My take on a discrete shunt voltage regulator

[ikoflexer]

Originally Posted by AndrewT

the quads can be used to give doubles and singles.
The problem at the moment is the 40 or so that want a +ve regulator only.
How do we satisfy that requirement cheaply?

If scoring is cheap, then quads could be the easiest and cheapest way to buy the PCBs. But what do we do about the +ve only guys?

If you split the order into single +ve and single -ve, then you have two smaller orders. That pushes the cost up for the rest of us.

If you can get the cost down to the CeToole, you have potentially >1000 regulators to source and sell on.

I'm thinking that there are two kind of people: those who need only the +ve version, and those who need both -ve and +ve. I can't think of any project that uses only -ve. So most sense would be to order a whole bunch of +ve boards, and a whole bunch of dual rail boards.

Andrew, what do you mean about the cost of the cetoole reg? The reason I don't see hundreds of people interested in this regulator is not the high price, since none has been set.

Edit: I should add, there are two ways you can get this regulator for the price of a few coffees:
1) You make your own boards. It's easy. I've done it. It works. I can guide you and help with that.
2) You're at at time in your life that money is very tight. Then let me know privately.

[Telstar]

Originally Posted by ikoflexer

The reason I don't see hundreds of people interested in this regulator is not the high price, since none has been set.

No, it's that they dont know how good it is

[ikoflexer]

He he Do you know that I've tried very hard to make the cetoole regulator work? Look back in the development thread, I really did put a lot of work into that, and just... the design had some trouble. Couldn't make it as stable as I wanted it to be. Then I saw this huge GB started on a design that wasn't even finished. It freaked me out in a way, because I couldn't understand the phenomenon. It's probably fixed now, I haven't kept track of the latest developments.

[Salas]

The current thread's CCT concept works smoothly in many guises. I am in good experience to know it. Then you invested much time and iterations, parts research, trial and error, measurements, did your very own board design and proofing. So you took V1 concept to V2 performance and stability standard. I feel that I maybe would not have shown this kind of dedication so to take it there, if I was set to. Thank you for highlighting the concept to its rightful elevated performance standard. I guess that your work should be counted in the price you are going to set for the boards, its in no way just a matter of resin sqInch and etched cooper thickness IMHO. Good work.

[ikoflexer]

Salas, you're too nice. But you know what's involved in making these so you speak from experience. If anyone should get something back from the hours on end of work, plus hours on end of customer support, it's you!

So I tried some quotes from pcb fab people and guess what? The cheapest way was to get the boards individually done, not several circuits on one bigger v-scored board. Andrew, are you listening? The difference in price was not large, but with the v-scored boards there is extra work involved in separating the circuits for those who want them separate. So, there goes that myth.

[stormsonic]

received logic level MOSFETs and assembled new board.

Picture without output cap, oscillations with 88 mV peak-to-peak


Output cap 3.3uF MKP Vishay


Zoomed trace, final version.
Output cap 47uF Panasonic FC + bypass 4.7uF MKT EPCOS
http://www.diyaudio.com/forums/attac...1&d=1261181853

Bottom trace (white-yellow) is noise floor.
Output wires & sense wires cut to 3 cm.
I can confirm that placing output cap at load introduced some oscillations, better to keep output cap close to MOSFET.
Tried different caps, good results with Panasonic FC and FM, Elna Silmic, Oscon. I think this cap should be low ESR type.
Another bypass with 100nF or 220nF and oscillations returns

[ikoflexer]

Good info. What load? Roughly what was the ripple on the input? 5V output?

The last image shows ripple close to the noise floor. What would you estimate it to be peak-to-peak?

stormsonic, did you receive the email I sent you a few days ago?

[stormsonic]

Yes, received email today (sent Dec. 15th)
I think problem was in different MOSFETs (IRFBC40 & CEB703AL).

Input 12V, shunt current set to 500 mA. Output 5V, load 100 mA (resistor).
Look at 2nd picture, lower left corner. Grid is 5 mV, output ripple 3.6 mV peak-to-peak.

[ikoflexer]

Thanks stormsonic. Now let's get to business. 3.6mV peak-to-peak is too much

The images I showed earlier were sub 30-40 uV peak-to-peak. This has happened with about 15cm or output wire (both output and sense). Both latest pcb prototypes same performance. And I don't cheat.

There might be something off with your implementation?

[Salas]

Can't be. The Simpler Simplistic HV which is ultra basic was shown with 1mV on an analogous TFT screen scope, long outs, no remote sensing, next to heater wires and valves, with rudimentary CRC input. V2 LV should show just thin line in a digital source environment. Is the Vref filtered with 220uF in your implementation Stormsonic? Or it may be the GND.