build up a very minimalistic fotobox with a white background for better contrast and even lighting. Next I took the compressor wheel and a 4mm bolt as a axis. I then used a small machining vise to secure the bolt vertical and perpendicular. I'm now able to rotate the compressor wheel with minimal runout. I now took several Images of a small rotation. For this using the cameras hands free capturing function lead to minimal movement that otherwise could have occured by pressing the shutter switch. The images then where postprocessed with a threshold filter and then overlayed. The result was, after cleaning up the noise, a crisp outline of the compressors shape. Finally I sketched a negative of the wheels shape and printed it with our 3d printer to ensure the tolerances where at the right spot. And the fit was way better then I would have tought. Even a bit to tight for my taste but that could also have been caused by the printers tolerances.

Mittwoch, 16. November 2016

Ricoh SP C250DN (Toner page couter reset) follow up

As described in an earlier Post, I hacked the Ricoh toner cartridge page counter in order to be able to refill them. Today I finally got around to draw a small box-shaped enclosure with four small pogo pins as the connecting element between the MCU and the cartridge. Because the flashing process does require external power, I added a standard USB-cable to connect to any 5V supply available. This really is handy, because the printer's USB port does also work just fine. As far as the circuit goes nothing really has changed. I added a 3mm yellow led to light up while flashing of the cartridge is in progress. In addition it does flash continuously on and off, if writing or reading back has failed. The Software is slowed down a lot to ensure good physical connection and a stable state inside of the cartridge's IC. Using the device works as follows:
  1. - Attach the device to a power supply
  2. - Align the device with the cartridge 
  3. - Wait for the yellow led to light up
  4. - Wait until the led goes off
  5. - done! If the led does flash instead of continuously off, start from step one again.


The finished "product"

The advantage of using pogo pins -> perfect contact for each pin!

Mittwoch, 28. September 2016

New power supply!

Say hello to my new precision laboratory power supply! I lusted after a good precision and adjustable power supply for years. Finally a week ago I decided to look after one that does fit my requirements. I was searching for at least two adjustable outputs, constant current and voltage mode, a sequencing funktion and a nice user interface. Of cause you'll find a lot of power supplys with these specs as they are really common, but on the user interface side most of them suck. The probably worst UI I've ever used was a Siglent unit and according to a few YouTube videos nothing has changed in the newer models, they still suck. For power supplys with that price to performance ratio a shame... Next brand on my list was GWInstek. Unfortunately these units are waaay out of my budget, so I continued with Korad units. Because I used these power supplys before I already knew their value.
Most of the smaller units only have one output and only 'for the price adequate' precision. But I still left them in mind in case I don't find something more suitable. Then I considered buing a used Rigol as they are a dream to use an really well build. You may have seen my hacked Rigol Oscilloscope on some of my other Blogs and I'm really in love with the usabillity and specs. Buying a new unit wasn't worth thinking about... About 900€ are really high for a hobby use powersupply. But after searching on eBay I found a auction for a DP831A. I placed it on my viewing list and stopped searching for that night. The next day the auction was going to end. I placed my bid and suprisingly got the unit for a reasonable price.

Just a few days later the unit arrived in good shape. I hope the supply will be worth the money I spend for my usecases.