23/03/2020
This project was sparked off by John's blog:
One of My Favorite Building Blocks
I've been thinking about power supply designs and read John's blog just after reading the power supply project in "The Art of Electronics, The X Chapters" (ISBN 978-1-108-49994-1).
The AOE design is rather complicated so I decided to try this board first. I bought mine from Amzon https://www.amazon.co.uk/gp/product/B077G8PK9Y/ref=ppx_yo_dt_b_asin_title_o06_s00?ie=UTF8&psc=1
They took a while to come and cost a bit more than John paid.
I built on the other day, the 56k resistor was missing and the resistors were possible the worst quality I've seen in 20 years but everything works. If you wanted one of these to be reliable it would be worth considering buying better resistors. But I wasn't interested in reliability but in how well the basic design works.
I started off with the power supply board powered by an AIM TTi bench supply supplying -4.5V limited to 50mA and +20V limited to 3.5A. For the load I used a Rigol Electronic load DL3031A, set for a steady current of 0.5A and a pulse current of 3A for 100ms. The power supply board was set for an output of 15V and the current limit at about 1A.
The scope trace shows the rather horrible result. The blue trace is the supply output, the purple trace is the current (actually voltage across R) and the yellow is the voltage on the output of U3.
It takes a bit of thought to work out what's happening, it helps to have John's nicely annotated schematic:
When steady with an output current of 0.5A, well below the limit setting, there is about 1/4V across R7 and the ouput of U3 idles at about +18V.
Once the load starts to draw extra current U3 's output starts to drop but takes a very long time to get below the voltage output demand on the anode of D9.
The load current peaks to more than the 3A I set it for, reaching a maximum of about 4.3A. The it drops back to the programmed 3A until current through D9 starts to reduce the voltage demand.
If the electronic load was a perfect current sink the current would not reduce at this point but it has limited slew rate.
The output of U3 goes -ve and eventually, about 5.5 divisions from the start, the load current drops to zero.
U3 output rises because there is no longer an over current and the output voltage rises a little. Then the Rigol load shows it's true colours and suddenly demands 5A or more.
I think this is because it winds up the internal current demand if there is no applied voltage.
Of course the power supply current limit action repeats and we get the horrible oscillation.
The next step is to prove this analysis, so I replaced the Rigol load with a 23.5R resistor in series with 3.3R resistor and a big MOSFET driven by a pulse generator to short the 23.5R.
This gave the scope trace below:
This is much more respectable, there is still a long time before the current stabilises at the limit level (230us) and it spends a scary 100us limited only by external resistance.
The really hideous second current pulse is clearly a problem with the load.
More testing to do, but so far the £9 power supply is probably doing its job a bit better than the £1200 electronic load.
I'm not quite sure why the output of U3 changes so slowly - first thing is to use a known real TL081 - I'm testing with the chips that came with the board.
Replace U3 with a known real TL081:
That's a dramatic improvement - overcurrent time down to 42us from 92us, U1 won't matter much but U2 will - I changed them both for good parts.
That gets the overcurrent time down to 36us and improves the current control a lot.
So John was right - throw away the "TL081"s that come with the kit !
And be very cautious about all the other bits too !
The basic design is not that bad - I'll see if I can think of a mod that will improve that dead time while U3 slews 20V to get the voltage down.
MK
Top Comments