DC-Link Capacitors

[Pasqualino31]

There exists these Metallized Polypropylene Film Capacitors (MKP) designed for DC-Link applications. I found these TDK Metallized polypropylene Film Capacitors(MKP) that seem too good to be true. They're p/o the B32320I_series (PN: B32320I820207K000) 200 uF @800VDC, ESR=3.6 mΩ ESL=75nH IRMS(max) = 27.5A with Ipeak = 1.5KA. Apparently they're discontinued, but Mouser still had some in stock, so I just had to buy 10-pack from the same lot.

I knew nothing when Ii first joined this group. Today, I now know slightly less than nothing about an actual functioning solar system. I watched Wills YT videos and some other people's stuff on YT and started by studying the most rudimentary block diagrams of off-grid and hybrid systems. The LiFePO4 batteries alone proved to be a series of rabbit holes, never mind the BMS systems. When I got to the DC (from the batteries) to the DC/DC converter -> the AC inverter, I was back to some familiar ground. It's conceptually the same as an EV.

So as an experiment I tried to design an SMPS using a cheapo COTS IC as a controller to get a DC/DC Boost converter for a 48VDC LiFePO4 boosted to ~ 170 VDC, the minimum voltage necessary to feed an inverter to get 120 VAC RMS. 170/sqrt(2) = 120. I managed to get a design to simulate in MicroCap, but when it came to the components, the inductor alone was like a 4lb toroid core with 40 turns of AWG #8. Not a practical experiment. I'm in the process of designing a full-bridge DC/DC converter and will use a cheap FPGA as the PWM controller. I can use the FPGA to control for the DC/DC converter and the Inverter.




When I decided to take a closer look at the DC-Link capacitor(s) I was amazed to find those TDK caps mentioned above. They may be a solution to a problem I had many years ago when trying to build a Fender Blackface style guitar amp. I apologize ahead of time for deviating from the purely Solar subject matter, but the power supply theory and the DC-Link caps are definitely part of all this good solar power subject matter. So at the risk of being sanctioned by the DIY Solar board of directors, I'll try and briefly describe the Gtr Amp's P/S circuit.

I have a toroidal power transformer for tubes (2x 6.3 VAC secondaries) that I got on the cheap. It has no secondary center-tap, but taps for 260 VAC & 280 VAC. The problem is that you need I'm trying to get ~ 460+ VDC to my tube plates and since there's no center-tap, I was forced to go for the dreaded half-wave capacitive voltage doubler. Such a circuit would require massive capacitance to prevent power supply 'sag' which is really undesirable for a guitar amp. I was forced to abandon this project because Electrolytic caps are polarized and non-polarized lytic caps are expensive and they're less than ideal for a doubler circuit for many reasons, inductance being one problem. The MKP caps used in these tubes (the "Orange Drops) are expensive as hell. I never imagined an 800V, 200uF version would ever be so affordable.

They're non-polarized, 200uF @ 800 VDC, so they can theoretically do the voltage doubling; however, these are designed for DC (positive or negative) with RMS ripple current rated for 27.5Arms. They aren't really designed for a reversing waveform which is what the first cap in series with the primary will get. It does have a pulse handling rating, though. (the taller the can, the lower the pulse handling capabilities.) In this case it's the 120mm (~4.7") tall can (it's a plastic can) rated for 7.5 V/us. The waveform should be ~ +/- 320V or so, clipping hard, so it should rise in 1/2 cycle ~ 8.33... ms. => 8,333.33 us. So that's 76.8 mV/us, well within the 7.5V/us rating for pulsed waveforms. I'll simulate it to check. this is the doubler circuit from the MicroCap project:



That component btw the AC source and the transformer primary is called an NTC thermistor. If you're not familiar with those an MOVs (Metal Oxide varistors), then you might want to check them out. They're crucial components for dealing with surge and transients (like lightening strikes). In this case the NTC is there to limit inrush current at startup time or for brownouts etc. I have a simulation of this circuit if anyone is seeing the behavior. I have pretty good models for MKPs and 'Lytic caps in there as well as the transformer core properties and the NTC. I also have simulations of the 48V to 170V DC/DC converter I mentioned earlier. Although it's not practical to build, it does work (in theory). The DC output is held at 170 VDC for up to 30% drop o the I/P. In the near future I'll post the simulation of the full-bridged DC/DC converter w Inverter that will have the same DC-Link caps from the Guitar Amp P/S.

So these crazy caps all seem too good to be true. Voltage doublers usually don't work out for DC supply for dynamic DC loads like motors tube amps. If there's anyone who sees any problems I've overlooked? Also, does anyone know why those TDK parts were discontinued? I understand Kemet has a pretty good line of these products. Any feedback would be appreciated.

 

[frankrudis65]

Hi Pasqualino,sorry haven't got much time to get really into it,but:
I installed in my single phase 4.5kw hybrid inverter in 2024 new Bus Capacitors including other mechanical things like fans etc to make my inverter fit for the next years .I kept two polarized caps ,the Kemet PEH200 series and added Cornell Dubilier Type 947D polypropylene dc link caps 2 pieces .You will be surprised how little capacity you actually need,as the usual electrolytes are solely based on the ripple current,they need to cover There's a very interesting paper on internet: PV Inverter Performance and Reliability :What is the Role of the Bus Capacitor.
Anyway I installed them,because of their size ,outside the cabinet and everything went well.Best regards Frank,northern italy

 

[efficientPV]

Voltage doublers aren't dreaded. That looks like a lot of work for nothing. I'd save those caps for something useful.

 

[Pasqualino31]

Hi Pasqualino,sorry haven't got much time to get really into it,but:
I installed in my single phase 4.5kw hybrid inverter in 2024 new Bus Capacitors including other mechanical things like fans etc to make my inverter fit for the next years .I kept two polarized caps ,the Kemet PEH200 series and added Cornell Dubilier Type 947D polypropylene dc link caps 2 pieces .You will be surprised how little capacity you actually need,as the usual electrolytes are solely based on the ripple current,they need to cover There's a very interesting paper on internet: PV Inverter Performance and Reliability :What is the Role of the Bus Capacitor.
Anyway I installed them,because of their size ,outside the cabinet and everything went well.Best regards Frank,northern italy

Thanks, Frank,

I agree regarding the caps acting as filters on the DC Bus. In my case, I need higher voltage than my transformer will allow, so I need caps (2) to act as voltage doublers so that I can achieve my desired plate voltage. I updated the schematic to its most recent and circled the Doubler caps in yellow. The symbol used for the part may make you think it's polarized, but it's not. It's just a way to show the center terminal from the four parallel terminals. (The center terminal is tied to the 260VAC tap of the secondary.) The doubler caps are the TDK DC-Link caps. All other DC filter caps in the supply are electrolytic. I used a little more capacitance than I would for an amp with no doubler circuit. The doubler caps are a beefy 200uF apiece. I had to double up on the filter caps since 450VDC were the highest affordable voltage and a better use of space. Unfortunately, it's half the capacitance since they're in series.

Below the schematic is a transient response simulation. It's a zoom-in of the first 4.6ms from startup time. Other than the crazy transient that screams out at you just after the first zero crossing, there is some other good information in there. Most important is the current through that first doubler cap. That's the black sinusoid in the second window of signals. That's a good indicator of how it might respond to a fast demand for current since there is a big inrush current at startup. The plate voltages at A thru E settle out nicely by 10s in and this simulation is under full load. There's an entire vacuum tube system attached.

https://drive.google.com/file/d/1CIMfGpu4WKi_qdkTMjDEYTBdb2pkKmgY/view?usp=sharing

Back to the schematic. You can see the snubber caps across one of the diodes. In the simulation, I monitor the current through one of the diodes (D4) the and the first doubler cap. You can see the transient current is moving in opposite directions. I was experiencing switching nose coming from the diodes, so I added the snubber caps. there are bleeder resistors for all large caps in the power supply.

I have a different series of TDK MKP capacitors designed for switching power supplies and EMI suppression. The effective inductance is much lower for these parts, so I'm going to put a .1 uF @1.6KV (peak) across each of the doubler caps and see how it goes. Other than that transient at startup, the DC voltages have very little ripple (~1%)

 

[Pasqualino31]

Voltage doublers aren't dreaded. That looks like a lot of work for nothing. I'd save those caps for something useful.

In general, voltage doublers are not "dreaded"; however, in the case of a guitar tube amp they most certainly are. This is especially true if you're forced to go with a half-wave doubler as I am here, since there is no center tap to do a full-wave or a bridge. I bought these caps with the intention of using them to get my plate voltages for this amp project, so *this* is the something else useful. There are 8 more caps left in the lot of 10 I bought, so if something "more useful" comes up they're available. I'm sure I'll use them for my Full bridge DC/DC converter when I get back to that project. The tube amps are old hat for me, it's this voltage doubler power supply that is something new. If it works out, then I'm sure I'll have a request to build another. That's a big if at the moment.

 

[Warpspeed]

The half wave doubler circuit you are using requires one end of the transformer to be grounded.
An ac to ground input voltage is often useful if not sometimes a necessity.

As you have a floating secondary winding available, its possible to use an alternative full wave voltage doubler circuit that requires only half final voltage electrolytics, which offers advantages of size and cost, especially for higher voltages.
A full wave doubler like this is much kinder to the transformer, because its equally loaded in each direction.
For +460v dc output a secondary voltage of about 165v will be required.
Standard aluminium electrolytics will work just fine.
You may need to think seriously about limiting the initial power up surge with some series resistance which can be switched out after a short delay.

The way I would do something like this would be with a heavy duty three position rotary switch. Power off, initial power up through a series resistor, main power on.

 

[efficientPV]

Anytime I see a pretty schematic on the internet my first thought is ....Watch out! The good news is it is a hobby application and nobody dies. Cap input is very limiting and should be avoided unless going more than double.

 

[Pasqualino31]

The half wave doubler circuit you are using requires one end of the transformer to be grounded.
An ac to ground input voltage is often useful if not sometimes a necessity.

As you have a floating secondary winding available, its possible to use an alternative full wave voltage doubler circuit that requires only half final voltage electrolytics, which offers advantages of size and cost, especially for higher voltages.
A full wave doubler like this is much kinder to the transformer, because its equally loaded in each direction.
For +460v dc output a secondary voltage of about 165v will be required.
Standard aluminium electrolytics will work just fine.
You may need to think seriously about limiting the initial power up surge with some series resistance which can be switched out after a short delay.

The way I would do something like this would be with a heavy duty three position rotary switch. Power off, initial power up through a series resistor, main power on.

The secondary is grounded to the DC Bus, you just need to look for the ground symbol, but yes, this circuit must have a ground, preferably at a center tap on the transformer secondary. Unfortunately, this is what I have to work with. The chassis of the amp is tied to earth ground and all DC voltages to the tubes (even the ~ -35VDC bias voltage). On the primary side there are safety caps from the Neutral (WHT wire) to Earth GND (GRN wire) and from the Hot (BLK wire) to Earth GND There is a 3A Slo-Blo fuse inn series w the Neutral. The MicroCap schematic does not show this, it shows the primary as GND, but I have to do that because MicroCap doesn't do multiple grounds. It's fine for simulation purposes.

The Voltage Doubler circuit in your diagram is a split supply. What we have here, DC supply-wise, is exactly what it would be without the doubler. A single sided supply with chassis ground reference. I have made this particular amp (the audio section and the P/S after the Schottky diode) a half dozen times in my life as well as many more with higher voltages, bigger tubes and bigger transformers. This transformer is a mere 50 VA and the power stage can deliver 22W of peak power to the speaker. That's not why I posted this here though. It's all about the power supply.

High voltages are nothing to mess around with. That goes for Home Solar, this guitar amp unit as well as our toasters or any household appliance. They are all capable of killing us. I have an isolation transformer, a variac and I made one of the DIY 150W incandescent light bulb limiter many moons ago. Long before there were YouTube videos. this paragraph isn't directed at you personally. It's just something I felt needed to be said just in case anyone gets any ideas when reading about this project.

As for the balanced loading of the secondary, check the simulation in the second window. That's where the currents are monitored. L21 is the secondary tap at 260 VAC. (X2.C1 is the current through the first DC-Link cap in series w the 260VAC tap) As you can see in the simulation, the current through both is pretty balanced at startup. The light blue line is the current through the NTC thermistor. It keeps surges under control. You are correct, though. A sudden load on the 6V6 tubes would tend to draw current during the positive cycle. With all the capacitance in this P/S, it "should" be stiff enough to handle the surges expected for those rare occasions when playing at full volume. (22W might not sound like much, but this is a pretty loud amp when turned up.)

So the next step is to see if the P/S sags under load or if any undesirable artifacts are generated. This will certainly a good test for these DC-Link capacitors.

 

[efficientPV]

I certainly like your enthusiasm and at first I found your explanations disturbing. But, this is a solar forum and after some thought I see it as great entertainment. There are certainly large gaps in your understanding and much of your drawing doesn't make sense to me. The solar world has a metric of better than nothing and this passes that. Certainly is not something anyone should emulate. What is this amp, 15W? Why don't you post that schematic on the electronics forum EEVblog.com if you want some additional opinions.

 

[OhYou_]

May i suggest instead simply using a 170v+ battery bank

 

[Pasqualino31]

May I suggest instead simply using a 170v+ battery bank

Sure you can. How much would tat cost and how much would that weigh? I'm no expert on battery banks, but a 170v+ battery bank sounds heavy and expensive. It's also kinda hot in that amp case. I'm all ears though.

 

[OhYou_]

Sure you can. How much would tat cost and how much would that weigh? I'm no expert on battery banks, but a 170v+ battery bank sounds heavy and expensive. It's also kinda hot in that amp case. I'm all ears though.

I didn't read your post thaaat much. At such a small scale, it's not worth it going in that direction, I meant more on the larger scale.

 

[Pasqualino31]

I didn't read your post thaaat much. At such a small scale, it's not worth it going in that direction, I meant more on the larger scale.

No worries, thanks for the response.

 

[Pasqualino31]

I certainly like your enthusiasm and at first I found your explanations disturbing. But, this is a solar forum and after some thought I see it as great entertainment. There are certainly large gaps in your understanding and much of your drawing doesn't make sense to me. The solar world has a metric of better than nothing and this passes that. Certainly is not something anyone should emulate. What is this amp, 15W? Why don't you post that schematic on the electronics forum EEVblog.com if you want some additional opinions.

I've got it posted on a DIY guitar amps forum, but I joined this group to learn more about DIY home solar and found Will's YT posts to be interesting and informative. I spent some time this winter working on a unique approach to making a "boutique" guitar amp using all toroidal transformers and choke(s). I also wanted to make my own inverter since I like take my Jeep out on the shore to fish, and it would be nice to have one to run some household appliances. I researched home solar and found this site from the YT videos. That's how I found the DC-Link capacitors and they seemed to fit a need for my amp project. The toroidal transformers are limited in how much voltage you can get out of them. I needed either a doubler circuit or to stack two toroidal Xfmrs (parallel the120VAC primaries and wire the secondaries in series.) The doubler turned out to be a more desirable solution. There is plenty of crossover here, please believe me, I'll get to it.

I currently only have a single 12V LiFePO4 battery with built-in BMS and someone stole my (crappy) little solar panel out of my Jeep last year along with a 12V marine battery. The weather in NY is just getting nice again, so I'll be purchasing another marine battery and a more proficient solar panel soon. It would be easier and probably cheaper to just buy an inverter, but that's not how I'm wired. I like to do electrical/electronic things myself. It keeps my skills sharp and I'm always learning new things.

That being said, I'll attempt to pull this thread clearly into the realm of DIY Solar by discussing my inverter project. For starters, we need 120Vrms, so that will require a minimum of ~ 170VDC:

Vpk = Vrms*sqrt(2) = 169.71 ~ 170Vpk

That would be for a 100% efficient inverter. Okay, we can do the 170v+ battery bank or a lower voltage with a DC/DC converter. I saw pure sine wave Inverters that claimed to take 12VDC -> 120VAC @ 1KW (or more). What BS! That's upwards of 100 Amps requiring AWG2 wire of better (the heaviest of heavy duty jumper cables) and many 12VDC batteries in parallel.
I sure AF don't need a 1KW inverter, anyway. I'm sure that's just their peak power rating, but why read further? They're obviously trying to deceive.

After a little thought, a 48V+ battery (at minimum) would do the job for A 120VAC inverter @20A. Just like one line in your home. I have some experience with boost and buck regulator design, but nothing nearly as big to get 170 VDC from 48VDC. So, just for fun I tried out a standard design using UC3842 Switch Mode Power Supply (SMPS) controller. It's an industry standard part and real cheap. A 4:1 step-up is about the limit for this type of part and this is especially true for an output of 170 VDC. (I wound up having to go with a 96+ VDC battery, but I'll get to that later.)

Just a quick model and description for how these SMPS' work:

The controller electronically turns the switch on & off. In the circuit, the electronic switch is a MOSFET. The controller turns it on and off by applying pulse to the gate of the MOSFET. When the switch is on, current flows through the inductor, L1 creating a magnetic field. The instant the switch is opened the magnetic field is still there and the current flow through an inductor cannot change instantaneously (like voltage across a capacitor.) So, current will continue to flow as the magnetic field decays, charging C1 though the rectifier diode, D1. Sparing everyone the derivation, the Voltage out is given below a a function of the duty cycle of the switching waveform AKA the PWM duty cycle, δ . Here's a link for beginners who want to know more.

Really simple demo:

conceptual video (no math):

equation derivation:

Let's do an easy case with nice round numbers, say Vin = 10V and Vout = 20V for a 100W Boost Converter.

We know Vin and Vout, solving for δ; δ = 1 - Vin/Vout => 1 - 1/2 = 1 - 1/2, or .5 This means the switch is on (and off) 50% of the time.
The switching frequency, fs of the UC3842 is ~ 50kHz; or the period, T = 1/f = 2us
We can easily calculate the the input current from the power equation, P = I•ΔV => I = P/V
However, Pin =Pout only if 100% efficient. Let's assume 90% efficiency, ղ = .9

Pout = ղ•Pin => Pin = Pout / ղ
Iin = Pout / ղ•Vin = 100W/(.9)(10) = 11.11A
Let's also assume 10% ripple (this is kinda reasonable because we can filter 10% ripple with our awesome DC-Link Capacitors)
For 10% ripple, ΔI = (.1) 11.11 = 1.11A
The required inductor value for a given ripple current can be given by the equation below.
Video w required inductance, L as a function of ripple ripple current equation derivation:

L = Vin δ / ΔI fs

L = (.5 )(10) / (1.11)( 50K) = 90uH


*DISCLAIMER:
The efficiency of the system is affected by many variables. The ripple current is more of a byproduct of the system elements which present a chicken or the egg paradox. Core losses of the inductor core will be the dominant contributor to efficiency. The current through the inductor and number of coil windings will contribute to core losses. The number of turns of the coil increases inductance with square proportionality => N2 ∝ L While a more inductive circuit is good for reducing ripple and storing energy, it also increases losses in the core.

Let’s make sure we can find or build this inductor ourselves.


IL = Iin +ΔI = 11.11A + 1.11A = 12.22A

Shopping around a little and I found this:

https://www.mouser.com/ProductDetail/Eaton-Electronics/CTX100-10-52LPR?qs=RvTAiNeQOggytKd0bHyw3g%3D%3D

100 uH @ 13A

Well isn't that nice? We don't have to wind our own and we don't have to calculate the max flux density and e don't have to hunt down a core and hunt down the core properties to see if it saturates with our value for max flux density...

Well, that's not how it worked out for boosting 96VDC to 170+ VDC, but I'll show that another time. In my next post, we'll configure the circuit for the UC3842. It will look something like this:



For the R1, R2 divider:

Choose R1=10kΩ,

R2 = R1⋅Vref / (Vout - Vref)

Where Vref is the reference voltage of the controller (e.g., 5 V).

R2 = (10k x 5) / (20-5) = 50k/15 = 3.333K => 3.32K (std. value)


For Oscillator timing:


The switching frequency is set by the timing resistor (RT) and capacitor (CT) connected to pin 4. For fsw= 50kHz:

fsw = 1.72RT⋅CT



Choose RT = 10 kΩ and CT = 3.3nF for fs ≈ 50kHz.





Note: Everything within the rectangular border is inside the UC3842 IC.

 

[MattiFin]

. I saw pure sine wave Inverters that claimed to take 12VDC -> 120VAC @ 1KW (or more). What BS! That's upwards of 100 Amps requiring AWG2 wire of better (the heaviest of heavy duty jumper cables) and many 12VDC batteries in parallel.
I sure AF don't need a 1KW inverter, anyway. I'm sure that's just their peak power rating, but why read further? They're obviously trying to deceive.

--------
Just a quick model and description for how these SMPS' work:

----------

Oh boy, I don't even know where to start with this nonsense.

1kW 12v inverters are real thing and 100A is still quite reasonable current, if you need substantially more than 1kW the general consensus is to go for 24 or 48v. ( even 4kW 12v inverters exist but wiring gets unreasonable.)

And nobody would seriously consider boost topology SMPS to step up 12v to 170v. Nor 12v to 340v in rest of the world.

 

[peufeu]

If you use a non isolated DC-DC converter between battery and inverter then the battery won't be isolated from the inverter's output. That's an issue because 12-48V gear, wiring, fuse holders, etc, are neither rated nor safe to use on something that is at mains potential. If you use it in a vehicle, and the battery negative is connected to chassis, then the AC output won't be isolated from chassis either. All pretty deadly.

That's why all inverters use transformer based boost converters, usually resonant for efficiency. It's also much more efficient when the boost ratio is high, but the main benefit is safety.

 

[efficientPV]

You really should take a serious look at schematics and anything else before you post them, like placement of D1 in the last schematic for just a start.

 

[MattiFin]

You really should take a serious look at schematics and anything else before you post them, like placement of D1 in the last schematic for just a start.

Care to explain?
Looks like bog standard boost converter schematic?

 

[peufeu]

The diode is in the wrong place

 

[Pasqualino31]

You really should take a serious look at schematics and anything else before you post them, like placement of D1 in the last schematic for just a start.

Yes, that's why I always post and ask for feedback. Thank you for pointing that out. It was actually the inductor that was in the wrong place.

Oh boy, I don't even know where to start with this nonsense.

1kW 12v inverters are real thing and 100A is still quite reasonable current, if you need substantially more than 1kW the general consensus is to go for 24 or 48v. ( even 4kW 12v inverters exist but wiring gets unreasonable.)

And nobody would seriously consider boost topology SMPS to step up 12v to 170v. Nor 12v to 340v in rest of the world.

Who said they don't exist? I'm just saying that they're useless. 12VDC is a terrible choice to drive a 1KW inverter, PERIOD! Ultimately, that 12V has to provide 1KW of power, that's 1KW/12V ~ 83.33A. Batteries have internal resistance, how hot do you thing tat battery will get? The ones advertised that claim 1KW of power really should put a disclaimer in a font twice the size of the claim that says ABSOLUTE MAXIMUM PEAK POWER RATING!!!

Care to explain?
Looks like bog standard boost converter schematic?

No shit sherlock.

 

[Pasqualino31]

The diode is in the wrong place

Yes, in that diagram it's in the wrong place. I don't care, the caption says, "looks something like this". Let me finish the thought, for Christ's sake! It wasn't even the Diode, it was the inductor.

I did a thorough analysis for the other diagram, which, is correct as is the finished and simulated circuit. I was doing a walk through for an example design. If I was to design it to that diagram, it would get picked up as soon as the circuit was drawn in schematic capture, so y'all shouldn't get your panties all in a bunch. It's not even my diagram, I clipped it from one of the many manuals out there, and the circuit does look "something like that". So there.

 

[Pasqualino31]

If you use a non isolated DC-DC converter between battery and inverter then the battery won't be isolated from the inverter's output. That's an issue because 12-48V gear, wiring, fuse holders, etc, are neither rated nor safe to use on something that is at mains potential. If you use it in a vehicle, and the battery negative is connected to chassis, then the AC output won't be isolated from chassis either. All pretty deadly.

That's why all inverters use transformer based boost converters, usually resonant for efficiency. It's also much more efficient when the boost ratio is high, but the main benefit is safety.

Yes, an SMPS is not what you want to use for this application. I mentioned this in the description, but went ahead with the design anyway for educational purposes.

Here's what I wrote:
" I have some experience with boost and buck regulator design, but nothing nearly as big to get 170 VDC from 48VDC. So, just for fun I tried out a standard design using UC3842 Switch Mode Power Supply (SMPS) controller. It's an industry standard part and real cheap. A 4:1 step-up is about the limit for this type of part and this is especially true for an output of 170 VDC. (I wound up having to go with a 96+ VDC battery, but I'll get to that later.)"

I did it just to see if it could be done. It pushes the SMPS to the very edge of it's theoretical limits.

So, I agree with you, but you're you're getting out in front of me here, though. I'll get to the Full Bridged design with transformer isolation after establishing the fundamentals of DC/DC Boost Regulation. This is a DIY post on a DIY Solar Forum. To some people DIY means buying mass produced products, and there's nothing wrong with that. You're not paying someone else to do it, so it's DIY. It makes sense to do it that way. Redesigning the wheel is not recommended.

I'm pressing on with this and will walk it to its completion.

 

[Pasqualino31]

Despite the kind of dismissive and sometimes a bit nasty tone of the commentary in this thread, I'm going to press on with this project. I'm not saying this to everyone who replied. It's a comment on the general tone of the replies. They don't have helpful nature about them, there mostly dismissive and border on confrontational.

I never claimed to be an expert on Solar Power, LiFePO4 batteries, or anything else on the level that Will presents in his videos. I've been very up front with that. I'm here to learn because I want to buy a house in Upstate NY and make it completely off grid.

My expertise is in low power, Electronic Hardware design. I have worked projects, like RADAR systems, where the system ultimately was very high power, but even the RF I worked on was on the back end of the receiver which was very low voltage. When I was in the NAVY, I diagnosed faults and affected repairs on a few occasions, but I was trained to maintain and repair, not build. Over the better part of last 40 years I've done mixed signal design (analog/digital) including signal integrity, but the last 20 years of my career have been focused on reconfigurable digital systems on FPGAs. These FPGA have evolved from a matrix rudimentary gates then gradually embedded higher functionality. First adders, then multipliers, then microprocessor cores. Now we have GPU processors with an insane amount of hard embedded cores for multiple functions. They're ALL built on FPGAs first. When I design digital systems on FPGAs, I do it in code. That's full 180 degree turn from where I started. I still have to design the support circuitry on the board: data acquisition (A/D D/A converters) BUS interfaces (e.g. PCIe, USB and others I'm sure most of you have never heard of) and the Power Distribution Networks (PDN) have gotten real complex. There are actually controllers that just handle the power distribution for a single FPGA system. Look at your computer MoBo. That's a great example of one of these systems. It's a smorgasbord of analog, digital, power, and other types of low power circuits.

So, If I'm going to learn DIY Solar systems, I'm going to start with what I know. That's from the bottom with low voltages. The SMPS is a great start. After doing one in the simulator, regardless of its practical applications, I can build one up and replace the controller IC (the UC3842 in this case) with a cheap FPGA module board. My choice was a CMOD A&-35T because I have 3 of them:

https://www.digikey.com/en/products/detail/digilent-inc/410-328-35/6133793

But you can just use a pic controller or whatever, not really the point. Once the FPGA controller is working within an SMPS configuration on LOW VOLTAGES, I can go to a more practical circuit, like a fully bridged and isolated DC/DC converter. It's much easier to scale the voltages up from such a circuit and I can scale the controller accordingly.

I welcome comments to help me gain insight into the finer points of Home Solar Power systems or EV applications of which I know even less than Home Solar. Alternatively, you can nitpick my diagrams and tell me thing like the diode is in the wrong place or just be a dismissive little prick if you want. It's your time.

 

[peufeu]

OK that explains the previous posts

I confess being a bit puzzled about the combination of 1) your lack of experience with power electronics and 2) "I'll just do it with FPGA" which either means total mastery or brash foolishness but that is now cleared up!

Doing RF and highspeed digital should provide you with some of the required skills that can be transposed to power conversion, not all of them of course, but I'll bet you already have the absolute most essential one, which is to not fuck up the layout. If you switch 10A in 10ns (or 1ns these days) e=L di/dt means there is a huge can of worms hidden inside tiny innocent looking traces even if only 2mm long. If you can do PDN for FPGA then... it's not that different... also the reason why it won't work on a breadboard.

Low voltage expertise exposes you to unforeseen risks and overconfidence: when the capacitors contain more energy than a carbine bullet, shit happens real fast. Of course you're building a defibrillator, which can kill you instantly. But besides electrocution, sometimes there is a bang and what used to be a multimeter probe tip ends up embedded deep into the safety glasses, a chunk of PCB turns into plasma, and bits of MOSFET shoot into the ceiling. The most important is to have the whole mess securely bolted on something heavy and sturdy enough so after you fall off your chair from the experience, the whole bunch of charged capacitors does not follow and fall on top of you. Likewise it should look like surgery, the whole area you're not operating on covered by an insulating sheet with a hole big enough to work on whatever you're going to work on without inadvertently touching anything else.

Now if you intend to build a small boost DC/DC to experiment on its control via FPGA, then I think it's an interesting project. Also hardcore difficulty.

However digital control has many pitfalls, one of them being protection: at low voltage it may take a really long time to blow a MOSFET, say 10µs, so an overcurrent cycle by cycle protection can be implemented in software or programmable logic. At high voltage things happen much faster so you may find the ADC too slow and need a high speed current sense comparator on top of that.

There are other details like: analog protection circuits still work when the programmable logic is frozen in the debugger, being reprogrammed, crashing, or otherwise.

I have a toroidal power transformer for tubes (2x 6.3 VAC secondaries) that I got on the cheap. It has no secondary center-tap, but taps for 260 VAC & 280 VAC. The problem is that you need I'm trying to get ~ 460+ VDC to my tube plates and since there's no center-tap, I was forced to go for the dreaded half-wave capacitive voltage doubler.

It's a toroid, so personally I'd add some more turns manually on the secondary to get the desired voltage and drop the doubler, mostly because of its dubious transient response which is not optimal for an amp.

 

[Pasqualino31]

OK that explains the previous posts

I confess being a bit puzzled about the combination of 1) your lack of experience with power electronics and 2) "I'll just do it with FPGA" which either means total mastery or brash foolishness but that is now cleared up!

Doing RF and highspeed digital should provide you with some of the required skills that can be transposed to power conversion, not all of them of course, but I'll bet you already have the absolute most essential one, which is to not fuck up the layout. If you switch 10A in 10ns (or 1ns these days) e=L di/dt means there is a huge can of worms hidden inside tiny innocent looking traces even if only 2mm long. If you can do PDN for FPGA then... it's not that different... also the reason why it won't work on a breadboard.

Low voltage expertise exposes you to unforeseen risks and overconfidence: when the capacitors contain more energy than a carbine bullet, shit happens real fast. Of course you're building a defibrillator, which can kill you instantly. But besides electrocution, sometimes there is a bang and what used to be a multimeter probe tip ends up embedded deep into the safety glasses, a chunk of PCB turns into plasma, and bits of MOSFET shoot into the ceiling. The most important is to have the whole mess securely bolted on something heavy and sturdy enough so after you fall off your chair from the experience, the whole bunch of charged capacitors does not follow and fall on top of you. Likewise it should look like surgery, the whole area you're not operating on covered by an insulating sheet with a hole big enough to work on whatever you're going to work on without inadvertently touching anything else.

Now if you intend to build a small boost DC/DC to experiment on its control via FPGA, then I think it's an interesting project. Also hardcore difficulty.

However digital control has many pitfalls, one of them being protection: at low voltage it may take a really long time to blow a MOSFET, say 10µs, so an overcurrent cycle by cycle protection can be implemented in software or programmable logic. At high voltage things happen much faster so you may find the ADC too slow and need a high speed current sense comparator on top of that.

There are other details like: analog protection circuits still work when the programmable logic is frozen in the debugger, being reprogrammed, crashing, or otherwise.


It's a toroid, so personally I'd add some more turns manually on the secondary to get the desired voltage and drop the doubler, mostly because of its dubious transient response which is not optimal for an amp.

It's already wound and sealed. If it were not pre-wound and I had to splice in more windings, I would have to be mindful of not saturating the core, which would more than likely to happen because that's the biggest delimiting factor with toroidal transformers. The core for a 120VAC primary with a 460VAC secondary would be huge.

Also, I'm not that green regarding power electronics and digital controllers using FPGAs are ubiquitous. They're used as controllers for the superconducting magnets in modern particle accelerators, like CERN and EIC at Brookhaven National Labs. Can you get much faster than accelerating a particle to 99.9999991% the speed of light. While the voltage drop across the main dipole superconductor magnets is extremely low, the current is in excess of 11KA!

https://accelconf.web.cern.ch/p05/papers/FPAT060.pdf

I agree that that digital controllers have their pitfalls, but I haven't even discussed any implementation schemes regarding an FPGA based digital controller, especially the A/D <=> D/A.

I designed the test system for the Quench Detection subsystem in Brookhaven's EIC collider using an FPGA. I know how to isolate the FPGA. My board alone had separate grounds for the FPGA and the interface circuitry. Opto isolators weren't enough. Grounding schemes alone for systems like that will give you vertigo. I think I can figure out how to isolate the FPGA from the driver MOSFETs in the Inverter bridge. As for speed and timing criticality? The quench system for a collider's superconductor magnets has to sequentially (and most elegantly) control the shut down of a chain of coils carrying 11KA+ in an 8.3T magnetic field. The Cern collider in 2017 used MAX1162 ADCs (16-Bit, +5V, 200ksps). They're about $8.13 apiece. I think I can swing that and it just might be able to do the job.

I take a little exception (not a whole lot) to the notion that I have made any outlandish or silly comments regarding anything I've posted here about power or otherwise. I'm no neophyte.