Isolated Power Supply

[Kingfish]

Greetings –

I am interested in building (or at least understand how to build) a clean (enough) PS on a 1:1 scale using common household (North America/CAN-USA) supply of 120-127 VAC. Purchasing an off-the-shelf solution is an option.

The goal is to simulate voltage/current supplied by a direct drive motor in regen. That’s it.

I will state up front that I’m not an EE, although I understand basic concepts at a high level view. That given from what I have read the isolation aspect is about winding a transformer that is shielded. I get that part, although I have no idea how heavy this transformer needs to be other than dependent upon current output. I am interested in studying 10A, and up to 50A service.

There is a second part to this question which begs to ask – is the transformer the best possible solution for isolation or are there more lightweight alternatives?

Thanks in advance, KF

 

[liveforphysics]

The core material choice needs to work with the frequency. It's not the voltage that matters when designing it (at least until you get to many hundreds of volts, than insulation starts to matter), it's all about the frequency. If you are just isolating 110vac at 60hz from a wall outlet, and for some reason do not wish to purchase an off-the-shelf isolation transformer (and I still don't understand it's purpose/function in your test to answer that), and you need something in the 1-2kW range, I would buy a microwave oven transformer from a scrapper (they are like $5-10, sold by weight generally), leave the primary windings, cut-off the secondary and re-wind it to match the primary windings, then you have a 1:1 isolation transformer that costs you a bunch of labor rather than cash.

 

[Kingfish]

Thanks Luke, you’ve provided enough crumbs for me to refine the request.

It appears what I am looking for is called a transformerless isolated circuit. The examples I have found online are for small current and do a poor job at isolation (or none at all).

Example: New topologies for Transformerless Single-Stage AC/DC Converter

Flipping over to an off-the-shelf solution, I dug around Digikey and could not source a transformer close to a 1:1 winding at say 1.5 kVa; instead there are several commercial products marketed under the pseudonym “UPS” :wink: : I found one on Amazon Tripp Lite IS1000 Isolation Transformer 1000W Surge 120V 4 Outlet 6ft Cord TAA GSA weighing in at a whopping 24.6 pounds for only $267.41. I gotta tell you – 1.8 Stone provides a lot of motivation to find a lighter weight solution. :lol:

The ultimate goal is something to replace my present charging system. For example if I went with a Meanwell RSP-1000 series requiring 3 units in series at 4.3 pounds each, that’s about half the weight of the Tripp Lite above. However it does not provide much flexibility for expansion.

In addition, we already have a sophisticated controller onboard that can accept reverse current (e.g. Regen). The part that prevents us from tasking it as a charger is “isolation” from the Mains.

What I am proposing is a way to test this concept of using the controller for charging. The first step is to source an isolated supply, and the second is to determine if it’s possible to accomplish this in a transformerless design.

I’ve asked this question before long time ago, just in a different way – although the goal is the same: Reuse existing hardware on-bike and save weight.

Maybe I should change the title of the post to better reflect upon the aim…

Ironless, KF

 

[liveforphysics]

The ultimate goal is something to replace my present charging system.

If you want isolation and low weight, the path is to increase switching frequency. There is some pretty wild isolated stuff out there now. 1kW per in^3 is actually in a handful of real available products now, and isolated. 3kW per in^3 has been claimed from lab prototype designs.



http://www.vicorpower.com/promotions/AC_to_PoL/ChiP_Technology/DCM/lp.php

 

[Ohbse]

Sounds like the need you're describing could be fulfilled by an off the shelf telecommunications front end rectifier. Cc/cv functionality, 42-58v adjustable output and up to 2kw output in a 10.5"x4.7"x1.6" footprint weighing ~3lb. Obviously depends on battery voltage, I know there has been some success modding these units for wider voltage range. 330,000 hr mtbf, vibration rated and can be dirt cheap second hand..

The regen charging you're describing already exists in the Adaptto controllers and is fantastic. I'm sure it's possible to achieve with other controllers but given all the other benefits of the Adaptto offering I couldnt justify the effort. I plug in one xt90, walk away for 15 minutes and come back to a balanced pack at whatever voltage I configured.

 

[Kingfish]

Awesome stuff! Really thought-provoking and presents a number of options 8)

Price checking the DCM290P138T600A40 unit, Digikey has it for $217.02 USD. Feeling the Midas pinch; then again, what would you be willing to pay for an equivalent Meanwell?

I imagine the trick would be to rectify the input AC and bump it to 200 VDC to enable the soft-start option, and would need to bump the output (or string units together in series) to get the final DC voltage.

Thanks for the tip Luke!
I'm off to play with Spice, KF

 

[liveforphysics]

What output voltage and current are you seeking my friend?

 

[Kingfish]

What output voltage and current are you seeking my friend?

30S -> 126 VDC; this is for the next ebike, er... eMoto build. For cross-country, figured I'd have to bring my own charger along: The Meanwell solution will be too bulky & heavy. Need a lightweight solution that can be integrated into the framework.

For practicality, the Amp service needs to be 15 to 50 to recharge overnight; less for commuting though more for cross-country. That DCM unit from Vicor can be paralleled up to 8; a nice option when stacking output.

This really gets me thinking out of the box - and I am grateful 8)

Inspired, KF

 

[liveforphysics]

What output voltage and current are you seeking my friend?

30S -> 126 VDC; this is for the next ebike, er... eMoto build. For cross-country, figured I'd have to bring my own charger along: The Meanwell solution will be too bulky & heavy. Need a lightweight solution that can be integrated into the framework.

For practicality, the Amp service needs to be 15 to 50 to recharge overnight; less for commuting though more for cross-country. That DCM unit from Vicor can be paralleled up to 8; a nice option when stacking output.

This really gets me thinking out of the box - and I am grateful 8)

Inspired, KF

If you can remember to be extremely responsible about charging, and/or setup a good HVC/LVC system, I think it's about impossible to beat the raw power density possible with the ghetto dirty charger diode+cap systems.

With just 1lbs of diode and capacitor, you could pump that 126v pack with 10's of amps easily if you found a breaker that wouldn't trip, and didn't mind an extremely dirty line power factor. Obviously not something you would want for a daily commuting charger, but as a touring opportunity charger it's a ton of charge power potential for like $20 in parts and 1lbs of mass.

 

[teslanv]

If you can remember to be extremely responsible about charging, and/or setup a good HVC/LVC system, I think it's about impossible to beat the raw power density possible with the ghetto dirty charger diode+cap systems.

What is this magical charger you speak of? Are there any threads on how to build one of these? or Search parameters?

 

[Kingfish]

Tes, here's an example, especially Figure-4 w/ the Full Bridge Resistive.

Transformerless AC to DC power Supply

Risky business without isolation.
Not for noobs. KF

 

[teslanv]

Ah. OK.

Guess I'll keep fiddling with my Meanwell, then. Bulk charging LiPo's with a MW is already risky enough for me.

I enjoy learning about electrical engineering, though.

I found this site, which is quite informative on general electrical elements.

http://www.facstaff.bucknell.edu/mastascu/eLessonsHTML/EEIndex.html

 

[Kingfish]

Greetings –

I’ve been working on the concept of utilizing the DCM290P138T600A40 device. Essentially there are 3 primary blocks to the design:

1. 120 VAC Rectification & voltage boost to about 315 DC @ 2A
2. DCM290P138T600A40 et al, consuming the 315 VDC_In, and creating an isolated 15.5 DC_Out @ 35A.
3. DC Boost from 15.5 DC to 126 DC... and with CC Regulation

I’m only focused on Item-1 right now just trying to understand how I can make this fly, and I have a few questions. Utilizing a Villard cascade voltage multiplier appeared to be the easiest solution. Example circuit given below.

View attachment SimpleAC-DC.Bridge.VoltageDoubler.zip

• Diodes are rated at 600V @ 4A; P/N MUR460, Mfr: GI
• Capacitors are more tricky. Using LTSpice I was able to simulate the 315 VDC output using a 2A load, however the Caps needed to be 1500uf x 4, plus another filter cap at the same rating, and even then I still had a 2.5A ripple. The part that I don’t know how to spec is the voltage rating and ripple current for the caps.
  • Is the voltage rating relative or absolute?
  • Does the ripple current calculation include the entire current spike? I will tell you now that C1 in the schematic above has a spike of -40 to +70A according to LTSpice.
• The last item that I’m dealing with is the weight and cost: These caps are not tiny, although I presume they will weigh less than a transformer. Still – this particular design is quite bulky if you figure each cap being about 2-inches in diameter and 3-inches long and costing $6 to $30 each.

Regardless, I appreciate the simplicity of the circuit, however dubious as to being useful for our particular needs.

Looking to be rectified by Four Pints tonight - the capacitance though is well understood :wink:
Cheers! KF

 

[Kingfish]

Well foo! :?

I put more time into this and discovered a little formula that was pretty dang on par with my LTSpice design. Quite useful really in helping me predict the output voltage + 20% for tolerance.

Villard cascade		Formula to calculate voltage drop	
						   ΔU = I/(fC)*(2/3*n^3 + 1/2*n^2-1/6*n)
	
I	Output Current (A)    2
f	Frequency (Hz)        60
C	Capacity of caps (F)  0.00224
n	# of stages		     2
ΔU  Voltage Drop		    104.167
			
Vi    Voltage Initial		120
Vf    Voltage Final		  480
Vdc	End DC Voltage		 375.833
Tf	 Tolerance factor	  0.2
      Safe Voltage		   378

Going to Digikey, I sourced Aluminum candidates. The prices threw my head back, but didn’t quite blow it off; I’m still here! Allow me to digress…

I modified the Circuit to allow for programmatic assignment of the capacitors in one single edit as opposed to 5 which speeded up the development.

According to the Jochen's High Voltage Page where I found the Villard cascade math formula, he writes that “All diodes and caps must be dimensioned for 2x Up” – and I’m pretty close:

• Diodes are 600V 4A
• Caps are 500V and roughly 2300uF

The cost of the caps would have been about $57 each. With a little digging I was able to get that down $10 by splitting up the caps as 4-in-parallel. However the weight of each is 136 gm ; 4 x 136 = 544 gm or 1.2 lbs per cap assembly (the single cap option did not give the weight on the datasheet). There are 5 caps total so we are looking at 1.2 * 5 = 6 lbs. @ $235 just in caps. Just as a reminder, we’re only putting out 620 Watts.

I don’t think this circuit is going to work at all for me because we’re already approaching the weight of a Meanwell RSP series unit which can deliver nearly 3X the wattage.

Reviewing my Meanwell options. DocBass had an interesting approach on this thread. The only problem I have is figuring out is the amp rating on the Lucent DC-DC converters; I can't see how thee is enough current allowed to pass through. How would I calculate that

Unplugged, KF

 

[liveforphysics]

If you're still working around the 60hz limitation, you will never have good power density or efficiency my friend. It's like the equivalent of trying to get something to be high power while working with a 1v bus voltage. Yes, it has been done before and obviously is possible, but at some point it must start to look like an exercise in absurdity.

All the high power density stuff takes line voltage input and turns it straight into DC to get switched at hundreds of kHz minimum, often MHz for the modern stuff. Then if you only want to carry a half-pound of mass in your transformer core, you can get 10's of kW through it rather than something abysmal at 60hz like 200w or whatever.

 

[parabellum]

How about those guys? Unregulated power supply for halogen lamps. This one is 250W, weight 200g. Output isolated, it is just wire wound on toroidal core. You can wind as many turns you need. Add 2 capacitors and some fast bridge rectifier and .....

Edit: It was about 9USD

 

[Kingfish]

If you're still working around the 60hz limitation, you will never have good power density or efficiency my friend. It's like the equivalent of trying to get something to be high power while working with a 1v bus voltage. Yes, it has been done before and obviously is possible, but at some point it must start to look like an exercise in absurdity.

I felt much better after reading this; put me in the right state of mind - thank you 8)

parabellum, the problem is I am looking at developing a setup that can provide up to 50A for a 30S battery.

Been doing some deep thinking and a lot of reading. The Meanwell RSP-2000 series is looking better and better; break open the wallet, purchase 3 units, save a lot of time. :wink: I have a full-plate of distractions already. The past couple of days I have considered giving up on 120 VAC @ 15/20A and just cut to the chase: Tap into the Dryer outlet and install a Type-2 EV Recharging tap. The NEMA connector is a 10-30 (Hot-Hot-Neutral). The way I figure it: The charger can be left at home; only need it when I go cross-country, and I don't need to worry about that right now. All the math I'm doing shows that the new ebike will recharge plenty quick with that setup.

Placing PSUs in series, the configuration will be RSP-2000-48 + RSP-2000-48 + RSP-2000-24.
I have the following questions:

1. When connecting these, I read on the RC Forum that they suggest breaking the EARTH GND from all but one unit and placing an inline Cap to eliminate the GND Loop as a safety precaution. I don't recall seeing that discussed previously. Is this a real concern?
2. I have read in several places that when placing PSUs in series that it is strongly suggested to place a reverse-bias diode in parallel to the outputs; on the Meanwell FAQ:
   

   ...basically you can do this to get the right output voltage, but be careful that the rated output current of the series system should be the rating of the minimum one in these series connected power supplies. Furthermore, we like you to parallel a diode at the output of power supply to prevent possible damage of internal capacitors.

   My question is: How do I spec the Diode?
3. The last question, and I'm just double-checking, the output of the RSP-2000-48 is 42A whereas the output of the RSP-2000-24 is 80A: Does it makes sense to keep with the same model or should I downgrade to the RSP-1000-24 which as an output of 40A?

Thanks in advance, KF

 

[Kingfish]

Answering #2: From reading, a source suggests to rate the blocking diode at 1.25 to 2.5 times the voltage & amperage of the device output. Using Digikey as a source I came up with this unit:

IR 100BGQ100, a Schottky 100V 100A, with a Vf of 1.04V @ 100A. Price: $4.42 USD each. Acceptable

Addendum:
RSP-1000 series
RSP-2000 series

~KF

 

[Kingfish]

Found a link to a generalized Meanwell User Manual (6.7 mb pdf):

Page-70, in Section 7.9 "Operation in Series", Part (2) states the following...

Increase the output voltage (current does not change). If there is no reverse blocking diode in the power supply, we should add an external blocking diode to prevent the damage of power supply while starting up. The voltage rating of the external diode should be larger than V1+V2 (as figure 7.11).

The closest device I found meeting that criteria is the Microsemi APT100S20B Schottky 200V 120A for $7.27 USD, evidently created expressly to be a parallel diode. Good enough

Addressing Item-1, I found nothing in the Meanwell doc which discusses unique grounding issues - other than to say

Using short and thick wiring connects the frame ground (FG) of the S.P.S. and earth ground on the case of the equipment to assure safety and prevent noise and leakage current.

I tested my HRP units that I have in series and the case is grounded to the AC GND (makes sense of course) although neither Vout + or - is grounded. I have no plans to ground them

As far as Item-3 goes, the rational for going to RSP-1000-24 is due to cost. I don't know what else to think; I'll have to wait until I recharge again to see if I can determine if the HRP-600-24 is running full-tilt with the HRP-600-48. :?

Lastly, something that came up in the reading (also mentioned in the Meanwell User Manual, Page 73/Section 7.16) is surge protection. The RSP-2000 series will have an initial surge of 50A when fired up, although the operational current is closer to 16A. Looking into surge protection... I found a device called a Delay Relay that could handle the load, but then I thought another way to handle that is with a soft-start remote ON/OFF circuit; the Meanwell already provides low current +5 & 12VDC local source, so maybe a 1 second delay circuit between each unit firing up would be enough to prevent a local brown-out

Getting there, KF

 

[liveforphysics]

Now that I know you are just looking for a 50A 30S charger, have you looked at the CC/CV conversion kits for the 3000w HP power supplies.

 

[Kingfish]

Now that I know you are just looking for a 50A 30S charger, have you looked at the CC/CV conversion kits for the 3000w HP power supplies.

No Luke, I haven't heard of this before. Are you talking about using the RCP-3K1U-48 units?

Intrigued, KF

 

[liveforphysics]

http://www.lifepohandel.de/epages/63129206.sf/de_DE/?ObjectPath=/Shops/63129206/Products/%22Steuerplatine+f.+HPNT+V2.2%22


These are unnecessarily tricky to purchase if you're not in Germany. I had to have a German friend order my kit and ship it to me after they delivered it to him.

If you want to do something super nice for the ES community that I wanted to do myself but never found time, I can get you the kit (it's just a small PCB with a handful of cheap parts) and you can improve it and do a run of boards with your improved design (maybe give it current and voltage digital readout? Maybe use one of the Ebay Ah/power/energy/Amps/volts digital meters they sell for like $20 (it will use the existing shunt in the power supply for current measurements, they are programable) and mount the adjustment pots for CC/CV in a place easy to adjust them.

That would let folks make pretty darn nice powerful cheap compact chargers. They are all isolated the moment you simply don't connect the chassis ground pin, so you can series them. The cool thing is, you only really need to have CC capabilities on one, the others in the series stack only need to be CV (server supplies meanwells or whatever someone may have).

If you use the right Ah/power/amp/volt meter thing, you can have it sense the whole series stack voltage so it gives you a nice graphical readout while it's charging.

Making a kit like that to help people get well instrumented adjustable >5kW charging for $100-150 in parts would be an extremely kind gift to contribute to ES that I wanted to give myself when I got time, but I think you're actively seeking this project yourself and I know you could do just as good of job my friend.

Looks like the cheapo ebay meters have relay control now too! So you could actually have some very useful safety control HVC/LVC/short circuit etc gained as well as good instrumentation.

http://m.ebay.com/itm/181430544783?cmd=VIDESC

 

[parabellum]

Looks like the cheapo ebay meters have relay control now too! So you could actually have some very useful safety control HVC/LVC/short circuit etc gained as well as good instrumentation.

http://m.ebay.com/itm/181430544783?cmd=VIDESC

Thanks for the link. It is the coolest thing ever
Universal CC control is only thing missing now.

 

[Kingfish]

OK, I see it now; thanks!

Here’s my deal…
I’m working on the replacement drive for my ebike, and actually it turned into a flat out new motor design. I have a model that I want to build; barebones for the test stand proof-of-concept. The target voltage is 126/30S. I have enough spare batteries to construct a 30S1P for the test run, however they will be depleted pretty fast if run at full-load. What I need a way to recharge quickly, or perhaps run a 126 VDC supply that is buffered by the 30S1P setup. -> Queue the Isolated Power Supply study.

Short story: I’m just doing this for me. I know that sounds selfish, but if I were a business man (and I yam) the first thing I’d do is validate the concept, figure out how to adapt it using off-the-shelf parts, review any custom work, and later – if time allows – evaluate the market. For the moment though, I am just serving a need and solving a problem by looking at all aspects of the system design. It is my feeling that solving how to recharge is as good a place to start as any other. :wink:

The sticky part of the problem is that the 126V 42A (~5.3kW) charger assembly is still going to weigh in at 5.85kg/12.9 lb.s (a concern if built-in on bike)… actually that number isn’t too terribly bad relative to battery weight. Stating the obvious: It’s not an integrated design; begs investigation into a single monolithic approach that accepts 2-Ph with proper high-current connections and safety features. Finally we need to evaluate how to integrate the J1772 EV connector at AC Level-2 (240V 32A); the assembly using the Meanwell RSP-2000 series is gated by my 240V 30A service to the Dryer… small challenge there, and I may need to de-rate the output to spoon it in.

Coming back to the question of CC/CV…
In my experience with the HRP-600 series, the assembly charges just fine without a CC Limiting circuit, however I will acknowledge that once the batteries are topped off, upward voltage creep continues past the POT settings… enough that I wouldn’t want to leave them on the charger for more than a couple of hours. (It wasn’t a problem for me on the Road Trip in 2011 where I was overtly attentive, though at home – sometimes I can’t hear when the charging is finished).

So I ask… what is the primary function that we need from an add-on:
Do we want absolute control on charging, shut down all current flow once we’ve achieved our terminal voltage? Or do we want to artificially limit the current (by POT, by hardware switch, or via software), and voltage too? I can see that being a valuable need for robust charging systems.

In reference to the Control Board kit from Germany, the specifications as follows:

• Voltage adjustable from about 28 to about 60V
• Current adjustable from 0 to 57A (max. 2900W)

To make this work for the Meanwell RSP series we’d have to make some adjustments. BTW - Luke, your ideas for features are insightful and exciting! I’m encouraged. It opens the door to possibilities in software monitoring. That Programmable Over Current Protection display on eBay is the bees-knees; definitely less expensive than what I found when looking for a Delay Relay. 8)

Would it be possible to adapt Fechter's Mini Meanwell Limiter design to fit these needs? There is a reference in his thread to using the Meanwell RSP-3000-24. Seems like this has been discussed before. What limits would appeal to you?

Glad for your input, KF

 

[liveforphysics]

Why mean wells and not $30 3KW ultra reliable supplies that had 10's of millions of development for the server industry.