Parallel or series 2 x Rigol dp1116a?

[FriedMule]

I have written to Rigol if it is possible to connect two Rigol dp1116a in series or parallel or (-16V — 0V — +16V) but the one who did reply wrote that the manual did not tell anything about it so "don't know".

Do any of you know if its possible, I understand that most PSU's can, but a few can absolutely not!! :-)

[FriedMule]

*bump*

Nobody knows?

[myf]

Hello,

Connecting two PSUs in series seems to me better than in parallel :

In parallel, without output diode, look at the schematic with one PSU at +10V and the other at +12V goes from 12V supply to 10V supply and can destroy it.
The problem may be almost the same is one PSU is at 11.99V and the other at 12.01V.

In parallel with an output diode there is no problem : The power supply with a too low voltage isn't used, and all the power (Watt) comes from the other PSU.

In series, the 2 PSU gives some power.

Last problem : in series or parallel the 2 PSU must be insulated from each other.

Others may confirm.

[Ian.M]

You *MUST* *NEVER* directly parallel them - see https://www.batronix.com/files/Rigol/Labornetzteile/DP8_DP1_ActiveLoads.pdf

Series connection of lab grade PSUs with floating outputs is entirely normal.  However one would normally expect to see a rating for the maximum floating voltage with respect to supply ground, and the DP1116A manual is deficient in this respect.  If  you already have the PSUs, check the resistance between Ground and -Out when powered off, powered on at minimum output voltage and powered on at maximum output voltage - if its open circuit in all cases it *may* be a floating supply.   Test with a 1K 1W resistor between each output in turn and Ground - the supply should operate normally and as long as its load isn't grounded, there should be no current through (i.e. voltage across) the 1K resistor.  In some cases, powering down one of a series connected pair of PSUs may cause damage, which can normally be avoided by connecting a reverse biased high current Schottky diode across the output of each PSU so the one that's on cant force significant current through the one that's powered down.  If you don't have them yet, contact Rigol for series connection instructions and restrictions. 

[alm]

You *MUST* *NEVER* directly parallel them - see https://www.batronix.com/files/Rigol/Labornetzteile/DP8_DP1_ActiveLoads.pdf

That seems like a very strong statement to me, just because Rigol could not be bothered to document it. It can be done, as long as you make sure that the one set to the slightly higher voltage is in constant current mode: https://edadocs.software.keysight.com/kkbopen/learn-to-connect-power-supplies-in-parallel-for-higher-current-output-589744099.html. It's also common for supplies to tolerate voltages that are a few percent higher than their set voltage, although they will of course drop out of regulation.

But I agree that series connection is the simpler option. I would not expect the maximum floating voltage to become an issue when raising it up to twice its maximum output voltage (assuming the power supply is floating like a bench supply should be). Common ratings are like 250V relative to ground.

In some cases, powering down one of a series connected pair of PSUs may cause damage, which can normally be avoided by connecting a reverse biased high current Schottky diode across the output of each PSU so the one that's on cant force significant current through the one that's powered down.

Are there standard bench supplies that don't have this antiparallel diode already at their output? I understand that they may not survive a car battery discharging through them, but another current-limited power supply should be fine.

[FriedMule]

As I understand it should the power supply be floating since it has a separate ground, see image:
I did try to contact Rigol about the power supply but as stated above (if it's not in the manual then = don't know)
I understand parallel may not be the best, but series with two diodes should be fine.

How about the (negative - zero - positive) configuration?

[Ian.M]

How about the (negative - zero - positive) configuration?

Yes, any variation of series with intermediate taps, with or without one node grounded should be fine, subject to the maximum floating voltage limit, as Alm has already discussed.   If you can conform the presence of an internal anti-parallel diode across the output (possibly with a fuse between it and the output), rated to carry the max. output current, you can omit the external diodes.

[TurboTom]

At least the DP800 series explicitely permits serial and parallel operation. See on page 2-8 of the manual. That's also valid for separate units. I assume Rigol warns to use "active loads" since they may not have any current limiting capabilities. If the situation is the same for the DP1116 still remains unclear...

[FriedMule]

Thanks yes I could take a look but I am wondering if it still would be better to build a small contraption with diodes and maybe some added safety?
Maybe an op amp to balance the output?

[FriedMule]

The 1116 is an older model but yest since most power supplies i capable to do it and the 800 series can do it, should the chances be good.
I would just wish that Rigol did a bit more than take a look in the manual and say "don't know".

[alm]

Thanks yes I could take a look but I am wondering if it still would be better to build a small contraption with diodes and maybe some added safety?
Maybe an op amp to balance the output?

Are you talking about parallel or series connection? For series connection the external antiparallel diodes could provide a bit of extra safety, especially if the internal diode is under-specified. And apart from the complexity they won't hurt performance in any meaningful way unless you're trying to measure tiny currents. But what would you want to balance? If you want to use the opamp as rail splitter, than the opamp needs to be able to source or sink the full current you are drawing.

For parallel connection, a diode in series would be safe if you're worried differences in output voltages causing problems (the Keysight link I quoted tells you to force one power supply in constant current mode for this reason), but would degrade voltage regulation, and I don't see the need to balance the current between the two power supplies. The one with the slightly higher voltage will probably be delivering full current before the other will start delivering current. Is that a problem?

[FriedMule]

I was thinking on both parallel and series, dependent on task but properly 90% series.
Yes you can use the current limiter but I am thinking, what if I need constant voltage and what if I do only turn one unit on at a time? Maybe it would be better to build an "only on if both is on" and balance output on both side of ground, maybe simply activated by a third small power?

[alm]

Don't connect two power supply in parallel and only turn on one of them without diodes preventing reverse current. Power supply will often tolerate the output being raised slightly above its set voltage, but I would not want to feed it 15V while being powered off.

For balancing two power supplies in series, do you really need better matching than what can be achieved by setting both power supplies to the same voltage? How much current would the midpoint need to source and sink?

[Ian.M]

Personally, in view of Rigol's warning about attempted second and fourth quadrant operation resulting in instability of the control loop and possible loss of regulation (output uncontrolled voltage), (stated in DP8_DP1_ActiveLoads.pdf I linked earlier),  I stand behind my statement earlier "You *MUST* *NEVER* directly parallel them". Concerns are:

• Damage to the load if your load has an upper voltage limit that isn't significantly greater than  PSUs'  max output voltage.
• If the instability causes very rapid oscillation of the load current between the PSUs,  IMHO there may be a risk of their control circuit and/or pass transistor overheating, and possibly failing.

Therefore, for parallel connection, blocking diodes are essential.  As even Schottky diodes seriously degrade regulation, consider using a pair of Ideal Diode modules (readily available from Amazon,  EBay and the usual far east marketplaces). Typically, such modules work over a 3V-28V positive supply voltage range, plenty for an 18V max PSU.   If you need them to operate right  down to 0V, you need to connect both module grounds to a negative bias voltage, e.g: -5V from an isolated 5V wallwart or -6V from 4 AA batteries, both with + connected to the commoned PSU negative terminal.

As for load power sequencing etc. - if your load is that fussy about *how* and in what sequence its rails come up, you should probably be building sufficient power management into your target board to guarantee the required powerup/powerdown sequence  For series PSUs the anti-parallel diodes will be sufficient to keep the rails in approximately the correct voltage order (which is usually enough to prevent damage to all but the fussiest modules and ICs) during individual manual switch-on/off of the PSUs in question.  However a breakout box with binding posts and a beefy multipole switch as a single control to isolate your target board's power without having to individually toggle the output of multiple PSUs can be handy if your development process involves a lot of 'thinking with your soldering iron' or if any of your PSUs are prone to uncontrolled output transients during power-up/down!

[alm]

Damage to the load if your load has an upper voltage limit that isn't significantly greater than  PSUs'  max output voltage.

Hardware overvoltage protection implement using a crowbar circuit that will short the output if the voltage exceeds a set level is designed for exactly this scenario: protecting the DUT in case of a fault in the power supply. The current limit or fuse will make sure the crowbar (thyristor) is not overloaded. That's the big difference between attaching a car battery and another bench supply. I have been troubleshooting a power supply where the control loop started oscillating (due to excessive inductance in the leads to the pass transistor), and the result was the crowbar shorting and getting hot, but no damage to the power supply or anything else. Does the DP1116A have hardware overvoltage protection, or is it just a software limit?

If the instability causes very rapid oscillation of the load current between the PSUs,  IMHO there may be a risk of their control circuit and/or pass transistor overheating, and possibly failing.

Any control loop can oscillate if you exceed its design specifications. Connecting bench supplies in parallel should be within its design specifications. How is this different from rapid down-programming with no load on the output, and the output cap still charged? In either case, the pass transistor should turn off and not oscillate, assuming these power supplies lack an active down-programmer.

Parallel and series connection has been documented in bench supply manuals for decades, sometimes more elaborate with tracking and remote sensing, sometimes just simple direct connection. This apparently includes the Rigol DP836. If these supplies are as fragile as you say, then I would argue this is a design flaw.

Having the ideal diodes in series will probably degrade the transient response (increase the output impedance at higher frequencies), but this may not matter for all applications, like ones with bulk capacitance and/or regulation on the DUT.

[FriedMule]

To answer some of your questions is the Rigol PSU's of high-ish accuracy and more than enough for my usage so no problem there. My reason for wanting to build a "balance" is so that i.e. a DUT is not only applied -15V for maybe 10 seconds until I get to press the other PSU.

And yes parallel is a bad idea if nothing is to protect the PSU's therefore my desire to build something. It could be nice with a simple box with an on / off switch that protects the PSU's. The two Rigol has sense wires and may be a great solution to assure correct voltage.

Regarding your advices about component selection am I wondering if a parallel / series box will be possible to build in one unit, since type and placement differs quite a lot?