EEVblog 1402 - Rohde & Schwarz NGA100 PSU Teardown + GIVEAWAY

[e-nighthawk]

Does anyone know how that works with import taxes? Do you have to pay 20% or something on that 1000+ USD value?

[Refrigerator]

Does anyone know how that works with import taxes? Do you have to pay 20% or something on that 1000+ USD value?

Pretty sure you can google it, it's not like we're supposed to know German import laws.

[Kleinstein]

The design with 2 large filter caps and 2 output transistors is likely with an electronik switch over between 2 transformer taps. Makes absolutely sense, as it saves on the heat sink and works without glitches and wear of relay switching.

The current measuring shunts look a bit disappointing - I had expected higher power version to get a really good dynamik range.

The overshoot on the current limit looks a bit strange, a bit like disabling the current limit just after changing the set votlage voltage. This may be with the idea to faster get to the new voltage. There is a slight chance to it could be fixed with a software update, as this looks like under software control.

With the rear ouput and seprate rear sense terminals / cables, I am slightly missing a way to switch between front and rear. The drive terminals may be in parallel, but the sense part would need some switching (front, rear with 2 wires and rear with 4 wires).

The regulations on import/sales taxes likely depend on the country - chances are you may have to pay in some countries. In the US you are even supposed to pay income tax for this, at least this is what KS tells you in there giveaways.

[Refrigerator]

Guys FIY there's a separate thread for the giveaway here:
https://www.eevblog.com/forum/contests/giveaway-rohde-schwarz-nga102-psu/
So stop posting "i'm in" in this thread.

You must have a decent number of posts on the forum over some time in order to enter. Let's say 20ish posts, Dave's decision is final.

Drive-by posters, you know where the door is. 
Like 80% posts here are giveaway "entries". 

Getting back on topic, that overshoot is quite a shame, really. Especially considering the price and the brand name behind the PSU you would really expect something better.
How easily would the overshoot in this R&S PSU blow an LED? (max output voltage with 20mA CC on an indicator LED)
Also is the voltage and current controlled via a SW loop? What benefits does that have over a regular op-amp based control?

Personally i don't find this PSU worthy of 1000+ bucks, especially with that overshoot. Putting expensive components into a product is like putting marble into an expensive NYC apartment, but that alone will not make it good. I have no attachment to R&S because I've never used any R&S equipment and my view point is not that "It's R&S so it's going be good" but rather "let's see how good it is". And so far the impression is not good.   
I want to see how R&S handles this overshoot situation and if it can be fixed. 

I think this warrants further testing of not only the R&S, but also other lab bench PSUs for comparison reasons.

[TheD312]

I`m in too!

[e-nighthawk]

I'd love to score one of these, here's the homemade one made ages ago and killed somewhat recently:

I like the idea of showing what the R&S would replace.

Here's what I put together actually not even that long ago:


It's based on a RD DPS5005 I had picked up about two years ago. I used it here and there, but while being quite okay, it never made it to be my preferred piece of equipment. I didn't like the hassle of always having to find a PSU that supplies enough DC for the DPS first before I'm able to power my actual project.

So this year I added an AC power supply to it which makes it much easier to use for me. While at it, I also upgraded the fan and made it temperature controlled. I'm not so sure yet, if it'll be able to keep the unit cool when it's running at full power. That's to be tested once I find something that draws enough energy.

[HattedSquirrel]

Getting back on topic, that overshoot is quite a shame, really. Especially considering the price and the brand name behind the PSU you would really expect something better.

It already has been that way many years back when it still was Hameg. R&S used to be reputeable for extremly high end RF gear (and still is). Then they bought Hameg to extend their portfolio with PSUs and scopes. Like many, I hoped it would make the PSUs better... but... well, they didn't.
They did some work on the scopes, though. User interface is now completely different and more in line with their RF stuff. (Showing the current frontend configuration as a block diagram for example.)

Also is the voltage and current controlled via a SW loop? What benefits does that have over a regular op-amp based control?

Better control and more complex control loops. One of the first steps in getting a good and fast control loop is linearizing the hardware's system response. That is easily done in software by applying the inverse (non-linear) transfer function. Creating the same inverse transfer function in HW is much harder. Another point is the ability to fine tune the loop control parameters via firmware updates or even during run time for different operational modes. Some PSUs have special settings for inductive loads or high / low capacitive loads. Numerous four-quadrant power supplies can be switched between current or voltage compliance mode. I have heared that it might be common to tune the speed of both control loops according to the mode in order to give you the best output response once you run into compliance. While there isn't much run-time tuning in the PSU in the video, the ability to retune via firmware updates is considered de-risking and also saves time during development.
And last but not least, it is easier & cheaper to get low drift / drift compensated DACs and ADCs, than to build the complete analog control loop in hardware with similarly low drift. You wouldn't think how important 0.1$ saved in a 1000+$ product is. Don't ask how I know.

[motorev16]

I would have preferred 2 sensing bananas on the front panel

[Refrigerator]

I would have preferred 2 sensing bananas on the front panel

But there's still connections for external sensing on the back side of the unit, no?

[Refrigerator]

Getting back on topic, that overshoot is quite a shame, really. Especially considering the price and the brand name behind the PSU you would really expect something better.

It already has been that way many years back when it still was Hameg. R&S used to be reputeable for extremly high end RF gear (and still is). Then they bought Hameg to extend their portfolio with PSUs and scopes. Like many, I hoped it would make the PSUs better... but... well, they didn't.
They did some work on the scopes, though. User interface is now completely different and more in line with their RF stuff. (Showing the current frontend configuration as a block diagram for example.)

Also is the voltage and current controlled via a SW loop? What benefits does that have over a regular op-amp based control?

Better control and more complex control loops. One of the first steps in getting a good and fast control loop is linearizing the hardware's system response. That is easily done in software by applying the inverse (non-linear) transfer function. Creating the same inverse transfer function in HW is much harder. Another point is the ability to fine tune the loop control parameters via firmware updates or even during run time for different operational modes. Some PSUs have special settings for inductive loads or high / low capacitive loads. Numerous four-quadrant power supplies can be switched between current or voltage compliance mode. I have heared that it might be common to tune the speed of both control loops according to the mode in order to give you the best output response once you run into compliance. While there isn't much run-time tuning in the PSU in the video, the ability to retune via firmware updates is considered de-risking and also saves time during development.
And last but not least, it is easier & cheaper to get low drift / drift compensated DACs and ADCs, than to build the complete analog control loop in hardware with similarly low drift. You wouldn't think how important 0.1$ saved in a 1000+$ product is. Don't ask how I know.

Thanks for the response, and i have to agree that tuning a circuit in software is quite advantageous, as i've found out with FIR filters in my projects.
But it makes me wonder why does it take the NGA100 some 10ms to correct the output. Surely the MCU should be plenty fast enough to do this in microseconds, especially with that slow ramp-up.  Are they spending too much resource driving that LCD? Also how does the PSU regulate a changing load, does the control loop lag behind in that case also, or is it just during output enable?
I wonder if that Keysight in Dave's video also uses a SW control loop, it seemed to behave a lot better. And the little spike (not an OV) in the output seemed like the control loop "catching" itself before an overshoot happened.
I really want to see how a Rigol and other brand PSU's perform on this test, i think Dave also had some ebay cheapies in his storage somewhere.

[taiteki]

i need one!! and this is a big one!!

[Kleinstein]

I don't think the normal control loop would be in software. The demands on the ADC / DAC would be quite high. Digital control starts to become popular with SMPS - these have lower requirements and usually don't need a DAC. They also tend to like getting closer to the limited with the control loop. A point that is relatively difficult to implement analog is anti windup.

For this lab supply it looks more like digital control on the set values.  The linear ramp part is likely from the set voltage DAC going up in steps to get a controlled ramp when the output is enabled. There is nothing wrong with this an I would not mind the little ripple seen.
The other point seems to be an increased current limit when the output voltage is changing (maybe only up) - this is likely controlled by software by adjusting the set point to the analog current limit. It looks like good intentions, but not really though to the end.
Best case there may be an option to turn off this feature.

It is a bit like the computer to solve problems that you would not have without a computer.

Separate sense terminals are not that often used with the 4 mm plugs. It absolutely makes sense to have 2 terminals at the front and 4 wires on the extra connector in the back. With only some 2 A the voltage drop on the wires is not such a big issue.

[nctnico]

Guys FIY there's a separate thread for the giveaway here:
https://www.eevblog.com/forum/contests/giveaway-rohde-schwarz-nga102-psu/
So stop posting "i'm in" in this thread.

You must have a decent number of posts on the forum over some time in order to enter. Let's say 20ish posts, Dave's decision is final.

Drive-by posters, you know where the door is. 
Like 80% posts here are giveaway "entries". 

Getting back on topic, that overshoot is quite a shame, really. Especially considering the price and the brand name behind the PSU you would really expect something better.
How easily would the overshoot in this R&S PSU blow an LED? (max output voltage with 20mA CC on an indicator LED)
Also is the voltage and current controlled via a SW loop? What benefits does that have over a regular op-amp based control?

Personally i don't find this PSU worthy of 1000+ bucks, especially with that overshoot. Putting expensive components into a product is like putting marble into an expensive NYC apartment, but that alone will not make it good. I have no attachment to R&S because I've never used any R&S equipment and my view point is not that "It's R&S so it's going be good" but rather "let's see how good it is". And so far the impression is not good.   
I want to see how R&S handles this overshoot situation and if it can be fixed. 

I think this warrants further testing of not only the R&S, but also other lab bench PSUs for comparison reasons.

Actually the power-on behaviour and crossover between constant voltage / constant current  is often tested when people do a review of lab PSU.

BTW: Keysight had to do a recall of their E36300 series lab PSUs because a spike appeared on the output when you plugged it in. But with power applied it already behaved way better than this R&S NGA100 PSU. Heads are going to roll in the main R&S office this Monday!

[kcozens]

That NGA unit is quite the beast in terms of size. I bet it must weigh a lot. As is often the case Dave quickly finds some questionable behaviour of a new piece of test gear. It would take up a fair amount of the space on my small workbench but it would be a nice upgrade to my existing power supply.

[splin]

[EDIT]: To make it clear the following is is an (obviously poor) attempt at satire - please don't take it seriously!

Why all the whinging about a mere 10ms or so to get the output under control? No doubt you're all hardware engineers who don't understand the extreme complexity and difficulties the firmware system has to deal with!

Integrating a unique new, state of the art, high-reliabilty RTOS, engineered for optimised PSU functionality, developed using a cutting-edge customised Agile based methodology helps manage the complexity of the large software development to minimise the distractions of constant inter-team interactive incompatibility disagreements, such as memory and CPU  resource allocations, which plague traditional homogenous software developments.

For example, this allowed the Wuppertal team to develop the overcurrent detection module, largely decoupled from the Stuttgart team developing the event management and distribution module, the Taipei team responsible for output current control functionality (with 1uV resolution),  the event logging object oriented distributed database team etc, and the somewhat undervalued 'Output Fusion' team. (Oddly, nobody seems to be able to recall where they are based).

With round-robin 1ms scheduling, 10ms response is actually world class leading especially given the need for triple checking the over current event is not a false noise abberation. And whilst the 550MHz processor might appear to be overkill,  consider that as the PSU could be used in safety crtical situations it needs to be fully MISRA compliant - appreciate how long it takes to do a full memory test of the hardware (including 256MB of Flash and 128MB DRAM) to be certain the overcurrent event wasn't due to a mains glitch or similar.

Personally I'd have included a speech processor:

"Dave. The output current seems to in excess of your expectational parameter tolerances... would you like me to attempt to constrain it?"
...
"I'm sorry, Dave. I'm afraid I can't do that."
...
"Dave, please put down the crowbar - it may invalidate your (limited) R&S warranty".

[gardner]

I would love to have one of those.  It would look so nice next to my crappy Yihua 30V/10A one.

With all those winding leads coming off the transformer, I was expecting to see some sort of switching between various windings to minimize the voltage drop on the pass transistors.  But you'd need 3 or 4 relays per channel to switch them and I don't see any of that.  Are there a raft of triacs hiding on one of those heatsinks to do the switching?

[Kleinstein]

With 2 main filter caps that are rated lower than the maximum output voltage, it is quite clear the fitler caps are in seires and very likely with a transformer tap in between. One can that use electronic switch over between the 2 raw DC voltages. This would be only 2 transformer taps, but much faster (e.g. µs instead of seconds). The 2 voltages with a fast corss over may be about as effective as 3 conventional AC swtiched transformer taps, as there is no need for much hysteresis and reserve for ripple and less estra headroom needed. The "switch" over could be by the 2 power transistors seen on the heat sink and a diode (maybe the little extra heat sink ?).

In addition to 3 wires for the dual main supply one may need another 2 or 3 wires to power the actual regulator circuit. Like most lab supplies they very likely use the floating regulator principle. Especially for such a high voltage this is the obvious choice.

Switching with relays also has some trouble - the relays usually work on the AC side and would see quite some current spike from charging the filter cap. This resuslts in quite some wear to the contacts and the possibility of sticking contacts which can result in failure. The swtiching can also cause some glitch reaching the ouput. Because of the time it takes for switching, it needs extra reserve in the voltage and thus more loss. To limit the wear it needs quite some hysteresis. So switching to a lower transformer tap would be only if the voltage is way higher than actually needed.

The overshoot on turn on is only with the current limit, likely not much with the voltage. One has to compare this to the charge stored in the output cap. Some 2 A for 5 ms are 10 mC, which would be equivalent to 10 V in a 1000 µF capacitor, like found with some lower grade lab supplies. It is not great and I have hope it would be improved / fixed, but it also is not a desaster. If in the low current range, the extra current is likely also lower.

[nctnico]

Why all the whinging about a mere 10ms or so to get the output under control? No doubt you're all hardware engineers who don't understand the extreme complexity and difficulties the firmware system has to deal with!

A 10ms spike is enough to kill your circuit which is why decent power supplies don't show this behaviour! It doesn't matter how difficult it is to get right, it simply must be right. You have to be able to rely on a power supply to do what you want and not push random voltages/currents into your circuit.

Dave should send it this PSU to me and I'll bring it where it belongs: at the recycling station. That will save other people the misfortune of having a circuit that suddenly doesn't work or lets the magic smoke out for no obvious reason. There is nothing worse than a PSU you can't rely on (been there, done that).

[Refrigerator]

Why all the whinging about a mere 10ms or so to get the output under control? No doubt you're all hardware engineers who don't understand the extreme complexity and difficulties the firmware system has to deal with!

A 10ms spike is enough to kill your circuit which is why decent power supplies don't show this behaviour! It doesn't matter how difficult it is to get right, it simply must be right. You have to be able to rely on a power supply to do what you want and not push random voltages/currents into your circuit.

Dave should send it this PSU to me and I'll bring it where it belongs: at the recycling station. That will save other people the misfortune of having a circuit that suddenly doesn't work or lets the magic smoke out for no obvious reason. There is nothing worse than a PSU you can't rely on (been there, done that).

I thought that post was sarcasm. 
Also don't be so hasty, that PSU could make a pretty ok car battery charger 

[igorstep]

I'm in, Nice PSU!!!

[Mick]

I'm in 

100v supply would be super useful.

[andybrandi]

Still lacking a nice PSU for my lab, those would make a nice addition for sure. 

[blackdog]

Hi

I often explain it this way:
The power supply must first be kind to itself.
And second, he must be very caring of his connected load.

Even though it seems that way, a linear power supply is not simple.
There are very many mistakes made by amateurs (me too in the past) and professionals, especially regarding the current limitng as Dave showed with the R&S power supply, among others.

Just because R&S designed this power supply does not mean it is good.
The same goes for a number of HP models. I have one here that has a low power rating, below 50Watt, nice and low noise and an output capacitor of 1000u! for less than 2 Amps.
1000uF is way to mutch in this HP Power Supply, to be very caring of his connected load.

For now, I don't know how the voltage loop and current loop are arranged in the R&S power supplies.
But digital is really a NoNo for a linear power supply for now.

If you design an Analog/Linear power supply you should in my opinion make sure that both loops are always stable without switching effects from voltage to current and vice versa.
It's nice that you have lots of options in the menus but loop stability is #1 for a linear power supply.

Also notice the noise or PARD when Dave puts the scope over the output of the power supply, you can sawing wood at the pulses the scope shows, now that's not exactly neat.
Those pulses may be a commonmode problem in Dave's measurement setup, but I have a sneaking suspicion that they are not.

Some opinions from me:
Good linear power supplies do not have digital loop control.
To solve this properly in the digital domain, you need very fast and good ADC/DAC and a processor just for this function with a very good "Watch Dog" setup.
This is I think, rather expensive.

The Power section will have been implemented with medium speed transistors of around 5Mhz Ft. (Faster is possible with modern audio power transistors, but it is harder to keep stable.)
If the compensation of your voltage and current loop is set up properly, with good opamps with large phase margin(AD4625), you will achieve 40uF capacitor per Ampere across the output with very good dymamic behavior.

Those were my two cents. 

Kind regards,
Bram

[Kleinstein]

The ouput cap is shown to be 22 µF electrolyic plus a little ceramic (more like <= 1 µF at 100 V max voltage) so it is some 12 µF per amp of output current.

I still expect the control loop to be analog, it only looks like there is some software action going in effecting the set points. The easy part to implement is a defined "slow" (some 5 ms) ramp up of the voltage on output enable. This is easy to implement and I see no problem with this.

The problematic part is the current limiting not working as expected when the voltage is changed (possibly just some changes, like up). My guess on this is a kind of extra "feature" to make the output faster following a changing set point and the current limit may be disabled / set up for a short time. This may be nice in some cases, but it kind of defeats the current limit idea. So it's a bad idea, at least as the default setting - maybe useful as an option.

There is no really fast connection from the high powered µC at the from to the ouput stages. The that µC / CPU is mainly for the UI, not the regulation loops (other than the fan).

[blackdog]

Hi Kleinstein, 

I'm not sure the electrolitic 22uF are the only capacitors across the output, I saw on the circuit board more electrolitic capacitors near the output wiring.

So it's possible they split the capacitance across the output.
And I'm not talking about ceramic capacitors at the output terminals.

But of course it remains speculation when I don't have the power supply in my hands.

Kind regards,
Bram