Programmable Power Supplies

[solid_liq]

Hey Dave,

Loved the video I just watched: #268.  I've been wondering about programmable power supplies for a while now, but I still don't know what they're good for, how you use them, what makes them programmable, etc.  How about doing a tutorial video on how to make use of them?

Cheers!

[eV1Te]

Also, how does the programming works? If it is a switched supply it can be done in software, but for high stability linear power supplies (with or without tracking pre-reg), how do you control them?

[free_electron]

programmable means they have an interface that allows you to control all the settings and read back what the supply is actually doing.
interfaces can be serial port, usb ethernet or GPIB. some older 'programmable' power suuplies actually have analog inputs. you can chain supplies together to make higher voltage, track each other or deliver more current.

one kind of programmability is 'sense' inputs. this allows you to overcome load induced oltage drop in the wiring.

What's it used for ? lab setups where you automate testing. Home use ? not useful ( apart form being able to chain supplies together using the analog program inputs )
typical prgram settings are voltage, current limit , output on/off. Advanced machines have controllable ramp times for on/off or dampening filters for current spike sensitivity.
you can read back the actual voltage and current being produced and drawn. Some power supplies can even record current versus time in the microsecond step range. ( keilthey and agilent communications power supplies can do that)

Other power supplies have an on board 'downprogrammer' that can be enabled or disabled.

Here is a bit of info about power supplies that most people have no idea about :
What is the output impedance of a power supply ? We are always assuming that , as a good voltage source the Ri (internal resistance) should be low.
The fact of the matter is that this is not true with 99% of the bench supplies out there. It is very low as long as you are DRAWING current. Try to inject current and the output impedance becomes infinite. A lab power supply can not absorb energy !

systems that expect to see a quasi constant supply impedance will not work right on a regular power supply. What are such systems you ask ? anything that is designed to be battery powered... a battery can take reverse current. So , expect differences in behavior while you are doing your tests.

It would not be the first time a system fails on bench because it expects to be able to get rid of excess energy back into the source... and a lab supply can not take this energy. I have seen switching regulators designed for battery operation fail when cnnected to a simple power supply. ( and vice versa )

So here is the problem. You make 12 volts with your lab supply and have a system with capacitors in it. these charge to 12 volts too. drop your voltage output of the supply to 11 volts and the capacitors have nowhere to discharge... The reason is that the regulating element in a power supply is a pass-gate. You start with a large voltage, send it through a pass element ( bipolar / mosfet) and look at the voltage behind the pass-element. and adjust the control of the pass element.
current can only flow in one direction !

Some power supplies put a diode across the pass-element not to let this reverse voltage across the pass element become too large ( reverse polarise the B-E juntion in a transistor with more than 6 volts and you get irreversible emitter damage... )

Your voltage is not behaivng as you expect. if i program something for 11 volts. it should be eleven volts. even if i apply external energy. enter the downprogrammer.
thsi si an additional transistor between the two output terminals. In such supplies the control system not only regulates the output up , but als down. if i set this supply for 12 volts and apply 12.1 externally this downprogrammer will kick in and start drawing current. such a supply will do all it can to keep the set voltage between its output terminals.
these are sometimes called two quadrant supplies.

Even more complicated supplies can reverse the polarity on their output. so they can deliver positive and negative volatage and source and sink current. This is a four quadrant supply.

[ejeffrey]

Also, how does the programming works? If it is a switched supply it can be done in software, but for high stability linear power supplies (with or without tracking pre-reg), how do you control them?

Good power supplies, whether switched or linear, use analog feedback not digital.  Programming in both cases is done with a DAC.  There is a voltage divider on the output that reduces the supply voltage to the DAC range and an error amplifier that compares the two and feeds the error back to the regulator.  Whether the regulator is linear or switch-mode doesn't really matter.

[free_electron]

I'd say even more : the entire regulation loop is done in analog domain.

a DAc sets the 'wanted value' . the analog system closes the loop.
an ADC digitizes the real value. the cpu is NOT active in the regulation mechanism. it only sets 'want' parameters and reads 'real' parameters. it does NOT act on 'real' . i only reports.
things like overcurrent are done by the analog block..

you don't want an unexpected CPU freeze when the supply needs to go in protection....

if you want to see what is inside a 'real' programmable PSU : download the service manual of an Agilent E3631A. they give the whole schematics.
If you want to see pictures of the inside of such a machine i can post these here. i got two of those beasts.

[solid_liq]

This is some great info!  Thank you!

Pictures of the insides would be very interesting to see.

[solid_liq]

Wow, now that I've googled that model number, I really want to know how to make use of them.  That looks like the exact same model of lab powersupply we have in our labs at my school.  The only thing I don't like about them: someone has somehow managed to ruin the Adjust rotary encoder on almost all of them.

[saturation]

Very interesting design considerations, I've not had to design stuff to this level, but aren't such battery operated systems operated using rechargeable batteries, so they were designed to source as well as sink current?  Primary batteries aren't expected to do so, and could be hazardous.

Does the Agilent E3631A allow you to sink current?

Here is a bit of info about power supplies that most people have no idea about :
What is the output impedance of a power supply ? We are always assuming that , as a good voltage source the Ri (internal resistance) should be low.
The fact of the matter is that this is not true with 99% of the bench supplies out there. It is very low as long as you are DRAWING current. Try to inject current and the output impedance becomes infinite. A lab power supply can not absorb energy !

,,,
Your voltage is not behaivng as you expect. if i program something for 11 volts. it should be eleven volts. even if i apply external energy. enter the downprogrammer.
thsi si an additional transistor between the two output terminals. In such supplies the control system not only regulates the output up , but als down. if i set this supply for 12 volts and apply 12.1 externally this downprogrammer will kick in and start drawing current. such a supply will do all it can to keep the set voltage between its output terminals.
these are sometimes called two quadrant supplies.

Even more complicated supplies can reverse the polarity on their output. so they can deliver positive and negative volatage and source and sink current. This is a four quadrant supply.

[free_electron]

the E36xx series does not have a downprogrammer. you need a 66xx series supplies for that, like a 6630b or a 6624a

http://electronicdesign.com/article/test-and-measurement/If-Your-Power-Supply-Needs-Fast-Rise-And-Fall-Times-Try-A-Down-Programmer-64725
http://powersupplyblog.tm.agilent.com/2012/03/if-you-need-fast-rise-and-fall-times.html

@solid_liq : the encoders all fail on the older models. that thing is optical: if it gets dust inside it dies. its easy to fix, pop the front panel off , slide the pcb out : a blast of air and you are done. the newer models have sealed encoders ( last 5 years of production )

@saturation: if you are designing battery operated systems you need to be awara that, when testing on a bench with a power supply instead of the battery, you may get unexpected behavior ... that is why companies like agilent and keithley make special 'battery emulating' power supplies. those machines have downprogrammers on board and you can tell the supply : i want you to behave like a nicd , nimh liion or other type battery with these parameters. and the machine will do that. especially when designin liion charger you need that stuff. you can't experiment on real liion cells ... way too dangerous. ( don't listen to people that claim 'i've done experiments on real li-ion cells and it never went wrong'. li-ion cells are dangerous. i have seen many blow up in the lab. whenever we were working on a charging system the cells were put in a blast proof container with a fire extinghuisher nearby. The temp chamber also had built in auto-extinghuishers. and there too i have seen many go off... ) i did designs for cell-phones... many blew up during the first trials. Li-ion was a new chemistry and very little was known... there were no dedicated charging ic's ( heh , i was designing them )

Primary batteries are not designed to eat current, but they will.... a power supply won't... so : different behavior again.
Another problem : you design car electronics.. that lead acid battery will eat whatever you feed it. so designing car electronics need to be done on a battery emulator , not on a regular bench supply !

as for the insides :

back end under the fan : GPIB and uart + control cpu through an optical link. the power supplies are galvanically isolated from the ground. communication is through built in optical link
big heatsink : 0--6 volts channel -0--5 amps. analog block on the board. the column of big fat 8 pins chips on the right hand side are analog optocouplers. this is done to be able to galvanically isolate the two channels on the supply as well. they are totally floating in respect to each other.


front


the +20 - 20 regulator and support electronics.

the big plcc's are a custom asic containing the control logic + digitizer and next to it is a 80188 with ram and eprom containing the supplies firmware. the frontpanel has its own cpu do control keyobard , rotary encoder and display.

the DAc is a 16 bit dac followed by a multiplexer and analog storage system ( Rc network followed by an opamp. the dac scans in 1 microseoncd increments and keeps recharging the capacitors. this save 5 expensive dacs.

the a/d system is the same principle as used in the 34401 multimeter: it is a charge balancing convertor. what they do is charge a capacitor in an integrator circuit for a known time from a known voltage. ( in an integrator capacitor charging is constant and does not follow the exponential curve ) then they discarge the capacitor with the voltage that needs to be measured. all they do is measure the time it takes to discharge. the voltage is a time-ratio of the reference voltage. the entire system is self calibrating.
all you need is a simple comparator ( they use an AD711 ) , two precision opamps ( AD706 ) and a reference source. the rest is a statemachine and counters.
you can breaboard such a convertor and get 20 effective bits out of it without blinking... if you do the layout right and pick the right parts you can pump it to 24 bits.

they are cheap to make and extremely precise. calibration is easy because it can be done in the numerical domain. no need for trimpots and other gimmicks.

[saturation]

Very nice, free_electron, I wasn't aware of this type of device.  These fast supplies are very helpful when working with pulsed power devices like e.g. cell phones, burst mode radio, or power lasers.

But one can emulate key downprogrammer functions or battery emulators with a linear supply, if the response time is within your need.

Linear PSU can operate in parallel with other DC loads, so it can deal with active loads.  But one can add a few protective elements rather than risk damaging the PSU, regardless.

Linear supplies often have bleeder resistors to discharge its output caps to set new dc adjustments; the bleeder will also sink any active load sourced to PSU, and that resistor is effectively its input impedance.  When in doubt, one may add a high watt resistor in parallel to the load to bleed or sink any current. Its wasteful, but will protect the linear supply to work.  One can also add a blocking diode to prevent any current flow into the PSU, leaving the designer added bleeder resistor after it to do all the dissipation.

If pulsed power is needed for brief intervals that a PSU can't deliver, a large cap in parallel could provide this short power burst.

Nonprogrammable supplies may not dynamically ramp output down or up to emulate circuit controlled sleep, power saving or trasmit modes, or, even switch to much higher output voltages for a burst RF transmitter, but it can be done manually to test each mode. 

Given Li batteries as a fire risk, I think working with them may require a true battery emulator. 

However, NiMH [or SLA] are safe, so one can use real cells during development with a linear supply.

Thanks for the Agilent tear down, as always in electronics, there is more beauty under the covers!
 

the E36xx series does not have a downprogrammer. you need a 66xx series supplies for that, like a 6630b or a 6624a

http://electronicdesign.com/article/test-and-measurement/If-Your-Power-Supply-Needs-Fast-Rise-And-Fall-Times-Try-A-Down-Programmer-64725
http://powersupplyblog.tm.agilent.com/2012/03/if-you-need-fast-rise-and-fall-times.html

@saturation: if you are designing battery operated systems you need to be awara that, when testing on a bench with a power supply instead of the battery, you may get unexpected behavior ... that is why companies like agilent and keithley make special 'battery emulating' power supplies. those machines have downprogrammers on board and you can tell the supply : i want you to behave like a nicd , nimh liion or other type battery with these parameters. and the machine will do that. especially when designin liion charger you need that stuff. you can't experiment on real liion cells ... way too dangerous. ( don't listen to people that claim 'i've done experiments on real li-ion cells and it never went wrong'. li-ion cells are dangerous. i have seen many blow up in the lab. whenever we were working on a charging system the cells were put in a blast proof container with a fire extinghuisher nearby. The temp chamber also had built in auto-extinghuishers. and there too i have seen many go off... ) i did designs for cell-phones... many blew up during the first trials. Li-ion was a new chemistry and very little was known... there were no dedicated charging ic's ( heh , i was designing them )

Primary batteries are not designed to eat current, but they will.... a power supply won't... so : different behavior again.
Another problem : you design car electronics.. that lead acid battery will eat whatever you feed it. so designing car electronics need to be done on a battery emulator , not on a regular bench supply !

as for the insides :

back end under the fan : GPIB and uart + control cpu through an optical link. the power supplies are galvanically isolated from the ground. communication is through built in optical link
big heatsink : 0--6 volts channel -0--5 amps. analog block on the board. the column of big fat 8 pins chips on the right hand side are analog optocouplers. this is done to be able to galvanically isolate the two channels on the supply as well. they are totally floating in respect to each other.


the +20 - 20 regulator and support electronics.

the big plcc's are a custom asic containing the control logic + digitizer and next to it is a 80188 with ram and eprom containing the supplies firmware. the frontpanel has its own cpu do control keyobard , rotary encoder and display.

the DAc is a 16 bit dac followed by a multiplexer and analog storage system ( Rc network followed by an opamp. the dac scans in 1 microseoncd increments and keeps recharging the capacitors. this save 5 expensive dacs.

the a/d system is the same principle as used in the 34401 multimeter: it is a charge balancing convertor. what they do is charge a capacitor in an integrator circuit for a known time from a known voltage. ( in an integrator capacitor charging is constant and does not follow the exponential curve ) then they discarge the capacitor with the voltage that needs to be measured. all they do is measure the time it takes to discharge. the voltage is a time-ratio of the reference voltage. the entire system is self calibrating.
all you need is a simple comparator ( they use an AD711 ) , two precision opamps ( AD706 ) and a reference source. the rest is a statemachine and counters.
you can breaboard such a convertor and get 20 effective bits out of it without blinking... if you do the layout right and pick the right parts you can pump it to 24 bits.

they are cheap to make and extremely precise. calibration is easy because it can be done in the numerical domain. no need for trimpots and other gimmicks.