0.13x vs 1.0x gain op amp regulated power supply - fail

[HendriXML]

I could just fit 2x parallel resistors of 15K. Without desoldering the previous ones.


The new results amaze me! The response has improved a lot! Wasn't expecting that  .

-- So I verified the test setup.... no magic to be seen here..

My circuit 2 not saturated!

0.13 V

0.12 V

0.10 V

0.08 V

[HendriXML]

Hmm made a mistake with the signal generator in these test runs: the load Mosfet did not become saturated..
The table with comparable conditions is:

Resistor	Siglent powersupply	My circuit	My circuit 2
5R1	2.08 V	0.60 V	0.78 V
6R8	1.90 V	0.50 V	0.65 V
10R	1.62 V	0.40 V	0.48 V
15R	1.36 V	0.30 V	0.29 V

It's worse as one would expect, but not much! With less amps its on par.

[HendriXML]

So I altered the circuit to have the lowest gain possible (by looking at the curve). Now I wanted to check whether it could go without an endcap (C7): it now seems to be stable at resistive and inductive loads.

I attached some scope images to show how it responds to sudden loads. It’s not good (the scales are very different!) but it stabilizes nicely.

These results should be discarded, reason down below.

[HendriXML]

I was doing some measurements and things didn't add up. Because of the better resolution I did the measurements in AC mode. But sometimes that hides stuff you would like to know. One thing that would have shown otherwise was that the capless measurements were taken when the circuit did not regulate properly. One of mod resistors had a loose contact.  The other measurements after the mod, seem to be fine. I did some retesting and got even (a bit) better results.

But this thread was not started to make the world’s finest supply, it is mostly about a circuit which can give some insight in how to model the amplification of a IRLZ44N. Because it was at first not stable without a cap, the needed measurement could not be done. I'm optimistic that there's now a situation in which some interesting conclusions can be draw. There're probably books which provide much better information  . But this is more fun.

Back to the –regulated- capless response:
I posted 2 curves, the difference are the reference caps (C1, C4, C5, C6).  The first one is with C4, C5, C6 on board, thus 300nF. The second is with only C4, thus 100nF.
It is clearly to see the caps need to charge up before a new balance is found. This takes about 200us per 100nF. Because that is a relatively long time, I’ll assume the ref voltage as being constant, when analysing the first few us. This will be done in following posts, but it might take a while.
Image 005 shows 2 curves

• Green: the drain of the pull down (load) Mosfet
• Yellow: VCC out

The difference between those is the voltage over a 5R1 resistor and should give us the current drawn at each moment. Will make some graphs to show this more clearly!

[HendriXML]

The question I want to answer is what range it can regulate without significant drops. Somehow I think the integration of the ratio curve in the other post has a relation to it. My thoughts are if it’s less than 1 (or 0.5 because of the resistive divider), it’s okay. If its more, the reference capacitor needs to charge a bit further, so the output voltage will drop.

From the measurements done above this does not seem the case.
That raises also the question about how important the raising of the ref voltage is.
Or wether it is possible to drive the Mosfet with even lower "gain".

To bad I've soldered the resistors down so that isn't just a quick check. The legs are also the traces...perfboard experimenting issue.

My initial "hypothesis" might still be correct! The endcap had such a huge impact, that no conclusions could be drawn. If the integration/integral stuff seems to be true, then I should somehow incorporate that in my "mental model" of "amplification". A empirical answer needs probably also some more (advanced) testing. Maybe one of you already knows a theoretical one?

The question is - given the curves in the posting below- how much gain is needed to keep a steady Vsd (and Vout from not dropping) in a given range of Vsg (which corresponds to a initial amperage - new amperage). Without the effect of the "delay" of the amplification/sensing, so keeping time out of the picture.
https://www.eevblog.com/forum/testgear/this-graph-brings-tears-to-my-eye/msg2189625/#msg2189625

[HendriXML]

Made the current graph (purple, right axis).

It shows that initially a small change in current, means a large drop in the output voltage. This because the initial Vsg results in a high Rsd resistance (about 2k). In this regard resistance means voltage rise/amp (-> output drops). A tiny 1 mA would thus mean a 2V rise! But Vsg is regulated and climbs, but it's clearly not enough:
This can be caused by at least 2 effects:

• The gain (sensing Vout drop -> rising Vsg) is to low
• There's a delay (between sensing en responding)

Anybody any ideas?

Will (also) try to unravel the problem myself.

[HendriXML]

I think I found a way to unravel the mystery of what gain is needed when an "amperage jump" is made.

Will do that in the other thread.

https://www.eevblog.com/forum/testgear/this-graph-brings-tears-to-my-eye/msg2203539/#msg2203539

[HendriXML]

Also, the op-amps won't have to be powered at the full input voltage.
The balanced amplifier would need to have near perfect CMRR to prevent input voltage variations affecting regulation voltage.

When I was done experimenting, I thought why not see wether this circuit can operate on an adapter. The adapter is about 9V and I added 2 smoothing capacitors of 2200uF.
The result was that the output became very “noisy” with loads, I don’t think there’s even a repetition going on.

So as an experiment and a way to check out idea’s it did a good job. But as a proper supply I would call it a fail.

As a battery voltage dropper its probably ok, I will have to check that. But I think its going to be recycled 

Could you give an advise on an op amp which can regulate well on about 5-7 V, without being to sensitive to variations? Would be nice if it is THT and can be bought on AliExpress. But I wonder if it’s not the time get stuff elsewhere.

[Kleinstein]

The circuit has a more general problem than just finding a suitable OP.  I think it would be a good idea to take a step back and look at other, more normal voltage regulator circuits or read some more about the theoretical background.

Electronic systems with feedback can be quite complicated to understand - the theoretical background is only from the first half of the 20th century - so much of it is less than 100 years old. I don't know your mathematical background - so it's hard to give a suitable way learn about it.

It would probably help to start with something like stability in OP circuits.  The simplest rule of thumb I know is aiming for a system that is slow at only one point and fast everywhere else. This leads to the simple case of dominant pole compensation.


For simple test on the bread board, I would keep it relatively slow, so OPs with a GBW of something like 0.3 - 2 MHz.  Though not ideal the LM358 is a good start for a voltage-regulator on the bread board.

[HendriXML]

The circuit has a more general problem than just finding a suitable OP.  I think it would be a good idea to take a step back and look at other, more normal voltage regulator circuits or read some more about the theoretical background.

Electronic systems with feedback can be quite complicated to understand - the theoretical background is only from the first half of the 20th century - so much of it is less than 100 years old. I don't know your mathematical background - so it's hard to give a suitable way learn about it.

It would probably help to start with something like stability in OP circuits.  The simplest rule of thumb I know is aiming for a system that is slow at only one point and fast everywhere else. This leads to the simple case of dominant pole compensation.


For simple test on the bread board, I would keep it relatively slow, so OPs with a GBW of something like 0.3 - 2 MHz.  Though not ideal the LM358 is a good start for a voltage-regulator on the bread board.

Thanks for the response, the LM358 seems a good one to experiment with.
As for my background, I’ve got me brain damaged, so I’m somewhat limited in my abilities to learn/comprehend stuff and the math stuff I did know has become a bit vague. So I’m just fine by doing experiments and to see how an idea turns out. As in this case: as a supply it may have failed. But checking my understanding about voltage rise/drop sensing and needed gain as an experiment was for me a success. If I now read something about this stuff I can link it to this practical experience (with blog!).

[xavier60]

Also, the op-amps won't have to be powered at the full input voltage.
The balanced amplifier would need to have near perfect CMRR to prevent input voltage variations affecting regulation voltage.

When I was done experimenting, I thought why not see wether this circuit can operate on an adapter. The adapter is about 9V and I added 2 smoothing capacitors of 2200uF.
The result was that the output became very “noisy” with loads, I don’t think there’s even a repetition going on.

So as an experiment and a way to check out idea’s it did a good job. But as a proper supply I would call it a fail.

As a battery voltage dropper its probably ok, I will have to check that. But I think its going to be recycled 

Could you give an advise on an op amp which can regulate well on about 5-7 V, without being to sensitive to variations? Would be nice if it is THT and can be bought on AliExpress. But I wonder if it’s not the time get stuff elsewhere.

I went through the same process a while back, trying my own ideas and failing because I didn't know enough to understand the challenges involved.
I settled on the well known floating type topology and successfully completed a  bench supply project
Although the performance is good, I found debugging to be awkward, having to take measurements with respect to the positive output terminal.
The LM358 is more than good enough for experimenting. It operates from 3 to 36 volts. One important op-amp parameter to be aware of is the input common-mode range. It includes ground voltage for the LM358 which can be very useful in certain applications.
You will lean a lot from data PDFs.

Are you just experimenting or intending to construct something?
 

[HendriXML]

Are you just experimenting or intending to construct something?

Just experimenting yet, at least until I got some “model” of how the components really work. On the way it is nice to end up with something that works, at which you can look back at. I already have a fume extracter for instance which I’m happy about. It’s starts when the you take the iron out of the stand, and it keeps blowing for a while after you put it back. This gives the smoke time to travel through a ventilation hose ending outside. In practice I don’t use the hose, but its much better than nothing. It has also dimmable LED’s (variable current).  Even used diffusion sheet to have less reflection on those tinned area’s.
I was making some boxed modules with logic gates, LED output, switches and stuff for my son (9) to play with. But he’s not really attracked to it. I think as a toy it needs to look more simple with only the printed symbols, ins and outputs. The prototypes are with lot of transistors, no ic’s.
For safety I want it battery powered, so if the circuit drops the voltage in a stable way I might use it. But I may also drop the whole idea. He plays this game Scrap Mechanic where he can do “virtually” much more cool stuff. And I don’t see any reason why playing with the boxes I (would) made would be more fun. Times have changed.

[HendriXML]

About the circuit: the next time I will use a larger perf board. Also first deciding which components might be swapped and use pinheaders for them. Using dupont female connecters to make those components swappable or directly solder them on the pins.

On breadboard's I've got mixed feelings. If you've got cheap resistors they are a bit loose (I bend them, doubling the thickness) adding components on heatsinks is also somewhat complicated.

[HendriXML]

The result was that the output became very “noisy” with loads, I don’t think there’s even a repetition going on.

I replaced the Mosfet, and indeed the strange noise went away. The circuit was somewhat stable using an adapter, but had a 100 hz ripple. Possibly the op amp needs a stable powersource. The used op amp is also not very suited for the usage, but I had it laying around.

Using a C1 capacitor: 10 nF made the circuit oscillate. 4.4 nF was OK, but did not enhance the performance.
So I won’t post any more graphs.

Working with header pins and dupont connectors worked fine. Having a wire with one dupont connecteron it makes it much easer to place header pins and keep them steady while soldering. The wire can for instance go down in another hole adding some pull to the pin.

[HendriXML]

Some conclusions that I draw from this adventure:

• A programmable power supply can be used to automate characterization
• Excel can make good graphs, giving the possibility to combine more traces and or do more complex math between them
• Using XML as a data format worked nice and fast. No asking for headers, data types etc. if you supply a XML schema.
• Maxima can be used to make graphs as well, can be saved as PNG or SVG, but doesn’t need Visio for that. It has less options to change the appearance.
• Maxima can be used to fit curves, having some idea of what function it should match with is important.
• The website zunzun.com can do curve fitting also, and has a lot of possibilities and also gives a function
• When using low gain error amplification it is important to not let the gate capacitance disturb the short feedback loop. However this also means sacrificing in gate drive speed.
• In a low current situation the Mosfet as a high Rsd resistance. Voltage drops are huge with only a small change in current. Because of the gate capacitance changing to a low resistance state takes some few 100ns of time. A end capacitor can slow down the voltage drop a bit, giving the circuit more time to adapt.
• The first stage gain needed to adapt is low, it even works with 0.13x gain, but no real advantages where found doing this. Except it was more stable without a endcap. However this situation was unusable.
• Even powermosfets can easily be zapped.
• Better to choose a large perf board, keeping space for modifications
• Using headers and dupont connectors to add swappable components works
• Don’t underestimate the usability of BJT’s
• A lm358 is a good and affordable op amp to experiment with
• There’s a difference in having a working concept and a working product. Both should be celebrated!