Picking a DSP MCU for Power Conversion - Experiences?

[NiHaoMike]

I'm not familiar with a PLC receiver.  This is about comms over the power line?

What sample rates do you expect to use for the various control aspects?

Yes, data communications using a RF carrier over the power line. The idea being that while every zero export inverter I'm aware of on the market requires the inverter to be connected to sensors using a dedicated cable in order to zero export to upstream loads, my design just requires installation of the sensors and a PLC transmitter module. The actual PLC modulation format I decided to use was a FM carrier at about 800kHz-1.4MHz (can be changed using a command over the reverse PLC channel) with an ASK modulated pilot tone for digital power readings and a subcarrier for the difference between the two phases. I went with that rather than a fully digital modulation since it would make the transmitter a lot simpler and also make it simpler to meet latency requirements. The same signal is sent over both phases and the coherent diversity receiver does some "beamforming" tricks to mitigate interference. (In the process, I also learned just how complex is the stuff my friend Tiffany Yep, a digital communications engineer, works on. Now if only she had more time to teach me DSP using Verilog...)

The voltages and currents are digitized at 768kHz, which is quite fast considering the PWM is on the order of 15kHz. The "core" frequency in the FPGA is likely to be the 36.864MHz bit clock for the ADCs or some multiple of it. The FPGA is a LX100, a bit dated and probably much bigger than necessary, but I have several boards that I got for cheap.

[Phoenix]

The idea being that while every zero export inverter I'm aware of on the market requires the inverter to be connected to sensors using a dedicated cable in order to zero export to upstream loads, my design just requires installation of the sensors and a PLC transmitter module.

That's a good idea. Is this a commercial zero export inverter you're working on, or more of a hobby project? Are you also making your own power meter, or can you interface to other meters that already have PLC? In guessing own power meter since it's home brew PLC.

[gnuarm]

I've worked a bit in comms so I am at least familiar with the terms you used. 

For an ADC the numbers you provide would give you only about 5.5 bits.  You can bump that up to 7 bits if you use a PLL to generate a ~100 MHz clock for the ADC circuit, ~8 bits with a 200 MHz clock.  The limitation will be the I/O cells themselves.  Rising and falling edge asymmetries and high/low drive level asymmetries will distort the average value of the feedback signal, at some point dominating as the rate increases.  You can also gain one bit of resolution for each halving of the data rate from the ADC.  So you might ask yourself if you really need a 50x ratio between the ADC sample rate and the control PWM rate.  I'm not sure what you would do with that data unless you are extrapolating curves. 

The ADC I am working on is sampling at 33.55 MHz but with a very low data rate, so lots of decimation and averaging.  The counter is 18 bits, so the other end of the scale from what you are doing. 

Just as a point of comparison, the audio output from the device I am working on uses 32 kHz to produce audio tones.  Seems they like their alarms to not be alarming, but rather musical.

[NiHaoMike]

That's a good idea. Is this a commercial zero export inverter you're working on, or more of a hobby project? Are you also making your own power meter, or can you interface to other meters that already have PLC? In guessing own power meter since it's home brew PLC.

A hobby/research project. I am making the power meter, there are ready made power meters specifically for zero export inverters but all of them use a low latency interconnect like Ethernet or high bitrate RS485/CAN. Because the latency of a "bare metal" Ethernet interface is incredibly low, even adding a "store and forward" switch in between will likely prevent it from working properly, let alone anything more complex like a pair of Homeplug adapters. (I did look at one inverter using such a power meter and the directions say to connect the meter directly to the inverter, no switches or routers in line.)

[Phoenix]

A hobby/research project. I am making the power meter, there are ready made power meters specifically for zero export inverters but all of them use a low latency interconnect like Ethernet or high bitrate RS485/CAN. Because the latency of a "bare metal" Ethernet interface is incredibly low, even adding a "store and forward" switch in between will likely prevent it from working properly, let alone anything more complex like a pair of Homeplug adapters. (I did look at one inverter using such a power meter and the directions say to connect the meter directly to the inverter, no switches or routers in line.)

As much as I like the PLC approach, I'm not sure I agree with the dire need for low latency. The response time requirements that I have seen are in seconds (VDE 4105 reduce to 10% of power excursion in 3s and AS/NZS 4777 draft talks about 15s) and you can do zero export using "slow" RMS measurements. Getting rid of the dedicated cable is the real benefit that I see. Could also consider an ISM band wireless which might be cheaper.

[NiHaoMike]

I think the "zero export" definition for some areas like Hawaii are relaxed for latency requirements provided the inverter has a graceful voltage limiting feature. I think that was done so that some older inverters (that have a production limiting feature but not a true zero export feature) can still be used for new installs by simply adding on an export limiter module. If such a loose definition of zero export was created specifically to make it easier to make zero export inverters with wireless sensors, I haven't found any on the market.

Other than working around limitations of older equipment that isn't really designed for it, it's not hard at all to make an inverter with wired sensors that can react to load changes in milliseconds. With some control loop tuning, a few hundred microseconds or less can be achieved. One big reason to use a zero export inverter is to not have to go through the hassle of getting the system to be allowed to export (which nowadays sometimes pays too little to be worth the hassle, especially for smaller setups) and therefore work with an existing electrical system including a utility meter that specifically looks for reverse power flow as a possible hacking attempt. Limit the reverse power flow to be a fraction of a cycle and it will basically act somewhat like a reactive load during that instant.

[Marco]

Regulation getting super uptight about microsecond level reversal of current doesn't make a whole lot of sense to me. A couple watt of poor power factor current getting sent upstream isn't a problem.

[nudge]

I’ve designed power converter controller boards using TMS320F2810 and TMS320F28069 parts. TI parts are pretty standard in the industry.

Pros are fast/hi-res & immediate trip PWMs, decent ADCs, fast trig math support, and tons of supporting documentation and resources for power conversion design.

Cons are propriety toolchain, and less advanced peripherals for applications processor type work — they’re aimed at real-time control applications. Having said that, some of the newer chips have both a C2000 + ARM Cortex-M4 core in a single chip, this is great for offloading the connectivity and applications work to the M4 core (eg TMS320F28386S).

[gnuarm]

I’ve designed power converter controller boards using TMS320F2810 and TMS320F28069 parts. TI parts are pretty standard in the industry.

Pros are fast/hi-res & immediate trip PWMs, decent ADCs, fast trig math support, and tons of supporting documentation and resources for power conversion design.

Cons are propriety toolchain, and less advanced peripherals for applications processor type work — they’re aimed at real-time control applications. Having said that, some of the newer chips have both a C2000 + ARM Cortex-M4 core in a single chip, this is great for offloading the connectivity and applications work to the M4 core (eg TMS320F28386S).

Sort of reminds me of Segal's law, "A man with one watch knows what time it is; a man with two watches is never sure." 

That said, I suppose two processors can be better than one... sometimes.  The trick is knowing when those times are.

[NiHaoMike]

Regulation getting super uptight about microsecond level reversal of current doesn't make a whole lot of sense to me. A couple watt of poor power factor current getting sent upstream isn't a problem.

Microseconds would indeed be overkill, but stopping reverse power flow in milliseconds or within a cycle or two is not difficult. As mentioned, the main reason is to not trip reverse power detection on utility meters.