Siglent SDL1000X/SDL1000X-E Electronic Load

[Timpert]

My SDL1020X-E arrived last Thursday. I bought it from Eleshop (http://www.eleshop.nl), and they promised me I could test it for a while, and return it if I didn't like it. So far I tested the DC performance. The burning question is of course: although it's an X-E, does it actually meet X spec? The X model has better specifications on the accuracy of current setpoint in CC mode, and on voltage readback. The specs I refer to are from the March 2019 datasheet, the interesting bits are:

• CC setpoint: (0.1% of value + 0.1% of range) for the X-E versus (0.05% of value + 0.05% of range) for the X
• Voltage readback: (0.05% of value + 0.02% of range) for the X-E versus (0.025 % of value + 0.025% of range) for the X

Apart from this, the X provides one more digit in the readout of voltage (low range only), and in the readout of current (both low and high range), but I am 99.9% certain that this is only a matter of software and not due to a hardware limitation.

Voltage readback was easy to check. My bench DMM has better DC voltage accuracy spec than the X, so a direct comparison should be sufficient. I can put down a long winded measurement report, but I'll just give you the executive summary. With voltage readback, the largest error I found in low range was 1/6 of the allowable error for the X. In high range, the maximum error was 1/3 of the allowed error (for the X) up to 120 Volts, at 150 Volts it was still 1/2. The measurement at 150 Volts was done with the DMM in 1000V range, so there is a good chance that actual error is smaller than the error I found. Conclusion: the voltage readback accuracy on my unit meets X spec. Yay!

CC setpoint was also easy to check. Although my bench DMM does not provide the required current range and accuracy, the current readback spec on the load itself allows testing.  The accuracy spec on current readback is (0.05% of value + 0.05% of range) for both X and X-E models. So if I set a current, and get the exact same value back on readback, I can be sure that the setpoint accuracy is within readback spec. For the X, setpoint and readback spec are identical. That makes sense, because the current setting and current readback share many components (shunt, PGA and whatnot), so actually many error sources are also shared. This means that a current readback that is bang-on the current setpoint proves that the instrument meets X spec, but only a deviation between setpoint and readback value larger than (0.05% of value + 0.05% of range) proves that it does not. The area in between is a grey area. But again, here comes the management summary: in both low and high current ranges, the current readback value and the current setpoint were identical. In low range I tested between 10 mA and 5 A, in high range I tested between 1 A and 30 A. So current setpoint accuracy on my unit meets X spec. Double yay!

I also tested the impedance of the short circuit, which represents the lower limit of usability of the load. For this test, I measured the voltage on the clamps of the load with my DMM using separate leads (i.e. a proper Kelvin connection). From 5 A to 30 A, I found the load presented a practically constant short circuit resistance of 26 milli-Ohms. The spec works out to 30 milli-Ohms, so that's good too.

I ordered some low-value resistors and other hardware so I can test dynamic behavior.  I'll keep you updated on my findings.

To wrap things up: my unit appears to meet X specifications in terms of current setpoint and voltage readback, although it is an X-E. 

I really wonder what Siglent wants with the X. The accuracy spec for voltage readback is (0.025 % of value + 0.025% of range). With a readout resolution of 0.1 mV for the 36 V range, this becomes (0.025 % of value + 90 digits). For the current ranges (0.05% of value + 0.05% of range) at a resulotion of 0.1 mA becomes (0.05% of value +25 digits) for the low range and (0.05% of value +150 digits) for the high range. With the findings above, does this really mean that they charge you around 160 Euros (excluding VAT), just for the privilege of looking at one or two random digits?     

[Timpert]

Siglent states in the datasheet that current and voltage input should not be less than 10% of the full scale. It would be quite unreasonable to test the load at really low currents of 1 to 100 mA and base a verdict on the outcome. So let's do it.

I stepped the setpoint in CC mode from 1mA to 100mA in 1-2-5 steps, and measured the current through the load with my DMM in 100 mA range. Input voltage was 10 V. The resulting current followed consistently, with a maximum error of 0.56 mA below the setpoint. The error remained negative (actual current between 0.34 and 0.64 mA lower than set). This is way better than specified.

I also set the current to zero, with the input on, and measured the resulting current when increasing the voltage to the maximum of 150 V. The load behaved essentially like a resistor with a value of 196 kOhms.

So all in all, DC performance looks quite good. Nowhere did the load go out of spec, and also at lower currents (100 mA or so, also see next post) it seems quite usable. The resistors have arrived, so it's time to to fire up the oscilloscope.   

[Timpert]

I set up the load to draw a 1 ms pulse of 30 A, at max dI/dt of 2.5 A/us. I have 4-wire shunts of 1, 10 and 100 milli-ohms, so I first used the 1 milli-ohms one. But the setup gave quite a bit of EMI in the measurement leads, which could be seen as ringing on the edges. Twisting the sense wires and leading them away from the setup would reduce ringing considerably, but it would not eliminate it. Replacing the shunt with the 10 milli-ohms one dramatically increased the signal-to-ringing ratio, and inserting the 100 milli-ohms one again gave a better signal. A current change of 30 A at 2.5A/us is quite a cookie to deal with...

Anyhow, Siglent specifies the slew rate between 10% and 90% of the set current, so I think that the load actually does this and is quite well behaved. The bit of ringing that can still be seen is acceptable for most purposes, and probably only partially (if at all) caused by the load itself.

I reduced dI/dt to 0.5 A/us, resulting in the third and fourth pictures attached.

During testing, supply voltage was a tad over 13 Volts, so at the current peaks, the dissipation in the load was 300 W (taking the drop across the 100 milli-ohms shunt into account). It didn't crap out, so it seems that the overpower protection has a bit of delay built into it to allow higher peak loads.



   

[Timpert]

I used a 10 Ohms current shunt for the next measurement. First I just set the current to a constant 15 mA. This gave the first scope picture. Although the average current is quite nicely at 15 mA, there is a 7 mA P-P ripple current on top. It looks like 50 Hz rectifier interference to me, as the conduction peaks and reverse recovery spikes are plainly visible. To be sure that it wasn't coming from somewhere else, I repeated the test with a battery as power supply, with the same results. A dynamic test between 10 and 20 mA with a pulse width of 1 ms looks like the second scope picture. I then increased the pulse width to pi milliseconds, to make the mains period a non-integer multiple of the test signal period. See picture 3. I then set the scope to 128 times average, to get rid of the interference and only see the load's own contribution. See pictures 4 and 5. As you can see, the instrument itself shows good settling behavior at these low currents.

This all looks rather dramatic, and it is indeed a bit of a bummer to see that the low range performance of the instrument is limited by what looks like a power supply issue. Capacitive coupling by the EMI filtering in the SMPS perhaps? Without the interference, low end performance would have been excellent, but now it is merely OK...

This needs to be put into perspective. A 200W electronic load is not the appropriate tool for (dynamic) testing of very small power supplies, just like a bone saw is not the appropriate tool for a nose job. A transistor, a few resistors and a signal generator do a far better job when dynamically testing such small things. I feel that I've explored the lower limits of usability, and here they are. The load is thus not ideal for testing the supply on your latest micropower thingamabob, and also not for capacity testing of small button cells. With anything over 100 mA or so, it is fine. But didn't we know that already just by looking at the specs?   


 

[BillB]

Nice review, Timpert!  It sounds like it's worth picking one of these up.     

[bitseeker]

Yes, nicely done review. Thanks, Timpert.

[Timpert]

I'm not done yet... In this episode: CV testing. CV mode is simple enough on paper: the load will draw the current required to keep its output at the set voltage. The load itself is still a current sink, but a feedback loop is wrapped around it to stabilize the voltage. There is potential for trouble here, because the impedance of the device under test now becomes part of the loop transfer function, influencing stability. In other words: with some doodads, the load will not be stable. The stability range is thus a design choice, and the designer can never make it unconditionally stable. So the test below is just a check to see if the designers at Siglent chose wisely...

So first I just set my lab PSU to a reasonable current limit, and set some constant voltages. All was well. I then connected a 4 Ohms power resistor in series with the load, and set the load to 10 Volts (supply voltage was 20 Volts). The voltage across the load can be seen in the first scope picture. It was not what I expected...

Upon checking what could be wrong, I noticed the current range being set to 30 A while only 2.5 A were going to flow in the stable situation. When I set the range to 5 A, I got the second attached picture. Much better! So apparently, the current range also influences the loop gain in CV mode. I switched back to 30 A range, and connected a 4700 uF capacitor in parallel with the load. Now the load was also stable, despite working low in its range (third pic). I tried several setpoints, and the load remained stable.

I have a power inductor (20 mH/12A), but that thing already has a body count in terms of fried electronics. With the above in mind, I didn't want to grant it the pleasure of another kill, so I didn't test stability with an inductive DUT. I think I already know how that would smell.

I also did some dynamic mode testing in CV mode, switching between 5V and 10V. Here, the load only allows steps of 1 second and longer, perhaps to allow for settling. From the scope pictures, the 1 second limit seems to be on the conservative side. DUT was the 4 Ohms series resistor, supplied with 20 Volts, low current range. It works as expected.

So the question is: did the engineers at Siglent choose wisely? I think they did. Most power converters will have an output capacitor, so stability with a capacitive DUT is a must in my opinion. The CV mode must be used with more caution than the CC mode. 

[tautech]

I have a power inductor (20 mH/12A), but that thing already has a body count in terms of fried electronics. With the above in mind, I didn't want to grant it the pleasure of another kill, so I didn't test stability with an inductive DUT. I think I already know how that would smell.

Thanks very much Tim for filling is some gaps of our knowledge and the wise use choices. 

A chat with one of the beta testers confirms some of your earlier findings that are already being addressed at the factory.
Please carry on. 

[Timpert]

Please carry on.

Ok, you asked for it. CR mode testing in this post. Here, I was primarily interested in how quickly the load would react to a voltage change. I could make the load one half of a resistive voltage divider, and it would settle in all current and voltage ranges. So I used a resistor of 16 Ohms, and made a voltage divider with the load also set to 16 Ohms. The load was the bottom resistor. I used the 0.1 Ohms shunt to measure the current through the load. The whole thing was fed with just under 70 Volts, the divider was biased with a good 2 A. Through a capacitor, I injected AC voltage from an audio amplifier to the midpoint of the divider. In this configuration, the amplifier sees a load that it can easily handle, while the load will have a tough job in dealing with the voltage changes.

First test: 50 Hz sine. See picture 1. Pink trace is the AC component of the load voltage (2V/div) while the blue trace is the current (100 mA/div). I think this looks excellent. The slight phase shift between current and voltage is unavoidable because the instrument sets the current based on a previously done voltage measurement, so it will always be a bit behind.

Second test: 400 Hz sine. See picture 2. The current has clearly dropped a bit, and you can actually see the steps in which the load adjusts the current in order to follow the voltage. This happens at about 12 kHz. Although the sine is looking a bit rough due to the discrete time current adjustment, the load is still behaving quite well.

Third test: 400 Hz sawtooth. See picture 3. It still works, although a wiggle at the reversal of dV/dt is visible. I presume this is due to a small capacitor connected at the output or due to some EMI filtering. I like how the load recovers quickly, and it remains stable.

Fourth test: 400 Hz square wave. See picture 4. This will be a tough cookie for it, and the influence of the coupling capacitor is visible on the pink trace. With this kind of signal, I find every outcome where the load remains stable to be good. There is a particularly nasty ring after each edge, caused by parasitic inductance and capacitance, so the load has every reason to go nuts. But it doesn't. It recovers quickly after each edge, and cheerfully starts adjusting current.

That inspired my to make a bode plot (picture 5). A resistive divider with 0.1 (the shunt) and 16 Ohms (the load) gives a response of -44.1 dB. The -3 dB point appears to lie somewhere around 300 Hz. So I think it is safe to conclude that CR mode is usable up to 300 Hz, and shows rather graceful degradation of its performance above that. The phase jump you can see at the end of the measurement (the plot goes to 1000 Hz) is probably due to the scope no longer being able to measure the delay between input and output properly.

So all in all, CR mode appears to work pretty well.   

[Timpert]

I could go on testing forever, but that would go a bit far for "just" a review. So I stop here. Some closing remarks.

Build quality is good, it feels rather solid. Threads of the power terminals are m6. The fan control is worth mentioning: upon power up, the fan is silent. The fan starts running when the internal heatsink heats up. It also starts running immediately when the input current exceeds 6 A, and then it proportionally ramps up with current. So the shunt is kept cool, even when the total dissipation is low. Nice touch!

The user interface has some rough edges. Having to press enter after each change of a numerical field is awkward, because (especially when editing a list or a program) it feels more natural to enter a number and then use the arrows to navigate to the next field. If you don't press enter after typing a value, the value you just typed is lost. Especially entering the IP address has a rather unnatural flow in it: entering 192.168.1.4 didn't work using the dot on the keypad. So it had to be <192><enter><arrow><168><enter><arrow> and so on. But I guess that future firmware releases will also address UI ergonomics.

After all this testing, I decided to keep the load. I hope it is possible to make CV mode stability a little less critical through firmware, and that the UI ergonomics are improved. But apart from that, I'm quite happy with it as it is now.   

[StillTrying]

I keep wasting time looking at this thread because of its bad subject changes.

[Timpert]

I had a play around with constant power mode, and I found something that IMHO should not happen:

So when the current gets too high the load switches off. This could cause some nasty surprises. It would be better when the load would simply remain at its max current and start beeping to signal that the power cannot be maintained.

I found that in most cases when a limit is exceeded, the load is dumped. I'd rather see it that the load tries to stay ON as much as possible, and beepwarns when the setpoint cannot be maintained  due to exceeding some range. This is my suggestion for a firmware update.

[Timpert]

I found a bug in the remote interface.

When I issue the *RST command, the device doesn't only reset the measurement system, it seems to perform a complete factory reset. That includes the TCP/IP settings, which are reverted to use DHCP (I use static IP addresses for my instruments).

A workaround would be to set my router to always give it the right IP address, but VISA doesn't like it when the connection is forcibly closed by the instrument after the *RST command. So right now, *RST cannot be used without breaking program flow.

I am using Pyvisa 1.9.1 and pyvisa-py 0.3.1. Instrument FW version is 1.1.1.19.

[bitseeker]

Yeah, that's a bit too much reset when the instrument's remote control configuration gets wiped out. No more automation.

[alexvg]

I've taken some pictures of my device before changing the output connectors to high current 4mm banana.

[tautech]

Thanks for the pics alexvg. 
Can I ask if it's the 200 or 300W model ?

[nctnico]

@Timpert: what did you use as a source to create the voltage waveforms?

[alexvg]

I've bought the model "SDL1020X-E 150V 30A 200W".

[tautech]

I've bought the model "SDL1020X-E 150V 30A 200W".

Ah, OK.
I wonder if the 300W model has the 4th shunt resistor. 

[Timpert]

I wonder if the 300W model has the 4th shunt resistor.

I'd be quite surprised about that, because the current ranges are identical for all models, and the short circuit resistance spec is the same too at 30 mOhm. Does anyone have a 300W model? The shunts are easily visible when you look through the air inlet in the right side of the instrument, so you can count them without breaking the seal.

@ntnico: I'll draw a schematic somewhere this week. I guess you mean the voltages from the CR test I did?

[nctnico]

Yes. I see now that you used an audio amplifier.

[plep]

Guess this is the "hardware issue"/delay

[flash2b]

I've taken some pictures of my device before changing the output connectors to high current 4mm banana.

I have ordered these as replacement for the original screw cap:




The red one has an 4mm banana female socket. I mainly use the load for small currents, so this is easier for me.

[tautech]

I've taken some pictures of my device before changing the output connectors to high current 4mm banana.

I have ordered these as replacement for the original screw cap:




The red one has an 4mm banana female socket. I mainly use the load for small currents, so this is easier for me.

Nice, what do the look like on the underside ?
Can you still insert a lead with these nuts wound full onto the stud ?

And a link to your source please.
TIA

[aheid]

And a link to your source please.

Yes please! Not elegant that spades or raw cable are the only options for connecting a source.