I am upgrading the power board of my Curtis 1231c DC PWM motor controller. It uses 18 MOSFETs to switch the power, and each MOSFET had a 47 ohm resistor on it’s gate input. The point of such a high resistance was to slow down the switching of the MOSFET’s so that they would all share the current somewhat equally and no single MOSFET would turn completely on before all of the others had a chance to start shouldering the load.
I’ve moved my SamKnows FCC broadband testing unit from one location to another in the middle of the month. The first location had one of the best home Internet plans available from BrightHouse networks (now Spectrum after the Charter purchase), the “Lighting 200” plan. The new location has the best available AT&T U-Verse plan, limited by the distance from a central office and older wires. AT&T did set me up with a bonded pair (using 2 phone lines, or 4 wires to provide better service). The U-Verse plan is much slower, on both downlink and uplink, but is perfectly serviceable. The graphs below do show the stark contrast in available bandwidth.
I am going to miss the 24 up / 200 down of BrightHouse, but the 5 up / 30 down on U-Verse is still enough bandwidth for most needs.
The quality of the service has also gone down slightly, with slightly higher rates of packet loss and slightly higher latencies.
The extra 12 ms of latency isn’t much to worry about, but it is a definite change. The jump from almost no packet loss to 0.5 to 1 percent packet loss is annoying, and you can see that even on the worst day Spectrum / BrightHouse is better than AT&T U-Verse on an average day.
I desoldered all of the main power components (diodes, MOSFETS, capacitors) from the power board of my failed Curtis 1231c PWM DC motor controller. The plan is to upgrade all of the components to give it higher capacity; while producing less heat. Of course, to replace them, I had to remove the old ones, which took around 6 hours of work with two different soldering irons and a solder sucker.
-Heat component legs (diodes/MOSFETS) from the top of the board (side with the component) while you solder-suck from the bottom. Get one leg completely free first, then work on the other. After you suck almost all the solder out, you may still need to re-heat the leg and push it away from the PCB with a small screwdriver so it doesn’t stick to the inside of the hole.
-For the capacitors, don’t be afraid to add a little solder to the smaller leg, and then use a 100 Watt super wide tip soldering iron to heat both legs up at the same time, and pull the capacitor straight out. Suck the solder from each hole individually later once all the components are gone.
-I heartily recommend the Engineers SS-02 Solder Sucker, the silicon tube it uses is great! I did get solder stuck inside the metal tip a few times, but nothing a 5/64th drill couldn’t fix right up.
Or, how I lost several hours of my life to a tricky and hard to diagnose networking issue.
I recently set up broadband Internet service with AT&T U-Verse. The internet appeared to be working correctly, except that I was not able to send or receive email (via IMAP and SMTP) from my work account, which is hosted by Microsoft on their office365.com platform.
I could send and receive email via IMAP/SMTP with two OTHER email providers, and if I used any other network (Brighthouse, T-Mobile, etc) the outlook 365 email would work just fine.
After calls to AT&T (Internet is working just fine, must be a problem with Office365), and my companies’ helpdesk (sorry, we don’t support Thunderbird or K-9 mail). I finally had to figure out the problem myself.
I thought that perhaps the issue was with the IP address I had from AT&T (perhaps Microsoft’s servers had blacklisted it for some reason, although I could still access the web based Office365 cloud with no problems), so I turned on my VPN software to tunnel all of my traffic through the work network (and get a new IP address).
Even with a new IP address via the VPN software, I still had the problem!
(BUT, if I used my t-mobile phone mobile hotspot, and THEN turned on the VPN, everything worked just fine). However, I was able to ping the smtp server just fine, so it wasn’t a network routing issue.
So the problem was definitely a symptom of the U-verse network. And if a tunneling VPN wasn’t fixing it, it had to be a packet problem (the tunneling VPN would get my packets out of the local network, but not change them).
So I found the MTU (maximum transmit unit) on my modem, which defaulted to 1500 (usually a safe number) and changed it down to 1472. Problem solved: Suddenly I was able to communicate with the Microsoft IMAP servers.
I did this with the online web interface for the modem. Settings -> Broadband -> Link Configuration, changed “Upstream MTU” to 1472 (from the default of 1500).
I found that I also had to disable IPv6 on my modem to be able to send email via SMTP.
My best guess is that Microsoft may be running IPv6 inside their server farm, but tunneling it over IPv4 connections, which means that any 1500 byte packets are too big to be tunneled. (And they are not supporting packet fragmentation correctly.)
So by artificially limiting my packets to 1472 on the sending end, it allowed them to make it all the way to the SMTP/IMAP servers.
[I’d like to point out that the other two email providers I use on a regular basis did not have this issue, so it is probably something specific to the Microsoft Cloud.]
So, now I can get back to work. I pity the casual user who doesn’t have a background in computer networking…..of course, they probably all use the Outlook application, which hopefully doesn’t have this issue.
I am looking to replace the MOSFETS, diodes, and capacitors in my Curtis 1231c with upgraded components. I unsoldered one of the existing TSR2402R (7103 K) diodes from the power board and tested it with my Fluke meter and bench power supply.
Here are my results:
Power Supply providing 3.2A, forward voltage drop: 0.776 volts
Power Supply providing 2.0A, forward voltage drop: 0.737 volts
Power Supply providing 1.0A, forward voltage drop: 0.697 volts
Fluke Diode Setting: 0.351 vdc
Average time for the button temperature to raise from 25 °C to 50 °C with a 3.2A current: 45 seconds
The replacement parts I purchased were from DIOTEC, specifically their DR7506FR model (the R at the end means “Reverse Polarity”, making them an exact drop in replacement in form factor and polarity). They were marked: “DT110 DR7506FR” plus a diode schematic. Here are my results for the upgraded component:
Power Supply providing 3.2A, forward voltage drop: 0.754 volts
Power Supply providing 2.0A, forward voltage drop: 0.700 volts
Power Supply providing 1.0A, forward voltage drop: 0.646 volts
Fluke Diode Setting: 0.399 vdc
Average time for the button temperature to raise from 25 °C to 50°C with a 3.2A current: 47.5 seconds
Of course, the original diode I’m measuring had been in use for many years (I estimate ~750 hours of driving time given the 22K miles) and was heated up as part of the soldering and unsoldering process, while the DR7506FR I tested was brand new straight from the manufacturer. After I unsolder a few more diodes I’ll check them to make sure their readings are similar. (I’ll probably also test a few other DR7506FR diodes from the bag as well.)
Of all the measurements, the temperature rise time measurement was the least scientific, as I was using an inexpensive non-contact IR thermometer and attempting to point it at a small button in each diode, waving it back and forth to find the hottest temperature. I took 4 measurements on each diode (alternating to let the other one cool down) and averaged them together. In general, the readings from the DR7506FR were longer than from the original TSR2402R with one exception. If I throw out that pair of readings, the averages would be 46 seconds vs 50 seconds. Given that the measured forward voltage drop for the DR7506FR was lower for any real amperage readings, it dissipating less power and taking longer to rise to 50 °C appears to be reasonable.
T-Mobile is offering a beta program where you can sign up for “Virtual” phone numbers, and/or use one number on multiple devices.
Basically, it’s a VOIP service that allows you to have one main (sim card) number on a device, plus one or more “virtual lines”, and then use their VOIP application (called DIGITS) to log into your account on any device and make/receive calls and send/receive SMS/MMS messages.
By default it allows you to “use minutes” to call in or out on your “virtual lines” which means that it redirects voice traffic over your regular phone number so you can use it even if your data service isn’t the best, (which would solve the “bad quality” VOIP issues I’ve seen with Freedompop), and it always sends messages via data. But, you can tell it to “use data” for voice calls, and go full VOIP (for example, on a wifi only device).
A select few Samsung phones have software built in that handles the virtual lines and VOIP calling, but for all other phones, you have to install the T-Mobile DIGITS application.
The concept is good, and the service works as it is supposed to, but the DIGITS application itself (at least as I tested it on a Nexus 6) has a few issues.
The first issue is that it burns through data like it’s streaming video. I’d accuse them of ex-filtrating data off my phone, except I don’t have anything valuable enough for t-mobile to try and steal. I suspect the application just has a bug somewhere that makes it check in with servers much, MUCH more frequently than it should.
For example, when using the service for a few days away from wifi, I racked up 320 MB of data, while IDLE. (T-mobile gave me extra data when I called to report this behavior, but I can see why it’s still in BETA).
Also, it refuses to work unless you have location services enabled. I guess this could be an enhanced 911 liability issue where they want to be sure to be able to give an accurate address to police if you call 911, but I already had to enter in my E911 home address just to activate the application. It seems like they could ask for permission to turn on location services if needed, and then NOT turn them on unless you called 911, as this drains the battery.
In addition to powering up your GPS all the time, the application itself uses a lot of battery power when idle in the background. (note, the graph below is just the application usage, and does not include the extra power sent to the GPS chipset to be always active…)
When you are using more data and more power than the Google App while sitting in the background idle, you know your app needs to go on a diet.
In summary, multiple virtual lines is an interesting and useful concept, but until the DIGITS app doesn’t kill your battery life and suck down your bandwidth when idle, I think a dual sim phone is a better solution for two phone lines.
My Curtis 1231C motor controller blew up some MOSFETs and died. I replaced it with a used unit to get my truck back on the road, but now I’m interested in repairing the one that died so that I’ll have a spare.
I might be able to replace the components that died with exact replacement parts (but the IXTH50N20 MOSFETs are hard to find nowadays, and the diodes are basically unobtainable) to get it working, but since I have it open and am doing all of this work, I am exploring alternative (new) components that will have higher ratings and possibly give my controller more capacity or at least more resistance to blowing up again.
Of course, if I replace one component (power switching MOSFET, freewheeling diode, or ripple controlling capacitors), I will probably need to upgrade the other two as well so that I don’t just move the weak link from the MOSFETS to the capacitors or diodes.
After replacing the Curtis 1231C-8601 motor controller that had failed, I opened the case up to figure out what had failed. The controller hardware is inside of an aluminum extrusion with both ends “potted” with some black semi-flexible material (hard silicon perhaps?) that could be cut using a razor knife and a lot of effort.
Inside, there is a Pi shaped piece of aluminum extrusion that acts as the heatsink for the MOSFETS and freewheeling diodes, as well as being electrically connected to the motor – terminal. It is held against a large thermally conductive, but electrically insulating pad, which separates it from the controller case, but allows heat to be dissipated. It is held in place with 8 screws that pass through insulating plastic brackets into the bottom of the case.
People online had told me that these screw holes were “potted”, but on my controller they were just filled with two rubber plugs.They also told me that you could not cut through the Curtis potting material with a razor knife. [This super hard potting material was also prone to cracking at the edges and letting moisture into the controller, so a flexible rubber like material is better anyways…]
After my high voltage fuse blew, I used a light bulb to make sure that the controller was not shorted and appeared to work (controlling 60 watts). [As it turns out, some of the MOSFETs in the controller had failed, but it would still work for low current draws; Where low is defined as 100A or less…more about that later…]
As I was accelerating across a road in my S-10 EV, I heard a pop, and I lost power. I was able to coast to the side of the road, and use my small “glovebox” multi-meter to determine that the HV fuse was blown. The real question was why did it blow?