@@yoeribolderdijk1257 that would greatly depend on the load, with a non-linear load the current is likely to go up to the maximum limit of the power supply, if I have to guess, is not likely to be the 2A maximum rated current but much higher, even the maximum current the supply is able to deliver or some hard limit the circuit has as protection, that will depend on the implementation, always assuming the load is capable to taking it, otherwise the load will just fail, and depending on the power supply phase on the start up, just keep going to 100V as there were no load (because it isn't once and if the load fauled open). It is also likely that the power supply could damage itself, but that would be more likely to get noticed in the development, but we may know once they hit the market and start to get abused.
The "X" notation is actually pretty common among other German manufacturers too. A lot of the CNC equipment I used to work on used Siemens controls. Siemens used the "X" prefix all over the place on various ports/connectors.
X refers to a electrical connection between two objects. Sometimes there are additional numbers indicating main and subassemblys and the direction. Often its just -X as prefix and an index.
Nice for a basic model. Like the toroidal transformer, looks like an audio quality transformer. Good on you for the giveaway, I'm a member but won't enter because there are others who would need it more. Love that you're giving one away to a hacker space school. 👍
About the output terminals... they can do way over 120V, so they absolutely shouldn't be used without shrouded banana plugs because you can accidentally touch bare contact and get a nasty shock. So those type of terminals are acceptable.
@@EEVblog It was a really old ebay 60V 5A kit with the green 1602 LCD. It worked well but I think the switching IC shorted when I was trying to draw high current and high power to run a heating element. It's probably fixable but I don't know how to protect it from happing again, I had already beefed up the heatsinks and added forced air cooling to the whole unit
I’m not sure about the intent here with the initial high speed fan; it could be lackadaisical engineering or R&S meant to do it. I used to work in a datacenter, and servers often have banks of multiple fans and two fan banks in series to allow for a fan failure in one bank. Anyway, the Dell servers we had sounded like a 747 jumbo jet for about five seconds when initially switched on. There would also be a puff of dust out of the back of the servers. The servers stayed pretty much spotless inside. I concluded that it was no accident that the fans were initialized at the highest speed - it was to blow out the dust. It also gives the auxiliary “lights-out-management” supervisory processor an opportunity to check the health of the fans to see if all fans can meet or exceed an acceptable RPM threshold.
great video as always - I was just wondering; will you ever do a tear down/walkthrough of your own little power supply? perhaps even a big video going through all the prototypes leading upto it?
IMO, the BEST test of current limiting on a PSU is... 1: Set output voltage to the maximum it can produce (100V in this case) 2: Set the current limit to 20mA 3: Connect a 'garden variety' LED directly across the output 4: Enable the output If you still have a working LED after that, then it's _possibly_ a decent device. (I had to make HUGE changes on my own bench PSU in order to have LEDs survive this test even though it only goes up to 30V)
Exactly my thought as well. I was thinking that with the current set to 10mA and 10V voltage the overshoot on this PSU would be bad enough to blow an indicator LED.
This seems like a silly bragging-rights test to me. The whole idea with CC output is that an ideal supply would have infinite voltage at zero load. If there's *ANY* output capacitance, you'll get a voltage spike when hot-connecting a load to a CC output. If there's *NO* output capacitance, your transient response will suffer. Here's the solution to avoid the spike: Don't connect a load with the VMax turned all the way up. Ramp up to a reasonable output voltage.
Yep totally agree! :) Capacitive load or not! Any good design will ramp slow enough for the regulation loop to keep up. Anything else is CRAP! ..yes I am talking to you overpriced R&S!
@@nickwallette6201 we are not talking about connecting the load while the supply is on, watch the video. With a load already connected the power supply overshoots when enabling the output, there's no capacitors being dumped here, this is purely the PSU overshooting when it is turned on.
@@stevenbliss989 It's not just R&S who 'fail' here... Quite a number of 'well known' brands will fail this seemingly simple test. You can make the test even harder by connecting the LED AFTER enabling the output on the PSU. (The ONLY PSU I have that's reliably passed this tougher test is an old Power Dynamics as it has basically no output capacitor after the current limiter)
Nice supplies, a 100V/200V supply would certainly come in handy when I'm fixing stuff. I really need to get hold of them again to see if they want to send me items to review, they almost did a couple of years ago but I hadn't reached 10K at that stage.
19:44 that Pmax of 0.36W was recorded when you first pressed the output on button because the output capacitors were empty and they simply charged, right?
About the current range overshoot, I use the r&s hmp4040 and did the same experiments a few months back. Sadly I observed the same thing during overcurrent conditions, the supply seems to reach the full voltage and needs 10-20 ms to start current control. Unfortunately that's not good enough with the GaN devices I am measuring. If the device is unstable, it usually can't handle the initial surge of current during that 10-20 ms and it destroyw itself ... I wonder why it is necessary to have that 10-20 ms, it doesn't seem needed in a classic analog constant current control loop like the one used in the μSupply...
Hi Dave! Really curious as to how the voltage overshoot was missed by the designteam at R&S. Also fun to see you pull out a Keysight EDU power supply (another can of worms regarding programming output values and actual output). Love these hands on and real world reviews. Did R&S correct the statistics function? Because for bench testing it would be essential to easily reset the stats.
Dave, did you check if there is a firmware update for the PSU? I had a thought that it’s just possible that they’ve corrected the over voltage problem in firmware but I’m no expert on these things.
Aww too bad. Hopefully they will take note of Dave’s analysis and release an update to correct the problem. That is assuming it is correctable in that manner and the PSU was designed to allow users to update the firmware.
You might want to have a look at these, they do actually work well, and NO overshoot crap on them either. KORAD and similar on eBay, AliExpress etc, like www.ebay.com.au/itm/120922717980
Oh! Nice tip about the output caps! I'm playing around in making a power supply and I'll keep that in mind! :) For an instrument of that price, I would expect a lot more polished functionality to be honest.
Yep, double that. I designed my own home brew semi precision PSU, and it uses a 1uF output cap, has no CC overshoots issues, and I am certainly NOT good at linear, I struggled with loop stability for the 1uf with much pain until I got it good for any load type. So for R&S to do this crap compared to what a hobbyist can do by a mix of guesswork and trial & error is very dissapointing. Where are we to look up to for great PSU design, Keysight maybe! :)
Generally a lot of manufacturers like to default to fan 100% on. I think it has several benefits. You don't know which state everything is in, so it might be hot as hell due to some power loss event or so. Fan spinup is easier at 100% and then throttling down than to try starting it at 5% or so. Users will earlier notice if something is wrong with the fan. Probably most people don't mind as they work with these and switch them on at most once per day
The ribbon cable to the front panel is touching the winding of the mains common-mode choke. A single-insulation ribbon cable is rated to 300V, magnetic wire insulation could withstand around 500V. I doubt that the device can comply with a breakdown voltage of 1500V.
I am not too keen on R&S equipment, especially the high end spectrum analyser and oscilloscope, they run too warm in my view and seem to boot off an hdd with internal Os. if a glitch occurs(and they do!) they have to be returned. Then what happens when warranty has expired? On the PSU front, Kenwood used to make very hardy and robust power supplies, 60v/10A which will run all day every day for years!
I believe they're doing current regulation in software, that's why we see the stepping (hunting) on turning on the output, and that's why switching from CV to CC is so bad. What is the current resolution worth, if it blows through the limit almost any time you adjust the voltage up? And this is a >1600€ PSU… Jeez.
Pretty much everything in the IT industry does this. The fan is usually controlled by a simple microcontroller- or ASIC-based supervisory circuit. Until the main device software loads completely, the supervisory circuit is running blind, and fails open with the fan at full speed. This is the safest failure mode, since you literally can't do anything more anyway if the unit is running hot. Noisy and annoying as it may be, it protects the equipment as much as possible, and alerts the user to a problem condition if it doesn't quiet down. But also, running a fan at 100% allows for a startup test to ensure the fan CAN run at full speed, and that full speed is at approximately the RPM you expect (if RPM is being sensed.) Consider it part of the POST.
I've used a lot of R&S test gear, and I am less than impressed with the reliability. This is one company whose extended warranty is value for money. As for their user manuals, all I can say is, they have a vacancy for a technical writer if anyone is interested.
@@EEVblog my only thought is that at 13:06 there appears to be tape on both sides of the heatsink so devices on both sides? They must have to have the devices bent outwards to get the heatsink between them unless they are already stuck on before soldering. If they are pushed on to the tape after the fact, then the leads are going to have some spring in them trying to pop them back off. When the tape gets old and crusty they are going to pop off.
@@tmmtmm I'd guess that the heatsinks are installed first, then the power devices are sticky-taped to them and only after that do they solder them by hand
It is too bad there is that chunky 21 uF capacitor across the output. The cap must be there to assure unconditional stability because who knows what ugly load might be attached. The shunt capacitance sort of negates the point of having six digits of resolution for the current display. The last two digits are essentially useless in many situations.
The current measurement used for display is past the capacitor surely?. There’s another local current stabilization loop I’d hope. The outer one would then cross over to the inner one at perhaps few hundred Hz to some kHz. But I haven’t looked inside one of those. Typically the state of the art has voltage feedback from both output terminals and point of load, and current measurement at the output terminals. The digital control loop would then have estimators to neutralize some of the load impedances. First of all it can get the inductive impedance of the leads going to the load. Then it would identify (continuously) at least 1 or 2 time constants and magnitudes in the impedance of the load, so that it could provide optimally stable output (with user option of a fallback to a fixed control response in case the AC-ish part of the load response varied wildly). I can’t quite understand how PS designs with so much computational power don’t use it to improve control loop performance. It’s basic stuff really these days. Undergrad students implement it in EE control theory courses.
@@absurdengineering Thanks for the insight. Just going on what is presented in this video, it lead me to believe the capacitor is across the output terminals after current sensing. I have seen electrolytic capacitors sometimes paralleled by a ceramic disc directly across the output terminals of those $75 - $150 switch mode bench supplies from generic manufacturers sold through Amazon, but those units are not meat for serious precision.
Could the Nonzero open circuit current be down to the leakage current of the capacitors NCC gives a leakage current spec for one of those caps of 0.02*CV+25, after five minutes, or 95 microamps. Seems the right order of magnitude, and it's probably conservative + you're not at the 160V spec. Would depend on where they'd mounted the current shunt I guess, seems like the capacitor is the last thing in the chain, which would make sense for minimising inductance/improving load regulation. Current might go down after some time under voltage though? Seems like a hard problem to solve if it changes/reduces over time- can't zero it out Another reason I guess to use smaller caps- less leakage You do seem to regain that second to last digit later in the video at 20v, maybe that last one is only viable at lower voltages, as the leakage current gets reduced. That or it takes a while for the caps to reform/stabilise, and you have to leave the machine on for an hour before using it.... Edit: would also explain the +500 microamps part- spec sheet says 0.04 CV+100 for leakage after 1 minute- 241 microamps. For 2 fused channels that would come quite close to 0.5ma Edit, transistor -> capacitor
I’d have hoped they have a current sensor after the capacitors, not before. The control loop should be compensated with those capacitors inside the loop. If need be they could add another local control loop on the board itself, but it’s generally not necessary. The capacitor impedance estimation is done trivially these days by injecting low amplitude (microvolt) probing currents into the output capacitor and synchronously demodulating the voltage and current feedback from the output terminals themselves (beyond the capacitor). This can be used to neutralize in large part this capacitance. It would require either a fully digital control loop, or a parametric adjustment of one or two time constants and magnitudes in the feedback loop. Two time constants would come into oaky if neutralization was done with a pair of roots instead of just one. A capacitance multiplier and a gyrator could be easily controlled with a digital potentiometer to generate the necessary “analog domain” poles/zeroes as needed.
There are not so many power supplies around with that high output voltage, i need at least 2x60v (symetrical) for audio repair, is there any alternative?
Interesting Power Supply. I have the NGE102 which is the two channel model in the NGE series. The operating software (firmware) is very similar and displays Min/Max values which I love. The Min/Max can be reset at any time by holding the arrow button for a couple of seconds under the “Menu” button. I have posted a short video demonstration of my NGE102 (“Rohde & Schwarz NGE100 Lab Power Supply Demonstration & RTB2004 Oscilloscope”) which was actually the “A” model upgraded by firmware update to the “B” model. I then believe it’s only the output binding posts and front colour that’s different. This video might be of interest if you are thinking of buying the two channel version. The three channel has a different firmware without Min/Max etc. I really love my NGE102, it’s a reliable compact power supply. It also had some great features such as “FUSE”, “ARB” and “RAMP” as standard. Unfortunately other functions like WiFi and Digital Switching are enabled by software licenses. Interesting video, especially the internals. Thanks.
How is the stuff pressed against the heatsink? I see no clamp or screw? The flaw with the constant current jumping during voltage change is shocking, seems to be a digital control stuff problem. Why not using analog control in a linear power supply. Of cource it could also be a digital input value with an analog control at the end. It also could be a problem with a range chage from the AC input voltage from the transformer taps, however. The 22 µF at the output is also Inconveniently large. Thanks for the insight and the tests, nice review like always.
Sorry, but this much overshoot should be a FAIL not just interesting. And the biggest fail is not having separate output button for each chanel. Because of price ot is more targeted toward production, where communication is used, but more hobby friendly companies are known for coping bad ideas.
It DOES have seeprate output buttons for each channel, those CH1 and CH2 buttons. You can choose to turng both on/off at once using the Output button, or each channel using the CH1/CH2 buttons.
Those terminal colors aren't really European. Following NEC it should be red & black, following IEC it should be brown & grey, blue is used for center tap (0V) if present. Australia & New Zealand apparantly do have red or brown for pos and black or light blue for neg.
That heatsink glue at 9:50 looks to me like the "yellow glue of death" that eventually turns conductive as I outlined here ruclips.net/video/9gqBzLNMFe4/видео.html. I'm a bit sad to see a high end product from R&S using it instead of silicone. To be fair it's not on conductive components, but still it can become corrosive and brittle and stop working as intended.
@@EEVblog It might just be PTSD from dealing with it on so much audio gear in my past, but I don't trust that type of glue at all. Since it usually takes years to chemically degrade, it's possible manufacturers don't know the aging characteristics. I don't know whether it's a characteristic of all glues based on synthetic rubber, or if it's just bad formulations that degrade. It's also possible for bad formulations to get into their supply chain. They only used it in one place, so not a big issue for this design, but just a general comment regarding if you see it in the future to maybe be suspect of it.
@@TheUbuntuGuy One audio repair guy here on YT calls this glue "gorilla snot" and in old amps it's always bad. It's also much more of a problem when it's gunked all over component leads, like caps and resistors, because electrolysis and stuff. Kind of surprised me to see it used in this PSU also, but let's hope the glue tech has advanced since the '90s.
I just love my TTi PL601 90w linear psu which was about 400$ brand new and does not overshoot at all. Dave, just stop showing us rubbish marketing for newly released "premium" test equipment!
Looks like that 6 digit ADC was so expensive they could only afford 1 per channel and multiplexed it between voltage and current. When the set voltage is changed, it only monitors that and forgets about current for awhile. If you want 2 ADC's you have to pay much more.
A bit better from one they did a few years ago which was what i'd expect from a college (not university) student project: an off-the-shelf switcher to provide 5V and then a bunch of switching regulators for other rails. Really, really noisy.
@@southerncharity7928 Compared to other "budget" supplies, their price says otherwise. I think it's the other features you are paying for, but they really could have got the basic right. :)
2 channel 100V model in the uk is £1400 BEFORE 20% tax and the digital triggers cost an extra £120 + 20% tax and the wireless another £180 + 20% tax... Thats £2040 inc tax...
From the rather quick tour inside, I am pretty sure the 100N06 mosfets do that. BTW tap switching is standard and very common for decades, at leat for linear PSUs.
It would be interesting to see a resistive load periodically turned on/off via transistor to see how much time it takes to switch from CC to CV and back, and see whether it looks like the 22uF of the output or much worse.
Hi Dave, I have a question. I have a PV array on my roof (UK), do I need to worry about dirt, dust, bird poo. etc. It's rainy here unlike your end I guess. It's been there for 10 years, no trees above it. Cheers dude.
They put a entire microprocessor with memory for just a power supply. I think it's a overkill for something displaying DC data. But kudos to current measurement dynamic range.
Great stuff, micro actuator, all golden info. Except, you better real that "tounge" in. Suddenly my girl friend likes your videos, I'm sure the business trip to Australia is a coincidence ! HA HA. Well, we love your presentations ,esp "trap for new players". Would you consider explaining V double circuit and conflict of pos power on neg side of cap. Reverse polarity?, How high?,How long etc. take care.
The software can be much faster than typical analog control loops, as long as there’s enough bandwidth in the A/D and D/A blocks, and you can do all sort of cool things like compensate for component aging, compensate capacitive and inductive loads, etc. Trivial in software, but would take a bunch of function blocks to do in analog domain, with their contribution to drift and noise. I bet though that the transistor protection is done in the analog realm. If they were fancy, they could implement SOA protection that integrated energy dissipated on the transistor. This can be done for same cost using a tiny 8-pin or 16-pin dedicated MCU with integrated analog functions, with single-purpose software that just has a fixed control loop and nothing else. Typically would be very robust, and may be even cheaper than some higher precision analog solutions. But most T&M gear companies aren’t too creative in that area.
@@absurdengineering What tiny MCU with integrated analog functions would you use for a voltage or current control loop? The only ones I know only have a comparator that can trigger an interrupt but this is next to useless in my opinion since the variable reference value needs to come from somewhere else.
@@donepearce I was not referring to you, R&S of course. :) Your experience with BPS-305 is exactly what I would expect from all but the most cheep and nasty PSU's.
I love to have it I am a fan of Round Transformer I love to have it keep it alive like my life and use it but how do I get it I don’t want to enter in that contest 😍
Anyone know why they'd use more than one bridge rectifier? I thought it would just use one rectifier and split off to other circuits further down the line. What's the benefit?
Keep the two channels electrically separate from each other. My Old UNI-T dual output bench PSU has the same thing for that reason. Separate windings, separate relays, and separate bridges, giving me two fully floating independent channels, ...like the R&S has.
The bridge rectifier adds impedance and would thus couple the two channels more than if there were two separate ones. The varying load current in one channel would present as a line variation in the input of another channel. Not good. And this coupling generally eats into margins in the control loops. But another good reason to separate the bridges is for reliability: the bridge will run with lower junction temperature and thus will be more reliable at least in terms of failures caused by mechanical fatigue.
now thats an awful psu. you pay 1300$ for the R&S name, not the hardware. i had a Owon 3ch linear psu that really looked nice for the price but had the same current overshoot issue but with a duration of almost 100ms or more. you couldnt hook up a led when having it set to 10v and 20mA limit, upon turning it on it allways blew the led before cc kicked in. software loops are just wrong in something like this. a dac that sets a properly tuned analog loop is alright but adc -> code -> dac is the worst
Unfortunately voltage and current controlled regulator circuits are horribly non-linear and need a lot of software correction. I designed one and I couldn't dream of reaching a resolution like R&S with my design. Also despite all lookup tables the voltage and current control never got really independent.
no offence to R&S but that 'looks' a bit like a kids toy from the opening shot..just something about the colors and cheap looking plastic.... on the positive side...looks quite lightweight the way you where moving it around...and the insides are sexy and well made/designed..and that current resolution is top notch!..the extra info on the screen would be quite useful too (if it worked right!) the overshoot and not so constant current mode are a bit how ya doin!
Not that light, but it's only an 80W PSU, so not a beast of a transformer required. Agreed on the plastic looking connectors, I really don't like them.
My thoughts exactly. Output C leakage and the load of the sensing itself. This is why, when I see micro-anything resolution, I always think... "what good is that, when it's just going to be swamped by noise anyway?"
@@stevenbliss989 Yeah, that's exactly what I did. If you have an output with anything parasitic on it, and you're measuring current down to tiny, tiny units, then any of those parasitic loads will show up. That would include - any active switching components (FET output switches), leakage in any of the output stage components, bleed resistors, voltage measurement ADC inputs ...
lmao "stingray", with that overshoot it makes it all the more hilarious
That's huuuge overshoot, with the 20V back to de CC mode, it might blow up quite a few loads...
Yes, concerning.
@@EEVblog And how much will the overshoot be at 100V output? :(
@@yoeribolderdijk1257 A gazillion amps.
How this got through QC is beyond me... Or whatever the equivalent term is for before the design is finalized.
@@yoeribolderdijk1257 that would greatly depend on the load, with a non-linear load the current is likely to go up to the maximum limit of the power supply, if I have to guess, is not likely to be the 2A maximum rated current but much higher, even the maximum current the supply is able to deliver or some hard limit the circuit has as protection, that will depend on the implementation, always assuming the load is capable to taking it, otherwise the load will just fail, and depending on the power supply phase on the start up, just keep going to 100V as there were no load (because it isn't once and if the load fauled open). It is also likely that the power supply could damage itself, but that would be more likely to get noticed in the development, but we may know once they hit the market and start to get abused.
A transient analysis of various power supplies would make a great video. Not only on startup/shutdown but CC shorts and recovery while in process.
The "X" notation is actually pretty common among other German manufacturers too. A lot of the CNC equipment I used to work on used Siemens controls. Siemens used the "X" prefix all over the place on various ports/connectors.
Electricians apprentice from Germany here, you are correct
X refers to a electrical connection between two objects. Sometimes there are additional numbers indicating main and subassemblys and the direction. Often its just -X as prefix and an index.
As a developer in at one of Germany‘s biggest electronics companies, I can also confirm that 😉
in the netherlands its also quite common in industrial gear
I think the X for connectors is inspired by the socket connector reference designator in IEEE 200-1975 resp. IEC/ISO/EN/DIN 81346
Pretty cool that there is no silkscreen but necessary designators and labels are directly in the soldermask and enig plated :)
Its a bit hidden in the manual but: "Long press on the Back key resets the channel history information."
Nice for a basic model. Like the toroidal transformer, looks like an audio quality transformer.
Good on you for the giveaway, I'm a member but won't enter because there are others who would need it more. Love that you're giving one away to a hacker space school. 👍
I had this in cart for the past few days. Glad I didn't pull the trigger before you uploaded this video. That overshoot is nasty. Thank you!
About the output terminals... they can do way over 120V, so they absolutely shouldn't be used without shrouded banana plugs because you can accidentally touch bare contact and get a nasty shock. So those type of terminals are acceptable.
I blew up my homemade bench supply and haven't had a working one since, hopefully I can nail the giveaway! Back to the vid!
Yikes! How did you fry it? Do tell!
Betcha it's the output pass trasistor blown.
@@EEVblog It was a really old ebay 60V 5A kit with the green 1602 LCD. It worked well but I think the switching IC shorted when I was trying to draw high current and high power to run a heating element.
It's probably fixable but I don't know how to protect it from happing again, I had already beefed up the heatsinks and added forced air cooling to the whole unit
It was 100% switching based, nothing linear
It worked fine for the most part but something cleaner would be nice
@@iamdarkyoshi , maybe you could get a used linear power supply on eBay? Those are usually pretty solid, has full schematics and everything.
Software driven CC is a bad idea IMO. Give me an old fashioned analog loop.
I agree, a low cost opamp will achieve great current regulation, and use much less resources at the same time.
Nothing wrong with digital regulation loops... as long as they're implemented right, unlike this one.
I’m not sure about the intent here with the initial high speed fan; it could be lackadaisical engineering or R&S meant to do it. I used to work in a datacenter, and servers often have banks of multiple fans and two fan banks in series to allow for a fan failure in one bank. Anyway, the Dell servers we had sounded like a 747 jumbo jet for about five seconds when initially switched on. There would also be a puff of dust out of the back of the servers. The servers stayed pretty much spotless inside. I concluded that it was no accident that the fans were initialized at the highest speed - it was to blow out the dust. It also gives the auxiliary “lights-out-management” supervisory processor an opportunity to check the health of the fans to see if all fans can meet or exceed an acceptable RPM threshold.
The 0.021A is probably a really quick spike in the measurement due to capacitance or such. Too fast to show up on the display but it's there.
or the output caps 7:40
great video as always - I was just wondering; will you ever do a tear down/walkthrough of your own little power supply? perhaps even a big video going through all the prototypes leading upto it?
IMO, the BEST test of current limiting on a PSU is...
1: Set output voltage to the maximum it can produce (100V in this case)
2: Set the current limit to 20mA
3: Connect a 'garden variety' LED directly across the output
4: Enable the output
If you still have a working LED after that, then it's _possibly_ a decent device.
(I had to make HUGE changes on my own bench PSU in order to have LEDs survive this test even though it only goes up to 30V)
Exactly my thought as well. I was thinking that with the current set to 10mA and 10V voltage the overshoot on this PSU would be bad enough to blow an indicator LED.
This seems like a silly bragging-rights test to me. The whole idea with CC output is that an ideal supply would have infinite voltage at zero load. If there's *ANY* output capacitance, you'll get a voltage spike when hot-connecting a load to a CC output. If there's *NO* output capacitance, your transient response will suffer.
Here's the solution to avoid the spike: Don't connect a load with the VMax turned all the way up. Ramp up to a reasonable output voltage.
Yep totally agree! :) Capacitive load or not!
Any good design will ramp slow enough for the regulation loop to keep up. Anything else is CRAP! ..yes I am talking to you overpriced R&S!
@@nickwallette6201 we are not talking about connecting the load while the supply is on, watch the video.
With a load already connected the power supply overshoots when enabling the output, there's no capacitors being dumped here, this is purely the PSU overshooting when it is turned on.
@@stevenbliss989
It's not just R&S who 'fail' here...
Quite a number of 'well known' brands will fail this seemingly simple test.
You can make the test even harder by connecting the LED AFTER enabling the output on the PSU.
(The ONLY PSU I have that's reliably passed this tougher test is an old Power Dynamics as it has basically no output capacitor after the current limiter)
Nice supplies, a 100V/200V supply would certainly come in handy when I'm fixing stuff. I really need to get hold of them again to see if they want to send me items to review, they almost did a couple of years ago but I hadn't reached 10K at that stage.
19:44 that Pmax of 0.36W was recorded when you first pressed the output on button because the output capacitors were empty and they simply charged, right?
About the current range overshoot, I use the r&s hmp4040 and did the same experiments a few months back. Sadly I observed the same thing during overcurrent conditions, the supply seems to reach the full voltage and needs 10-20 ms to start current control. Unfortunately that's not good enough with the GaN devices I am measuring. If the device is unstable, it usually can't handle the initial surge of current during that 10-20 ms and it destroyw itself ... I wonder why it is necessary to have that 10-20 ms, it doesn't seem needed in a classic analog constant current control loop like the one used in the μSupply...
I assume they blast the fan at 100% on startup to blast out some dust.
Some OEM PCs do something similar.
Still annoying, though.
I believe it's someone's idea to warn the operator of a failed fan.
I think it's just to make it fail safe - If the fan controller functionality fails, it won't then overheat.
its actually a good design... the over current is far more concerning
Dave, could the overvoltage issue be a result of charging the output capacitor? Would be nice to see the actual current drawn by the load.
I believe the probe is directly on the resistive load. You ARE seeing the current.
@@IvanStepaniuk Ok, that would make sense.
Hi Dave! Really curious as to how the voltage overshoot was missed by the designteam at R&S. Also fun to see you pull out a Keysight EDU power supply (another can of worms regarding programming output values and actual output).
Love these hands on and real world reviews. Did R&S correct the statistics function? Because for bench testing it would be essential to easily reset the stats.
Part of the 100% on fan at boot may be for testing, I know my little BK Precision does this and immediately says "FAN OK" after.
Oh, so I can stay with my 40ye old one with no issues ... Ok, it won't have waveforms and data logging but as a power supply it is better :/
Dave, did you check if there is a firmware update for the PSU? I had a thought that it’s just possible that they’ve corrected the over voltage problem in firmware but I’m no expert on these things.
There is no firmware for NGA100 on their website.
Aww too bad. Hopefully they will take note of Dave’s analysis and release an update to correct the problem. That is assuming it is correctable in that manner and the PSU was designed to allow users to update the firmware.
Looks like there was firmware released on 29 June. The first improvement listed is current overshoot.
Bummed I wasn't on the forum earlier, always been looking for a benchtop supply, oh well, maybe next time, might as well start posting haha
You might want to have a look at these, they do actually work well, and NO overshoot crap on them either.
KORAD and similar on eBay, AliExpress etc, like www.ebay.com.au/itm/120922717980
Oh! Nice tip about the output caps! I'm playing around in making a power supply and I'll keep that in mind! :)
For an instrument of that price, I would expect a lot more polished functionality to be honest.
Yep, double that. I designed my own home brew semi precision PSU, and it uses a 1uF output cap, has no CC overshoots issues, and I am certainly NOT good at linear, I struggled with loop stability for the 1uf with much pain until I got it good for any load type. So for R&S to do this crap compared to what a hobbyist can do by a mix of guesswork and trial & error is very dissapointing. Where are we to look up to for great PSU design, Keysight maybe! :)
Generally a lot of manufacturers like to default to fan 100% on. I think it has several benefits. You don't know which state everything is in, so it might be hot as hell due to some power loss event or so. Fan spinup is easier at 100% and then throttling down than to try starting it at 5% or so. Users will earlier notice if something is wrong with the fan. Probably most people don't mind as they work with these and switch them on at most once per day
The ribbon cable to the front panel is touching the winding of the mains common-mode choke. A single-insulation ribbon cable is rated to 300V, magnetic wire insulation could withstand around 500V. I doubt that the device can comply with a breakdown voltage of 1500V.
I am not too keen on R&S equipment, especially the high end spectrum analyser and oscilloscope, they run too warm in my view and seem to boot off an hdd with internal Os. if a glitch occurs(and they do!) they have to be returned. Then what happens when warranty has expired? On the PSU front, Kenwood used to make very hardy and robust power supplies, 60v/10A which will run all day every day for years!
I believe they're doing current regulation in software, that's why we see the stepping (hunting) on turning on the output, and that's why switching from CV to CC is so bad.
What is the current resolution worth, if it blows through the limit almost any time you adjust the voltage up? And this is a >1600€ PSU… Jeez.
Wondering how power components fastened to the heatsinks. There's no screws. Is that yellow stuff a double-sided adhesive?
Yes, thermally conductive double sided adhesive.
The fan on full blast on start helps unstick fans
Pretty much everything in the IT industry does this. The fan is usually controlled by a simple microcontroller- or ASIC-based supervisory circuit. Until the main device software loads completely, the supervisory circuit is running blind, and fails open with the fan at full speed.
This is the safest failure mode, since you literally can't do anything more anyway if the unit is running hot. Noisy and annoying as it may be, it protects the equipment as much as possible, and alerts the user to a problem condition if it doesn't quiet down.
But also, running a fan at 100% allows for a startup test to ensure the fan CAN run at full speed, and that full speed is at approximately the RPM you expect (if RPM is being sensed.) Consider it part of the POST.
I've used a lot of R&S test gear, and I am less than impressed with the reliability. This is one company whose extended warranty is value for money. As for their user manuals, all I can say is, they have a vacancy for a technical writer if anyone is interested.
@28:40 That initial spike is probably from the capacitors discharging and then the supply kicks in.
Would be interesting to discharge them first, then try again and see if it still acts with a overshoot..
No. ...think about it. :)
13:19 are the power devices just double-stick taped to the heatsink? I don't see any bolts. Seems a bit how ya doin.
Common industry technique. You can get special thermally conductive double sided tape designed for this purpose.
@@EEVblog my only thought is that at 13:06 there appears to be tape on both sides of the heatsink so devices on both sides? They must have to have the devices bent outwards to get the heatsink between them unless they are already stuck on before soldering. If they are pushed on to the tape after the fact, then the leads are going to have some spring in them trying to pop them back off. When the tape gets old and crusty they are going to pop off.
Yeah, I don't know how long it's going to last.
@@tmmtmm I'd guess that the heatsinks are installed first, then the power devices are sticky-taped to them and only after that do they solder them by hand
Well, for a device at this budget price-point, it's hard to justify tapping the aluminum for screws. * cough *
Thanks for sharing the awesome content, allways amazed at your knowledge and attitude!
The overshoot has been taken care off according to information from R&S to mij dealer. How about an update about that on this lovely channel?
It is too bad there is that chunky 21 uF capacitor across the output. The cap must be there to assure unconditional stability because who knows what ugly load might be attached. The shunt capacitance sort of negates the point of having six digits of resolution for the current display. The last two digits are essentially useless in many situations.
The current measurement used for display is past the capacitor surely?. There’s another local current stabilization loop I’d hope. The outer one would then cross over to the inner one at perhaps few hundred Hz to some kHz. But I haven’t looked inside one of those.
Typically the state of the art has voltage feedback from both output terminals and point of load, and current measurement at the output terminals. The digital control loop would then have estimators to neutralize some of the load impedances. First of all it can get the inductive impedance of the leads going to the load. Then it would identify (continuously) at least 1 or 2 time constants and magnitudes in the impedance of the load, so that it could provide optimally stable output (with user option of a fallback to a fixed control response in case the AC-ish part of the load response varied wildly). I can’t quite understand how PS designs with so much computational power don’t use it to improve control loop performance. It’s basic stuff really these days. Undergrad students implement it in EE control theory courses.
@@absurdengineering Thanks for the insight. Just going on what is presented in this video, it lead me to believe the capacitor is across the output terminals after current sensing. I have seen electrolytic capacitors sometimes paralleled by a ceramic disc directly across the output terminals of those $75 - $150 switch mode bench supplies from generic manufacturers sold through Amazon, but those units are not meat for serious precision.
Could the Nonzero open circuit current be down to the leakage current of the capacitors NCC gives a leakage current spec for one of those caps of 0.02*CV+25, after five minutes, or 95 microamps.
Seems the right order of magnitude, and it's probably conservative + you're not at the 160V spec.
Would depend on where they'd mounted the current shunt I guess, seems like the capacitor is the last thing in the chain, which would make sense for minimising inductance/improving load regulation.
Current might go down after some time under voltage though? Seems like a hard problem to solve if it changes/reduces over time- can't zero it out
Another reason I guess to use smaller caps- less leakage
You do seem to regain that second to last digit later in the video at 20v, maybe that last one is only viable at lower voltages, as the leakage current gets reduced.
That or it takes a while for the caps to reform/stabilise, and you have to leave the machine on for an hour before using it....
Edit: would also explain the +500 microamps part- spec sheet says 0.04 CV+100 for leakage after 1 minute- 241 microamps.
For 2 fused channels that would come quite close to 0.5ma
Edit, transistor -> capacitor
I’d have hoped they have a current sensor after the capacitors, not before. The control loop should be compensated with those capacitors inside the loop. If need be they could add another local control loop on the board itself, but it’s generally not necessary. The capacitor impedance estimation is done trivially these days by injecting low amplitude (microvolt) probing currents into the output capacitor and synchronously demodulating the voltage and current feedback from the output terminals themselves (beyond the capacitor). This can be used to neutralize in large part this capacitance. It would require either a fully digital control loop, or a parametric adjustment of one or two time constants and magnitudes in the feedback loop. Two time constants would come into oaky if neutralization was done with a pair of roots instead of just one. A capacitance multiplier and a gyrator could be easily controlled with a digital potentiometer to generate the necessary “analog domain” poles/zeroes as needed.
Wonderful show , even i don't get it all , but using that yellow pointer ! makes sense now ,
thanks
There are not so many power supplies around with that high output voltage, i need at least 2x60v (symetrical) for audio repair, is there any alternative?
When did blue become the color of the low side of a power supply? Is blue plastic cheaper than black plastic?
Interesting Power Supply. I have the NGE102 which is the two channel model in the NGE series. The operating software (firmware) is very similar and displays Min/Max values which I love. The Min/Max can be reset at any time by holding the arrow button for a couple of seconds under the “Menu” button.
I have posted a short video demonstration of my NGE102 (“Rohde & Schwarz NGE100 Lab Power Supply Demonstration & RTB2004 Oscilloscope”) which was actually the “A” model upgraded by firmware update to the “B” model. I then believe it’s only the output binding posts and front colour that’s different. This video might be of interest if you are thinking of buying the two channel version. The three channel has a different firmware without Min/Max etc.
I really love my NGE102, it’s a reliable compact power supply. It also had some great features such as “FUSE”, “ARB” and “RAMP” as standard. Unfortunately other functions like WiFi and Digital Switching are enabled by software licenses.
Interesting video, especially the internals.
Thanks.
How is the stuff pressed against the heatsink? I see no clamp or screw?
The flaw with the constant current jumping during voltage change is shocking, seems to be a digital control stuff problem. Why not using analog control in a linear power supply. Of cource it could also be a digital input value with an analog control at the end. It also could be a problem with a range chage from the AC input voltage from the transformer taps, however. The 22 µF at the output is also Inconveniently large.
Thanks for the insight and the tests, nice review like always.
Could the 20µA measurement at open circuit be the leakage current of the output capacitor??? Does it rise with voltage? Does it settle after time?
Sorry, but this much overshoot should be a FAIL not just interesting.
And the biggest fail is not having separate output button for each chanel. Because of price ot is more targeted toward production, where communication is used, but more hobby friendly companies are known for coping bad ideas.
It DOES have seeprate output buttons for each channel, those CH1 and CH2 buttons. You can choose to turng both on/off at once using the Output button, or each channel using the CH1/CH2 buttons.
Those terminal colors aren't really European. Following NEC it should be red & black, following IEC it should be brown & grey, blue is used for center tap (0V) if present.
Australia & New Zealand apparantly do have red or brown for pos and black or light blue for neg.
Has the problem been resolved by the new firmware? What about the NGE and the HMC804x which also have an overshoot in CC?
Thanks for exposing the quality.
That heatsink glue at 9:50 looks to me like the "yellow glue of death" that eventually turns conductive as I outlined here ruclips.net/video/9gqBzLNMFe4/видео.html. I'm a bit sad to see a high end product from R&S using it instead of silicone. To be fair it's not on conductive components, but still it can become corrosive and brittle and stop working as intended.
I'm sure there are many manufacturers of such stuff, how do you know it's the same type and/or all of them do that?
@@EEVblog It might just be PTSD from dealing with it on so much audio gear in my past, but I don't trust that type of glue at all. Since it usually takes years to chemically degrade, it's possible manufacturers don't know the aging characteristics. I don't know whether it's a characteristic of all glues based on synthetic rubber, or if it's just bad formulations that degrade. It's also possible for bad formulations to get into their supply chain. They only used it in one place, so not a big issue for this design, but just a general comment regarding if you see it in the future to maybe be suspect of it.
That is just plain BS....Not all yellow goo is the stuff you talk about...
@@TheUbuntuGuy One audio repair guy here on YT calls this glue "gorilla snot" and in old amps it's always bad. It's also much more of a problem when it's gunked all over component leads, like caps and resistors, because electrolysis and stuff.
Kind of surprised me to see it used in this PSU also, but let's hope the glue tech has advanced since the '90s.
Dave himself found that happening in the studio monitor (speaker) teardown and repair. :-)
You forgot to call it 'The money shot' once you opened the case and showed the internals.
I just love my TTi PL601 90w linear psu which was about 400$ brand new and does not overshoot at all. Dave, just stop showing us rubbish marketing for newly released "premium" test equipment!
Looks like that 6 digit ADC was so expensive they could only afford 1 per channel and multiplexed it between voltage and current. When the set voltage is changed, it only monitors that and forgets about current for awhile. If you want 2 ADC's you have to pay much more.
10:01 Code Name, "Stingray". Makes your humble electronic workshop feel like it's part of some covert activity!
A bit better from one they did a few years ago which was what i'd expect from a college (not university) student project: an off-the-shelf switcher to provide 5V and then a bunch of switching regulators for other rails. Really, really noisy.
Sure shows up my Elenco (build it yourself) power supply kit.
At that price, I would expect the power components to be screwed on and not mounted with adhesive tape like this.
Makes me wonder if their other equipment is similarly overprice mediocre stuff.
To be fair, this is their "budget" model
@@southerncharity7928 Compared to other "budget" supplies, their price says otherwise. I think it's the other features you are paying for, but they really could have got the basic right. :)
My 'bench power supply' is just one of those wall plug, switchable voltage with a screwdriver thingies. It has about 10.00% accuracy at best 😆
Good enough for Austraila
Who needs constant current mode when you have circuit breakers?
I feel you
2 channel 100V model in the uk is £1400 BEFORE 20% tax and the digital triggers cost an extra £120 + 20% tax and the wireless another £180 + 20% tax... Thats £2040 inc tax...
Hello David That is a beautiful power supply and I also am not fan of those output jacks .
Always fun to watch your vlogs
Can you show me as a beginner how you constructed your 2 amp load? Thank you sir!
Hello Dave
The transformer have a lot of strings. Did and How R&S swichtes between them?
Is a nice idea to save some Power
From the rather quick tour inside, I am pretty sure the 100N06 mosfets do that. BTW tap switching is standard and very common for decades, at leat for linear PSUs.
It would be interesting to see a resistive load periodically turned on/off via transistor to see how much time it takes to switch from CC to CV and back, and see whether it looks like the 22uF of the output or much worse.
The 22uF output cap makes me think the R&S design is crap, maybe with stability issues being covered up.
I am sure this video will trigger a meeting or two at corporate. "Okay scratch EEVBLOG".
Hi Dave, I have a question. I have a PV array on my roof (UK), do I need to worry about dirt, dust, bird poo. etc. It's rainy here unlike your end I guess. It's been there for 10 years, no trees above it. Cheers dude.
Worth cleaning periodically, but I've noticed minimal degredation with dirt.
@@EEVblog Thx, that's a weight off my mind
They put a entire microprocessor with memory for just a power supply. I think it's a overkill for something displaying DC data. But kudos to current measurement dynamic range.
That's not uncommon nowadays. With all the remote control, USB and color display features, there's not really another way.
Teasing that uSupply in the background...
Great stuff, micro actuator, all golden info. Except, you better real that "tounge" in. Suddenly my girl friend likes your videos, I'm sure the business trip to Australia is a coincidence ! HA HA. Well, we love your presentations ,esp "trap for new players". Would you consider explaining V double circuit and conflict of pos power on neg side of cap. Reverse polarity?, How high?,How long etc. take care.
I'm wondering what the current measurement across a sense resistor on a scope would look like...
That is a seriously cool and functional PSU
The 100V model is perfect for valve projects.
Is it enought for valves? Maybe better something like 400V?
@@VolodymyrTorkalo Yes, you can put the outputs in series for 200V. It's perfect for prototyping.
Thanks for the teardown, what was the op amp part number?
It looks like a TL082. The TI datasheet says the TSSOP package is marked T082.
See page 36 of the datasheet... www.ti.com/lit/ds/symlink/tl082.pdf.
@@stevenbliss989 Thank you.
I wonder how it is acceptable at all to do the current limit in software. How can this be fast enough to prevent transistors from burning?
The software can be much faster than typical analog control loops, as long as there’s enough bandwidth in the A/D and D/A blocks, and you can do all sort of cool things like compensate for component aging, compensate capacitive and inductive loads, etc. Trivial in software, but would take a bunch of function blocks to do in analog domain, with their contribution to drift and noise. I bet though that the transistor protection is done in the analog realm. If they were fancy, they could implement SOA protection that integrated energy dissipated on the transistor. This can be done for same cost using a tiny 8-pin or 16-pin dedicated MCU with integrated analog functions, with single-purpose software that just has a fixed control loop and nothing else. Typically would be very robust, and may be even cheaper than some higher precision analog solutions. But most T&M gear companies aren’t too creative in that area.
@@absurdengineering What tiny MCU with integrated analog functions would you use for a voltage or current control loop?
The only ones I know only have a comparator that can trigger an interrupt but this is next to useless in my opinion since the variable reference value needs to come from somewhere else.
I'm to busy to do the give-out so keep up the good work dave... guess next time for me ...
The max speed fan is a function of cleaning.
Whenever someone opens some piece of electronics I'm thinking now: *"OOIii, what a surpROOOIIiise"* hehe
It looks more like Rohde & Bluez...
Rohde obviously don't use their own oscilloscopes to test their equipment haha Besides don't expect getting their best stuff for that entry level sum.
Those overcurrent glitches almost certainly happen as it switches taps on that mains transformer when you change voltage..
Maybe, but I doubt it, ...still no excuse, crap design!
@@stevenbliss989 Not an excuse - just an explanation. I checked my cheapo Lavolta BPS-305 bench supply. No problem at all - not a trace of overshoot.
@@donepearce I was not referring to you, R&S of course. :)
Your experience with BPS-305 is exactly what I would expect from all but the most cheep and nasty PSU's.
@@stevenbliss989 It's OK, I got what you meant.
I love to have it I am a fan of Round Transformer I love to have it keep it alive like my life and use it but how do I get it I don’t want to enter in that contest 😍
It would be so awesome to win one of these for my birthday coming up!
Nice review video
3:10 ....No....gotta love the Ozzie crap colorcode.....That sells out the house....
Great resulation 👏👏👏👏
22 uf @ 100V sounds like more than enough to ruin your day to me! (output caps....)
Such bug discovered by one person in 15 minutes whereas in RS several people was involved in the project. Blamage.
Would be nice to get the psu for a review and then give it to someone else 😁
Was expecting the hand with screwdriver with that statement :P
Anyone know why they'd use more than one bridge rectifier?
I thought it would just use one rectifier and split off to other circuits further down the line. What's the benefit?
Keep the two channels electrically separate from each other.
My Old UNI-T dual output bench PSU has the same thing for that reason.
Separate windings, separate relays, and separate bridges, giving me two fully floating independent channels, ...like the R&S has.
The bridge rectifier adds impedance and would thus couple the two channels more than if there were two separate ones. The varying load current in one channel would present as a line variation in the input of another channel. Not good. And this coupling generally eats into margins in the control loops. But another good reason to separate the bridges is for reliability: the bridge will run with lower junction temperature and thus will be more reliable at least in terms of failures caused by mechanical fatigue.
And the Nichicon fan boys go wild.
can you do one on the rp2040 board. Maybe make a board around it using parts
now thats an awful psu. you pay 1300$ for the R&S name, not the hardware.
i had a Owon 3ch linear psu that really looked nice for the price but had the same current overshoot issue but with a duration of almost 100ms or more.
you couldnt hook up a led when having it set to 10v and 20mA limit, upon turning it on it allways blew the led before cc kicked in.
software loops are just wrong in something like this. a dac that sets a properly tuned analog loop is alright but adc -> code -> dac is the worst
Unfortunately voltage and current controlled regulator circuits are horribly non-linear and need a lot of software correction. I designed one and I couldn't dream of reaching a resolution like R&S with my design. Also despite all lookup tables the voltage and current control never got really independent.
Where are the screws for those power transistors?
looks like they use bond ply as we do
no offence to R&S but that 'looks' a bit like a kids toy from the opening shot..just something about the colors and cheap looking plastic....
on the positive side...looks quite lightweight the way you where moving it around...and the insides are sexy and well made/designed..and that current resolution is top notch!..the extra info on the screen would be quite useful too (if it worked right!)
the overshoot and not so constant current mode are a bit how ya doin!
Not that light, but it's only an 80W PSU, so not a beast of a transformer required. Agreed on the plastic looking connectors, I really don't like them.
@@EEVblog i forgot youre a gym junkie! :P
Nice one !
Your videos are great!!! Informative
R&S output cap too big?
Yes, 1uf or less would be more appropriate.
Near unconditional stability, and tight regulation is difficult with
you had me at "feelovision", lol
It has output capacitance, doesnt it? I bet thats the "load" you are measuring
My thoughts exactly. Output C leakage and the load of the sensing itself. This is why, when I see micro-anything resolution, I always think... "what good is that, when it's just going to be swamped by noise anyway?"
No.
@@nickwallette6201 No.
Have a think about how the current is being measured. i.e. where is it flowing from and to, and you will have your answer.
@@stevenbliss989 Yeah, that's exactly what I did. If you have an output with anything parasitic on it, and you're measuring current down to tiny, tiny units, then any of those parasitic loads will show up.
That would include - any active switching components (FET output switches), leakage in any of the output stage components, bleed resistors, voltage measurement ADC inputs ...