RoadTest: R&S NGP814 Power Supply
Evaluation Type: Power Supplies
Did you receive all parts the manufacturer stated would be included in the package?: True
What other parts do you consider comparable to this product?: Various other power supplies from Keithley, GW-Instek, Aim-TTi, B&K Precision, Keysight, Rigol, Amrel, ELC ore home brew, mainly not containing four channels. Or from R&S the HMP4040.04 which is a four channel PSU.
What were the biggest problems encountered?: Despite the unit has not had the carrier's attention to respectfully deliver the package as it deserved, I found no damage to it.
document version history:
First, I want to thank Element14 and Rohde & Schwarz for giving me the opportunity to review the NGP814 Power Supply. I am really happy that my applications was selected. This is quite an impressive power supply.
For equipment like this, with a lot of interest, 44 applicants in this case, it is not easy to write a convincing application. To be transparent, and also to inspire other members, for future roadtests, here are the relevant parts of my application:
(c) Why did you apply for this particular roadtest?
A power supply is key for all electronics projects. In the past all power supplies were linear, meaning a transformer at 50 or 60 hz transferring the mains voltage to a lower value. Currently the majority of PSU's are switching, which is much more efficient and as such better for the environment, but it also comes at its costs. I do have a lot of experience with all kind of power supplies for my projects, and really like to test one more thoroughly.
(d) What is your testing procedure or project plan (Be as specific as you can)?
- As a ham radio amateur, I'm very interested in well performing power supplies since bad designed models often emit noice that disturbs reception of weak signals. For the R&S NGP814 I don't expect problems as explained in the video below. Nevertheless I would rather like to test this using my ham radio receiver, as well as with the R&S FPC1000 for which I was a lucky roadtester (3GHz Spectrum Analyzer - R&S® FPC1000 - Review).
You don't have permission to edit metadata of this video.
(Courtesy from Scubasky)
- Recently I came across a very interesting article about running the shortest path algorithm on an analog computer that simulates the total network as strings of leds. (Analog Computing: Shortest Path | Analog Zoo ). For testing a large network you need quite some leds, and as every led has a forward voltage of 2.2V, this needs a high voltage power supply. This power supply also needs to be used as current source. On the website mentioned above a current source is built, but when using the the R&S NGP814 in current mode no additional electronics are needed. Furthermore the maximum voltage of the the R&S NGP814 with four channels in serie is 192V. This makes it possible to test large networks, up to 87 LEDs in the shortest path. Would be great to be able to perform this experiment.
You don't have permission to edit metadata of this video.
(Courtesy from devttys0)
- In addition to the above mentioned experiments I will look at general features, such as user interface, fan noise, interface possibilities, software (specifically on macOS), robustness, safety, weight and everything else that pops up when playing with the device.
Planning: When selected I expect the device to be shipped before the end of September, this means that the review period is until the end of November. Du to Covid-19 this might be delayed, but that is not a problem for me.It will give me enough time to perform the planned experiments. Furthermore I don't have to wait building the LED maze until the device is shipped. Detailed planning:
- building a LED maze: 5 days (duration, not building time )
- playing and testing with the shortest path application: 10 days
- measuring and listening for RF emissions: 10 days
- testing general features: 5
- testing connection and software: 10
- wrap up and writing the review: 10 days
- extra time for unforeseen circumstances: 10
- In total: 60 days
Pictures of the unboxing are generally not that informative, but I can assure you it is this time. This is what the mailman delivered to my front door one day:
Unfortunately, the carrier did not pay attention to deliver the package with the respect it deserved. As you can see the power supply together with the packing material was put in a plastic bag. The power supply loose on the bottom of the box with the foam loose on top. It looks like someone (customs???) took it out and did not put it back properly. I expect originally the unit was in the plastic bag and together firmly placed in the foam pad as was the case with my fellow roadtester waleedelmughrabi .
In any case, given that the cost of transportation was over $ 700, it is a shame that the equipment was handled that way.
The NGP814 comes with a small Getting Started guide. It is hilarious to read what is written on page 14:
I carefully inspected the unit and luckily there seems to be no visible damage.
Next to the getting started guide, the box contained Basic Safety Instructions, a calibration certificate and a reference for packing material.
Power cords with plugs for the different regions were also in the box, and plugs for the rear pannel connectors.
The video of baldengineer on the readiest page already gives a good impression on the instrument. What stands out is the large size and heavy weight. The footprint is 360 x 440 mm (W x D).
For me this is a bit inconvenient as my shelves are 400 mm deep, so the power supply doesn't fit on my shelves and will consume place on my desk. The height is 100 mm which is quite nice, but for me a device with half the depth and twice the height would be more practical.
The picture above is from baldengineers video review. baldengineer mentions the weight as 17.6 pounds, which is 7.5 kg in metric units, this was a mistake with regard to the type number as stated in the specifications, where the weight of 2 channel power supplies are listed as 7.5 kg and the 4 channel versions as 8.0 kg.
Still for the NGP814 this is not true. As waleedelmughrabi already mentioned, the unit really is almost 10 kg (22 pound).
I tested this by measuring the weight with a bathroom scale. Since this scale may deviate a bit in the low range, I also measured the difference in weight between myself and myself carrying the power supply. Both measures confirmed that the weight is really close to 10 kg.
The front feet can be folded out to set the instrument in an inclined position.
The output connectors are Red , Blue and Black for the sense input. I agree with baldengineer that this is not logical. All power supplies that I'v seen up till now (and that is a lot) had Red and Black .
The rear panel contains the AC inlet with a power switch, USB and ethernet (LAN), real pannel connectors for all output and sense terminals, connectors for analog input and digital I/O, place for an optional GPIB bus and a Kensington lock for protecting the power supply against theft.
Although the plugs for the digital I/O and analog input are provided, the option for using them is disabled by default. In the menu these options are greyed out.
When looking at the R&S website there is also a software option for wireless LAN remote control. The WIFI module itself is already installed. I don't have an idea on costs, as they only will be provided when asking for an official quote.
As can be seen in the pictures above, there is a power switch on the rear and a power button on the front. Right after unboxing (and inspection for any damage) I turned on the device with the switch on the rear panel. The device immediately started to boot, and after a few seconds it was fully functional. Then after upgrading the firmware this behaviour changed. after powering up with the rear switch, the device starts and immediately shuts down, while lighting up the front power button in red. After that it can be started with the front button which lights up in red.
As you can see in this video, the first power/shutdown sequence takes 10 seconds, while the second one is 13 seconds. Not bad, compared to other equipment but this booting to stand-by mode looks a bit weird. On other equipment, for instance the R&S FPC1000 you can change this behaviour in the settings, but on the NGP814 I could not find a setting for this.
As already mentioned, one of my first actions was to upgrade the firmware. Latest firmware can be downloaded from the R&S NGP800 Product Webpage.
The update procedure is easy and well documented.
Here is a screenshot of the device information before the update, as you can see version 02.004.
When looking at the release notes interestingly this version is never shipped to end customers!
The latest version at the start of my review was 2.011, as can be seen in the next screenshot:
Compared to my old AMREL LPS 301 power supply the NGP 814 is rather quiet. As can be seen in the screenshot above, at startup the fan works at 20%.
Interestingly during my tests with 150W halogen lamps, the fan stays at 20% for a long time. After turning it on for a while the temperature of channel one gradually increases to 60 degrees, with the fan still at 20%.
When the temperature reaches 61 degrees, the fan speed is increased to 28%.
When the temperature reaches 62 degrees, the fan starts to alternate between 28% and 36%, which sounds a bit strange. It looks like it oscillates around this value, the temperature doesn't increase anymore. Isn't it an idea to build in a bit of hysteresis?
Wouldn't it also be possible to set the fan even lower when not much power is supplied, making the device even more quiet?
Continuing my experiments, the fan occasionally had a somewhat unpredictable behavior. For example during my experiments for ripple measurements; after supplying 10W to the lamp for a long time, the fan speed increased, and a little later it increased again, then slowly oscillated between 28% and 36%. Then I increased the voltage to supply 50W to the lamp, then the fan speed decreased again to 20%. Apparently the power supply dissipates more energy at lower voltages. This is certainly the case for a linear power supply, apparently for the NGP814, which is a switched mode version, this is also the case to some extent.
The R&S NGP814 power supply provides a maximum output power of 200 W for each channel, so 800 W in total for 4 channels.
Channel one and two are rated at 32V and 20A max, while three and four can provide 64V and 10A max.
Note that although the max ratings of voltage and current can not be set at the same time, since that implies an output of P=V*I = 32*20 (64*10) = 640W.
Here the following graph comes in, showing the limiting curve for voltage and current for each channel.
The power supply itself prevents you for entering the wrong settings. For instance when the voltage is set at 20V, the current can not be set higher then 10A.
Be aware that in the graph below, which is figure 5-11 from the manual, the channels are mixed up. Blue is for Ch1 and Ch2, Red for Ch3 and Ch4.
This is fixed in firmware version 2.015.
Lets first play a bit with the device, before diving into the planned activities for the roadtest.
You need a hefty load to test an 800W power supply. Unfortunately, I didn't have a fancy electronics load. That's why I use some power resistors (5ohm 25 Watts) on a heatsink and halogen lamps of 12V, 75W and 21V, 150W that I mounted on a provisional frame, along with a fan to cool them down.
First test with the 150W 21V halogen lamp. Lamp on channel 1 and fan (12V) on channel 2.
Voltage is set on 21V, and current at maximum, which is 9.524A in this case ( 200W in total).
When switching the output on, the numbers turn red, which indicates that current limiting is active. In less than a second the numbers turn green and the display shows the Voltage (21V), current (7.224A) and power (152.4W), and their minimum, maximum and average value. Also the total amount of energy is displayed.
Another nice feature is that these values can be displayed in a graphical view. From these screens you can see that after turning on the lamp, it's resistance is apparently low, so the current is limited at the max. After less then a second the filament is warmed up which increases the resistance and lowers the current to its steady value.
The graphical view is very informative, but it has some limitations, for real measurements you need an oscilloscope.
The timebase of the graphical view is fixed at 5s/div (one minute for a full screen) and the vertical scale is automatically adjusted.
In the screen above the power output is added to the graph. Unfortunately the automatic scaling is wrongly set at 10W/div, which is far too less for the 150W output
This problem is fixed in firmware 2.015:
But now the zero - origin of the graph is at 50% of the y-axis, which is weird, as we never expect negative values from the power supply.
The NGP814 makes it also possibility to connect the channels in parallel to increase the current, or in serial to increase the voltage.
Since I was curious how much current the lamp draws at startup, which value was limited to the max at one channel, I put channel one and two parallel. Before doing so I enabled the tracking for these two channels, such that changes made on one channel are applied to the tracked channel.
Here is how this looks like:
Not the best quality connectors are used for the lead between the two channels, so the majority of the current comes from channel 1.
Interestingly the graphical view shows that at the start both channels clips at the maximum value, which means that almost 20A is drawn from the power supply.
After exchanging the wires, channel two becomes the most dominant one. In order to try to share the current a bit between the channels, I limited the current to 4A. Interestingly the main channel now jumps to constant current mode while the voltage drops a little bit and the remainder of the current needed comes from channel 1.
I suspect this behaviour is due to the resistance of the extra lead between channel 1 and 2 and from the not so good connectors. I tried to solve this by using the sense connections, but in that case the main current providing channel is the one with the 'best' sense connection to the lamp. In the mean time it's a mess of wires on my desk.
With the Ramp function, the power supply provides the operating condition to ramp up the supply voltage within a defined timeframe up to 10 s, to give the load a soft start.
Here you see the result for a 1 s and a 2 s ramp.
with one second the current is still going to the max, but with two seconds this is not anymore the case. Looks like the proper way to power up these halogen lamps.
When going to the extreme of 10 seconds the graph becomes:
For setting the ramp time the device shows a number input screen equal to the screen for setting current or voltage values.
Unfortunately the button does not work in this case, contrary to setting the current and voltage values.
The first topic I planned to have a look at for this roadtest is disturbance by the power supply of weak signal reception for instance for ham radio signals. The NGP814 power supply is based on a primary switched-mode regulator with power factor correction. This concept allows the instrument to achieve highest accuracy and lowest residual ripple. So let's first have a look at the ripple. According to the specs the ripple should be less than 3 mV RMS, or less than 30mV peak to peak. So I measured the ripple using an oscilloscope while varying the output power from 10 - 150W by adjusting the voltage over the halogen lamp. The screen dump of the scope for 10 and 150W shows:
As you can see the ripple is a bit more as in the specs, also more than measured by baldengineer .
Here is a graph of all values measured.
As I was curious about the NGP814's ripple compared to other power supplies, I also measured two cheap 12V adapters, an Amrel linear power supply, the LPS301, and an old homemade linear power supply. The first 12V adapter I tested was a disaster, it has very sharp peaks of 7V and an RMS of 128mV, when tested with an output of 5W.
The second one was much better, 50mV peak-peak and 9 mV RMS at 5W, but still give 5 times more ripple than the NGP814 at 5W.
The disaster power supply, which was used as adapter for an external harddisk, is rated at 2.5A, while the other is 2A. As I like to throw the disaster away and replace it by the other, I need to know whether 2A is enough for my harddisk. To test this the harddisk was connected to the NGP814, and I logged the current and voltage while performing a backup. This is a nice feature of the NGP814. From the menu I enabled logging, with an interval of 1s to the internal storage. After the backup I stopped logging, and copied the csv logfile to an external USB stick and plotted the data in a spreadsheet.
As you can see the current was never higher than 1.6A, which proves that I easily can replace the bad adapter with a 2A version.
The linear power supplies does a great job as can be seen in the pictures below. As you might know a linear power supply directly transforms the 230V 50 Hz AC (Europe) to low voltage, using a bulky transformer, this means that there might be 50 Hz ripple on the output. In both cases I could not measure 50 Hz ripple above the signal which was picked up from the surroundings by the test wires.
The NGP814 is equipped with a number of interfaces to connect it to the outside world:
Regarding software R&S provides HMExplorer, unfortunately only for Windows PC's and not for macOS or Linux. I didn't test it.
There is a nice option though, which I could not find documented in the manuals or the R&S website.
If you connect the NGP814 to your local area network, with DHCP, you can connect to it from a browser.
The web address is rs-instrument.local (or another one if you change that in the device settings).
This opens a webpage with several options:
You can change network settings and a password to access this webpage and also go to the documentation. The documentation page contains a link to the R&S NGP800 Product Webpage, and to the Getting Started Manual, from which the latter is locally stored in the power supply, so you can also read it when your network is not connected to the internet.
the most interesting is the second menu option 'SCPI Device Control':
From this page you can send SCPI commands, but also view and download the screen data, which is very convenient when working on a roadtest review.
With the SCPI commands, which all are well described in the user manual you are able to list and look into the files in the internal file system:
(Input command is denoted with <- and response from the power supply with ->)
<- DATA:Data? '/int/logging/log-20201121T092609.952.csv'
While it is possible to download data from the device this way, I strongly recommend R&S to build a file manager in the device's web interface in future software releases. With this file manager, files in the internal file system can then be uploaded and downloaded via the browser.
One of the consequences of the high frequency ripple, caused by the switch mode circuit of the power supply is that it can generate high frequency (HF) noice that possibly disturbs radio communication. This is well known. In the user manual, on page 16 and page 9 of the getting started guide, R&S warns for the Risk of radio interference.
The CISPR 32 standard is dedicated to multimedia devices, to me it is a bit strange that a high value lab power supply is tested with multimedia device standards. For more information follow this link: https://interferencetechnology.com/cispr-32-what-is-it-why-was-it-written-and-where-is-it-going
As one of the goals of my roadtest plan I would like to investigate this further. Unfortunately I don't have a faraday cage, so I'm not able to test the device against the standards. I do have radio receivers and a spectrum analyser, so I could do some qualitative tests. To give an impression of the performance I compared the results with the two switch mode power adapters mentioned in the previous paragraph. In this video the spectrum picked up from the power lines with a small piece of wire is shown on the spectrum analyser. In order to get rid of other disturbing signals I saved the full measured spectrum shortly before I turned on the power supply. On the display the difference of the spectrum after powering on the device with the saved spectrum is shown.
Wat stands out in the video is that the 'disaster' adapter (adapter 2) really emits a very load and broad interference. When listening on the radio the signal is all over the place, increasing the noise signal to more than S9 over the whole shortwave spectrum.
The other adapter is much better, although there are also disturbing noise peaks in the HF spectrum, they are not covering the whole spectrum.
The NGP814 has the cleanest signal, the interference consist of small band 'whistles' in the lower part of the HF spectrum.
You can find the spectra as measured on the spectrum analyser below:
As mentioned in my application, the second experiment I had in mind for the NGP814 is to use it for calculating a shortest path in a maze using an analog circuit. You may coin it an analog computer or analog twin as well. The circuit works by building the maze with strings of leds (Analog Computing: Shortest Path | Analog Zoo ). For this application I was triggered by the possibility of the NGP814 to act as a constant current source. Red or Green leds mainly needs 20mA to light them up, resulting in 2.2V over the LED. For testing a large network you need quite some leds, and thus needs a high voltage power supply. On the website mentioned in my application a current source is built, but with the R&S NGP814 in constant current mode no additional electronics are needed. Furthermore the maximum voltage of the R&S NGP814 with four channels in serie is 192V. This makes it possible to test large networks, up to 87 LEDs in the shortest path. Would be great to be able to perform this experiment. To start with this idea in order to be able to build large circuits/maps/mazes/street plans I designed a circuit and PCB with led strings and connectors to build large circuits using a few of these PCBs.
As you can see, the strings on the border can be used in two directions, while the cross connected strings in the center are 'one way roads'.
The ordering of the PCB's took longer than expected, just a few days for the due date of the roadtest my PCBs arrived.
Due to the lack of the PCB's I started building some LED strings on a breadboard.
An important question now is how fast the power supply will switch between constant voltage mode to constant current mode, or change the voltage when the current starts to increase, for instance when the shortest path is changed from 10 to 1 led. For 10 leds the voltage is 20 V, for 1 led 2 V. I made a quick test with a 180 ohm resistor, and the power supply at 20V, current limit 10 mA. This is what you see when switching on the power:
This will certainly destroy my single led with 20 V for about 50 ms. Another weird phenomenon is the fast voltage decay halfway the exponential decay.
I expect this has something to do with the switch mode operating principle.
I also tested this using a ramp of 1 s. This makes it a lot better, but I doubt whether my LED will survive 5.8 V amplitude.
Next experiment is with the power supply on, and a 2k2 resistor (9 mA at 20 V) which was shorted with the 180 ohm resistor. This gives the following signal:
Again too much for my single LED.
After switching back to the 2k2 resistor, the voltage slowly increase back to 20 V in about 2 seconds.
Finally I ended up with the current limit at 3 mA and a 330 ohm resistor in serie with the circuit. On the breadboard I made two strings of leds in parallel, first one with 12 leds, and second one with 15 leds and a shortcut with a switch at 5 leds. In the video you nicely see the shortest path lighting up in different configurations. Note that when switching the shortcut on the current briefly goes to twice the set current.
The main conclusion is that for this application the current control of the power supply is too slow, but I think it is in accordance with the specifications (fall time < 50ms with no load).
The NGP814 is a rather unique and powerful four channel power supply with a lot of capabilities. Despite the heavy-handed shipping method the device arrived without any visible damage at my desk. The front panel with a large touchscreen and just a few buttons looks great, and the the operation is very intuitive. There are a lot of handy functions, such as ramp, arbitrary function, and buit-in measurements, even with a graphical display. The unit actually consists of four independent power supplies of 200W each, which can be connected in parallel or in series.
The device is heavy, with a weight of 10kg and the footprint is also large, in my case it doesn't fit on my shelves. There are a lot of possibilities for connecting it to a PC, such as LAN, WLAN, USB and GPIB. Unfortunately WiFi is an option you have to pay for, so I sticked to the LAN connection, which worked great. I liked the built in webserver a lot.
The device is reasonably quit, the fan speed depends on the heat production, with a minimum of 20%. Sometimes the fan speed showed a bit of unpredictable behaviour.
The ripple as measured with a scope, was a bit more than defined in the specifications. I also found some radio interference, measured with a spectrum analyser, and listened to with a radio, but I assume it is all compliant with the Class A CISPR 32 standard. Controlling the output in constant current mode was too slow for low power, low current applications like the shortest path led experiment. This was solved by setting the current in a very safe low value, and adding a resistor. I expect the switching time of 50 ms to be suitable for more 'normal' applications.
For lab applications that require high power and a lot of flexibility I can really recommend this power supply.
Here is a list of issues I found that could potentially be resolved in the next release of the documentation and firmware.
Just updated the review according to firmware update 2.015.
To be sure I measured the weight again, now by measuring the difference in weight between myself and myself carrying the power supply. This again confirmed that the weight is really close to 10 kg.
So with the wires it is like with the T-Ford all colours are possible, as long as it is black? (Red in this case).
The graphical view is indeed great, I hope the autoscale for power will be fixed soon.
Great road test report.
That was an interesting review to read. Thanks for that.
to me it is a bit strange that a high value lab power supply is tested with multimedia device standards.
It conforms to multiple standards, including EN 61326 for lab equipment, FCC, and so on [the datasheet has a list].
I think 61326 incorporates EN 55011 for conducted and radiated limits and methods [equivalent to CISPR 11].
But I would think that the later CISPR 32 is then actually used because it can cover multiple I/O and output leads (a complicated setup for conducted emissions that 55011 wouldn't deal with properly on the measurement method side). But I'm very rusty with all this stuff now, so I might be wrong.
I'm sure that R&S, of all people in the world, know what they're doing with these standards.
So it seems that the 7.5kg weight comes from the NGP802/822 models.
As for 10kg ... hmm. R&S are claiming the NGP814 weighs in at 8.0kg ... so there's 2kg discrepancy somewhere. Perhaps not the first time R&S have had some datasheet discrepancies (based on my experience) ... but perhaps scales for human body weight aren't so accurate at the low end either.
Regardless ... the red and blue jacks are very much an R&S colour scheme. All of their supplies follow that convention and it's never occurred to me that it was illogical ... however, now that you mention it, it is non-standard. But seeing as I use red wires for everything (power, ground, sense) as element14 just so happened to have a special on red banana plug leads and no stock of anything else... I can't say colour is all that important in the end.
Perhaps most impressive is the graphical view capabilities which would be a first in the R&S power supply line to my knowledge and is a feature often seen more on DMMs and SMUs. While it's not going to be fast like an oscilloscope, it's a better way of seeing the log data rather than just a screen of numbers. It seems to take the interface from the NGL/NGM series, so I hope that the graphical view feature gets backported to the NGM series where FastLog can record at 500kS/s as it would be quite useful even if for post-visualising a FastLog recording.
The picture above is from baldengineers video review. baldengineer mentions the weight as 17.6 pounds, which is 7.5 kg in metric units, but this is not true. As waleedelmughrabi already mentioned, the unit really is 10 kg (22 pound).
Thank you for the correction. I wonder how I got that wrong. Either I used the wrong datasheet, or maybe I saw the weight for a 2-channel unit?
Great Road Test. Very detailed!