Fixing Rigol DS2000 design/manufacturing flaws

[Behemot]

I've run into this wonderfull stuff at EEV forums http://www.eevblog.com/forum/project...sign-problems/

Basically it shows how screwed the DS2000 is, as well as the DS1000. Well I would expect slightly better for my 1000 bucks than 8 of 8 regulators oscillating Another near-future project. As I'll be getting components for my next major project from Mouser I think I'll order the SMD components for the O-scope as well.

With so much stuff to work on you can start betting if I'll first manage to learn how to operate the diff-probe or fix the O-scope itself

[ITelektro]

There's a reason why Keysight, Tek ...... costs more. Keysight 2000x which is in a class to this Rigol costs 30% more. Rigol's not bad, but final inspection of the product as it should be, so mistakes like this happen .... too often.

It tickles my imagination, what major project?

[Behemot]

I am afraid those cost a lot more than just 30 %. Here for sure.

Load tester 2.0. Won't be as good as what I have planned so that version will be 3.0. On the other hand it will act as a testing prototype so in the 3.0 I could implements tweaks if needed. Who knows, maybe the 3.0 will already be good enough to provide cheap (non-programable) alternative to things like Sunmoon programmable loads.

The projected load capability is for 1.5kW rated units with nominal voltages.

[ITelektro]

Yes, you're right. I intended to buy 2000x series. The offer was somewhere around 1500e. I had him short on the test. I think it's worth the money.

If well designed I see no reason why would not be as good as professional solutions. It's not rocket technology. The principle of operation of both is identical. Just be careful when selecting mosfert and operational amplifiers.

If you own a programmable power supply and signal generator you can operate with dc load through these instruments.

What components did you pick?

[Behemot]

You are right, I would gladly pay 10 % more if they cared to check it so there are not so incredible flaws like this. Though it seems that only some are affected, maybe just a single batch as some people reported their units were running fine, no regulator oscillations. But well, it seems that Rigol is still the better of many chinese brands…and than if I'd want some 4channel with 1 GSa/channel the price difference for Tek or something similar is getting really painfull…

As for my parts, besides some usable current shunts also TO-247 thick-film resistors. That's for the passive part, for the active part I hope to tweak the circuit I have here (and which does not work ATM ) using some of the shitloads of MOSFETs I have from UPSes etc. ditched for parts. So don't need much in this area. Basically a linear load driven by reference so the load stays more constant. Basically there will be just one more or less preciselly driven active load dialer per rail. That is enough, to increase the load in steps I could easily just turn on the resistors one by one. Why do it more complicated than it needs to be? Resistors survive almost anything. Transistors can be blasted and it would be pain in ass to replace them if I only used active load for the whole load.

As for panel voltmeters it turned out very interesting. I'll go for some 5digit from fleebay. While the real accuracy may be way worse than rated 0.3 %, as it has one digit more it is still way more accurate than all the 4digit ones from big-ass resellers. And I will get them for abour 1/8 the price than even from Conrad which carries some nice and cheaper ones. Thats frelling huge difference!

[ITelektro]

Resistors survive almost anything. Transistors can be blasted and it would be pain in ass to replace them if I only used active load for the whole load.

Well .... it depends. If you design for yourself you can oversize. Trust me, it will be still far away from the purchase price of Sunmoon.
I decided to MOSFET design. Each MOSFET theoretically can sustain about 80A @ 60c in the linear mode(As I remember). I made a modular design, each module by two MOSFETs. For the 12V rail have 4 modules, ie 8 MOSFET which provides theoretical power over 7kw. Clearly, I can not cool down with CPU cooler so I limited the whole thing to 120A. Theoretically, one module could work on that power. Four modules on that power is practically indestructible.

This is a picture while it was still under construction.
Attachment 2433

As for panel voltmeters...

I've try some from ebay and just given up. They are so inaccurate that it simply there is no point to use those. It crossed my mind to buy a cheap uni-t multimeter and put it in case. If you find something acceptable, paste link.

[Behemot]

Even if they will be 1 % (and I do not think it will be SO bad) it is still less than about 3.5 % the clamp meter has. So I should still be finally able to measure efficiency with reasonable accuracy, not like now (calculated about +-5 % ).

Ya really have small heatsink for that. I'll be using CPU boxes, those AMD you have also in a line. But mounted from the outside (so the heat will go through the case wall and than to the heatsinks, plus the case could also conduct some heat away). About 20 of them and with some high-speed industrial 12cm fans

The problem with transistors is they are rated high current, but that's for moving the power elsewhere. Here you burn the power on the transistors, they act as dynamic resistors and have to disipate all the power away, not only the losses. But TO-247 is only good for something like 200 W, depending on the heat resistance. So you can never burn 80 A@12 V on them for more than a very short while as they would melt.

[ITelektro]

Even if they will be 1 % (and I do not think it will be SO bad) it is still less than about 3.5 % the clamp meter has. So I should still be finally able to measure efficiency with reasonable accuracy, not like now (calculated about +-5 % ).

Accuracy is not a problem, but precision is. Accuracy is not something that worries me, for this application 1% is quite acceptable, while large differences in precision, at least for me, is never acceptable.

Ya really have small heatsink for that. I'll be using CPU boxes, those AMD you have also in a line. But mounted from the outside (so the heat will go through the case wall and than to the heatsinks, plus the case could also conduct some heat away). About 20 of them and with some high-speed industrial 12cm fans

It will be enough.

The problem with transistors is they are rated high current, but that's for moving the power elsewhere. Here you burn the power on the transistors, they act as dynamic resistors and have to disipate all the power away, not only the losses.

That's why I wrote at the beginning that it is important to choose right components, FET that is designed to work in a linear mode. They are designed to work in these conditions.
The disadvantage is actually the operating temperature that affects the total power FET can dissipate.

Benefit is much larger, the possibility of precise regulation, price, handling, the ability to easily perform transient tests, etc.

[Behemot]

With high power resistors the transient test can be done equaly good. It also applies almost instant load for the PSU.

I generaly consider accuracy and precision the same but it seems it has different meanings regarding to measurement errors. In such case, I mean precision of course.

[quest for silence]

https://en.wikipedia.org/wiki/Accuracy_and_precision