Electronic load and ripple measurements

[salomonander]

Hey there,
Im often testing linear power supplies in audio equipment. I would love to have an electronic load for convenience. But would a dc load mess up my ripple measurements by introducing its own ripple to the measurement somehow? Ai told me its fine. But i never trust ai in case it is a problem, does it only apply to budget units? I see that some high end manufacturers list ripple current in their datasheets. Is that a critical value? The psus im testing are low noise - often lm317 circuits

Thanks!

Dont know how to post without this poll… oh well

[David Hess]

No, the electronic load should not be introducing even measurable amounts of ripple.

[twospoons]

*Should not*, but do. At least the Tenma one we have seems to have stability issues sometimes. Bad enough that I built a load bank out of manually switched power resistors for our load testing.

[lhk5195]

Code: [Select]

In fact, electronic loads already include ripple testing functionality. For example, the UNT8512+ I purchased recently has a manual that clearly specifies the measurement accuracy, and you need to determine whether it is suitable for your particular application.

[epitaxial]

There is a nice simple design published using the LM10 and a large pass transistor as a variable load. It's self powered and only uses a handful of passives besides those two parts. I've built it numerous times. The original schematic seems to have vanished from the internet but I have a copy saved. Let me try and locate the thing.

[zike]

There's a commercial product based on this open source design. The schematic and BOM are shown, as is a link to the original EDN Magazine article. I am currently testing one (Saelig carries them in stock). Very simple to DIY your own, too. Can't vouch for noise yet but there's not much to go wrong.

I use an East Tester 400W load and it's great, but it absolutely DOES put switch-mode ripple back on the DUT inputs. My HP6060B is significantly quieter if you are testing for microvolt-level ripple and noise. But I suspect these self-powered designs are best of all, short of a simple power resistor.

There is a nice simple design published using the LM10 and a large pass transistor as a variable load. It's self powered and only uses a handful of passives besides those two parts. I've built it numerous times. The original schematic seems to have vanished from the internet but I have a copy saved. Let me try and locate the thing.

[epitaxial]

Are the new cheap loads using a mosfet and doing PWM or something to achieve an average?

[zike]

Are the new cheap loads using a mosfet and doing PWM or something to achieve an average?

There are some reverse-engineering posts here and on YouTube; the East Tester output stage is a bank of IRF250's driven by LM324's in linear mode (this much I have confirmed by inspection). I don't know how the reference voltage is generated, if it's PWM or a static DAC. Either way, my guess would be the hash comes from an internal (low-power) switching supply that provides its operating voltages. IMO for all it needs to do a small linear supply would be a better choice, but I do not see any mains transformer inside.

[blackdog]

Hi,

Scroll down a bit and you'll see “my” design, which is based on the EDN article/Tekbox version.
Look @ this topic: https://www.eevblog.com/forum/projects/dummy-load-problem/
.

This is a low-noise design; many commercial designs use PWM or other techniques that are not low-noise.
Try measuring the noise characteristics of a analog PSU using a Rigol/Siglent DC load, there is so much noise and interference that the analog PSU is practically impossible to measure.

There is a small error in the schematic; I haven’t corrected it yet, and it concerns C6 in the top right corner of the schematic.
This value is much too high and should be replaced with a small Zobel network.
The values of the resistor and capacitor depend on the bandwidth you want to measure.

Capacitor C5 at 2.7 nF and resistor R10 at 2.7 kΩ provide a bandwidth of approximately 20 kHz.
Usually, C5 can be reduced to 1 nF to achieve a wider bandwidth.

To maximize the range of applications, the Zobel network that is to be placed across the output should be made switchable along with R10 and C5.
This way, for AC testing, you can select the widest bandwidth that suits your tests.

Kind regards,
Bram

[Grandchuck]

Uni-T UDP3305S set at 12 volts, 5 A current limiting
multicomp MP710259 set at 4.9 A CC mode

[tszaboo]

This is a low-noise design; many commercial designs use PWM or other techniques that are not low-noise.
Try measuring the noise characteristics of a analog PSU using a Rigol/Siglent DC load, there is so much noise and interference that the analog PSU is practically impossible to measure.

I the rigol/siglent loads are analog, there is no switching in them.
IMHO the most difficult part of measuring the true noise is the oscilloscope probe noise and enviromental noise.

[nctnico]

No, the electronic load should not be introducing even measurable amounts of ripple.

But in the real world many do. The problem is in the mains transformer. The magnetic field gets coupled into the current sink circuitry and presents itself as a ripple. From my experience this is hard to get rid off. Especially if a DC load has a compact design.

[Kleinstein]

The quality of electronic loads varies. Especially cheap ones can introduce ripple current. I good (regulated) supply should handle ripple current quite well. A simple supply may however show more ripple voltage.

Another possible issue with an electronic load is possible oscillation - some electronic loads (especially DIY plans) can be tricky and not compatible with all supplies.

If in doubt a simple resistive load is lower ripple.

[zike]

@blackdog Thanks very much for linking this, I'm playing with a similar design (also inspired by the EDN article). I'm trying to make it self-powered, as they have, to further limit noise or ground loop opportunities. In this case it's critical to minimize dropout; have you determined that the Sziklai-Darlington is strictly necessary, or could one achieve adequate gain with a simple Sziklai connection (i.e., eliminate Q2 in your design)?  A higher-beta part for Q1 (e.g., 2SD1273) should in principle saturate the outputs with ~ 10mA from the opamp (in my case an LM10) and save a full Vbe of dropout (so it could characterize, say, a 2V supply at 10A).

To maximize the range of applications, the Zobel network that is to be placed across the output should be made switchable along with R10 and C5.
This way, for AC testing, you can select the widest bandwidth that suits your tests.

Is there a "conventional" test impedance/frequency for characterizing load regulation? I see some lab suppllies are specified for settling time to within X % after, say, a 90% output current step, and some offer output Z vs. frequency,  but to my inexperienced eye it all seems to be 'dealer's choice'...

[nctnico]

The quality of electronic loads varies. Especially cheap ones can introduce ripple current. I good (regulated) supply should handle ripple current quite well. A simple supply may however show more ripple voltage.

It is not a supply problem. Remember opamps have a very good PSRR at 50Hz. I took a deep dive in a Korad KEL-2010 DC load and the ripple on the output had two sources: the magnetic fields of the fan and the mains transformer. These fields induced ripples into the control circuitry. https://www.eevblog.com/forum/testgear/korad-kel2010-multicomp-mp710771-review/msg5039434/#msg5039434

[Martin72]

That reminds me—I still wanted to do something about my Siglent load...

[Doctorandus_P]

Sometimes there is an instability caused by an interaction between the load, and what it is connected to. An this even happens every now and then by expensive / commercial loads.

Take for example a power supply with a current limit at it's output. What will happen if you increase the load until the power supply trips into current limit mode?

Some loads can be switched between either a constant current mode, or a variable resistance mode. This may help with preventing such problems.

Nearly all loads also have a shunt resistor to measure the current. You can amplify the voltage over that shunt and then do an ac Voltage measurement  to see if there is a ripple in the load current.

When I was young, I built an LM317 based power supply, and I wanted the maximum output voltage, but with a heavy load, the transformer, bridge and output elco's sagged several volts (which is normal), but I still wanted to use the higher voltage for low current applications. So I added an AC amplifier to the output voltage and had it trigger a buzzer when AC was detected.

[tautech]

That reminds me—I still wanted to do something about my Siglent load...

Like what ? 

[Martin72]

Sometimes there is an instability caused by an interaction between the load, and what it is connected to. An this even happens every now and then by expensive / commercial loads.

This is a not insignificant aspect: the interaction between the control loops of the source and the load.
We've had recurring problems with this in the past, so it was ultimately decided which electronic load should be used to test each power supply.
In that context, I find a Statron load interesting—we have one of those, too:
It involves connecting a chain of passive resistors, and it’s quite sensitive.

[Martin72]

Like what ? 

The SDL1000X loads also have ripple; I discovered that during a test using a battery as the power source.
I wanted to look into it further , but I ended up putting it aside for now.
I’ll definitely get back to it soon.

[blackdog]

Hi,

I’ve created several designs for DC loads, and I’ve designed most of them to be suitable for AC testing as well, since that was important to me at the time.

Tomorrow I’ll show you another design using a NE5534a op-amp, and yes, I know, it’s an op-amp intended for audio!
Why shouldn’t I take advantage of those special features for a measuring instrument that can utilize the drive current and wide bandwidth?
Dynamic Loads are not simple te use and build.

A lot of people who are into electronics want to build a power supply, and most of what they build is junk.
Don't worry about it—I've built a lot of junk, too :-)

DC loads, just like power supplies, are difficult to design properly.
They must be phase and gain stable.

And what do you think happens when you connect these two?
This happens when the current drawn by your load is higher than the power supply’s set value.
The two will fight for control.

This is just one problem; the other arises when the problem described above doesn’t occur,
because the tester has configured the settings correctly and then wants to test the power supply’s dynamic behavior.

For example, the power supply has banana jack connectors, just like the DC load being used.
The tester takes a pair of banana cables and connects the two devices, say, two 1-meter cables.
You now you have about 2.5 uH of inductance between these devices...

The faster the pulse you use for testing, the stranger the result you’ll get.
There are many factors you need to pay attention to if you want to take accurate AC measurements.

The first step is to keep the connection cables as short as possible and twist them!
Mayby tomorrow more about this, this old man has to sleep now. 

Just for some inspiration this design...
Sorry for the colors.

.

And the second part.


Kind regards,
Bram

[Hydron]

In my experience the *magnetic* field from the mains transformer OR cooling fan can get into the current sense shunt and cause ripple. This is separate from control loop instability.
I've seen this on a Korad load (KEL103) where the (shielded EI) transformer field is always present, but you get an additional ripple at currents over 2A as the current shunt cooling fan kicks on, plus a "Dingchen" load where the main fan is right in front of the shunt itself. In either case physically stopping the fan will kill that component of the ripple, and swapping the transformer for a toroidal (low external field) or DC/DC fixes the other component. The Dingchen had less (or none?) from the distantly mounted transformer, but also had some poor layout in the PSU section that causes some 2*mains frequency ripple, fixed by cutting a track and bodging it to be more of a star ground config.
My GW Instek PEL-3032E and GPP-4323 (in load mode) do not show this sort of ripple, thought the latter has an unusable constant-resistance mode (very slow SW control loop, oscillates like crazy with the wrong input).

[nctnico]

Sometimes there is an instability caused by an interaction between the load, and what it is connected to. An this even happens every now and then by expensive / commercial loads.

This is a not insignificant aspect: the interaction between the control loops of the source and the load.

This is why I have a switched resistor DC load. There is no control loop. In the past I also had an open-loop transistor based DC load (the collector of a transistor is more or less a current source) but that died and I never rebuild it. If I ever do, I'll add a low frequency DC wander tracker to compensate for temperature related drift. The open-loop DC load was quite nice to test PSUs for stability.

[Martin72]

This is the Statron switched resistive load, 95 steps:
https://asset.conrad.com/media10/add/160267/c1/-/de/000512905DS01/datenblatt-512905-elektronische-last-statron-32241-300-vdc-13-a-2200-w.pdf

[Doctorandus_P]

This is why I have a switched resistor DC load.

On Hackaday there is a 10kW version of such a thing.
And in an australian magazine I saw an "arduino" version which used a bunch of FET's to switch the power resistors.

It's also a very easy DIY project. Either with big switches that can handle the current, or a rotary (signal level) switch (with some diodes) and one of those cheap 8 relay boards. Very simple DIY project.

But such a project has it's own limitations. power resistors are often wire wound, which makes them a bit inductive, and they often (especially the cheaper versions) have a high TCR.