uSupply USB Rev B Update

[EEVblog]

For those following along, here is the Rev B USB µSupply.
Still preliminary, and not 100% checked & tidied up, but close.

Key points:
- Back to my original concept of dual 3 digit LED display (nice big easy to read 0.56") and dual (right angle) knobs to make it look'n'feel like a real power supply. No squinting at a small LCD, see the values across the room.
- One function push button for basic modes or set V/I display etc, same as original designs.
- 120mm x 50mm
- Battery powered from a standard 18650 cell
- Selectable battery or mains input with jumper. Allows 1A charging + full output operation.
- Max output 15.25V @ up to 500mA. Capable of higher current at lower voltages if really needed.
- Right angle riser board for the banana plugs, soldered onto the main board (no wiring).
- TO-220 LT3080 with bottom plate PCB used as heatsink.
- Construction is kinda like the attached, open style sandwich but with a PCB heatsink bottom plate. The top is clear acrylic.
- Micro USB charging (400mA) and isolated serial comms for, eh, whaever.
- Faster 1A charging using a 5V DC plugpack, which switches automagically
- Self test capability
- Small expansion board for, whatever. Maybe a fixed 5V USB output connector or something.

Schematic:
http://eevblog.com/files/uSupplyUSBrevB.pdf

Anything I've goofed please let me know.
I don't want to do any more videos on it until I actually have a the new board built and working, then I'll likely do a flood. That prevents any more videos that lead in the wrong direction.
No correspondence will be entered into on the design or feature set 

It's almost 1:30am, today's achievement ticked off, need sleep. 

Dave.

[Tooms]

So the adapted the Sick of Beige for the case ?

if yes then it will fit into the stanards sizes ?

http://dangerousprototypes.com/docs/Sick_of_Beige

http://dangerousprototypes.com/docs/Sick_of_Beige_standard_PCB_sizes_v1.0

 


Tooms

[fenclu]

Wow, top job. That's the best USB PSU of all your designs. As long as you keep it thin it'll be very portable because footprint ain't really that important in this case, I think.

[mikeselectricstuff]

I'd be inclined to put a high resistance (e.g. 5x100k 1206 in series) across the isolation barrier to drain any charge accumulated on the cap.
I don't see any reverse polarity or overvoltage protection on the ext DC input. Polyfuse+zener perhaps?
Selfttest probably not very useful.
R11 2K5 is an oddball non E24 value - I'd use 2K4, 2K7, or a combination if you really care. 
TXD/RXD polarity across the isolators appears to be wrong unless there's some non-obvious inversion going on. To minimise power draw you want the optos off when the bus is idle. So just have the LEDs pulled down instead of up, which should fix both issues.
I've not looked at the FT230X data but on all the other FTDI chips, status LEDs are pulled down, not up.
Is the DC-DC quiet enough without a cap across the output?
Is C1 big enough as an input decoupler to the boost reg?

 


 

[baoshi]

I'm looking at the hostile/dumb user case. If BATT/EXT power is set at EXT side but no external 5v is applied, the output will be powered by USB with 2W isolator. Shall some protection be included to limit the output?

[Mr Smiley]

Hi,

Sorry to nit-pick Dave, but middle right of your schematic, you have constant current load  I = 0.6/R14.

shouldn't it be R25.

Maybe some spare pads for possible hardware de-bouncing (R/C ) for you encoders, just in case you have problems, not populated as such, but there in case. They could become problematic with use.

 

[mikeselectricstuff]

Maybe some spare pads for possible hardware de-bouncing (R/C ) for you encoders, just in case you have problems, not populated as such, but there in case. They could become problematic with use.

 

I'd agree - the edges from some contacting encoders can be too fast to reliably keep track of and can lead to erratic behaviour -  some simple RC filtering can save a lot of grief. 

[Jad.z]

Dave is probably using some software debounce routine.

[fenclu]

Yep, looks like it's designed to tweak it in the software.

[EEVblog]

Dave is probably using some software debounce routine.

Same encoders as previous bench design. You can see it working in previous videos. No problems.

Dave.

[EEVblog]

So the adapted the Sick of Beige for the case ?
if yes then it will fit into the stanards sizes ?
http://dangerousprototypes.com/docs/Sick_of_Beige

I looked at those.
Would probably fit in the 100x80 version, but I didn't like that square form factor.

Dave.

[EEVblog]

Wow, top job. That's the best USB PSU of all your designs. As long as you keep it thin it'll be very portable because footprint ain't really that important in this case, I think.

I wanted it thin, but there is only so much you can do using an 18650 battery.
I contemplated cutting the cell into the board (as an idea I had for the bench version), but it's just not workable.
So I decided to just let it be what it turns out to be, rather than get all fancy.

Dave.

[EEVblog]

Selfttest probably not very useful.

It's for essential for production.

Dave.

[mikeselectricstuff]

Selfttest probably not very useful.

It's for essential for production.

Dave.

Surely part of the test/calibration process needs a calibrated  external load or current reference?

[EEVblog]

Surely part of the test/calibration process needs a calibrated  external load or current reference?

Haven't decided if I'll simply leave it up to the user to tweak the cal values if they want the best accuracy.
Either way, having the self test stuff in there allows an almost complete power on self test of the major features (reference, output voltage, output current, current limiting) without the need for a bed of nails or other dicky test jig connection. Combined with more precise batch testing that pretty much ensures units are quick and easy to test with high confidence.
All for the cost a transistor and a few resistors I already had in the BOM

Dave.

[JoannaK]

Nice schematic, with the helpful pictures.


AFAIK Those serial data isolation leds are indeed put non optimal (as Mike posted) since TTL-level (or cmos level) RS-232 idles at high voltage.

I noticed that PA7 is used as a high current pulsed led driver .. Perhaps not a problem as is, but I do hope it won't cause any loss of A/D accuracy since it's on the same 8-bit port as the Analog inputs? (wild speculation, yes, I'm good at it)

I'm not sure how Dave has decided to do the 7-segment muxing, so my assumptions may be wrong.  But anyhow, I'd add cheap fet (or bc846) buffers at the common (kathode?) side for safety. At the moment those Digit selector pins (Dig1-Dig6) need each to be able to sink all the current that high side delivers. Most likely the real current will be a lot less than calculated (5.8mA * 8 ), but >40mA is quite a lot for a digit selecting pin.

May well be a stupid question, but what are those CN3 and CN4 used for? And having  2*3 pin headers with 5 pins each?

[hans]

I really think the boost converter could use some extra capacitance. Although.. ceramic caps are good , I don't know if 1 can handle switching currents of 2-3App at 1.25MHz.
The value itself doesn't sound too odd, because the frequency is quite high.
Mike mentioned the input cap, but in a boost converter the input cap has to smooth out the ripple current through the inductor. The output capacitor is much more 'stressed'.
For a buck this is vice versa if I recall correctly.

What is the exact reason for 2.5V Logic by the way? Lower power consumption? Because it seems harder for people to start hacking the firmware because I am pretty certain a lot of 5$ AVR programmers work on fixed voltages. Maybe some series resistors in the input lines of the program header will protect/fix the problem.

Wouldn't it also be cheaper to replace the LM324 with for example a LM358, because you only use 2 opamps?

Other than that, it looks good.

[mikeselectricstuff]

I'm not sure how Dave has decided to do the 7-segment muxing, so my assumptions may be wrong.  But anyhow, I'd add cheap fet (or bc846) buffers at the common (kathode?) side for safety. At the moment those Digit selector pins (Dig1-Dig6) need each to be able to sink all the current that high side delivers. Most likely the real current will be a lot less than calculated (5.8mA * , but >40mA is quite a lot for a digit selecting pin.

At the very least it would probably be wise to keep the display mux synchrounous with  the PWM to avoid ripple artifacts - if the MCU is sourcing 40mA, I'd bet the voltage on the PWMs will shift noticeably. You might even see some variation in PWM voltage depending on the number of segments lit.

One possible issue running the MCU at 2.5V vs. 3.3v is the output FETs on the ports may not source/sink as hard as they do at higher voltages, sourcing is probably worse than sinking due to the P-channel output FET.
 

[Bored@Work]

since TTL-level (or cmos level) RS-232 idles at high voltage.

It idles at low voltage. A line driver like a MAX232 then inverts and the real RS232 signal then idles at space, i.e. the high voltage of RS232. But for conversion to USB one would not add a line driver, but stick with the logic levels going to the USB interface chip.

[EEVblog]

I really think the boost converter could use some extra capacitance. Although.. ceramic caps are good , I don't know if 1 can handle switching currents of 2-3App at 1.25MHz.
The value itself doesn't sound too odd, because the frequency is quite high.

The values are as recommended in the datasheet app for a max power output.
I have not gone through additional calculations.

What is the exact reason for 2.5V Logic by the way?

Direct power from the battery.

Wouldn't it also be cheaper to replace the LM324 with for example a LM358, because you only use 2 opamps?

Not really. Same price. The LM324 is more common. Arbitrary choice.

Dave.

[HLA-27b]

Dave, I hereby solemnly take the credit for badgering you to continue with this. 

Now then, can we have the mounting hole positions / diameters and top and bottom clearances please. I'd like to see how it fits into that other project being designed right now.

[hlavac]

I like the 2.5V rail idea, with regard to losses on the LED current limiting resistors. Cuts it to somewhat acceptable levels.

Too bad you use a linear regulator to get the 2.5V, negating all the savings

I was experimenting with minimizing the LED display current limiting losses on a 5V rail, and had to resort to kind of high efficiency DCDC converter driving the segments. Works, but is complex to drive and adds components to the BOM.
I basically connected the LED segments in place of inductor free-wheeling diode to maximize efficiency

[robint91]

Why have you chosen for two optocouplers and not of those digital isolators? Which I find better for isolating datacommunications. Optocouplers are sometimes too slow if the pull-up circuitry isn't properly chosen.

https://www.silabs.com/Support%20Documents/TechnicalDocs/Si8410.pdf
http://www.digikey.com/product-detail/en/SI8421AB-D-IS/336-1755-5-ND/2170662

[EEVblog]

Why have you chosen for two optocouplers and not of those digital isolators? Which I find better for isolating datacommunications. Optocouplers are sometimes too slow if the pull-up circuitry isn't properly chosen.

Cheaper, and more jelly bean.

Dave.

[mikeselectricstuff]

since TTL-level (or cmos level) RS-232 idles at high voltage.

It idles at low voltage. A line driver like a MAX232 then inverts and the real RS232 signal then idles at space, i.e. the high voltage of RS232. But for conversion to USB one would not add a line driver, but stick with the logic levels going to the USB interface chip.

Wrong. TTL UART signals idle high. RS232 idles low.

[xani]

A possibility with having "too big" micro would be adding a bunch of extra features, like "battery charger" mode, ramp-up/ramp-down voltage  ( to test "brown-out" detection) or adding artificial noise/50Hz hum to output for testing audio circuits. But that's easy to for modders . Kinda sucks to have LED (no decent menus, hard to display other units like watts etc). but I guess one can always desolder it and connect some LCD

It would be nice to have some connectors/jumpers for isolated side of RS232, it would allow for connecting cheapo bluetooth<>RS232 converter and use it from tablet or phone. (or laptop few meters away).

[Bored@Work]

Wrong. TTL UART signals idle high. RS232 idles low.

I think I should take up plumbing instead of EE. Would keep me out of embarrassment and would help fixing the broken heat pipe we have since yesterday. Win-win situation.

[mariush]

I'm curious how - and if - the Vr voltage will affect the opamps.... the LM324 seems to need 3v at least and if i remember correctly it's not a rail-to-rail opamp, and you're feeding it from the tracking pre-regulator... it's not clear from the schematics what's the lowest voltage of that regulator. I assume it has to be larger than the lt3080, since that's a ldo and assuming this psu is a 0-15v therefore the tracking pre-reg needs to be as close as possible to keep efficiency high...

Would the accuracy of the opamp change if you give it 5v or 10v, is it an accuracy/stability issue for the current/voltage limiting?

[hlavac]

Couldn't you use something like MCP1603T-250 or REG711EA-2.5/250 instead of the lossy linear MCP1700-2502E for the 2.5V rail?

[EEVblog]

Couldn't you use something like MCP1603T-250 or REG711EA-2.5/250 instead of the lossy linear MCP1700-2502E for the 2.5V rail?

a) You wouldn't be saving much if anything in terms of efficiency at say a nominal 3.7V input.
b) Extra cost, complexity, and noise.
c) The linear reg has a low temp co. I'm using it as a quasi ADC reference.

Dave.

[hlavac]

I see, power efficiency is a bit of a lost cause there when powered from USB.
First the isolating DCDC converter steals 20% of the power, then the linear battery charger, then the linear 2.5V reg, at this point whatever is powered from the 2.5V has already power efficiency of 40% even before the actual losses are considered...

[EEVblog]

I see, power efficiency is a bit of a lost cause there when powered from USB.
First the isolating DCDC converter steals 20% of the power, then the linear battery charger, then the linear 2.5V reg, at this point whatever is powered from the 2.5V has already power efficiency of 40% even before the actual losses are considered...

Nothing to do with the isolating DC-DC converter as such, that just for charging. It's simply the low input battery voltage.
Assume 3.7V batt with a linear reg @ 50mA. That's 0.185W input power.
A 2.5V DC-DC converter with an ideal 100% efficiency at the same 50mA output current is an ideal 0.125W input power.
So any DC-DC converter over 68% efficiency will gain you an advantage over the linear reg.
For say an 80% efficiency, it's now about 0.16W input power, saving you not a huge amount in the scheme of things.
And when you compare it to the fact the PSU will usually be supplying a load, any saving becomes pretty insignificant.

Dave.

[hlavac]

You are right, if the USB is used just for charging, and not as in the original uSupply concept of completely USB powered lab supply.
I guess the merge between the battery powered supply and the old uSupply left us closer to the battery powered one...

Speaking of USB charging, the FT230XS-R has functionality for detecting 5V/1A chargers (which have D+ and D- shorted together),
maybe you want to use it to enable fast charging?
Is it worth an extra optocoupler?

[fenclu]

Still not sure if I get the idea...it's a battery powered, portable PSU and the USB is just for charging? Why not get rid of the DC-DC converter and just note on the silkscreen that the PSU cannot be used while charging? That would be a significant cost advantage...

[robrenz]

Still not sure if I get the idea...it's a battery powered, portable PSU and the USB is just for charging? Why not get rid of the DC-DC converter and just note on the silkscreen that the PSU cannot be used while charging? That would be a significant cost advantage...

And a significant loss of usefulness.

[klinkerstein]

Dave, you need an output clamp or similar on the 2W SIP7 DC/DC Converter or use a regulated one. Without any load, you easily get 20V out of a 5V/5V sucker.

Secondly, I think the Couplers are very battery inefficient. have you considered using an ADUM12xx from Analog Devices for this purpose? They are pretty easy to handle and usually easy to get.
http://www.analog.com/en/interface-isolation/digital-isolators/adum1200/products/product.html
But it doesnt seem as they would support 2.5V.. Hm

Other than that, it really looks pretty good. See, that you choose the right ATmega, which supports 2.5V when buying. And choose a ceramic resonator, that is accurate (drift, tolerance and tempco) enough for RS232.

[mikeselectricstuff]

Secondly, I think the Couplers are very battery inefficient. have you considered using an ADUM12xx from Analog Devices for this purpose? They are pretty easy to handle and usually easy to get.

Once they get reconnected the right way round, they will only draw power when communicating, which will be rarely if ever, so current draw is not an issue.

[klinkerstein]

Secondly, I think the Couplers are very battery inefficient. have you considered using an ADUM12xx from Analog Devices for this purpose? They are pretty easy to handle and usually easy to get.

Once they get reconnected the right way round, they will only draw power when communicating, which will be rarely if ever, so current draw is not an issue.

Yeah, that is absolutely correct.

EDIT: Oh, about the 120R resistor at AVCC on the microcontroller: Do you know, that the entire Port A is supplied through AVCC and not only the internal ADC circuitry?

[EEVblog]

Speaking of USB charging, the FT230XS-R has functionality for detecting 5V/1A chargers (which have D+ and D- shorted together),
maybe you want to use it to enable fast charging?
Is it worth an extra optocoupler?

The DC-DC converter is only 2W.

Dave.

[KeithBrown]

I have a real problem with the "insert tab A into slot B and solder" connection method. My problem is based on the experience of several of these becoming open circuits with no visual cues. I'd feel a lot better if there were some plated through holes on both the main board and the riser board so that solid jumpers could be soldered in as well. This is especially a risk if the banana jack board is not to be fastened VERY firmly mechanically to the main board.

[mikeselectricstuff]

I have a real problem with the "insert tab A into slot B and solder" connection method. My problem is based on the experience of several of these becoming open circuits with no visual cues.

Were the slots plated and the tabs double-sided ?

[KeithBrown]

Were the slots plated and the tabs double-sided ?

The tabs were double sided but different signals on both sides. Not sure if the slots had plating through, however I guess they could not have because of the last sentence.
I understand that  Dave's design is a much better situation, however I still cringe at using solder as a mechanical connection and expecting it to never crack. I suppose the clearances would be critical: allowing just the right amount of solder to flow into the slot beside the tabs. Too much space (and solder) and the mechanical connection would be soft, too little and there could be no solder there at all!
Anybody know if there are guidelines (IPC or somebody else) in regard to this connection method? I was taught a looong time ago that the first rule of soldering is a good mechanical connection, and this violates that teaching, again especially if it is also the mechanical connection. I understand that all SMD soldering also violates this--unless the components are glued.

[EEVblog]

I was taught a looong time ago that the first rule of soldering is a good mechanical connection, and this violates that teaching, again especially if it is also the mechanical connection. I understand that all SMD soldering also violates this--unless the components are glued.

Correct. By definition, a solder joint is a mechanical method of holding components together. That's one of it's jobs!
It's used every day on big bulky components, like PCB mount transformers for but one of countless examples.
If it makes you feel any better, there are also tabs in the top side of the vertical board that will also slot into the top case cover. The paranoid could also glue those to create a very rigid box type structure.

Dave.

[baoshi]

Speaking of USB charging, the FT230XS-R has functionality for detecting 5V/1A chargers (which have D+ and D- shorted together),
maybe you want to use it to enable fast charging?
Is it worth an extra optocoupler?

The DC-DC converter is only 2W.

Dave.

On the contrary, I think the "USB enumerated" signal can be used to stop charging if the PSU is connected to a bus powered hub

[KeithBrown]

Correct. By definition, a solder joint is a mechanical method of holding components together. That's one of it's jobs!

But I would not attach a trailer to my car with solder: sure cold solder is a mechanical connection, but not a very good one. We have all come across very tight banana plug/socket combinations where a lot of insertion or extraction force is required. And, if the sockets are exposed, as I expect they must be, a drop on the floor will possibly impart a large impact force on the solder joint...of course you would never drop any piece of electronics on the floor on purpose, would you!

[konfu]

Hi Dave,

it might be a nice feature to allow the LT3080 to go down right to zero volts output. To do this there is the need for a negative feed into the constant current load ground connection as shown in one of my last posts How to bring down the LT3080 to 0V output. Easy to do and just a nice feature and not too expensive (about 0.88 EUR plus 2 elkos).

I'd like to understand what value and type you are using for F1 - the ferrite bead between the pre-regulator and the LT3080.

Are you using precise values for R38, R39, R40 and R41 (Self Test Dummy Load) so the load is somewhat stable on it's current draw?

Nice work though!

Cheers

Florian

[Hobgoblin612]

EDIT: Oh, about the 120R resistor at AVCC on the microcontroller: Do you know, that the entire Port A is supplied through AVCC and not only the internal ADC circuitry?

In any case all of port A is used as analog inputs anyway. (except maybe Spare2)

[amspire]

Hi Dave,

it might be a nice feature to allow the LT3080 to go down right to zero volts output. To do this there is the need for a negative feed into the constant current load ground connection as shown in one of my last posts How to bring down the LT3080 to 0V output. Easy to do and just a nice feature and not too expensive (about 0.88 EUR plus 2 elkos).

If you are going to add a negative supply rail, there are better or cheaper ways to do it then using the temperature sensitive LM334. Like a 2N2222A transistor with collector connected to the output, base to 0V and emitter connected to a resistor to the negative supply. The LM334 has a 0.3%/C temp coefficient whereas the transistor/resistor pair have a 0.05%/C temp coefficient if you have a stable -5V rail. Also the transistor will not let the output voltage fall below -0.7V. With an extra diode and resistor to bias the transistor base at 0.6V, you can reduce this minimum to less then -0.1V. This is important as you cannot let most opamp inputs (eg the voltage regulator) get below -0.3V if they are using 0V as the opamp negative supply.

[konfu]

Hi amspire

I really like that hint. Gonna test it out during the next days and smash a rough circuit together for it. Thanks a lot!

[rec]

Greetings --

I stumbled onto this project yesterday.   I was looking for a way to make a USB recharged battery supply for a radio.  Google led me to the Turnigy charger tear down episode, which led to the uSupply project, or at least parts of it.

(It would be really nice if http://www.eevblog.com/projects/usupply/ were updated so it actually linked to all the videos about this project and the forum topics which continue the project.)

But I'm dubious whether the uSupply can actually be used with a radio as it's designed.  I expect it's going to have ample noise from the switching frequency, harmonics, and mixing products.  But it will be interesting to try.