Announcement

Collapse
No announcement yet.

Designing and building a simple 3 phase VFD - ideas and collaboration welcome :)

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Yes, that was a major consideration. This power board uses a discontinued IGBT module that I bought 3 pieces of on eBay for about $10 each. The seller has more than 10 still available and once I prove this design at realistic power levels I may order a few more. They should be able to handle, conservatively, about 10 amps per phase and 240 VAC, which would be about 4200 watts or about 5.5 HP. So I think my original goal of about 3 HP should be realistic, although I think I would need to use double sided power tracks and maybe 2 oz or greater copper. It might even be better to use external components for the input power circuitry and bus capacitors, and wire connections to pads and short tracks to the IGBT module. I am using 30 mil wide tracks which should handle 4 amps with 45o C temperature rise. That should handle 1660 watts or about 2.2 HP.

    [edit] I checked the vendor of the IGBT module and saw that he has some IR2175S current monitor ICs, 10 pieces for $20. They are designed to monitor motor phase currents using series resistors, and the output is a PWM signal with 130 kHz carrier. The duty cycle is 9% to 91% for voltages of +/- 276 mV. So 50 milliohm resistors would work for 5 amps P-P or about 3 amps RMS. A low pass R-C filter could be used to provide a linear signal of 0.15 to 14.85 volts with a 15 volt logic supply. This could then be measured by a PIC with a suitable voltage divider. However, these devices are not strictly isolated, so just like the MOSFET drivers that failed in the previous design and damaged the PWM outputs of the dsPIC, the same thing could happen with these. But the PWM signal could instead drive the LED of a high speed opto-isolator, or perhaps a digital isolator, and the output would also be a PWM signal. Instead of an ADC, it may also be possible to use digital processing to determine the duty cycle and relate that to phase current. The dsPIC16F has three input capture peripherals that can be used for this purpose. But I think the ADCs may be better suited to this purpose. I think I may order some of these for a future design.
    Last edited by PStechPaul; 03-16-2022, 02:29 AM.
    http://pauleschoen.com/pix/PM08_P76_P54.png
    Paul , P S Technology, Inc. and MrTibbs
    USA Maryland 21030

    Comment


    • I completed the Power Board design and sent the files to JLCpcb.com Everything seemed OK but then I got a message that an Excellon drill file was missing. I made sure it was in the zipfile but it was rejected again. And when I sent the files to FreeDFM.com it came back with some errors that didn't make sense, and the drill locations were offset from the rest of the board. But then as I was checking things and looking more closely, I found at first a minor error, and then a more major error, so I changed what was needed and resubmitted the design, which now seems to have been accepted and is in process. The cost for five pieces was just $2, plus about $20 for shipping, and then I used a $3 introductory coupon so the total cost is just $17.63.

      This board is the first one in a long time that I laid out 100% by hand, without the aid of the autorouter, and I think it looks pretty good:

      Click image for larger version

Name:	Simple_VFD_Power_PCB_R1.jpg
Views:	210
Size:	161.8 KB
ID:	1994082

      Here is the schematic, snipped from a PDF which is too large to be uploaded to the forum. If anyone is interested I can send any of the design files wanted. Now I will be working on the control board, and getting additional parts on order.
      Click image for larger version

Name:	Simple_VFD_Power_PCB_R1_Schematic.jpg
Views:	206
Size:	147.5 KB
ID:	1994083
      http://pauleschoen.com/pix/PM08_P76_P54.png
      Paul , P S Technology, Inc. and MrTibbs
      USA Maryland 21030

      Comment


      • Looks good
        Helder Ferreira
        Setubal, Portugal

        Comment


        • Just got word that the board is completed and shipped. I'm working on the control board now.
          http://pauleschoen.com/pix/PM08_P76_P54.png
          Paul , P S Technology, Inc. and MrTibbs
          USA Maryland 21030

          Comment


          • I have been out of town for a week, only looking at the forum while eating meals or just before turning in. So I have not been keeping up with the layout progress.

            I just noticed some things They include "R30", and also a set of resistors from R7 through R16.

            You have a safety exclusion zone including a number of apparently separating the "hot" side from the control inputs. It includes bidirectional LED isolators. But R30 and the others cross that barrier. That seems to defeat all the effort involved in creating the exclusion zone.

            The other resistors seem to be set up to read voltages from the hot side, presumably going from the "hot" side to the processor on the "safe" side. But their existence effectively makes the entire "safe" side into an extension of the "hot" side.

            Was that intended? Is the exclusion zone not intended as a separation between a "hot" and a "safe" area?

            CNC machines only go through the motions

            Comment


            • There are eight 100-150k resistors which mate up with another set of similar resistors on the control PCB, which comprise four differential amplifiers. And an additional 100-150k resistor which connects the analog grounds of the two boards to avoid a totally floating condition. With 200-300k total resistance, and peak voltage of about 320 volts, RMS leakage current will be only about 1 mA each, or 8 mA total, and that may be considered a "safe" level. This is a low cost means of monitoring the three phase to phase voltages, as well as the bus voltage. The lowest cost true isolation amplifiers are in the order of $6 each, and a linear optocoupler like the LOC111 or IL300 (one of which I am using for bus current) are in the order of $2.50 - $4 each.

              I intend to use the DRV5053 Hall devices, about $1 each, with a loop of wire to measure phase currents. I may use the IR2175 for future designs, but they are not in stock at the usual distributors, although I can get them for about $2 each on eBay.

              There is a lot of extra room on the control board, so I may add circuitry for IR serial communication. The 28 pin dsPIC has limited I/O so I need to make some design decisions such as only monitoring output phase voltages 1-2 and 3-1, and current in phase 1. However, the circuitry is there to monitor all six parameters, which can be done if I use a 44 pin TQFP version, which I will incorporate on an enhanced design for the control section. It may also be able to monitor temperature. And I might consider using a full color graphics display, like the perhaps also with a touchscreen interface like the ILI9341. But I'd better get this initial design working before I go down that rabbit hole.
              http://pauleschoen.com/pix/PM08_P76_P54.png
              Paul , P S Technology, Inc. and MrTibbs
              USA Maryland 21030

              Comment


              • There is not that high a limit on the isolation......... and it would never "pass" as a commercial product. Resistors are not regarded as good isolation, especially when there is just one in series. Some suitably rated capacitors are in fact accepted for usage type "Y" running from hot to ground. There is often one in an SMPS from the hot side to the low voltage side, for instance.

                Usually, the actual controller for the IGBTs (a uP) is on the power board, and the input board has a second uP that handles all the analog inputs etc, and has communication to the power controller via one bidirectional line through an optocoupler.

                The power controller handles all the output side measurements, calculations, and switching. That way there is good isolation, and the input uP can do all the parameter storage, etc, passing them to and reading them from, the control uP. The reduced number of optocouplers and board space can often pay for the extra uP, plus the improved function is a plus.
                Last edited by J Tiers; 03-31-2022, 05:58 PM.
                CNC machines only go through the motions

                Comment


                • I understand your concern, but I think this design will be OK for what it is meant for. There actually will be two resistors in series for each voltage sensor, one on the power PCB and one on the control board, both 150k. The main purpose for the isolation in this case is to avoid damage to the control circuitry as happened on the first version, with individual IGBTs and gate drivers. My design is loosely based on that of Microchip's ACIM demo board which used a PIC18F4431 instead of the dsPIC33FJxxMCxxx series I am using. Their MC board uses three 560k resistors on each leg as well as the bus voltage. It appears that they also use TLP2630 optoisolators for these voltages, although it looks like they are using them only for fault interrupts. They are not meant for analog signals. They are using an LOC111 isolator for bus current. Maybe I will go with a fully isolated version if it seems really necessary.

                  I have completed the control PCB layout so I can get those on order soon.
                  http://pauleschoen.com/pix/PM08_P76_P54.png
                  Paul , P S Technology, Inc. and MrTibbs
                  USA Maryland 21030

                  Comment


                  • Oh, the non-isolated design works, look at all the DC drives from various makers such as KB that are not isolated. You just have to move the isolation to mechanical isolation of controls and buttons. It's done all the time on light dimmers and other things of that nature.

                    And, analog optos are not particularly good, nor cheap, certainly not both at once. The usual cheaper solution is to digitize and send the info on the single bidirectional line I mentioned. That's if the isolation is intended.

                    The odd part here was the apparent concern for almost perfect CE isolation through the optos and trace spacings, and then having resistors cross that border. There is not much need for the isolation area if there are things crossing it, so it looked inconsistent.
                    CNC machines only go through the motions

                    Comment


                    • I have added the IR transmitter and receiver for the serial connection, so it will be possible to implement a custom terminal for basic monitor and control functions. It would also support a simple Bluetooth terminal which could be done on a PC or smartphone. Either one would provide total isolation. I used the "keepout" on the PC board to make sure there was adequate spacing between the power circuitry and the interface to the external control board. And having a "standard" interface connector simplifies the design variations such as individual IGBTs or MOSFETs, and more fully featured control boards using a 44 pin dsPIC and perhaps a full color TFT display with programmable touch-screen control functions. I have already implemented a small and cheap 1.8" display for an EV battery charger I was working on a few years ago, but it used an Arduino with its easy-to-use graphics libraries.

                      https://www.youtube.com/watch?v=-A8L1Q07kmU (Just the demo sketch)

                      http://enginuitysystems.com/files/EM...arger_2080.AVI

                      http://enginuitysystems.com/files/EM...arger_2085.AVI

                      Click image for larger version

Name:	Simple_Charger_2081.jpg
Views:	119
Size:	111.9 KB
ID:	1995227
                      Click image for larger version

Name:	Simple_Charger_2082.jpg
Views:	112
Size:	148.1 KB
ID:	1995229
                      Click image for larger version

Name:	Simple_Charger_2083.jpg
Views:	112
Size:	98.7 KB
ID:	1995230
                      Click image for larger version

Name:	Simple_Charger_2086.jpg
Views:	113
Size:	122.7 KB
ID:	1995228
                      http://pauleschoen.com/pix/PM08_P76_P54.png
                      Paul , P S Technology, Inc. and MrTibbs
                      USA Maryland 21030

                      Comment


                      • Wow! I just received the 5 PC boards from JLCPCB. They were actually delivered last night:

                        Click image for larger version

Name:	DHL_Delivery.jpg
Views:	140
Size:	176.4 KB
ID:	1995588
                        I had not expected them until next week, but I saw a link on an email that showed the delivery in my mailbox. However, when I went out to check, there was mail in the box, and the package was on the ground by the side of the road.

                        So here are the boards. I was just about to order some parts from Mouser and DigiKey, so I need to do that now.

                        Click image for larger version

Name:	Simple_VFD_Power_PCB_20220405_133743213_HDR.jpg
Views:	136
Size:	396.1 KB
ID:	1995587
                        http://pauleschoen.com/pix/PM08_P76_P54.png
                        Paul , P S Technology, Inc. and MrTibbs
                        USA Maryland 21030

                        Comment


                        • I received the second set of PC boards yesterday from JLCPCB via DHL express, and I also got my orders from Digi-Key and Mouser. I had ordered 10 pieces of a high power white LED, which was shown as a size 1608 surface mount package, and I had made the PCB with that size decal, expecting a part 0.16 x 0.08 inches. But it was actually a METRIC 1608, which is 1.6 x 0.8 mm, or 0.06 x 0.03 inches - way too small. So I contacted Mouser and they will credit my Discover card for the $2.55 and I ordered 10 pieces of a more suitably sized part 3.5 x 2.8 mm for $2.53. I did have to pay for shipping, but they have an economy option which is only $3.49.

                          Click image for larger version

Name:	Mouser_Parts_White_LED_1608_20220415_171245502_HDR.jpg
Views:	111
Size:	339.1 KB
ID:	1996981
                          Click image for larger version

Name:	Mouser_Parts_White_LED_1608_20220413_210727960_HDR.jpg
Views:	105
Size:	475.1 KB
ID:	1996983
                          Attached Files
                          http://pauleschoen.com/pix/PM08_P76_P54.png
                          Paul , P S Technology, Inc. and MrTibbs
                          USA Maryland 21030

                          Comment


                          • Believe it or not, I have been making some progress on this thing. Some issues I have encountered and dealt with:
                            1. There were several board layout errors, such as missing connections and incorrect part decals. Some cuts and jumpers were required.
                            2. I thought I had a large quantity of precision dual op-amps TLE2142 or TLC272, but I discovered that most of the devices were actually RC4558. They do not work on a single sided supply, and actually exhibit phase reversal when the input approaches the negative supply (GND). The design uses four of these so I had to order 25 at $1.44 each. Fortunately they were in stock. The TLE2142 in stock are over $4 each and cheaper versions expected in about a year!
                            3. The HV7800K1-G 450V high side current sensor was not in stock at the usual distributors but I found an eBay vendor where I got 15 pieces for $20. It's a Microchip product and no direct equivalent is available, so I may need to specify something else for future designs.
                            4. I have implemented an algorithm to derive the true RMS value of phase voltages and currents as well as bus voltage and current, including an "auto-zero" function.
                            5. I am now working on the power PCB, which has a relay that connects the output of the mains through a full wave bridge to the 560 uF bus link capacitor, but I am concerned that the initial charging surge current might damage the contacts. The ESR (equivalent series resistance) is only about 1/2 ohm so with 300 volts peak there could be as much as 600 amps. So I am considering a current limiter circuit to be added:
                            This simple circuit limits charge current to about 8 amps and allows an output current of 3.5 amps at 348 volts (1200 watts) with losses of 3.3 watts in Q1 and 6 watts in R4. More efficient means of current limiting could be done using a TRIAC or SCRs with phase firing or PWM with a MOSFET on the high side, but those require greater complexity. I'm not really sure how much such charging current might damage the relay, which is Omron G2RL-2-HA-DC12,

                            Click image for larger version

Name:	Current_Limiter_NPN_PNP.jpg
Views:	82
Size:	206.7 KB
ID:	2005687
                            http://pauleschoen.com/pix/PM08_P76_P54.png
                            Paul , P S Technology, Inc. and MrTibbs
                            USA Maryland 21030

                            Comment


                            • Many inverter welders just use a power resistor to limit the inrush current and a a relay to bypass the resistor with a simple voltage level detector to energise the relay coil. The normally open contacts just short the resistor and since the capacitor is almost charged, the switching current is minimal.
                              Helder Ferreira
                              Setubal, Portugal

                              Comment


                              • I might just solder a couple resistors across the normally open contacts of the relay, and wait until the capacitor charges up sufficiently. I don't have a voltage sensor for bus voltage on this board, but I could repurpose the ADC input currently used for the ITRIP sensor on the IKCS module. I do have a BUS_OC signal that should go low if bus current exceeds about 1 amp with the present 1.3 ohm resistor and HCPL817A opto that should turn on with about 1.3 volts. Good enough for low level testing, and easy enough to use something like 0.5 ohms for heavier motor loads.

                                I think a couple 100 ohm resistors should provide an RC time constant of 200 * 560 E-6 = 112 mSec or about 500 mSec to "close enough", and about 2 amps maximum charge current. Maybe sample every 100 mSec and turn on the relay when the samples indicate full charge, without knowing what the input voltage is. The problem here is that the resistors would dissipate about 300 watts each if the bus is shorted. I could use two 5k resistors which would be a TC of 10k * 560e-6 = 5.6 seconds and maximum power of 6 watts each. But that would mean about 15-20 seconds before closing the relay.

                                My current limiter circuit has a similar problem. If the capacitor is shorted, the MOSFET will see a peak power of about 2.2 kW and will quickly be destroyed. Perhaps a solid state relay using a TRIAC or SCRs would be best,especially if it is zero crossing activated. Using that approach, the peak input current would be about 67 amps with a total input resistance of 0.2 ohms, followed by 23 amp surges for capacitor charging with a nominal 100 ohm 3.5 amp load. Here is a simulation:

                                Click image for larger version

Name:	Current_Limiter_NPN_PNP_ACin.jpg
Views:	68
Size:	220.9 KB
ID:	2005708
                                Of course, applying power with a shorted capacitor would be a challenge for any simple protection scheme. My circuit also includes a 6.3 amp medium blow fuse, which would open within a few cycles for a solid short. A more likely fault scenario would be a short circuit in a motor lead or one of the H-bridge components. I have circuitry that is supposed to disconnect the bus link from the H-bridge if overcurrent is detected, but it may not act quickly enough to prevent damage. At some point one must accept the possibility of having to repair a PCB or replace damaged components. Hopefully, protection will be sufficient to prevent fire or operator injury, where a good fuse should do the job.
                                http://pauleschoen.com/pix/PM08_P76_P54.png
                                Paul , P S Technology, Inc. and MrTibbs
                                USA Maryland 21030

                                Comment

                                Working...
                                X