Super Regulator, new ideas

[peranders]

A fact is that this type of regulator, due to it's speed, is sensitive against capacitive load. AD797 is like a wild stallion and AD825 is one tiny bit better, according to my experience now. One thing I notice is that the inductance of 1" wire (normal connection wire) is enough to make the regulator much more stable.

One idea came up: How about using an MuRata-filter (10 nF cap with two ferrite beads) and a 100 nF 0805 after this filter towards the load?

This filter starts to work at 100 kHz - 1 MHz, and after the regulator you will have a well defined output impedance. I'm thinking of using the regulator for digital circuits. This was just an idea of combinding the best of two worlds, low impedance in audioband AND in the >1 MHz band.

[fmak]

[QUOTE]Originally posted by peranders
[B]A fact is that this type of regulator, due to it's speed, is sensitive against capacitive load. AD797 is like a wild stallion and AD825 is one tiny bit better, according to my experience now. One thing I notice is that the inductance of 1" wire (normal connection wire) is enough to make the regulator much more stable.

-------------------------------------------------------------------------
I have just built and tested several ALW regulators with AD825, AD817 and Ad797. the AD 797 measures best (around 1uV/1MHz rms noise) and no oscillations powering XOs to 45.xx MHz using 5 cm or so twisted cable and 150 uF OSCON local decoupling.

I dread making your regulator with all the fiddly SMT parts and excess components I have to buy but will get round to it one day. This regulator should ideally be preaassmbled.

[peranders]

My idea was really to block HF and let "audio" out and also make the regulator less sensitive against EMC

I know your "fears" about SMD but if you can borrow a second soldering iron, it's really easy to solder all 0805 parts. I'm not joking.

I would love to manufacture the SMD version but the interest is zero and if it was more than zero people aren't willing to pay anything.

Audiocom takes 51 GBP (including shipping) for one sulzer regulator and I think it's OK if the series is small but if it were about 1000's of units, then it's a bit too much.

[fmak]

I know your "fears" about SMD but if you can borrow a second soldering iron, it's really easy to solder all 0805 parts. I'm not joking.

-----------------------------------------------------------------------------

I don't fear SMD; I just don't like to fiddle about with magnifying glasses and handle storage and retrieval for use.

Also you can only order many SMT parts by 10 or 100. A waste unless you are mehtodical in labelling and storage. Most of the time I waste them!!

[Jocko Homo]

Maybe for you it is.

Jocko

[peranders]

Originally posted by fmak
I don't fear SMD; I just don't like to fiddle about with magnifying glasses and handle storage and retrieval for use.

Also you can only order many SMT parts by 10 or 100. A waste unless you are mehtodical in labelling and storage. Most of the time I waste them!!

Listen to me: Solder one type at the time, don't mix them! If you do so you won't need any magnifying glasses. Order, order order. that's the keyword. Also make copies of the schematics and partslist and check EVERY item when you have soldered. Order....take no shortcuts. Check LED's for polarity before you solder them and use Diode-meter and check that they shine.

The other thing, www.elfa.se sells even 1 pcs. of resistors! But they are so cheap so you can buy 100 of them. Elfa has very good marking in plastic bags. Don't mess around with those. Just keep them. Keep all parts in their tapes until you really shall solder them.

[peranders]

Originally posted by Jocko Homo
Maybe for you it is.

Nothing else to say?

The idea occured to me because the Super Regulator is the most sensitive regulator I have come across.

Jocko, do you think it was a good idea?

[jam]

Super Regulator, new ideas Post #1


A fact is that this type of regulator, due to it's speed, is sensitive against capacitive load. AD797 is like a wild stallion and AD825 is one tiny bit better, according to my experience now. One thing I notice is that the inductance of 1" wire (normal connection wire) is enough to make the regulator much more stable.

One idea came up: How about using an MuRata-filter (10 nF cap with two ferrite beads) and a 100 nF 0805 after this filter towards the load?

This filter starts to work at 100 kHz - 1 MHz, and after the regulator you will have a well defined output impedance. I'm thinking of using the regulator for digital circuits. This was just an idea of combinding the best of two worlds, low impedance in audioband AND in the >1 MHz band.


__________________
/P-A

Post #3


My idea was really to block HF and let "audio" out and also make the regulator less sensitive against EMC

I know your "fears" about SMD but if you can borrow a second soldering iron, it's really easy to solder all 0805 parts. I'm not joking.

I would love to manufacture the SMD version but the interest is zero and if it was more than zero people aren't willing to pay anything.

Audiocom takes 51 GBP (including shipping) for one sulzer regulator and I think it's OK if the series is small but if it were about 1000's of units, then it's a bit too much.


__________________
/P-A
Only my mother calls me Per-Anders.



Now compare to Fred's earlier post.





Digital clock supply design Post #294


It is helpful to remember that there are three parameters that are of major importance in a regulator. Power supply rejection of input noise on the the supply to the regulator, low noise from the regulator's reference and error amplifier, and low output impedance (small changes in output voltage from changes in load current). It quite possible to optimize all three of these for a clock supply. I think it is a very reasonable assumption that audio frequency load current changes are negligible in a fixed frequency clock and very very small in even a VCO (voltage controlled oscillator). The load currents at the clock frequency and it's harmonics can be pretty high in amplitude. As such, I think that a low ESR cap (or caps) are doing the decoupling since the regulator output impedance has risen open loop output impedances at these RF frequencies. I wonder if remote sensing is even useful much less practical in this application. My line of attack would be to use local sensing on the supply and an RC network on the output of the supply of 0.5 ohms and 10 times the capacitance of the local RF/digital decoupling capacitance at the clock circuit. You can probably even get away with a small resistor (a couple of ohms down to maybe 0.5 ohms) or resistor damped air-core inductor in series with regulator output. I think I might start with a value of inductance that will form resonance at about 100KHz with the regulator output capacitor (not the digital decoupling cap which should be a much smaller value than the output cap at the regulator) and damp it with 1 to 10 ohms. This should achieve good stability along with the op amp based regulator's excellent low noise and PSSR, two things that should be very benificial to a low jitter clock circuit. Knowing what the current demands of your load are will help you optimize all these parameters simultaneously. These suggestions are a starting point and I recommend a bit of experimentation for your particular circuit.

Fred



Our trusty moderator is at it again.......................I bet he never came up with an original idea in his life!

[fmak]

Originally posted by peranders

Nothing else to say?

The idea occured to me because the Super Regulator is the most sensitive regulator I have come across.

Jocko, do you think it was a good idea?

------------------------------------------------------------------
I find the ALW implementation very stable even with AD797. With the AD817 driving an XO there was absolutely no problem with a load decoupling capacitance of 620 uF OSCON. Noise is very low measured and studied with 400MHz scope.

[peranders]

Originally posted by fmak

------------------------------------------------------------------
I find the ALW implementation very stable even with AD797.

If you take a polyester or ceramic cap, straight at the output, how many nanofarads can the regulator take?