EEVblog #259 - PSU Rev C Schematic - Part 12

[caroper]

I like it, living in Africa and, as I do, in a village that suffers fairly frequent power outages, a battery operated variable supply would be worth its weight in gold. And that would be from Murphy's Pot of Gold too as Murphy generally arranges the Power outages right at a critical phase of the project.


I think the improved Voltage range at the cost of precision is a viable trade off.
Seeing as most of the measurements in modern electronics that require low noise and high precision would probably happen in design's that are running between 2V0 and 3V8 anyway, so the switching regulator is off.


Two things I would like to see, to make it really shine.


1) Ability to supply burst currents of up to 2 Amps, common when working with Cellular modules and other Radio devices.


2) A Analogue or Digital link between two boards to allow for Serial/Parallel/Independent operation of 2 supply's. Many modern designed need dual supply, i.e. 3V3 for the MCU but 3V8 for Cellular/GPS or possibly separate supplies for the Digital and Analogue portions. Serial tracking would also be a great benefit for those who still wish to experiment with split rail op amps etc.


All in all an excellent device, versatile and expandable.


You beauty 


Cheers
Chris

[BravoV]

Great series, learned a lot of in design change and modification in the building a circuit, thanks a lot Dave.

Noob questions :

- I saw there are 2 x LM358 op-amps there, will this save even for few cents by replacing it with LM324 ? Since they're almost identical by spec. Just checked Digikey, the cost of single LM358 is almost identical with LM324 which has 4 op-amps instead of 2.
- Correct me if I'm misinterpreting Dave's explanation, is that pre-regulator control loop is now handled by the mcu (cpu cycles) instead of linearly controlled like the front end linear section ?

[FreeThinker]

Hi Dave
You asked for input on the latest design so here goes.
First I love it, the increased range is a winner for me.
During vid you mention mounting the pcb on the top and using a pcb to mount the lcd. Please don't do what I did and forget that this would make the display upside down  . I can laugh about it now....Just

[markus_b]

What is the link to the pdf file of the schematic ?

[firewalker]

If you haven't already thought it, place a female header for the Tx,Rx pins on the board, not only to the rear of the enclosure.

Alexander.

[EEVblog]

1) Ability to supply burst currents of up to 2 Amps, common when working with Cellular modules and other Radio devices.

You can never satisfy everyone's power requirements. Give them 2A and they want 3A...
This is deliberately a low power PSU.

2) A Analogue or Digital link between two boards to allow for Serial/Parallel/Independent operation of 2 supply's. Many modern designed need dual supply, i.e. 3V3 for the MCU but 3V8 for Cellular/GPS or possibly separate supplies for the Digital and Analogue portions. Serial tracking would also be a great benefit for those who still wish to experiment with split rail op amps etc.

I don't see much point in tracking.
If you want dual supplies, just use two supplies.

Dave.

[caroper]

You can never satisfy everyone's power requirements. Give them 2A and they want 3A...
This is deliberately a low power PSU.

This is true, and if anything I worded it incorrectly, it is more a matter off can the device tolerate such a spike rather than current limit at it or sustain 2 Amps and it looks like the specifications of the parts can indeed handle it for short peaks. It would be a software option to turn off current limiting. In reality I find that just having a 1000uF 16V cap across the Celuller modules supply pins is sufficient to meet that occasional 2A spike.

I don't see much point in tracking.
If you want dual supplies, just use two supplies.

Dave.

Also valid, and with the expansion capability's you have provided or allowed for already with SPI, ICSP and optional Ethernet, Tracking etc. becomes a software rather than hardware feature anyway.

I haven't seen the schematic yet but it should be easy enough for a daughter board, that plugs in place of the Ethernet Module, to provide optically isolated communications with 2 or more of your boards. So that, allong with any other wild ideas can become a challenge for the community rather than you having to satisfy every possible request.

Keep it up Dave,

I wan't 2 at least (Just for myself) and I think several of my mates will want them too, so I look forward to final design and production.


 

[tnt]

I don't really get how the battery is connected ... it seems like it's connected in "series" on the schematics.

[sacherjj]

Timing of this is funny.  Last week I was looking at linear regulators and wondering how tough it would be to make a hybrid switching and linear to minimize cooling required for a large voltage range supply.  The difficulty seemed to be making it variable.  I'll be learning a bunch from rev C.

[Bored@Work]

- Correct me if I'm misinterpreting Dave's explanation, is that pre-regulator control loop is now handled by the mcu (cpu cycles) instead of linearly controlled like the front end linear section ?

I am also surprised about that. Actually I was surprised on several accounts.

While a slow digital control loop should work for a per-regulator I was excepting an analog feedback loop. More precisely, something along the examples in the LT3080 data sheet. There they use a simple transistor to keep the pre-regulator output above 0.6V of the LT3080 output. And the transistor should be cheaper compared to the digital pot. But that feedback loop would offset the current measurement.

Still, the digital pot bugs me. Simple DACs have been used instead for controlling regulators for ages For amusement I have attached a 37 year old application note I rediscovered on some datasheet collection site. They use a current DAC in a little bit convoluted way. Kids these days always think digital stuff was just invented yesterday

This more recent article http://electronicdesign.com/article/power/an-easy-way-to-roll-your-own-programmable-power-su uses a voltage DAC, some math (which they got wrong), and a resistor to do the same with a switching regulator. The schematic (Figure 1)  http://www.electronicdesign.com/files/29/20515/fig_01.gif is dead simple. Although startup can be a problem in some application, it starts with full output. But when used as a pre-regulator this shouldn't be too big of a problem.

I didn't check if Dave has a spare (PWM) DAC in the circuit. If yes, this would be an additional way to shave of some cost.

[Rufus]

To me the power input circuitry and battery have given a power supply with a continuous power limit through the input of about 4W which will be more like 3W at the output.

You are going to have to jump through mechanical and thermal hoops to prevent U2 running steaming hot and in thermal current limit all the time which would further reduce the power limit.

Maybe the ?Supply title is more descriptive than you would think.

[FreeThinker]

Schematic C up now here http://www.eevblog.com/files/uSupplyBenchRevC.pdf
Battery connector fixed and 5v Reg mysteriously has reappeared 

[kkp]

U12A (buffer LM358) will be running outside of common mode input range. Which will ruin your happiness.
Suggestion:
Add 1 ohm in the short vertical wire between R1 and R19. Then remove the buffer.
With the 1 ohm added, the current in R17 and R23 will cause equal drop on the inputs to the INA219, and also cancel in the differential amplifier.
Now you have a free opamp.

/Kasper Pedersen

[hans]

Just my thoughts:

I don't remember battery support was mentioned earlier. It's unique, but very specific. I would have no use for it, and rather power it via a plug pack only. Doesn't matter, it's Dave's power supply

I would have used a DC-DC converter with wider input range, though. Maybe even buck-boost so really low voltage applications can be powered effieciently.

Oh by the way, a high frequency regulator is by essence not more efficient. There is a trade off between that as higher frequency equals in more switching losses. Take a look at the datasheet of LT3505, it clearly says that the programmable frequency set should be used so the power supply can be very efficient or very small (in PCB dimensions).

Also, you mentoined in the video that the tracking regulator will be +2V of the output voltage. What happens in a long period short circuit operation? All the output power would go into the LT3080, which would be dissapating 12W (12V 1A) or more.. maybe make it so that it follows the live output voltage instead (or only if the output voltage deviates alot from the input).
I wonder how stable the power supply is with a constantly adjusting input voltage on the LT3080. I also wonder if the update speed is good enough , because now all is being controlled through I2C.

I like I2C for it's bus capability, but it's so slow. C'mon 400kHz, you can't even get 30kB/s over that. Whatever, it saves a ton of I/O space and makes the design simpler.

I see you also power the whole system with a LDO. How much would the screen+backlight+control circuitry draw? If it's like 100mA, you're probably dissipating 5V*100mA=500mW in the LDO there..

I really like the new INA209 chip. I think it's a great choice over the microcurrent. Also measuring high-side current is so much better than putting a 1 ohm resistance in series. It also makes the power supply stable for applications with low-current sleep but high current bursts when operational (like sensor terminals sending data over GSM network).
The LM358 opamp is not rail-to-rail btw, you have to pick one which is.
I also have concerns regarding the maximum output voltage of a typical R-R opamp, as it sometimes is 10mV or even more under V+, which means that opamp is constantly reading 10mA current draw, on top of the input offset voltage!

I find it disappointing to see the 'ultra high resolution' thing removed though. I really was looking forward to modding or expanding the PSU design with an even more precise measurement capability (as I have a design back from project that had 16-bit resolution auto-ranging from several mA up to 2Amps). But I guess that would be my own spin-off, as I hate Arduino and still don't understand you're putting up with after all the trouble you've had with it

The ethernet control thing is cool feature tho, especially an expansion board is a great idea to integrate.

Going for complete SMD kinda spoils the fun as it's not a kit anymore in sense you'd have to solder the PCB, but I guess you're right for being able to pick so much more and better components.

[sacherjj]

After looking through the INA219 data sheet, I'm wondering if it would be possible to move the shunt to the output.  The sense would still compensate for the voltage drop.  However, you would have to go with a 0.25 ohm shunt to get full range current.  It would allow you to measure Vin- as the actual output voltage, with the 12 bit DAC in the ship.  It would also calculate output power.  This sacrifices 1/4 of your resolution with current measurement, but is probably more accurate than ADC on the Atmel for output voltage.  Hard to say without a schematic to pour over, I'm probably missing something.

[FreeThinker]

After looking through the INA219 data sheet, I'm wondering if it would be possible to move the shunt to the output.  The sense would still compensate for the voltage drop.  However, you would have to go with a 0.25 ohm shunt to get full range current.  It would allow you to measure Vin- as the actual output voltage, with the 12 bit DAC in the ship.  It would also calculate output power.  This sacrifices 1/4 of your resolution with current measurement, but is probably more accurate than ADC on the Atmel for output voltage.  Hard to say without a schematic to pour over, I'm probably missing something.

See my previous post for link to schematic

[McMonster]

I didn't see anyone mentioning it, but there are actually more pins on AVR available than you think Dave. The Rev C schematic shows a DIP-28 ATmega168A-PU chip, while the SMD version, ATmega168A-AU is in TQFP-32 package. The "missing" 4 pins go for additional power pins and two dedicated ADC pins (no I/O capability, internally tied to the ADC). So you may replace the chip in the schematic with the proper part, move two of your channels to the dedicated pins and still have two free I/Os for greater flexibility or you can just route them to some I/O header for people to use as they want.

Magic!

[McMonster]

By the way, I think Rev C is much better than previous versions, much simplified, much less ICs (like an entry for a "how many digital/mixed ICs can you stuff into a linear PSU?" contest). Having a mixed SMPS/linear kind of design sounds a bit strange, so I think first thing you should do is to fully characterize the PSU once you build it (and post a video of course), I'm really curious what's the total efficiency of this device. And now that we really know the "big secret" (not that the battery connector wasn't screaming it out in previous revisions) I must say I'll reconsider buying it. I was absolutely sure I can't justify it before, but I could use a battery powered PSU of that size actually. So I'll wait until the price is known and just hope the shipping to Poland won't double it.

Now waiting for Rev D.

[sacherjj]

I didn't see anyone mentioning it, but there are actually more pins on AVR available than you think Dave. The Rev C schematic shows a DIP-28 ATmega168A-PU chip, while the SMD version, ATmega168A-AU is in TQFP-32 package. The "missing" 4 pins go for additional power pins and two dedicated ADC pins (no I/O capability, internally tied to the ADC). So you may replace the chip in the schematic with the proper part, move two of your channels to the dedicated pins and still have two free I/Os for greater flexibility or you can just route them to some I/O header for people to use as they want.

Magic!

Based on Dave's comments on YouTube for this video, I believe he is sticking with a socket DIP-28 based processor, to allow replacement if someone screws up programming it.

[ModernRonin]

I realize Dave said he didn't want to do this in the video, but I'm seriously thinking that using a 0.1 ohm current shut (10 x 1 ohm 0.1%) might be a good idea. Several reasons...

First, with a 12 bit ADC on the INA219B (*great* choice on that chip, BTW) you're not hurting for resolution. Go ahead and use a smaller shunt!

Second, that would allow reading up to 320 / 0.1 = 3.2 amps. Makes it super easy for someone to drop the LT3083 reg in there and get a full 3 amp PSU with very little trouble.

Seems like the only problem would be that the voltage going into U12A would be ten times smaller. But to fix that, U12A could act as a x10 voltage multiplier (just add 2 resistors around the negative feedback path). Or we can change the gain resistor on U12B (R24) to make it an x10 diff amp. Everything else can stay the same.

The only downside to using opamp gain is that the output voltage error is (input offset) * (gain). So we might have to go to a more precise opamp for U12A or U12B, depending on where the gain is done.

[McMonster]

I didn't see anyone mentioning it, but there are actually more pins on AVR available than you think Dave. The Rev C schematic shows a DIP-28 ATmega168A-PU chip, while the SMD version, ATmega168A-AU is in TQFP-32 package. The "missing" 4 pins go for additional power pins and two dedicated ADC pins (no I/O capability, internally tied to the ADC). So you may replace the chip in the schematic with the proper part, move two of your channels to the dedicated pins and still have two free I/Os for greater flexibility or you can just route them to some I/O header for people to use as they want.

Magic!

Based on Dave's comments on YouTube for this video, I believe he is sticking with a socket DIP-28 based processor, to allow replacement if someone screws up programming it.

Didn't catch that from the video, thanks. But then... Most of the people would probably go with using Arduino or some other bootloader, so there's no room for screwing up the chip. So I think using SMD micro is worth it if it's going to be SMD desing after all.

[m12lrpv]

So I think using SMD micro is worth it if it's going to be SMD desing after all.

Yes I would agree go for the SMD version and the extra pins. I don't see how you can screw up the chip, they are pretty robust and it would take a lot less PCB space. Not to mention that they are probably cheaper too.

Having screwed up a chip in my own adventures and reading about others having done the same I believe the current choice is better than the smd version.

It's also a hell of a lot easier to program if you can just swap the chip in and out of a commercial board.

You can guarantee that someone is going to want to reprogram the fuses to clock it differently and that's another great way of bricking the chip.

[EEVblog]

U12A (buffer LM358) will be running outside of common mode input range. Which will ruin your happiness.

Damn, you are quite right. I must have been thinking I was still using the rail-rail opamp.
Thanks.

Suggestion:
Add 1 ohm in the short vertical wire between R1 and R19. Then remove the buffer.
With the 1 ohm added, the current in R17 and R23 will cause equal drop on the inputs to the INA219, and also cancel in the differential amplifier.
Now you have a free opamp.

I can't see how that would work. There would only be a fixed extra current current flowing through that extra 1ohm resistor, so it can't compensate the variable current flowing through R17 and R18 as the load current changes.

Dave.

[EEVblog]

Also, you mentoined in the video that the tracking regulator will be +2V of the output voltage. What happens in a long period short circuit operation? All the output power would go into the LT3080, which would be dissapating 12W (12V 1A) or more.. maybe make it so that it follows the live output voltage instead (or only if the output voltage deviates alot from the input).
I wonder how stable the power supply is with a constantly adjusting input voltage on the LT3080. I also wonder if the update speed is good enough , because now all is being controlled through I2C.

I like I2C for it's bus capability, but it's so slow. C'mon 400kHz, you can't even get 30kB/s over that. Whatever, it saves a ton of I/O space and makes the design simpler.

The update speed of the tracking pre-regular doesn't really matter.

I see you also power the whole system with a LDO. How much would the screen+backlight+control circuitry draw? If it's like 100mA, you're probably dissipating 5V*100mA=500mW in the LDO there..

The whole design takes about 15mA I think, and the backlight will only add 5-10mA on top of that. I would not be silly enough to run it at 100mA per LED.

Dave.

[EEVblog]

After looking through the INA219 data sheet, I'm wondering if it would be possible to move the shunt to the output.  The sense would still compensate for the voltage drop.  However, you would have to go with a 0.25 ohm shunt to get full range current.  It would allow you to measure Vin- as the actual output voltage, with the 12 bit DAC in the ship.  It would also calculate output power.  This sacrifices 1/4 of your resolution with current measurement, but is probably more accurate than ADC on the Atmel for output voltage.  Hard to say without a schematic to pour over, I'm probably missing something.

I briefly considered that, but the LT3080 driving sense would need to be on the output to compensate, and that might screw up the stability somehow. It would have to be tested.
But that would now fix the LM358 issue I've got...

EDIT: 2nd thought, no it wouldn't work the regulator response time would now to terrible due to the RC on the set pin.

Dave.