Project Logs

Collapse

• Testing the 50A Mag Switch Prototypes

  Bud Bennett • 02/23/2020 at 22:35 • 0 comments

  Assembled the components onto the PCBs without a hitch. I soldered the back side first because I heat the back side of the board whenever I have exposed-pad parts. This gives a better chance to melt the solder under the part to properly connect the exposed pad. Hopefully, the parts on the back side won't fall off when the top side is soldered.

  I found a small gotcha when checking for shorts on the back side. If you touch the DVM to the gate of the power FET and then measure resistance across the drain-source you will get a short circuit. This occurs because the DVM charges the gate of the FET above the threshold voltage and it will then appear to be a short circuit across the drain-source. I had to short the gate-source terminals prior to measuring for short circuits between drain-source terminals.

  At this point the design appears to be fully functional. The prototypes look like this:

  Those are 14AWG wires. I believe that 12AWG is the max size that the board will accommodate. 14AWG is rated at 34A for chassis wiring. If you think you will be averaging more than that, then the 12AWG is rated at 41A. It's all relative to power and heat -- use your discretion. 

  Test Results:

  There aren't many parameters to test. 

  Off-state current vs. Voltage:

  50µA @ 30V

  46µA @ 25.2V

  41µA @ 21V

  37µA @ 16.8V

  33µA @ 12.6V

  29µA @ 8.4V

  On-state Current:

  9.5mA @ 25.2V

  I soldered two 30AWG Kelvin leads to the PCB pads near where the wires were connected. The location of the Kelvin leads makes a big difference in the measured voltage drop across the FETs. Since I don't have any precision/programmable load that can generate 40A I used my Tundra airplane's motor and ESC to provide the load. The Tundra draws very close to 40A during full throttle from a near fully charged hefty (5Ah) 3S LiPo battery. 

  Voltage drop across FETs during 40A load: 13.5mV. That's 0.34mΩ switch resistance. With a 40A load the switch will only dissipate about 0.54W. It should be able to handle this load continuously without overheating.

  The 40A load current did not create enough magnetic field to turn off the magnetic switch. The switch seemed rock-solid to me during testing.

  Next Step:

  Test the switch with a >50A load with a 5S or 6S battery. I'll need a friends help for that.

• More is Better?

  Bud Bennett • 02/02/2020 at 05:22 • 2 comments

  Two years ago I did not think that anybody really needed 6S LiPo batteries. Now they're everywhere. So I decided to capitulate and build a magnetic switch that would accommodate the new technology. I have a friend, not an engineer, that keeps insisting that 5S LiPo with a 40A or 50A ESC is the way to go for his latest high speed demon airplane. And I have watched as the smoke pours from the cockpit as he destroys his latest high dollar brushless motor.

  The Schematic:

  Not much is different. The change is to the FETs and how they are driven. Previous magnetic switches were driven with 5V gate voltage. This circuit level shifts the gate voltage of the output switching FETs to 12V to extract every bit of RDSon that the FET is capable of providing. 

  I searched Digikey for the FETs that would offer sub-mΩ RDSon with logic level gate voltages. Yes, I could find them, but there was always a 50% improvement when the gates were driven by 10V. A 6S LiPo battery offers 25V, so the extra gate voltage is not difficult to come by. The best FET that i could find for this application was Toshiba's TPHR6503PL -- about $1.40/each from Digikey in low quantities. It can withstand 30V (VDS) with an RDSon < 0.65mΩ if driven with VGS > 10V. Two of them in parallel should provide < 0.38mΩ. But these FETs are not the wimpy 3mm x 3mm powerPAK devices...they come in 5mm x 6mm powerPAKs, requiring a lot more PCB area. This is not a bad thing -- the more PCB area the lower the theta-jA ( the junction-to-ambient coefficient). I also figure that any plane requiring 40A/6S will have a reasonably large canopy volume to safely install this switch.

  In order to get the reduced RDSon from the TPHR6503 I added M4, R2 and D1. The gates of M1-M2 are now driven to 12V when the magnetic switch is activated. The PCB dimensions increased to 16mm x 22.7mm. Still pretty small by most standards. In fact, it should be larger to withstand the increased power dissipation. The layout has components on both sides.

  Most of the extra area is for the larger area required by M1 and M2. I added space between U2, the Hall effect switch, and the high current ground trace -- hoping 3.5mm is adequate distance to prevent the switch from tripping off during high current events. If that doesn't work, then the Hall effect switch will have to be moved off the board. The landing pads for B+, B-, OUT+ and OUT- are not through hole -- I decided that the large 12-14 AWG stranded wire would undergo a severe 90° bend if forced into a hole, so now the wires are simply soldered to a landing pad. The landing pads will accommodate up to 120mil diameter wire (12 AWG). In addition, there are numerous vias to help spread the heat and conductance between the top and bottom layers. The board will use 2 oz. copper for improved heat dissipation.
  

  The switch should dissipate a maximum of 0.6W for an applied current of 40A. Maximum current can exceed 1000A for short durations ( less than 100µs). That should be adequate. 

  The PCBs are still small, and therefore inexpensive -- 3 boards are only about $2.80 from OSH Park.

  [Edit 2020-02-02: BHarbour pointed out my laziness in his comment. The following is a bit of extra work to solidify the design.]

  Reducing the standby current:

  I had originally set the value of R2 = 10kΩ. I was sure it would work and did not really consider the impact on standby current. I don't usually leave the battery connected more than 3-4 hours; when the model is returned to its resting place the battery is removed. It would be better if the standby current was low enough that there was no need to disconnect the battery. This can be accomplished by raising the value of R2 to 1MegΩ. (A MegOhm is my personal limit -- enforced by a few years in the automotive industry. High values of resistance are subject to degradation over time by crap that collects on the PCB surface.)

  I spent a few minutes with a calculator and LTspice to...

  Read more »

• All's well that ends.

  Bud Bennett • 04/29/2018 at 22:07 • 0 comments

  At this point I consider the project completed. I have been flying my AXN Floater for a while now without any issues. The 30A mag switch works without any problems. It is nice to connect the battery and place the cover then just touch the magnet to the fuse to turn on all of the electronics. After landing I just pick up the plane and disable the power with the magnet. I don't have to open the cockpit and disconnect the battery until I get it back to my workbench. Very nice. 

  This summer I plan to install the 30A switch into two other powered gliders -- I don't expect any problems. I also want to install the BEC version into at least one unpowered glider.

• 3rd Pass Low Current Switch Results

  Bud Bennett • 04/24/2018 at 21:09 • 0 comments

  The third pass low current switch design was to lower the switch resistance from 12mΩ to around 4mΩ by replacing the 3x AO3400 FETs with a single TSM038N03PQ33 FET. I built 5 prototypes. The measured switch resistance at 1A load current was between 4.2Ω and 4.8Ω, as expected. 

  All of the other measured parameters were consistent with previous measurements.

• Second Pass Results for Mag Switch + BEC

  Bud Bennett • 04/24/2018 at 20:53 • 0 comments

  The second pass modifications were only to improve the input transient response to prevent catastrophic failure when connecting 4S (16.8V) batteries. Here's the scope trace of the input transient from a Multistar 3S battery (with a 25mΩ resistance):

  This was a HUGE improvement from the first pass:

  It appears that the input transient problem is solved. The only downside is the additional power dissipation from the 75mΩ input resistor.

  All of the other parameters that I measured were consistent with first pass results, as expected. There  won't be a third pass. I still need to evaluate the circuit using a 6A switcher IC, but that is pretty low priority.

• Value Engineering

  Bud Bennett • 03/29/2018 at 00:14 • 0 comments

  I did not build any of the 2nd pass low current mag switch prototypes. They just didn't measure up against the Zepsus product (and I was too cheap when designing them). My decision to use three AO3400 FETs instead of a more capable power FET seemed wrongheaded. Spending $0.50 more for a better power FET would  add capability and robustness. So why not?

  I found a relatively cheap ($0.61/each in QTY=10) FET in a 3.3mm x 3.3mm DFL-8 package -- TSM038N03PQ33. Its RDSon = 4mΩ typ. @ VDS=4.5V, which is 3x lower than the combined AO3400 FETs, with the same 30V VDS max rating. But it cost $0.61 (Digikey) vs. $0.026 (eBay) for the AO3400.

  I bit the bullet and scrapped the second pass low current switch boards. The single big FET will increase the current handling to at least 7A, while allowing a short term burst current capability of 10A. Another consideration in making the switch is that these DFL packages, while seemingly difficult to solder no-lead packages, were surprisingly easy to solder (if you know the trick) with consistent results.

  The latest schematic for the low current version is posted in the details section. I've ordered 9 PCBs from OSH Park in 2 oz. copper. They should arrive in a few weeks.

• 2nd Pass 30A Mag Switch Test Results

  Bud Bennett • 03/28/2018 at 23:42 • 0 comments

  I assembled one of the second pass 30A prototypes using different FET switches than the first pass -- substituting a  Vishay SiSS28DN for the IRLHM620. The Vishay part claimed slightly lower RDSon (1.6mΩ vs. 1.8mΩ @ VGS = 4.5V) and a higher max VDS of 25V instead of 20V. I thought, erroneously,  that the higher voltage rating might improve the survivability of input transients when batteries are plugged into the input. The addition of R1 between the input and the LDO solves the problem by eliminating the capacitor (C2) that caused the ringing so a higher voltage rated FET is not required to support 2S-4S LiPo batteries.

  I was able to use a different FET because the 3.3mm x 3.3mm DFL-8 package is apparently a standard. 

  Functionality:

  The switch no longer is sensitive to disconnection by high currents. I ran more than 35A through it and it remained closed. 

  It is now tolerant of the input transient caused by plugging a 4S LiPo battery into the inputs -- there is no measurable ringing. 

  Those were the two major items that had to be fixed on this pass.

  The new Hall-effect device trips at 45 Gauss and requires a stronger magnet. I ordered some 6mm x 10 mm N52 magnets and now the switch trips when the magnet is about 5/8 inch (16mm) when directly over the Hall-effect device. Since the Hall-effect device is active high, the switch activates and deactivates when the magnet is applied, rather than removed, which makes this measurement simpler (and the operation of the switch more straightforward).

  Measurements:

  Off-state current :

  5.3µA @ Vin = 6V

  5.7µA @ Vin = 21V

  On-state Current: 7.2mA @ Vin = 16.8V

  Switch resistance: 

  1.22mΩ @ 8.2A

  1.23mΩ @ 14A

  1.3mΩ @ 33A

  1.3mΩ @ 35A (45mV across switch)

  EDIT 2018-03-29:

  The switch resistance measured above was taken from the two pads on the board. While testing the other 5 boards I noticed that the resistance was 25% lower when measured from the copper landing that I had placed near the sources of M1 and M2. That's significantly lower, which I did not expect. I connected up two more boards, using the landing pad for the B- lead. Here's what I measured with that configuration:

  1.0mΩ @ 1A for the board using SS28 FETs.

  1.2mΩ @ 1A for the board using IRLHM620 FETs.

  In the end, I did get a pretty good improvement in switch resistance from the SS28 devices.

  Memory hold time is between 30 seconds and 45 seconds. The lower off-state current drain has caused this to lengthen. C6 should be reduced to get this back into the 15-20 second range.

  Conclusions:

  Meets both functional and parametric requirements. No circuit changes necessary. I'm going to build a total of 6 units -- 3 or 4 for my use and the rest will be sold. I'll report back on how they perform in my electric powered planes.

• Power Corrupts (Absolutely)

  Bud Bennett • 03/27/2018 at 22:31 • 0 comments

  I've been traveling a bit the last few weeks and came home to a bunch of PCBs and a few components that finally arrived from China. I had to prioritize. The BEC version was a first pass effort, so that got the highest priority. I knew that the BEC was going to need a second pass to fix the input transient problem so i only assembled a single unit. It was larger than I expected. I got accustomed to dealing with very small boards and this one is huge by comparison.

  Crappy Input Caps:

  To keep my options open, I had ordered several ceramic SMD capacitors from Chinese eBay vendors that I hoped would work as input caps:  
  

  • 10µF 50V 1210 (no temp rating)
  • 22µF 25V 1206 X7R (not really)
  • 22uF 50V 1210 X7R (unbelievable!!!)

  I tested the capacitance change vs. voltage and was disappointed with the result. The 10µF/50V caps decreased to less than 1µF @ 17V. The 22µF/25V measured <3µF at 10V. The only hope was the 22µF/50V, which measured 6µF @ 17V. I could have ordered some X7R 1210 from Digikey, but it was going to cost $3.00/each in my quantities -- not an option. 

  The Data:

  I measured a few parameters to make sure that things were working:

  • Off-state current draw was pretty good -- 7.7µA @6V increasing to 10.3µA @16.8V.
  • On-state current maxed out at 14.2mA @ 16.8V.
  • Line regulation is so-so: 5.05V @Vin=6V, to 5.17V @Vin=16.8V.
  • Load regulation is hard to measure without an electronic load, but seems to be better than the line regulation.

  I also measured input ripple voltage and output ripple voltage:

  No Load Output Ripple Voltage:

  Output Ripple Voltage with 5A load:

  Input ripple voltage (Vin ~ 3S LiPo Battery):

  This all looked pretty good. I decided to see what the input transient voltage waveform looked like for a couple of different battery types:

  Input voltage transient for 2S LiPo battery:

  The battery had a resistance of about 100mΩ so the resulting transient doesn't ring much.

  Input voltage transient for 3S LiPo battery:

  This 3S battery had a much lower internal resistance and there is a bit more ring as a result (And there was quite a bit of wire length between the battery and the board). Even so, the peak is about 17V, which is not a problem for the switching regulator (which has a 20V abs. max. input voltage rating). Though it is doubtful that the board will survive a 4S LiPo battery.

  Load Tests:

  I used a 22Ω 1W resistor for an easy load test. The BEC did not have a problem with this load and the output decreased only about 20mV over the entire input voltage range. 

  Things got more interesting when I attached a 1Ω/20W resistor to the output. The output voltage dropped to 4.97V and then steadily dropped over the next minute or so until the output voltage dropped below 4.9V where the buck converter hit its thermal limit and it began to disconnect the load to limit its junction temperature. 

  At this point I was pretty bummed out. It was obvious that the BEC was not capable of handling 5A as a continuous load. This result was not unexpected. I was boning up on thermal resistance and power dissipation of PCBs while traveling and had determined that a PCB this size could only handle about 0.5W of power dissipation to keep the junction temperature of the switcher IC less than 150°C.

  So what load current was it capable of handling continuously? I kluged together three 4Ω resistors in parallel to yield a load resistance of 1.33Ω - 1.35Ω, which would draw 3.75A at Vout = 5V. When I connected this reduced load current the BEC was able to keep the output voltage steady at 4.95V until the battery dropped below 6.5V and the input voltage crashed pretty quickly. My calibrated thumb determined...

  Read more »

• Preventing Poof!

  Bud Bennett • 03/04/2018 at 02:18 • 0 comments

  After destroying the LDO on the 30A mag switch board I decided to revisit the protection circuitry for all of the other boards. The 4A circuit is a copy of the 30A board so it should be OK. The problem is with the BEC version. The switching buck converter requires low ESR ceramic caps at the input. This causes ringing when the battery is connected to the board -- potentially damaging or destroying the circuits that are connected to the battery terminal.

  I investigated protection devices like TVS diodes or Zener diodes to clamp the inputs, but they did not have close tolerances that would prevent the input voltage from exceeding the 20V abs. max. VIN value for the RT6255 switcher. I decided to dust off LTSpice to see what was happening and look for possibly a simpler, less expensive solution. Here's the test schematic that I used:

  The inductance is just the wire leads from the battery to its connecter and then from the BEC connector to the PCB. I measured a fairly constant battery wire length of 4 inches for all of my batteries. If I add 1.5 inches for the mag switch leads then I get a total length of wire connected to the input that is approximately 11 inches (doubling the wire lengths to accommodate both the positive and negative leads.) I then visited this website to calculate the approximate inductance of the wires leading from the battery to the PCB -- which is about 350nH, represented by L1 above. Of course, there is some variation in these values, so I varied it over a reasonably wide range (300-500nH) to accommodate reasonable scenarios

  I also measured the resistance of nearly all of my stock of LiPo batteries. I found that it ranged from about 25mΩ to over 100mΩ. If you're interested, the Multistar 1.6A 3S 40C LiPo batteries had the lowest resistance and the cheap Chinese NoName batteries tended to be at the top end of the resistance range. I figure the middle of the pack is about 50-60mΩ. This is represented by a portion of R2.

  C1-C3 are the input capacitors required by the RT2255 buck switching regulator. I gave each of them 5mΩ ESR ( variation around this value did not have much effect).
  

  This is what the input transient looks like without any circuit changes made for protection and R2 = 25mΩ for the battery resistance:

  That first peak of nearly 30V would smoke the switcher IC. The peak transient waveform ranged up to 24V-27V, depending upon the values of C1-C2 and the associated ESR, the inductance in the lead wires, and the battery resistance.

  My first thoughts were to add a snubber -- a series R and C across the inputs. This could be just a tantalum capacitor with an appropriate ESR value . The problem I found was that the capacitor had to be greater than 100µF and the series resistance (or ESR) around 100mΩ. I could not find a reasonable size ceramic 100µF capacitor with a 25-50V rating, or a tantalum with <500mΩ ESR. The solution for the snubber to work was to put 15x 10µF ceramic caps in parallel -- so I abandoned this idea in favor of adding resistance in series with the input.

  R2 now represents the battery resistance and also a discrete resistance added in the circuit, R10, for the specific purpose of transient suppression. I won't bore you with the details, but I found that the value of R2 needed to be between 100mΩ and 150mΩ to keep the peak of the transient waveform below 20V (the absolute maximum VIN value for the RT2255 IC). C1 and C2 need to be increased as well -- even with a 50V rating the capacitor value decreases by 40% with 17V across it -- so the simplest solution is to stack two 10uF/50V caps in the place of one. This would not be a solution for high volume production but I can accommodate it easily when populating the board manually.

  Here's a typical simulation waveform result showing the improvement:

  R10 needs to be a power resistor. When you insert 100mΩ of resistance into the input...

  Read more »

• 30A Mag Switch Problems

  Bud Bennett • 03/01/2018 at 23:29 • 0 comments

  I assembled the 30A mag switch PCBs without issues. I expected some problems with the M1 & M2 PowerPak packages, but for some reason they always solder up just fine. I measured about 22µA of off-state current with 6V < VIN < 20V. Pretty much the same as the 4A boards. With 5A applied load current I measured 1.1mΩ switch resistance -- spot on the design target. 

  The bad stuff happened when I hooked the switch up to a 30A_ESC/Radio/Motor. The switch disconnected itself when the current exceeded 16A. Bummer. I remembered a warning from a colleague at LTC/ADI about "scary currents" that could affect the hall-effect switch, so I went online and found this website calculator for magnetic field strength vs. current through a wire. It turns out that the magnetic field strength 1mm distant from a wire carrying 16A is about 32 Gauss -- very nearly the trip point of the hall effect switch. Stupid me.

  In order to test out this theory, I desoldered the hall-effect switch from the circuit and located it onto a separate board with 2 inches of wire. Problem solved. The switch is now apparently immune to current flow.

  But I could not reproduce the same problem as before by holding the hall-effect sensor near the current carrying wire as high currents flow. That was somewhat disconcerting, but par for the course.

  POOF!

  I decided to use a 4S battery and a 60A ESC to generate a current above 30A to test the circuit. When I connected a fully charged (16.8V) 4S LiPo to the input the circuit lets some of its magic smoke out. After all the cussing ended I was able to determine that the TPS70950 LDO had apparently latched up -- the current was enough to fuse the VIN lead open:

  It is apparent that a 30V rating is not enough to prevent inductive transients damaging the LDO. The second pass, and all the other mag switch circuits, will have a protection resistor to prevent this in the future. After replacing the LDO and the hall-effect device, which was also damaged by the event, I jury-rigged a 330Ω protection resistor in between the LDO and the battery. I plugged/unplugged the battery many times to see if there was any remaining tendency to smoke, but the circuit doesn't have that problem anymore. 

  High Current measurements:

  With load current 35A, the voltage across the switch was 40mV (that's 1.1mΩ). Power dissipation is only 1.3W, which should not be a problem for a board even this small. I repeated the measurement with 20A load current and got 22mV across the switch, and the same switch resistance. No changes required.

  So here's the plan:

  1. Add a 330Ω protection resistor between B+ and the LDO input.
  2. Change the layout to move U2 as far from the current carrying traces and wires as possible. Here's the proposed topside layout (the bottom side did not change much):
     See that U2 (the hall-effect switch) moved to the bottom right side of the PCB. I measured 4mm from the high current traces. A wire carrying 50A generates 25 Gauss at a distance of 4mm.  The board area increased, but not materially.
  3. Change the hall-effect switch to increase the trip threshold. The S-5716ACDH2 component is specified with a typical threshold of 45 Gauss vs. 35 Gauss for the AH180 part. The minimum specified threshold is 25 Gauss.
  4. Use a magnet with higher field strength to compensate for the increased hall-effect threshold. I ordered a few 6mm dia. x 10mm length N52 Neodymium magnets to use in the application going forward. These new magnets have significantly higher B-field to compensate for the increased threshold. Another website yields useful calculations.
  5. There may be some users that want to locate the hall-effect sensor away from the main circuit board. I added pads to allow leads to be soldered instead of mounting the sensor on the board.

  The change from the AH180 hall-effect sensor to the S-5716 sensor changes the way that the switch works. Previously, the switch would change state when the magnet was removed. Now...

  Read more »

View all 17 project logs