Another issue with the Rigol DP832 power supply: sense wires carrying 3 amps!

[rs20]

Did my subject line catch your attention? Yes, that's right -- if you source current from Ch3+ and sink it into Ch2-, the current flows through the voltage sense wires and induces a drop (and therefore error in voltage readout/regulation) of up to 57 mV, which is well outside spec (0.05% + 10mV = 10.5mV). Sense wires appear to have a resistance of 19 milliohm. Here's a specific example:

Configure both outputs to 1V, 3A limits. Attach voltmeter to Ch2.
A. Measure voltage. You get 1.000V. Win!
B. Short out Ch3+ to Ch3-. No worries, you still get 1.000V. Win!
C. Now, instead, short out Ch3+ to Ch2-. Now you get 0.943V, even though the display on the power supply still reports 1.00V. Boo!

Now you might be saying I'm being a bit silly, returning the current to the "wrong common terminal". Now I'm of the opinion that common terminals should be just that, common, completely identical in every way. But more to the point, if you configure a system with +/-12V rails for your power audio output (for example), and +5V for some other logic, then you've got a common ground all around and you should be able to just have one ground wire running from power supply to circuit-under-test. Having to carefully account for and make sure that every electron has an easy path back to its "home" negative terminal seems a bit nuts.

I mean, the whole point of sense wires is that they're supposed to have (practically) zero current, and therefore zero voltage drop, right?

What do you guys think? Am I being pedantic and making a fuss, is this standard practice for non-isolated channels on power supplies? Is it generally known in the profession to be bad practice to rely on common terminals being, well, common? Or is this a real problem particular to this power supply?

Update: Threw together a video to demonstrate this:

[Psi]

What do you guys think? Am I being pedantic

na, i dont think your being pedantic.

They should never have common'ed ch2 and ch3.
3 Isolated outputs is more useful and eliminates these sort of unexpected issues.

[c4757p]

The longer this goes on, the more and more comfortable I become with my clunky, ancient, non-programmable, all-analog power supplies.*

*Before the torches and pitchforks come out.... yes, I know whether it's analog or digital has nothing to do with how buggy it is. I think it's more about time-tested designs, really - the crap doesn't last, so when you buy old stuff you know it's the one that was good enough to survive. I have yet to see an old analog power supply with control loop bugs like this.

[BravoV]

The longer this goes on, the more and more comfortable I become with my clunky, ancient, non-programmable, all-analog power supplies.

+1

About the non-programmable part, if the ps has remote terminals for external adjustment for the voltage/curent loop, coupled with a digital controlled voltage reference, I wouldn't call it clunky & ancient anymore. 

[robrenz]

+2

[rs20]

What do you guys think? Am I being pedantic

na, i dont think your being pedantic.

They should never have common'ed ch2 and ch3.
3 Isolated outputs is more useful and eliminates these sort of unexpected issues.

True. The trouble is, though, that to do that you need board space for isolation and well-separated heatsinks and so on, there's always going to be a considerable compromise in power/weight/size/cost when making truly independent supplies. What I don't get, though, is why they didn't use proper, thick (less than 38 milliohm!) wire to common the terminals directly, rather than via two sense wires and a crimp. Even 24AWG wire, at a length of 5cm, is 1.3 milliohms or so, low enough to make this problem reduce to being within spec. Incidentally, for those that don't mind voiding your warranty, that's exactly how you can fix this problem on your own DP832.

[Psi]

Yeah, i agree. cost savings will be the reason.

But a power supply is so critical for electronic design that it's better to have less features that are all done properly than more features where some have weird issues that might catch you out.

It's not like an advanced piece of test equipment where you might tolerate some differences in operation/usage between brands.

[orin]

 Having to carefully account for and make sure that every electron has an easy path back to its "home" negative terminal seems a bit nuts.

Strongly disagree.

Considering the return path of a current is part of good design.  And it's even more important at RF.  And these days most every circuit is an RF circuit.  (If you think not, see if it works with a GSM cellphone next to it.  Fluke 87 V anyone?)

Even at DC, you have to consider the I*R voltages in your "ground" connections and decide if you can ignore them.

Orin.

[c4757p]

If you built the circuit, you presumably know if you can ignore them. Basic testing of some random audio circuit (like the given example) isn't going to care about grounding. (Sure, it might need more attention to return paths to make it work well, but 99% of the time on the bench you just want to see that it works.) And I can make the decision to ignore that much faster than I can figure out exactly how I should caress the power supply to convince it to give me what I want.

[mickpah]

Lets face it, RIGOL stuffed up in a big way with the LM317 . The spot light is now on  and everyone is going microanalyse this bit of gear.
This is the price of not getting it right first time.

Frankly as the owner of one this is good, so before it gets sent back for whatever rework it is going to have I can decide if I just want to cut my losses and have a refund.
And IT WILL get sent back our consumer laws here in OZ put the responsibility on the seller/distributor so there is no hiding. I'm sure the US has even better "lemon" laws is some states.
 
http://www.accc.gov.au/consumers/consumer-rights-guarantees/repair-replace-refund

[orin]

If you built the circuit, you presumably know if you can ignore them. Basic testing of some random audio circuit (like the given example) isn't going to care about grounding. (Sure, it might need more attention to return paths to make it work well, but 99% of the time on the bench you just want to see that it works.) And I can make the decision to ignore that much faster than I can figure out exactly how I should caress the power supply to convince it to give me what I want.

"If you built the circuit..."?  Seriously, given the crap that is shipped, building a circuit means nothing.  Remember, we are preaching to those that might not be circuit design gods.

Then you can't just design a circuit then slap quality (the last 1%) on as an afterthought.  You have to design it in.  And that includes how you draw power from your power supply.

Perhaps you have the intuition to get it right.  I don't.  I guarantee that any reasonably well designed circuit* will fail if I build it**.  And the reason it fails is always one of these kind of gotchas.  Call me on it... please... I'd like my projects to actually work right for a change just to prove me wrong.  I have a stack of expensive test equipment just to work out why my projects don't work.

Orin Murphy.

*analog.  The digital ones usually work.
**in some cases, they don't work right for anyone else, but they don't have the test equipment to realize it.

[c4757p]

Sometimes I just want to see if it'll release the magic smoke...

I have a stack of expensive test equipment just to work out why my projects don't work.

Dangerous. You have a "stack of expensive test equipment" to make sure you'll never be able to use an "assortment of affordable equipment".

[Corporate666]

What is the isolation supposed to be between the 3 channels?

[rs20]

 Having to carefully account for and make sure that every electron has an easy path back to its "home" negative terminal seems a bit nuts.

Strongly disagree.

Considering the return path of a current is part of good design.  And it's even more important at RF.  And these days most every circuit is an RF circuit.  (If you think not, see if it works with a GSM cellphone next to it.  Fluke 87 V anyone?)

Even at DC, you have to consider the I*R voltages in your "ground" connections and decide if you can ignore them.

Orin.

So of course, when you're designing a PCB, thinking about the ground return paths within the PCB is super-important. And when you're dealing with RF, the supply-decoupling cap-ground loop is super important, yes yes yes. But we're talking about an external power supply, if your circuit is subjecting your power supply to RF-specific conditions, then you've done your decoupling on your PCB wrong.

And at DC, I can calculate my I*R drop in supply cables, and I *should* be able to subtract those drops from the figures displayed on the DP832 readout. Like Dave said in the video, the great thing about these precision supplies is that they essentially have a multimeter built in -- but this one specific condition breaks this multimeter. And all of this is completely avoidable, if only they had properly shorted out the negative terminals with a strong, thick copper wire, this problem would just vanish. Even just 24AWG would massively help.

So yes, the way I worded that sentence is misleading, I'm not trying to claim that assuming ground is ground is the same voltage everywhere should be allowed, that's obviously a terrible idea in the context of designing and laying out PCBs -- but having two terminals marked as "common" which are so ridiculously far (38 milliohms) from being common, and so unnecessarily, is crazy in my books.

[rs20]

What is the isolation supposed to be between the 3 channels?

All channels are isolated from mains earth. CH1 is isolated from CH2 & and CH3, but CH2 and CH3 are *not* isolated from each other -- they have a common negative terminal. For completeness, all terminals are marked with "max +/- 240V to earth".

[rs20]

Lets face it, RIGOL stuffed up in a big way with the LM317 . The spot light is now on  and everyone is going microanalyse this bit of gear.
This is the price of not getting it right first time.

The funny thing is, the reason I found this out is because of the first, non-critical teardown eevblog episode. I saw that the the current sense resistors were on the negative rails, and was wondering how that could possibly work.  Turns out it's totally legit, except I noticed this unrelated problem!

[free_electron]

if this is true ( current flowing through the sense wires) then this is a catstrophic design failure !

the normal sense configuration in a system is like this:

Code: [Select]

output

(+S)--o-----[50k]--------------------------
      |                                    |
     [5k]                                  |
      |               _________        ____|___
(+)---o--------------|         |      /        |
                     |driver   |_____/sense    |
                     |         |     \control  |
(-)---o--------------|_________|      \________|
      |                                    |
     [5k]                                  |
      |                                    |
(-S)--o-----[50k]--------------------------
if you leave the sense wires floating the supply sense through the 5k and 50k resistors ( the values i give are average). most supplies i know have something in the range of 1k to 10k between sense and power terminals. from the sense lines towards to sense inputs themselves this is more in the order of 20k to 1meg and beyond ( you are looking ino an opamp so essentially it should be almost infinite impedance)

if you connect sense wires  you create a short across the 5k resistor. the cable resistance is so low compared to the parallel sense resistance it has no noticable impact.

so, if you can indeed pull current from the sense wires ... please doublecheck it ... then this is a catastrophical failure.

now, it is possible there is a shunt diode !

Code: [Select]

output

(+S)--o----o---o------[50k]-----------------
      |   _|_  |                            |
     [5k] /_\ _V_                           |
      |    |   |       _________        ____|___
(+)---o----o---o-----|         |      /        |
                     |driver   |_____/sense    |
                     |         |     \control  |
(-)---o----o---o-----|_________|      \________|
      |    |  _|_                           |
     [5k] _V_ /_\                           |
      |    |   |                            |
(-S)--o----o---0-----[50k]------------------
in which case this behavior could be normal. there should be no more than 0.5 volts drop between sense and power terminals. the diodes will clamp that.

in that case you can pull current from the sense wires but there should be hal f avolt drop (0.3 if they used shottkys)

this is done to speed recovery of the sense amplifier and protect the control oop. ideally thse diodes should connect after the 50k resistor but i have seen them before at the sense pin itself ( mainly when they were inserted as an afterthought)

so, doublechekc that stuff before drawing conclusions. the diode may be there.

if it is really possible to suck the full current withouth the diodes ... baaaaad design !

[orin]

Sometimes I just want to see if it'll release the magic smoke...

I have a stack of expensive test equipment just to work out why my projects don't work.

Dangerous. You have a "stack of expensive test equipment" to make sure you'll never be able to use an "assortment of affordable equipment".

OK, I'm guilty of seeing if a project lets out the magic smoke too

I should have left out the adjective 'expensive'.  In reality, almost all* of the stack of test gear is used and I fix it myself.  It would be expensive if I bought it new.  It is in fact affordable or I wouldn't have it.  I have no objection to affordable equipment if it works.

Orin.

*so I got an Agilent 34461A.  First new equipment in about 15 years.  Dave's review had no influence on that.  No, none at all.

[orin]

And all of this is completely avoidable, if only they had properly shorted out the negative terminals with a strong, thick copper wire, this problem would just vanish. Even just 24AWG would massively help.

But wouldn't that invalidate the readings from the low side (as you note in a later post) current shunts?  They'd have to use a high side sense in that case with all the problems that entails.

Orin.

[free_electron]

And all of this is completely avoidable, if only they had properly shorted out the negative terminals with a strong, thick copper wire, this problem would just vanish. Even just 24AWG would massively help.

But wouldn't that invalidate the readings from the low side (as you note in a later post) current shunts?  They'd have to use a high side sense in that case with all the problems that entails.

Orin.

ehhh.. sensing is supposed to happen on BOTH sides ! or it's not sensing at all !

[rs20]

if this is true ( current flowing through the sense wires) then this is a catstrophic design failure !

So, to clarify, this supply does not have externally accessible voltage sense terminals, so the circuitry you're suggesting doesn't apply. What's happening with the DP832 is that the internal voltage sense wires, which are soldered directly to the output terminals (thereby measure voltage at the terminal rather than at the regulation PCB), are actually carrying 3A if you connect CH3+ to CH2-. Specifically, the CH2- and CH3- sense wires are tied together at the PCB (this is "fine" because the - terminals are common), but the point is that one connection point is the only connection between CH2- and CH3-, hence it has to carry any current returning on the "wrong" terminal -- not that I think there should be any such thing as the "wrong" terminal. I'll upload a video soon that might give a touch more context.

[rs20]

But wouldn't that invalidate the readings from the low side (as you note in a later post) current shunts?  They'd have to use a high side sense in that case with all the problems that entails.

Orin.

Forgive me, I'm not sure which later post you're referring to. So to answer this question;
a) The negative terminals are already shorted together by the sense wires, it's just a very high-resistance short (at 38 millohm).
b) In the instances where the two channels are used independently to power independent circuits, the voltage-sense wires and my proposed thick wire do/would both carry zero current, so there'd be no change; this scenario would continue to behave perfectly.
c) In the instance where the current is returned to the "wrong" terminal, the voltage drop between terminals is due to the high resistance of the sense wire. Reduce this resistance, and the problem reduces. There's no point in forcing the current to travel all the way to the PCB, to the point where the voltage sense wires are crimped together, and then all the way back again.

I'll post a reverse-engineered schematic in a sec.

[rs20]

Just to clarify things, I did some other testing and threw together and attached a rough reverse-engineered schematic.

Also, here's a video demonstrating the issue:

[Spikee]

Shouldn't that be an obvious error for the designers ?
It seems like they didn't even bother to test the device before full on production...

[sync]

Just to clarify things, I did some other testing and threw together and attached a rough reverse-engineered schematic.

 

I think i wouldn't make such a mistake. And i'm just a hobbyist. If i would do then i surely found it while testing.

PS: The boiling water demonstration is great!

[Rufus]

if you source current from Ch3+ and sink it into Ch2-, the current flows through the voltage sense wires and induces a drop (and therefore error in voltage readout/regulation)

The real fail (ignoring lack of isolation between the two supplies) is having two terminals for the common connection. It doesn't provide any useful function, just tends to confuse the user.

Having significant resistance between the two terminals is a secondary failure.

[rs20]

You wouldnt think remote sense would be needed for a 3A supply. 1 + 1 =2 Feet of AWG18 would be about 40mV. Remote sense can really hamper dynamic response and is more likely to cause stability issues.

Nono, this power supply doesn't have remote sense wires. Sorry, I should have been clearer. I'm referring to the sense wires internal to the supply -- they measure the voltage right at the terminals, so the voltage drop on the leads from PCB to terminals aren't a problem.

In any case, 40mV exceeds the 10mV accuracy spec for the supply, so the sense wires are totally legit.

[rs20]

The accuracy spec is probably for at the terminals. You can’t really cant really spec accuracy any other way then where it is sensed.

Yep, it's good that the spec is for the voltage at the terminals; it's very bad that the reading can be way out of spec (the topic of this thread)!

[AndersAnd]

Recommended Connections: DP832 CH2 and CH3 COM Terminals

http://beyondmeasure.rigoltech.com/acton/attachment/1579/f-034c/1/-/-/-/-/DP832%20Proper%20Connections.pdf

[rs20]

Yeah, I saw that... I think it's pretty poor, really. "Hey, our power supply has a massive flaw where the voltage sense wires are carrying 3 amps if the user uses a common ground return lead, let's just declare that sort of connection to be 'incorrect'. Problem solved!"

[rs20]

I threatened it in the original video, and it's only been several months, so I figured it was time! I soldered a bit of 3mm (!) copper wire between the two negative terminals, and the problem is fixed, no surprises. Post-fix, voltage sag on channel 2 is about a millivolt or so (noise floor of my cheapo DMM, compare to 57mV dip pre-fix) when injecting 3A up where it doesn't expect it. A video demonstrating the (more-or-less ideal) behaviour of the supply after the fix is below.  No, there's no video of the fix itself, sorry! I don't have Dave's talk-and-fix-and-operate-camera-at-same-time skillz...

[eurofox]

I threatened it in the original video, and it's only been several months, so I figured it was time! I soldered a bit of 3mm (!) copper wire between the two negative terminals, and the problem is fixed, no surprises. Post-fix, voltage sag on channel 2 is about a millivolt or so (noise floor of my cheapo DMM, compare to 57mV dip pre-fix) when injecting 3A up where it doesn't expect it. A video demonstrating the (more-or-less ideal) behaviour of the supply after the fix is below.  No, there's no video of the fix itself, sorry! I don't have Dave's talk-and-fix-and-operate-camera-at-same-time skillz...

Hi,

I intend to buy such a power supply, maybe someone who bought it recently could confirm or not that it is fixed in the mean time?

eurofox

[sacherjj]

I intend to buy such a power supply, maybe someone who bought it recently could confirm or not that it is fixed in the mean time?

I will be checking mine out when it arrives.  But the ship notice for my DSO4054 also includes a ship date of 2014-05-02 for the DP832.        Looks like the DM3068 is also 2014-05-02.   

Maybe they are reworking the DP832 again.  Or TEquipment just has stock issues.

[tequipment]

We.. tequipment.. are backordered.. we cant keep up right now with depand.  I hope to get ahead of it soon.

Thanks
Evan

[grego]

No worries Evan. You guys always treat us right.

[Sparky]

I intend to buy such a power supply, maybe someone who bought it recently could confirm or not that it is fixed in the mean time?

Hi eurofox, there were some similar questions about the DP832 status, which I replied to here.

Also, you may wish to follow my thread on DP832 firmware bugs and hardware issues here.  I will try to keep the thread up-to-date.

As for the specific issue you mention (design fault related to current sense between GND terminals of channels 2 and 3), we don't know whether Rigol will correct it in a future hardware revision, but so far there are no reports of them fixing this issue.  If I have to guess, I would say it is not likely they will provide a fix --- they had a chance to do it when the LM317 heatsink problem was addressed, and also Rigol has replied directly about this issue in "DP832 Recommended Connections.pdf" (see my thread linked above, or post #30 above by AndersAnd) in which they suggest installing a heavy duty GND wire yourself.

Hope that helps!

[eurofox]

I intend to buy such a power supply, maybe someone who bought it recently could confirm or not that it is fixed in the mean time?

Hi eurofox, there were some similar questions about the DP832 status, which I replied to here.

Also, you may wish to follow my thread on DP832 firmware bugs and hardware issues here.  I will try to keep the thread up-to-date.

As for the specific issue you mention (design fault related to current sense between GND terminals of channels 2 and 3), we don't know whether Rigol will correct it in a future hardware revision, but so far there are no reports of them fixing this issue.  If I have to guess, I would say it is not likely they will provide a fix --- they had a chance to do it when the LM317 heatsink problem was addressed, and also Rigol has replied directly about this issue in "DP832 Recommended Connections.pdf" (see my thread linked above, or post #30 above by AndersAnd) in which they suggest installing a heavy duty GND wire yourself.

Hope that helps!

Thanks Sparky

[Macbeth]

I intend to buy such a power supply, maybe someone who bought it recently could confirm or not that it is fixed in the mean time?

I will be checking mine out when it arrives.  But the ship notice for my DSO4054 also includes a ship date of 2014-05-02 for the DP832.        Looks like the DM3068 is also 2014-05-02.   

Maybe they are reworking the DP832 again.  Or TEquipment just has stock issues.

Snap!

I've got a DP832 and a DM3068E on order with TEquipment with an estimated shipping date of 2014-05-06! I placed my order on 2014-02-28 and only a couple of days before could swear both were in stock.

Oh well, I've checked other Rigol suppliers both closer to home and China, but none of them have any real stock either.

My problem is I paid paypal to make things easier for us. But it just means my cash is in TEquipments pockets until despatch, whereas if I paid with credit card then the charge would be authorised but not taken until despatch.

Having worked in telephone sales/accounting for a small company, I can tell you there was never a deliberate policy, but people who only paid deposits or cash on delivery got the goods quicker than anyone who paid upfront and left us to it without nagging us. Anyone who paid bank transfer or charge card in full - welcome to the back of the queue (paypal didn't exist then).

So, I feel this will put me at the back of the queue. Can I reverse my paypal and then give credit card details seeing as this cash is tied up for so long? Problem is paypal rips people off on everything and reverse transactions incur inflated forex fees so not so good.

I appreciate the delay is none of TEquipments fault but would like to be reassured that first come first served is somewhat the policy?

The only good thing I can think of is Rigol are improving their products, hence the manufacturing delay? (Clutching at straws!)

[echen1024]

Wait, Macbeth, so you ordered in the future?

[Sparky]

I appreciate the delay is none of TEquipments fault but would like to be reassured that first come first served is somewhat the policy?

If you read above, Evan from TEquipment.NET has said stock is simply backordered.  TEquipment.NET is a huge distributor with a solid reputation.  I've never had any problem with them, and cannot imagine them pulling any dodgy business like you're thinking --- the queue is FIFO.  So yes, you're at the back of the queue, behind everyone who ordered before you.

[tequipment]

we have a few hundred coming in on this shipment with about 100 left so we are good.. ( when they finally get there )
Thanks again for the business.

I just got 5 kids to bed.  After 11 hours of work I'm dead tired. 
Cheers and many thanks for the business!
Evan Cirelli
TEquipment.NET

[Macbeth]

we have a few hundred coming in on this shipment with about 100 left so we are good.. ( when they finally get there )
Thanks again for the business.

I just got 5 kids to bed.  After 11 hours of work I'm dead tired. 
Cheers and many thanks for the business!
Evan Cirelli
TEquipment.NET

      

Thanks Evan and Patrick for sorting out a switch from paypal to credit card for my order. I really look forward to it! 

[tequipment]

The credit should have been done today..

Thanks
Evan

[Rigby]

(design fault related to current sense between GND terminals of channels 2 and 3), we don't know whether Rigol will correct it in a future hardware revision, but so far there are no reports of them fixing this issue.  If I have to guess, I would say it is not likely they will provide a fix --- they had a chance to do it when the LM317 heatsink problem was addressed, and also Rigol has replied directly about this issue in "DP832 Recommended Connections.pdf" (see my thread linked above, or post #30 above by AndersAnd) in which they suggest installing a heavy duty GND wire yourself.

It's a design flaw; it won't be corrected until a new model is released, if then.  A fix to the DP832 would require more work than they'd be willing to put into a model that costs as little as the DP832, given that there is a workaround.

More costly models would get the attention that their cost demands.  This supply isn't worth the effort, likely. 

Just be aware of the issue this supply has and you've armed yourself with everything you need to use it effectively.

[neslekkim]

So I found this thread while I'm reading up on this psu, so I guess the situation still is the same on this?
Not sure in what situation I would combine ch2 and ch3, but I would use an combination of ch1 and ch2 to provide -15/+15 power, but I guess I would use ch3 for powering the 5v part of the same circuit, would I be in this situation then?

on the DP831, did they change this there?, the ch1 is the low-voltage channel, and ch2+ch3 is the ones with higher voltage, and common ground it seems from the pictures.

Maybe just continue looking...

[rs20]

So I found this thread while I'm reading up on this psu, so I guess the situation still is the same on this?
Not sure in what situation I would combine ch2 and ch3, but I would use an combination of ch1 and ch2 to provide -15/+15 power, but I guess I would use ch3 for powering the 5v part of the same circuit, would I be in this situation then?

Yes, that's exactly the problem use-case I suggested in my video, +/- 12V with a 5V logic rail on the same ground.

on the DP831, did they change this there?, the ch1 is the low-voltage channel, and ch2+ch3 is the ones with higher voltage, and common ground it seems from the pictures.

Don't know; the DP832 says it has common grounds on CH2 and CH3, and it does: but only via the voltage sense lines.  The fact that the DP831 says it has commoned grounds doesn't imply much.

Maybe just continue looking...

You're not going to match the features and accuracy of the DP832 for the cost. The issue is A) limited to a 57mV error which is insignificant in most cases, B) easy to avoid by returning the current to the ground from which most of the current is coming, and C) easy to permanently fix as well as I outlined in the following video. So overall, I'd still choose the DP832 any day, unless someone can suggest a concrete alternative (ironically, there's a lot of "my old analog power supply was fine" sentiment, where we're talking about a 0.057V error!!!)

[neslekkim]

Thanks!, that is good to know, 57mV is of no concern , I was more worried about the 3Amps on the sense wires, but doing it "properly" according to the guides that Rigol gave out earlier it should be ok. Will check out your video also.

I checked with the Rigol pusher in Norway, and find that the dp832 is very affordable.. even for this country.

[PioB]

So if I understand correctly, if one wants to drive dual rail Op Amps and some digital logic (with split analog/digital grounds), one would be better off with something like the GW Instek GPS-3303S or an QL564T by Aim-TTi, without common grounds, but then one would be looking at two to three times the cost of the rigol?
 (Power supplies taken arbitrarily from a Swiss distributor website, I checked for 3 outputs and > 195 W and the price not in the stratospheric range)

[rs20]

So if I understand correctly, if one wants to drive dual rail Op Amps and some digital logic (with split analog/digital grounds), one would be better off with something like the GW Instek GPS-3303S or an QL564T by Aim-TTi, without common grounds, but then one would be looking at two to three times the cost of the rigol?
 (Power supplies taken arbitrarily from a Swiss distributor website, I checked for 3 outputs and > 195 W and the price not in the stratospheric range)

If you look at my more recent videos, I demonstrate how to fix the problem with a single wire once and for all. Also, you can install a wire externally (shorting the two common rails) if you don't want to open up the supply. Finally, in the example you cite, the current flowing to the 5V logic will be very small, so if you return the ground wire to the negative terminal corresponding to the 12V supply (i.e., channel 2 negative), then it will appear to work fine. Also, even in the worst case, when 3A is flowing, the errors introduced are quite negligble for most applications (57mV max error).

So no, there are plenty of fixes, workarounds and reasons why this isn't actually a deal-breaker.

[PioB]

Thank you very much for the information. I am still on the fence.
How well does the current limiting work, as the output capacitor is on the terminals after all the regulation circuits? With the 3633A at work, I put in 20 V, max current 20 mA, turn the output on and  connect A LED. It immediately regulates to 20 mA and the LED (or sensitive circuit) survives.
Is this a problem while operating the 832? If I understood the information from Dave's Video correctly, by the time the (I don't remember the exact capacity from the video) 1 or 10 mFD dumped its 30 Volts on the circuit, it is too late for the regulation?

I always have to remember that a 3631A is four times the cost and only delivers 1 amp on the +-25 v rails... (20 year old design with GPIB/RS232 instead of USB/LAN... The common on the high voltage channels and the isolated 6V are an advantage though.
As I said, the more I read, the less I know whether the holdups are relevant in practice or imagined... One would need to "testdrive" one

[rs20]

You should open up the 3633A at work and see how much capacitance it has it its output terminals, I'd be very interested to find out! I'm sure it's possible to make a supply with minimal output capacitance, but it's certainly not schoolkid-trivial to make a supply that can instantaneously transform between being constant voltage (i.e. infinite capacitance) to constant current (i.e., infinite inductance). Believe me, I spent a whole 30 minutes in LTSpice once, got nothing but oscillation in one mode or the other (or indeed, both). But like I say, there will be topologies out there that work, but they must be more expensive somehow.

In practice, I rarely find myself (deliberately) using CC mode. There's one exception, which is my massive 6 amp LED, but in that case I keep the CV setting close to the nominal forward voltage of the LED. If I was driving a simple red LED, I'd set the voltage to 5V or so. Setting it to 20V and relying entirely on the CC mode is a perverse thing to do. If you're working with something ridiculously sensitive, plug it in first and then enable the output on the PSU. Oddly I don't have any LEDs with me, it'd be interesting to see just how high I'd have to bring the CV voltage before it started killing LEDs, I'm not entirely convinced it'd do so even at 30V.

I don't think you need to test drive, what more would you learn by having a real DP832 in front of you? You need to think about what settings you'd sensibly use and when/why you'd be relying on CC mode. The energy that'll be delivered to faults in these conditions can be calculated.

[PioB]

Hi,

Ok, I was wrong, thank you very much for the explanation! At 20 Volts, the LEDs burn without fail, at 8 Volts it's better (no statistics there ) According to the service manual, the 3633 has 470 uF and 1 uF on the output.
So I got convinced and got probably the last DP832 from Conrad in Switzerland before the price increase today. (470 -> 700 Fr.  )

You should open up the 3633A at work and see how much capacitance it has it its output terminals, I'd be very interested to find out! I'm sure it's possible to make a supply with minimal output capacitance, but it's certainly not schoolkid-trivial to make a supply that can instantaneously transform between being constant voltage (i.e. infinite capacitance) to constant current (i.e., infinite inductance). Believe me, I spent a whole 30 minutes in LTSpice once, got nothing but oscillation in one mode or the other (or indeed, both). But like I say, there will be topologies out there that work, but they must be more expensive somehow.

[rs20]

Ok, I was wrong, thank you very much for the explanation! At 20 Volts, the LEDs burn without fail, at 8 Volts it's better (no statistics there )

Which power supply was this? What was the current limit you chose?