R&S NGE100 voltage jumping around

[justanothername]

Hi.
I have here a brand new R&S NGE100 from work. My colleague gave it to me to verify. He said that the voltage is slowly jumping around..
I can confirm this, look at the chart I recorded. Anyboy knows whats going on here? Should I send it back?

[Alex P]

Is it set to supply 4V? And what is the current? You don't have EasyArb or EasyRamp enabled?

[justanothername]

Supply is set to 4V. No current (directly into the multimeter). No arb or anything.

[Cloud]

What if you try to load it up a bit and measure then?

[justanothername]

Ok I connected a 390R ceramic resistor to the channel and the jumpy behavior seems to go away.
But look at the charts! It is a really terrible noisy and drifty power supply!
with 11mA@4V it has a range of 38.7mV!! This PSU has a price of over 1000€!
The jumpy drifting on light loads is also unaceptable if you ask me!
What should I do? Do you think this device is faulty?

[maginnovision]

You're certain it's the PS and not the meter? If you've had same results with another meter and compared your measured stability with the spec in the datasheet and it's outside of that tolerance... Then something is wrong.

[ogden]

What should I do? Do you think this device is faulty?

It is faulty when does not meet the specs:

Voltage noise specification:
typ. < 1.5 mV (RMS)
< 20 mV (peak-to-peak) (meas.)

There's catch: noise specified for frequency range 20Hz-20MHz. If during 0.05s voltage change does not exceed 20mV, then everything is fine with your supply

[YetAnotherTechie]

I used a 100ohm resistor and got the data in the picture.
I tried to match your scales as close as possible, BW is limited to 20Mhz, the blue line is a low pass filter of 100Hz.
According to this, mine is within spec, is it reasonable? i don't know, I've read here in the forum that HP's are much quieter.

[RBBVNL9]

But look at the charts! It is a really terrible noisy and drifty power supply

Interesting! But your post does not say much about the measurement setup. How confident are you that the noise and spikes you see are actually coming from the NGE?

In a recent threat https://www.eevblog.com/forum/testgear/seeking-psu-advice-rohde-schwarz-nge103b-com3-or-keysight-e36313a/msg3227650/#msg3227650, several people discussed the the NGE series, several of its shortcomings, and also shared ripple and noise measurements. Looking at these, receiving input of others, and measuring several PSU’s myself, i learned how critical the measurement setup is. A regular scope probe with a 4cm ground lead, for instance, can already result in picking up a lot of induced signals, and do what you then see may not come from the PSU but from elsewhere (like dimmer circuits closeby). Using just two banana cable leads from PSU to meter (instead of coaxial cable) will even make things much worse.

See also the below video of Dave, where he advises a differential probe setup for PSU ripple and noise measurements.

Voltage noise specification:
typ. < 1.5 mV (RMS)
< 20 mV (peak-to-peak) (meas.)

The NGE series are certainly not having the lowest ripple and noise specs around. Many power supplies in this price range (and also cheaper ones) have considerably better specs. But measuring whether a NGE is within its specs is not a trivial matter…

[justanothername]

I've redone this test with a test setup as proper as possible (see attached pictures).
I've verified the performance of my multimeter before testing with the PDVS2mini voltage source. As you can see, the multimeter works fine. All units had more than 30 minutes to warm up
I've made recordings for over 10 minutes with and without a light load of 390 Ohm. It is still very bad, having a voltage range over 30mV.
I can do tests with an oscilloscope, I got a Picoscope4000 for noise measurement. But I think this alone proves that the unit is defective or what do you think?

Edit: I hooked it up to the picoscope4000. It shows >70mV peak-peak noise. With an USB isolator (necessary!) i could measure the noise. Yes, within a window of 50ms, it is less than 20mV noise. But it looks not good in longer time frames. Note that the measurement is ac-coupled, so voltage changes would look like spikes.
In comparison, the PDVS2mini in battery mode shows 1.6mV p-p noise.

Edit2: attached some faster readings with the sourcemeter as 100mA load. ~ 29mV range.

[maginnovision]

Smu shows it's in spec, right?

[justanothername]

Smu shows it's in spec, right?

In my opinion not. It has a peak-to-peak-range of ~29mV. A problem is that this is not defined in the data sheet since the noise measurement is from 20Hz upwards.
I'm still waiting for a reaction of our local representative. But if they tell me it is in spec and I should f*ck off, then people beware of this PSU!

Edit: Lets wait what they say.

[maginnovision]

Well that's fine but 1 unit is not representative of the product in it's entirety. It's possible that your unit is faulty. I'd rather see more examples before condemning it. It's entirely possible that it's not suitable for low noise or that unit isn't performing well.

[RBBVNL9]

But I think this alone proves that the unit is defective or what do you think?

I find that hard to say. Let us remind us again what the R&S specs are:

Voltage ripple and noise (20 Hz to 20 MHz):
-   typ. < 1.5 mV (RMS)
-   < 20 mV (peak-to-peak) (meas.)

Firstly, like noted before, good noise and ripple measurements are difficult to make. Potential issues are:

• Induced signals resulting in common mode noise (the reason some are suggesting differential probes to address this)
• Mismatching loads resulting in reflections and ringing; note the PSU output impedance is in the mOhm range (milli), and that of the scope is typically 1 MOhms (Mega) (there are impedance matching solutions to address this)
• Ensuring measurements are under different voltage and load conditions that those used for the specifications (unfortunately often not provided by the manufacturer),
• Impact of temperature
• Possible impact of other devices used in the measurement chain. For instance, an electronic load may be cause of problems (but may be the only practical way to measure a PSU a specific series of voltages and loads). Or, if a wire-wound resistor is used as load, its parasitic inductance may have an effect

Your new measurement setup is probably better than your first one, bt may not address all above issues.

Having that said, lets now look at your Picoscope screenshots. I do not know that instrument well, but the screenshots you shared seem to show only peak-to-peak measurements, not RMS measurements. That limits us in answering your question. I also cannot see whether the bandwidth of the measurement was limited to 20MHz (which is the way the specs are specified). Especially because there are switching elements in this PSU, this bandwidth may really make a difference. The peak to peak measurements on your screenshot seem to be close to the specs, though.

Yesterday, I did some more ripple and noise measurements on my NGE103B (and acknowledging these are probably far from perfect for reasons mentioned above). My results are roughly speaking in line with those posted by Fixpoint a short while ago, and those posted by YetAnotherTechie in this threat. Depending on the voltage and load, my measurements are close to the manufacturer’s specifications: not much better, not much worse either.

Despite the possible imperfections, let me share my measurements:

First measurement uses a 19Ohm resistor as a load, PSU set at 10V, and a resulting 0.54A current flowing. I do this to be able to see whether the electronic load, used for the following measurements, makes a lot of difference. PP and RMS results are close to R&S specifications. (See pictures Ripple 1 and Ripple 4)

The second measurement is basically the same, but now the 19Ohm load is an electronic load. The waveform is not identical (may have several reasons) but the resulting PP and RMS results are again close to R&S specifications. Using an electronic load instead of a resistor does not seem to affect findings a lot (see picture Ripple 2). (Note that Dave, in a video above in the threat, finds that the electronic load is a source of measurement problems in a similar setting.)

Thirdly a measurement at full load (10V, 3A, which is the maximum for that voltage because of the 33W per channel). Now we go above R&S specifications by a factor of 2-3, but again, I cannot be confident the measurement is perfect so cannot claim lack of compliance. (see picture Ripple 3)

In comparison, the PDVS2mini in battery mode shows 1.6mV p-p noise.

Yes, I see that is lower than the noise you measure yourself from your NGE. But it may not be a fair comparison. The NGE is a power supply, designed to significant current over a wide range of voltages. To do so in a compact housing R&S chose to use a switching regulator (offering efficiency and compactness) followed by a linear control circuitry (for stabilisation). Having switching elements inside (like many modern designs do) inevitably will result in some noise on the outputs. Comparing that to a device solely design to be a close-to-perfect voltage source may not be fair.

IMHO, the bottom line is that the NGE100 series may not be one's first choice when shopping for a low ripple PSU, its specs in this regards are certainly not stellar. But whether a specific unit meets those specs is not that easy to measure, and the measurements of NGEs we have seen up to now seem to suggest they are close to or meeting the specs, despite probably measurement imperfections.

[justanothername]

@RBBVNL9 I have problems in a complete different time base than you. In my case it is drifting, over tens of seconds or minutes! My logging setup with the HP 34401A is one measurement per second.

[ogden]

@RBBVNL9 I have problems in a complete different time base than you. In my case it is drifting, over tens of seconds or minutes!

Exacty. You are trying to prove that your supply does not meet specs of unspecified long term drift

[justanothername]

Yes that is why I need the help of the community to have a measurement to compare to. I need anything to put this back to R&S since yes, I made the decision to upgrade to this PSU und I myself have to fix this problem, my coworkers threw this unit back to me. At the moment nobody wants to work with this and its B2B, so no returns.
But I'm sure that this is not a normal behavior for a power supply that expensive.

[RBBVNL9]

@RBBVNL9 I have problems in a complete different time base than you. In my case it is drifting, over tens of seconds or minutes!

OK. Maybe I did not realise this well enough as you talked about noise (with I associate with short time spans) and this threat is talking about meeting the published specs (and the only specs in this respect are those on ripple and noise and given for 20Hz and higher only).

If the issue is longer term voltage drifting, then I really would be fascinated to see whether this is somehow related to the design of having a switching regulator first and a a linear control circuitry next. I don't know the details of the NGE design, but I wonder whether it involves some steps in the process, and these could possibly explain the behaviour you seem to notice. If so, then such issues may occur at certain (threshold) voltages and not at others. This all is pure speculation, however.

Note that the measurement is ac-coupled, so voltage changes would look like spikes.

I guess that if we are to measure multi-second voltage drifts, AC coupling is not the way to go (especially if we don't have precise information on the ac decoupling characteristics of our measurement device). Rather DC measurements, and perhaps also ensuring that unwanted higher frequency AC components are filtered out effectively. Perhaps even charging a relatively large capacitor when longer time spans are of interest.

If you know what you're exactly looking for, I'd be willing to make the same measurement here using the same set-up and share with you what I get.

[justanothername]

If you know what you're exactly looking for, I'd be willing to make the same measurement here using the same set-up and share with you what I get.

Thank you for that offer. I guess it would just help if you connect a 5.5 digid DMM or better to the PSU and measure the voltage with 1 reading per second over a span of 2 minutes or longer.
Without load and optionally with 1mA and 100mA load. I have this behavior at all voltage ranges, but in this case the measurements are all done at 4V. I can repeat the measurements with any voltage if necessary.

[ogden]

But I'm sure that this is not a normal behavior for a power supply that expensive.

Nah. In this price range you get similar specs (20mVpp) from Keysight as well. Even worse - just 1 channel and much lower current capacity supply (E36104B). If you want low lower noise, then look for something like E36313A or NGL202. Also read the funny datasheets

[RBBVNL9]

@justanothername

I now did the measurements you asked for. They are done with the NGE103B set to 4V output and the following conditions:
-   No load
-   1mA load (using a 3.9 kΩ resistor as load)
-   100mA load (using three 120 Ω resistors in parallel as load)

Measurements are done with a Siglent SDM3045X DMM. That is only a 4.5 digit DMM, but given the voltage differences we are interested in, it seems fit for this task. I also repeated the measurements with a set of 3 new AA battery as voltage source, to check there is no short-term drift in the DMM itself, or inaccuracy in repeated measurements (there is not).

This DMM also has an AC filter. According the manufacturer: “The analog filter of the Multimeter can be used to reduce the influence of the AC component in DC measurement. Generally the filter may not be required, but sometimes it can improve the DC measurement. For example, if the measured DC electrical source has a big AC ripple, the analog filter can reduce the influence.”

Every measurement I did consists of 120 samples, with an in-between delay set to 1.0 second. Because the measurement itself also takes some time, the total time span is a bit longer than 2 minutes (2 min 26 seconds to be precise).

The data file is attached at the end of this post, as well as some graphs.

My observations are:
-   The AC filter does not seem to affect measurements a lot, but it does seem to need some time to stabilize. So for PP calculations, I’ll ignore the first 20 samples (where this stabilisation takes place).
-   Overall, in all my measurements (of 2.5 minutes each), the maximum PP difference is 20 mV or lower.
-   In two cases, I do observe a voltage step (see the red arrows in the graphs), both of about 10mV. Both times this happens during measurements with the AC filter on, but I take that as coincidence (N is way too low to make any inferences).

As discussed above, R&S only has published specs for ripple and noise for 20Hz to 20MHz, and not for slower voltage drifts. During the measurements I did (which are not infinitive or exhaustive), the average drift over 2.5 minute periods was 12mV, and the maximum drift I saw was 20mV. I guess this is good enough for my purposes, but others may have other needs.

Yet, the steps I observe are fascinating, and they would be compatible with the idea of the design (a switching regulator followed by a linear control circuitry) indeed has discrete steps when doing its work. But of course this is speculation, not evidence.

[rf-messkopf]

If the issue is longer term voltage drifting, then I really would be fascinated to see whether this is somehow related to the design of having a switching regulator first and a a linear control circuitry next. I don't know the details of the NGE design, but I wonder whether it involves some steps in the process, and these could possibly explain the behaviour you seem to notice. If so, then such issues may occur at certain (threshold) voltages and not at others. This all is pure speculation, however.

I don't know whether this is of interest in the present thread, but by coincidence I noticed a similar jumping of the output voltage on an Agilent E3646A. See the attached picture: The output is set to 5V and directly connected to the multimeter (no load). It does jumps of about 500µV. When I noticed that I just shrugged my shoulders as it is within spec (<0.01% of output voltage + 3mV), even though it's not exactly nice.

The Agilent E3646A has no switching regulator, and the jumps are present on both channels. I haven't seen any jumps on a R&S HMP4040 that is sitting next to it, which does have a switching pre-regulator, but is somewhat more noisy than the Agilent.

[justanothername]

Thank you for the measurements RBBNVL9!
Your measurements seem to be way better than mine!
You have 7.7mV at no load min to max, the worst value is 11mV at 100mA. This is the performance I'm expecting.
My NGE has around 30mV at all loads to 100mA, so it is definitively worse! I will return the device and lets see what happens.

[Fixpoint]

I am skeptical that all this is really meaningful.

(a) Firstly, keep in mind that your measurements are not necessarily comparable with each other because you don't have the same conditions. For example, the performance of your power supply is influenced by your mains characteristics. If somewhere in your house a fridge, a vacuum cleaner, or an LED lamp turns on, this could change the behavior of the power supply. Some of you may have a very clean mains, others a very noisy or "ripply" one. If one of you measures "30 mV" and another "10 mV", what does that mean? I don't think that it means much as long as we don't know all the parameters. In my opinion, performance measurements must be done under very well-controlled conditions, otherwise they are of very limited value because we don't really know how to interpret them. We can only use them as ballpark indicators.

(b) Secondly, keep in mind that we don't buy equipment "just because" (except we are collectors) but for using it for something. I am using the NGE103B on a daily basis for hours (and yes, I am doing precision stuff with VCOs), it does its job very well, and I never think about 10 mV or 20 mV ripple just because I have better things to do. So, is all this talk really worth the time? Just build something ...

[justanothername]

Fixpoint, we bought this PSU as replacement for the old analog Hameg 3ch PSUs. As such it was presented to us by R&S when we visited them. Whatever my coworker was trying to do, he searched for an error in his circuit awhile before pinning it down to the new PSU and throwing it back to me. Hes is using the old one again.
And I don't know why a properly built PSU should jump 30mV around at 4V constantly, even when mains is unstable (which is not, of course). I just needed to know if others have the same flaws, seems not, so I will send it back!
I would overlook this when it would be a cheap aliexpress PSU but not for a >1000€ unit.