DC-Load wanted(Siglent SDL1020X-E found)

[nctnico]

I doubt that is truly possible as the 1030X lists a better accuracy compared to the 1020X-E. Unless the accuracy is crippled on purpose, there must be better parts used in a real 1030X.

Why would you think it not possible?
EDIT: You said accuracy, while I meant resolution. Still not aware of any difference between these.

Check the datasheet really carefully and you'll see there are differences in the accuracy specifications between the models.

And even though the board may look the same, it is possible the higher end models use more precise versions of the same components. Look at the LM4040 voltage reference as a random example. It is available in 4 grades going from 0.1% to 1%.

[Martin72]

In all the values listed in the data sheet, there are only differences in this area ("Measuring Range") between the more expensive X model and the cheaper X-E variant.
However, I am not yet able to clearly assign the term "measuring range".
It can mean that the measurement and display are more accurate, but it can also mean that the better accuracy is not only used for display purposes, but is included in the overall accuracy.
As I write this, I think the latter will be the case.

[nctnico]

Look at the accuracy specs. I never wrote about resolution

[Martin72]

Ah, ahh...I didn't see that...

Well, the only solution would be to actually and directly compare an X model with an X-E via teardown.
Or, I would take and save various measurements, then "optimize" and repeat the measurements.
What I find suspicious is this jump in price.
However, if you can convert a Siglent 2.1Ghz SA to a 3.2Ghz SA/VNA, why should it be any different in the case of loads?
The only way to find out is to test more thoroughly.

[egonotto]

Hallo,

The accuracy figures for readback voltage are somewhat strange. Since the X-E model has only 0.02% at FS, while the X-Model has 0.025% and thus less accurate in this range.

Best regards
egonotto

[delvo]

One downside to the siglent ones is that there is no public calibration/adjustment procedure (Id love to be proven wrong on this)

[Martin72]

We (at work) had this topic in general a few years ago, regarding power supplies/loads also calibrating (having them calibrated).
If you record/measure the voltages/currents with external (calibrated) measuring equipment anyway, then you don't need to have it calibrated.

[alm]

We (at work) had this topic in general a few years ago, regarding power supplies/loads also calibrating (having them calibrated).
If you record/measure the voltages/currents with external (calibrated) measuring equipment anyway, then you don't need to have it calibrated.

Sure, but isn't the value in a modern load / power supply the built-in measurement and programming features? Otherwise you could also use an old load or power supply with analog panel meters. You could of course do a software correction yourself before programming / after readback, so if you want it to sink 1 A and you know it's 1% low, you tell it to sink 1.01A. And if you know it reads back 1% high and it reads 1.01A, you calculate this to be 1A. But this is a hassle.

I don't understand the trend of particularly the upcoming brands like Siglent to not publish adjustment information. It's not like adjusting a power supply or electronic load is difficult or requires special equipment. A good DMM with an external current shunt for higher currents should be sufficient for 95% of the adjustments. The only exception would probably be rise time tests which would require a scope.

[Martin72]

you tell it to sink 1.01A

That's another question, whether you have to have a flowing current of 1A with an accuracy of less than 10mA.
Nevertheless, I can try to get more detailed information, which I would then share accordingly.

[Martin72]

Maybe build something yourself where you can control some properties yourself?

Hi Bram,

In progress...

Martin

[orzel]

Hi Bram,
In progress...

What a teasing !

[nctnico]

I'd put the opamp much closer to the transistors and make sure high impedance nets (especially the feedback loop) have the smallest possible loop area.

[Martin72]

Good hint, I´ll place  a part of  the circuitry on the bottom side.

[nctnico]

Ah, ahh...I didn't see that...

Well, the only solution would be to actually and directly compare an X model with an X-E via teardown.
Or, I would take and save various measurements, then "optimize" and repeat the measurements.
What I find suspicious is this jump in price.
However, if you can convert a Siglent 2.1Ghz SA to a 3.2Ghz SA/VNA, why should it be any different in the case of loads?
The only way to find out is to test more thoroughly.

You can't really test it. You need to identify the components that determine the accuracy. But even if the sample of the lesser model you have has the higher spec components, you can't be sure a lesser model always has the better components in them. Old story from S.E.D.: One of the members over there produced a circuit that needed 1% capacitors. But since these where expensive, they bought the 5% ones and binned them. The excess capacitors where used elsewhere, so nothing got wasted. However, one day they received a batch where all the capacitors within 1% tolerance where missing. In the end you can't be sure unless you specifically buy a part of piece of equipment that is specified to certain accuracy / tolerance levels.

[Martin72]

Hi,

But even if the sample of the lesser model you have has the higher spec components, you can't be sure a lesser model always has the better components in them.

In this case, I wouldn't care if another batch was equipped differently.
I have the "good" one.
But I know what you mean and how you mean it.
The load is already on its way, I can test it at the weekend.

[Martin72]

The load is already on its way

Must correct this...
Let it acclimatise first...

[5U4GB]

A vaguely-related question, I know the OP asked for mA-level loads but for general-purpose low-voltage testing are the Chinesium ones that bad?  I've got one of the near-infinite number of copies that consist mostly of an Intel-style CPU heatsink and fan strapped onto the load FET, but with the great feature that it has every style of USB connector in existence, as well as barrel jacks and screw terminals, which means I can do immediate no-hassle testing of 5V/12V power supplies, cables, and whatnot to diagnose power problems with various devices, and accessories like solar panels used to drive some of said devices, for example to check whether the claimed 5W panel really is 5W (no, it's 0.75W).  I wouldn't use it for precise testing, but for sheer convenience it's hard to beat.

[orzel]

What you describe definitely sounds like an "Atorch DL24". I got one, and I find it very useful as a "better than nothing" electronic load. I got a 180W model and I use it up to 150W very often, and it hasn't burnt yet. Though some other people were less lucky it seems.

Of course it's not a proper/genuine electronic load, it's not even sold as is (it's officially a "battery tester").  But still, for the price, I'm very happy so far. I got a version with bluetooth, but it's buggy like hell and I don't use it.

Obviously, I don't design power supplies, so my needs are very modest.

[5U4GB]

Yup, it's from that general family, I just got one with coarse and fine pots rather than the pushbutton one which makes it easier to vary the load slightly over time, but that's just a personal preference.  As with your use case, it does what I need even if it's not terribly fancy and it's incredibly convenient to have all the required connectors built in.

[blackdog]

Hi,

To make things a little clearer here is the diagram Martin72 is talking about.


.
The BNC connector attached to R1 is the DC input, which allows you to set the DC current via e.g. a LAP Powersupply, DAC, 9V battery.
The BNC connector below it can be used with a function generator that gives you a negative DC signal modulated with an AC signal.

A combination of both inputs is also possible, it's your party. :-)
Keep in mind the eventual interaction and the large RC time of C1 of 47uF.
If you don't want to modulate at low frequencies, you can also make this capacitor C1 4.7uf, adjust for what makes sense for you.
I added a Stand By switch and that is S2, if the contact is made then the -input of the NE5534a is pulled to the positive power supply by the orange or the red LED depending on the function.
With large signals on the inputs maybe R11 of 6K8 is not good enough, make it 3K3 or so.

The 80C Clikson take from a temperature value that fits your cooling element for the MOSFets and the power you want to dissipate.
I have good results with the modern PC Processor coolers, mount the MOSFets directly on the cooling element and mount the cooling element isolated in the case.
This gives the best cooling for your MOSFets.

I have tried to squeeze as much performance out of this circuit as possible and that the circuit also still maintains good stability.
But there are plenty of "tune" possibilities to experiment with.
The limitations are mainly in the speed, that is the speed of the NE5534a, the compensation by means of C7 of 560pF and the Drain/Gate capacities at a large AC swing on the Drains.

Not every load is a power supply with a good capacitor across its output terminals.
I also tested this Dynamic Load with a 4 Ohm and an 8 Ohm resistor without Snubber circuitry.
And taking that into account, I determined C7 to be 560pF.

But you can still test some faster Power Supply's, and then there are several possibilities.
Use the input that R1 is connected to, it does not have a 100Khz low pass filter.
And second, use the normal modulation input through C1 and then make C7 smaller but not lower than 180pF.
Otherwise the phase margin becomes too small and you could sometimes get generation problems.
The purpose of C3 of 330pF is, to not feed the NE5534a too fast edges that it cannot handle.

As a tip, don't go too far with boosting the speed.
This circuit is intended for my normal wiring length of about 30cm between the Dynamic Load and the D.U.T.
And of course twist the wires, about 1 twist per Inch is OK.

This then still results in a paracitic inductance, that if this inductance is driven too fast by steep edge from the Dynamic Load can produce resonance.

How do you measure the properties of e.g. a linear power Supply.
I use the following signal for this purpose, 400Hz with 10% Duty Cycle.
The slopes are then 50 to 1uSec this depends on my application.
The minimum current I draw is then 10% of the maximum current of the Power Supply and the maximum current is 95 to 100%.

I use a battery scope to measure the error signal on the power supply terminals.
Clean your banana plug sockets just like your banana plugs with e.g. IPA!
And I don't mean beer by that! :-)
Connect the scope via a coax cable to the + and - connection, so e.g. the holes for wire connection in the banana plugs.
And connect de Dynamic Load via Banana plugs.

Don't use banana plugs you can stack! WRONG!
Or actually do it once the right way as I describe and then once the wrong way which you can see a lot via Youtube, they really know what they are doing :-) and see the difference on the scoop.

A different and verry fast Dynamic Load
For very fast test pulses, I use a circuit by Jim Williams that was in one of his Application Notes.
Then I do not use a connecting cable but use an almost direct connection via a bannaan coupler, this is to keep the inductance as low as possible.
Then I can test up to 1MHz block pulse, but every mm of wire already gives abberations on the signal to be measured.
So increasing the speed comes with al kind of different limitations and it is good to learn and realize this. :-)

I designed this circuit about 10 years ago, now I would do some experimentation with MOSFets that have smaller internal paracitic capacitances and are also suitable for linear operation.

But I still think this circuit is suitable for the hobbyist to have fun experimenting with and get good results.

Oh well, I chose the NE5534a specifically for this application!
Low Noise and sufficient output current with reasonably good bandwidth and it costs little.
And the offset is the trim and it remains reasonably stable.
So do not think I put in a uA741 , TL071, LM324, no, no,no. :-)

With kind regards,
Bram

Translated with DeepL.com (free version)  <= So I hope you will take that and my dyslexia into account.

PS

Some test pictures
This is the current pulse measured at the BNC output, that is, the voltage across the Source resistors from a measurement at a linear HP6237A power Supply.
The Dynamic Load draws a clean pulse from the Power Supply under test, no abberations are visible.(twisted pair wiring used).
The maximum current is slightly above the maximum current the Power Supply can deliver, this is possible due to the energy contained in the capacitor across the output terminals of the Power Supply.
By making the current pulse even larger or making the duty cycle longer you can gain insight about the current behavior of your D.U.T.


.
This photo shows the result of the current pulse used.
This is the voltage across the output terminals of the HP Power Supply.
The Phase Margin and Gain margin of this power Supply is optimally set, by the way, this is the 18V output that can deliver a maximum of 1-Ampere.

If the pulse drawn by the Dynamic Load becomes even faster, then the peak amplitude will become even higher.
The reason is largely the inductance of the capacitor across the output terminals and the measurement point of the scope.
I usually mount my measurement cable on the back of the banana terminals to avoid suffering from the inductance of the connection terminals.

[mikeselectricstuff]

If you're on a budget, this is hard to beat price-wise.
https://vi.aliexpress.com/item/1005005997230178.html?spm=a2g0o.order_list.order_list_main.5.3ac21802x8cQJa&gatewayAdapt=glo2vnm

I've only given it a cursory test so far but seems to basically work. rate of change of load is quite heavily damped so probably no good for impulse/load-step testing.

 

[Martin72]

I haven't really tested the siglent yet, but I couldn't contain my curiosity and unscrewed it earlier.
The clean build quality is very good, nothing to complain about.
12 IRFP250 FETs do their job there, which should easily be enough to turn the 200W load into a 300W one.
A very small detail also convinced me that the "hack" should not cause any major problems...

[Martin72]

After I quickly and successfully "upgraded" the load to a 1030X, it's time to test it too.
However, something came up in between (testing the Micsig CP1003 B), but I'll soon get started.
I will stick to what nctnico tested, it seems to make the most sense to me.
And today the boards for the DIY load arrived, still with the "old version" of my load board, where they had (rightly) criticized the too long feedback paths.
But it was already in production and that's why it's here now.
The temptation is there to fit both revisions and test whether and what difference this makes in practice.
I will soon have completed the design for the new revision.

Martin

[oz2cpu]

about siglent SDL 1020
did anyone figure out the internal programming modes ?
like you can make a lot of steps that do any type of mode and feature, and delays and all that,
BUT i am missing a REPEAT or LOOP or goto x step feature ??
the program only run once, and that is far from what i need,

I can ofcourse use a pc to control it, but such simply thing as a loop feature must be there , right ??
forgot to say the DC supplies from siglent wich i also own, dont have loop feature in their internal programming either..

SOLVED : dont use PROGRAM no loop feature to find there..
use the less obvious LIST in here you make a list of settings, and the repeat number can go below 1
if you dial it down there, it will go on forever

[5U4GB]

Of course it's not a proper/genuine electronic load, it's not even sold as is (it's officially a "battery tester").

I may have discovered why, I wanted to measure noise from a random SMPS at different power levels and decided to use one of these to do it.  The result was waaay noisier than even a relatively crappy SMPS should be, while a resistor kludge load didn't exhibit the problem.  I haven't looked at the loa... battery tester in any detail but it's possible it's using PWM control from the micro that runs it, or at least doing something that makes it impossible to get a reliable noise reading.