Programmable dc electronic load

[Spikee]

It has been a while but i finally started on the programmable part of my dc load and in this topic I will post my progress and questions.
Old topic:
https://www.eevblog.com/forum/projects/did-i-forget-anything-constant-current-load/

Basic idea:
Cypress Psoc 5Lp + dc load = Programmable dc load

I already made most of the processor part of the design:

Schematic(pcb is not done): https://dl.dropboxusercontent.com/u/100819959/DCDCTESTER/dcdc_tester.pdf

The board has:
Psoc 5LP - 20bit delta sig adc , 2x 12bit SAR adc , rtc ,full speed usb ...  http://www.cypress.com/?docID=42882
3.3V and 5V ultra low noise ldo's http://www.ti.com/product/lp38798
128x64 Newhaven GLCD
Micro SD card socket
High resolution ALPS rotary encoder

Things that I will add:
Support for multiple temperature sensors (thermistors or somthing like LM35 ...)
External high resolution dac
The load stuff
The current measurement stuff

As Dave has explained in his dummy load video you can have a :
-Pulse load
-constant current load
-constant resistance load
-constant voltage load

And test / plot a lot of cool things

I also incorporated input voltage and current measurement so you can test / plot the efficacy of linear / dc-dc converters over load.
And it measures the output/load voltage so it can also do low voltage cutoff to protect batteries under test.

My basic design for the input measurement:


And for the load:


TODO:
-Determine which external dac to use
-Determine the best opamp for this situation (adc buffer , voltage follower for mosfet etc..)
-Use IC1 (INA168) yes or no
-Adc / DAC Input/output protection
-Reverse voltage protection
-Overvoltage protection
-Power INA168 externally (problem when DUT < 2.7V)
...

I would like to have your help to select the components to use and other things to implement in the design.

Thanks!

[TerminalJack505]

I see one potential problem.  You are powering the INA168 from the input but what happens when the input voltage is less than 2.7V, the minimum voltage required by the INA168?

[Spikee]

I see one potential problem.  You are powering the INA168 from the input but what happens when the input voltage is less than 2.7V, the minimum voltage required by the INA168?

I missed that  ... thank you for mentioning it and I will add it to the todo list.

[Jay_Diddy_B]

Hi,

I would not use the INA168. I would just measure the voltage across the 10x 1 Ohm (0.1 ohm) resistor in the source of the MOSFET.

I started another load thread yesterday which concentrates on the Analog portion of the load design. Our efforts could be blended together to make a nice project.

https://www.eevblog.com/forum/projects/dynamic-electronic-load-project/msg288313/#msg288313

You could replace the pots, in my design, with DACs.

Let me know if I can help.


Jay_Diddy_B

[tszaboo]

Is there any particular reason why you need two set of shunt resistors? I would go with one set. Also, connect the INA168 power supply to 5V, or VCC.

[Spikee]

Hi,

I would not use the INA168. I would just measure the voltage across the 10x 1 Ohm (0.1 ohm) resistor in the source of the MOSFET.

The INA168 is only used for the high side input measurement , not the low side (at load).
At the load I do exactly as you described.
And it would be great if we could help each other.

Is there any particular reason why you need two set of shunt resistors? I would go with one set. Also, connect the INA168 power supply to 5V, or VCC.

There is some confusion about what does what so I made this diagram to explain it.


When you want to test a dc-dc converter board you made you do the top measurement. Because it measures the input voltage / current you can make a efficacy plot.

When you just want to test a bench power supply you do the one on the bottom of the diagram.

[tszaboo]

In that case you really need that two set. To be honest, I like it, thumbs up!

[Harvs]

I'll be interested to see what sort of accuracy you end up getting using the PSOC on-chip ADC.

Lately I've been using a cheap microchip 18-bit D/S ADC with a OAR 5W 50mOhm shunt for a +/- 2.5A current measurement in the power supply conversion I was talking about the other day.  It tracks +/- 100uA to my bench meter up to 1A, then after that heat effects make it drift up to about 400uA off @ 2.5A.  Overall not bad and good enough for the job. Could probably do some non-linear correction to improve the >1A measurements slightly.

It would be good to be able to compare, particularly in terms of measurement noise.

[Spikee]

According to some forum posts i found it should be about 18bit useful resolution.
But time will tell how accurate it will be.

[firewalker]

mA resolution down to zero?

Alexander.

[Spikee]

mA resolution down to zero?

Alexander.

That would be the end goal , mA and mV resolution to zero. This will require alot of fiddling and a good pcb layout.
Guess I'll have to start spice to simulate some things now ...

[firewalker]

And a better opamp. Something like LT1013 should be ok.

Alexander.

[Spikee]

I have tried the LT1013 in LTspice simulation but the results were not that good. I saw a big offset of about 30mV when the input was zero.
It is fixable by using a negative rail but that is just to much work.

So i went to the linear website -> opamps -> precision opamps and just chose one at random that had low offset voltage: LTC2054 But it costs quite alot (3.50€)

When 1.00000 is applied to the voltage follower 1.00050 mV comes out so that is good. I also chose a IXYS mosfet because it is designed for extended linear range : IXYS IXTH75N10L2 (see application at the bottom "Programmable DC load")


The LTC2054 only can handle 6V so i need to change this part (VGS max of mosfet is 20V).

For the DAC I probably need/want a 16/18 bit dac , does anyone have a suggestion for a good resonably priced one ?

This is a good start , i requested a sample of the IXYS IXTH75N10L2 because none are in stock @ Mouser.

[Harvs]

It is fixable by using a negative rail but that is just to much work.

The one piece of advice I have is to drop this assumption.  It's trivial to add an off the shelf sub $1 charge pump voltage inverter IC that'll give you a negative rail good for < 10mA.

Trying to get something to truly go to zero output without a negative rail is far far more work than adding this.

In the low cost DAC arena, this is even more problematic.  Most 16bit DAC's have buffered outputs that'll only get within a couple of mV's of ground.  So my solution in the past has simply been to feed this into another opamp stage with about a 20mV negative offset, then trim to zero in software.

With a negative rail your options for low offset, low noise op-amps become much easier.  As much as some may singer at using a 30yr old op-amp, I often find the good old OP07 to be excellent as long as you operate it's inputs close to center of the rail.  I just bought a heap of TI OP07's element14 @ $0.40ea, and conducted tests with +/- 12V rails.  Typically the offset when the inputs were at the center of the rails was <2uV.  However that climbed to 14uV once you were 5V away from center.

[Spikee]

That is probably the best option for this problem.
I'll have a look at the LTC1983 and simulate the offset circuit soon.

Thanks for the info.

[TerminalJack505]

The op amp driving the MOSFET won't need to go to 0V.  It only needs to go as low as necessary to turn the MOSFET off and 50uV is fine for that.  Yes, the input offset probably looks terrible when you set the op amp up as a voltage follower and ground the input but that isn't a factor here.

Your op amp also might not need to go above 6V.  Its output only needs to go as high as necessary to drive the MOSFET for the maximum programmed current.  The simulator and datasheet will give you some idea of what will be required.

You will want an op amp with a fairly low input offset voltage since your sense resistor will be 0.1R.  See this post as to why.  (It also demonstrates that the op amp not going to 0V won't be a problem.)

[BravoV]

The op amp driving the MOSFET won't need to go to 0V.  It only needs to go as low as necessary to turn the MOSFET off and 50uV is fine for that.  Yes, the input offset probably looks terrible when you set the op amp up as a voltage follower and ground the input but that isn't a factor here.

Hey Jack,

Hope you don't mind answer few questions regarding this topic :

1. Regarding the op amp for driving mosfet, looking at the mosfet gate capacitance, is it "generally" safe to say that its "better" to use op amp that has an output that capable of driving load with quite big capacitance ? or its not a big deal at all ?

2. In the case driving the mosfet at most of the time within linear region in this specific application, will high slew rate op amp (high speed) helps ?

3. By using choper op-amp as OP planned, will the switching noise inside the op-amp affect the regulation loop ? I mean will it amplify/propagate it at the final current loop on loading the DC supply in form of noise/ripple ?


@Spikee, sorry for hijacking, hope you don't mind. 

[Spikee]

...
Your op amp also might not need to go above 6V.  Its output only needs to go as high as necessary to drive the MOSFET for the maximum programmed current.  The simulator and datasheet will give you some idea of what will be required.
...

I simulated fast rise time etc and it worked good using a 5/6 volt rail. But I might have a look in the datasheet to be sure.

Also i was wondering when I use a inverting charge pump woulden't the noise of it have influence on the precision and regulation loop of the other parts of the circuit ?

@Spikee, sorry for hijacking, hope you don't mind. 

These are valid questions and will be useful for me and others.

[BravoV]

Also i was wondering when I use a inverting charge pump woulden't the noise of it have influence on the precision and regulation loop of the other parts of the circuit ?

When it comes to negative rail generator using capacitance charge pump, check this thread HERE, Jay_Diddy_B posted a really simple charge pump using common discrete components, the interesting part about the design, it uses parallel fed and full wave multiplier, the switching noise residue is quite minimal imo, compared to only common half wave charge pump topology. Btw, I'm not an expert, its just my humble ee wannabe observation, so take it with a grain of salt. 

[Spikee]

Also i was wondering when I use a inverting charge pump woulden't the noise of it have influence on the precision and regulation loop of the other parts of the circuit ?

When it comes to negative rail generator using capacitance charge pump, check this thread HERE, Jay_Diddy_B posted a really simple charge pump using common discrete components, the interesting part about the design, it uses parallel fed and full wave multiplier, the switching noise residue is quite minimal imo, compared to only common half wave charge pump topology. Btw, I'm not an expert, its just my humble ee wannabe observation, so take it with a grain of salt. 

If you count and lookup the price of the components It is cheaper and easyer to buy a IC that does it. Less pcb space , smaller BOM , and probably more stable. Something like a LTC1550 is quite reasonably priced: http://www.linear.com/product/LTC1550

The question now is do you really need neg rail or not. All depends on dac / opamp / mosfet selection.

[BravoV]

That LTC1550 looks promising, especially with the built in linear post regulator, thanks. 

[Jay_Diddy_B]

Spikee,

Once you have made the decision to have a negative rail. I would then arrange all the op-amps to be in the inverting configuration. This has several advantages:

1) The inputs are mid rail which optimises their performance.

2) You can balance the resistance at the input to reduce the effects of bias currents.

3) You have lots of places you can add signals for modulation etc.


Jay_Diddy_B

[Harvs]

Also i was wondering when I use a inverting charge pump woulden't the noise of it have influence on the precision and regulation loop of the other parts of the circuit ?

As long as you have a linear post reg it'll be fine.  Ok, probably not as good as having a transformer with all the secondaries you want then simple rectifier and linear reg (which is the strategy I've taken lately, modding transformers to my req.) But it certainly wont be bad, a good op-amp should have 80 odd db of PSRR.

If you count and lookup the price of the components It is cheaper and easyer to buy a IC that does it. Less pcb space , smaller BOM , and probably more stable. Something like a LTC1550 is quite reasonably priced: http://www.linear.com/product/LTC1550

Yeah ok.  Where are you getting your prices from?  I see a $4.80 part there in small qu.  Perhaps you've got a good place for LT parts?
http://www.digikey.com.au/product-search/en?vendor=0&keywords=LTC1550

Where as I've used something simple like this (different brand but same part):
http://www.digikey.com.au/product-detail/en/TC7660COA/TC7660COA-ND/267395

Followed by a standard $0.25 three terminal linear reg and had quite excellent results.  You can add a small PI filter with an 0805 ferrite bead between the charge pump and reg to further kill the noise.  This can be surprisingly effective for small currents.

The question now is do you really need neg rail or not. All depends on dac / opamp / mosfet selection.

Well it's your design, so your choice.  I would use the term need though, you can do it either way. 

But IMO there's some compelling reasons to use a negative rail.
e.g. how do you plan on doing the input voltage measurement?  There's some clever circuit designs in the old service manuals from HP and Fluke for input circuits with analog mux switching the attenuators, but result in the mux resistance being irrelevant.  This is not possible without a negative rail.

[Spikee]

Farnell has a secret discounted stash of LTC1550 : http://nl.farnell.com/linear-technology/ltc1550cgn-pbf/volt-inverter-smd-ssop16-1550/dp/1273881
1.24€ at single qty.

As for the voltage measurement i just plan to use a single voltage divider from +60V to 2.000V using the 20bit adc.
I'll have to do the math if it is possible too get my 1 mV resolution using that. Else I could use multiple voltage dividers for multiple ranges.
And use the PSOC5LP internal mux to switch between the ranges. And when the voltage reaches the top of a range it would be diode clamped to protect the adc input. But i have to test if that is actually possible.

I should get some samples and make a quick 1 layer pcb or breadboard and test some configurations of negative rail vs positive rail.
I'm not yet sure which to pick.

[Harvs]

Farnell has a secret discounted stash of LTC1550 : http://nl.farnell.com/linear-technology/ltc1550cgn-pbf/volt-inverter-smd-ssop16-1550/dp/1273881
1.24€ at single qty.

Good old Farnell, they never cease to amaze me how over-priced they are on one thing, then completely under-cut everyone on something else.

As for the voltage measurement i just plan to use a single voltage divider from +60V to 2.000V using the 20bit adc.
I'll have to do the math if it is possible too get my 1 mV resolution using that. Else I could use multiple voltage dividers for multiple ranges.
And use the PSOC5LP internal mux to switch between the ranges. And when the voltage reaches the top of a range it would be diode clamped to protect the adc input. But i have to test if that is actually possible.

I should get some samples and make a quick 1 layer pcb or breadboard and test some configurations of negative rail vs positive rail.
I'm not yet sure which to pick.

I'd suggest doing some tests would be a good idea. I haven't used the PSoC parts so it might be alright, but you also may find the analog performance of things like the mux disappointing as you get close to the rails.  This is generally the case with jelly bean mux's that otherwise perform admirably with a decent rail voltage headroom.