New Regulator

[Bemused]

Well not so new but all the Flea tricks on one board and in need of both a name and some further tweaking.

Its called the Mite

Provisional schematic for consideration below.

Greg Erskine has kindly offered to spend the time on the board layouts.

Some notes.

A. I added R3 in parallel with R2 as a CCS trimming resistor.

B. R2 would be nice as an option for a multi-turn trimmer instead of the fixed resistor.

C. I added R6 as a get out of jail Vref divider.

D. I added Q3 as a Darlington option for lower rail operation. IE 1v2.

D. I added R10 as an optional choice to R11, R10 is probably preferred and R11 is a no fit.
R11 would probably be fitted if the Darlington array was used to increase turn off time.

E. R7 is now just a single through hole rather than a parallel pair of 0805 SMD.

F. The 0805 SMD could be replaced with through hole or maybe the board could have options.

G. The board needs to be ground planed.

H. Be nice to accommodate physically larger capacitors C2, C3, C6 and C9

I. Be nice to have an option to replace Q2 BD139 with a small signal type for faster operation when on lower loads.

J. I added provision for upto three Vref LEDs

K. A nice option would be for some spare pads around the CCS to enable use of other CCS devices, for example the Jung cascode JFET current source.

L. Load is in the region of extreamly low to a few hundered milliamps depending on the devices fitted at the current dumping stage and the gyrator.

M. The board should have screw terminal headers for input power, output and feedback.

N. The board should an option not to use the feedback terminal (on board link) not worth the risk of open circuit feedback if not required for the application.

O. A really wild option of a method to feed the 797 with an external symetrical supply is open to suggestion / interest. Usefull for extreamly low output voltages.

P. If sufficient interest this would be a group buy with ‘proper’ boards.

 

[Tenson]

I'm no electronics expert but I wonder why you used LEDs for the voltage ref. rather than something like LM431? You could use LM334 and LM431 together.

 

[Bemused]

I'm no electronics expert but I wonder why you used LEDs for the voltage ref. rather than something like LM431? You could use LM334 and LM431 together.

Neither am I
However LEDs are extreamly quite and have yet to dissapoint me.
The method of feeding an LED with a CCS is well documented and there are many regulators using this method. I would expect the CCS to be the limiting factor not the LED.

Some clever chaps will no doubt offer comment on the suitability of the design.

 

[timH]

Well not so new but all the Flea tricks on one board and in need of both a name

tick?
nit?

fish-tick?
fish-nit?

 

[Bemused]

tick?
nit?

fish-tick?
fish-nit?

Fishnet as in phwoer and seldom seen and stocking

 

[martin clark]

Nice idea Tony; a few quick random thoughts:

'431 is a pretty horrible (noisy) part. LM329s or other Vrefs can be used but leds are cheaper - and quieter, though not so good on temperature stability - which is a minor issue most of the time.

I built a few Flea-alike regs using LM329s - no benefit overall and I reverted to LEDs. (CD player sound better afterwards too). Remember the LM329 is a buffered reference and as an active device it can be provoked (eg supply rejection at HF isn't all that you might expect IME). Simple passive LEDs and a lowpass filter = difficult to get wrong.

The reason I suggested 3 leds on the other thread is flexibility. with that many hole in the PCB you can fit two LEDs and a link for 3.3v uses or 3 LEDs for 5v use or indeed make any other reference you like fit somehow.

The longest lived flea in use at home has a cascode CCS driving a 3-led reference, which means you can build the whole thing using small-signal BC547s / 550s and about 5 LEDS for low-current use.

Side note: I've never been keen on promoting this augmented flea approach for loads over say 100mA simply because the flea is so voltage-inefficient - requiring 16-18VDC for low-volt output means a lot of wasted heat. I've one running at 250mA + for some years now in my CD2 for the digital filter but that uses one (regulated +20v) rail for the error amp and an existing regulated 9v rail as a supply for the output transistor - which minimises disspation and prevents unwanted interactions.

ETA: I was going to suggest 'Tick' for another crack at a clock reg. Perhaps Mite - play on 'mighty' flea / a DIY annoyance that gets under the skin

 

[GWM]

If this turned into a groups buy I would be interested the siren call of a pair of Buff111SE is getting to me been spending to much time on the TP site

 

[Sparts155]

Just some thoughts ...

Why not feed the CCS from down stream of the 7812, and take advantage of the extra regulation?

Spot on about the LEDs, plus of course the thing lights up like a little Christmas tree when you power it up, adding to the feel good factor

Also, if you are using a tant at C6 to prevent impedance peaking, you might want to use one at C9 too (?)

 

[martin clark]

Also, if you are using a tant at C6 to prevent impedance peaking, you might want to use one at C9 too (?)

Spot-on.

I've played a lot wth the 797-based regs into some funny loads and - although I've had complete stability into large film caps and oscons* -a little bit of ESR here does make things nicely predictable. This does still allow high-q bypass caps at the load, offboard.

The 250mA variant at home has 47uF oscon and a film bypass cap on the output... not something I'd recommend for a general DIY offer. At the very least it makes R10 a thoroughly recommended part, and C6 as a tant a very good choice by adding damping into two of the three 'ports ' of the output emitter follower.


Tony - a last thought for lunchtime: I'd lay Q1 and Q2 out for T0220 package transistors, it widens the choice a lot.

 

[Bemused]

Thanks all, specific values don't really matter at this early stage we can sort some BOMs out later the main concerns are getting a suitably flexible layout and a layout that will contain any BOM variations without such severe hacking as the standrd Flea involves.
However it is a good point regarding Tants and I sort of covered this in note H.

Would it be OK to layout Q1, Q2 for T0220 and fit small signals in the same pad spacing, if so thats one headache less.

What are the thoughts on having linking for a choice of CCS feed, either before the 7812 or after. Although I understand a half decent gyrator as an ideal feed for the CCS.

I have built some 200mA plus versions with BD139s and they do get very hot but that has always been the price to pay for having a small Flea to play with. We could always adjust the stage dropouts for severe cases. Should we consider options for heatsinking?

I like the idea of film bypass on the output capacitor, should we include pads, my thoughts were this is never going to be a 'blindly stuff the board and forget' type of regulator but a board to give those with patience some scope.

Regarding the name.
ARIB also pinged a light, Another Regulator Is Born

ETA: I was going to suggest 'Tick' for another crack at a clock reg. Perhaps Mite - play on 'mighty' flea / a DIY annoyance that gets under the skin

Mite was also in my mind could we settle on Martins choice?

Tony

 

[martin clark]

You don't actually need an output bypass on the board at all, certainly not a film cap - load HF bypass belongs at the load.

If anything, rather than C9 perhaps fit pads for a resistor here - say nominally 1K - because its useful to have a test load before final connection, and it could be swapped to ensure a certain minimum (say 25ma or more) is drawn to get best performance

 

[Bemused]

You don't actually need an output bypass on the board at all, certainly not a film cap - load HF bypass belongs at the load.

Yes getting carried away.

If anything, rather than C9 perhaps fit pads for a resistor here - say nominally 1K - because its useful to have a test load before final connection, and it could be swapped to ensure a certain minimum (say 25ma or more) is drawn to get best performance

So some through hole resistor pads in // with C9 and 1/4W should be fine.
And C9 is always better of board at the load, no capacitance on the end of the boards at all?

 

[john.luckins]

Interesting thread Tony. I have been playing around with three ALWSR's in use on my Buffalo II and have found a few things that may be interesting and perhaps worth a try on your Flea(?) build. They mostly revolve around the feed for the Vref and the opamp. it's a bit of a long story so please bear with me.

On the ALWSR the Vref is fed from the output, which is 3.3 volts for my applications. For the 1.2 volt reg the ALWSR vref feed has to come from the pre-regulator output, a very similar set up to that of the flea. To get enough current for the Vref from the 3.3 volts output meant using a 150R resistor to feed the LED and this was limiting the filtering effect of the LED which has a resistance of 10's of ohms. I tried a few alternatives. Firstly a jfet CCS (2SK117) fed from the pre regulator. This sounded a tiny bit better than the resistive feed from the output. Then I tried using a simple 9 volt battery to feed the same jfet CCS and LED reference. This was the best yet, giving a sizeable improvement to the vocals in particular. It was nice and I wanted to keep it. Then I realised that there might be some load related components getting through the jfet CCS from the tracking regulator that were causing that feed to sound worse than the battery, and that these same load related components were probably getting through the 150 ohm-LED filter as well. To solve all of this I went back to using the 3.3 volt output to feed the LED but filtered the 150 ohm feed by splitting it into a 100ohm and a 47 ohm and using a 220uF to 0v from the junction.

Let's just say that the improvement was not small when feeding the Avcc input. I've since made the same mod on all the 3.3 volt regs, and even added a series choke to the 150R resistance string, improvements were obvious each time. I'm now convinced that noise, particularly load correlated noise, feeding the reference LED(s) should be minimised at all costs and that this is critical for performance and can get past a CCS. I've also read of people using LM334's as loads in valve amps and having trouble with the noise they produce. They are more complex and noisy devices than jfets current souces, a price you pay to get their versatility. The ALWSR gets a big reduction in its noise by using its own output to feed the reference, but even then my additional filter has improved this perhaps something similar can be done for your LM334.


I would therefore suggest that you try a 220uF capacitor to 0v on the negative terminal of the temp compensating diode and a small loading resistor, say of 47 ohms between the junction of the cap and diode and the upper LED. You can't lose anything by trying it and I think there may be some gains.

There are some other interesting things when comparing the Flea to the ALWSR. The ALWSR opamp is sinking current to 0volts and as a result the opamps +ve supply input is just feeding the constant current sinks that bias the opamp into operation. That makes isolating them from their +ve supply rail more straightforward, allowing the use of a larger series resistance (your R7) and extra filtration. The Flea opamp on the other hand is sourcing current into its load and its +ve supply current will vary with load current. This to my mind makes the good isolation of its opamps supply from the loads effects on the 7812 output much more critical. An extra layer of RC decoupling before your R7 (with a larger R and C, say 4.7R 470uF) would do no harm IMHO.

Try these out if you can/want to before you commit to etching if not too late and let me know if I'm going OTT. I found the Avvc reg to be the most revealing test location for these changes.

John

 

[martin clark]

Completely agree John with your reference comments - it tallies closely with my own experience with Jung regs and others. I am not a fan of feeding references even bootstrapped from the final reg output. And funnily enough I've just emailed an idea to Tony... (can't load attachments from here)

Q2 can be fed from +ve connection of C3 to completely kill interaction with the error amp; or even a separate supply (-yes, I have done that...)

If you feed Q2 from the output of Q1 then it makes more sense again to feed the references' CCS from the output of IC2 - which essentially has constant current loading. Say using the discrete ccs I've just emailed Tony, which is happy with the lower voltage headroom. (WIthout such justification given LF PSRR of an AD797 one might realise IC2 feeding it alone is looking a bit redundant...)

 

[Bemused]

You want me to post the sketch Martin?

 

[martin clark]

It was just for you to play with but feel free if it's legible!

BTW if my previous post was unclear - the ccs was only moved to be fed from the input filter to save a bit of voltage headroom originally in one of Tony's applications, to use the LM334. In summary I suggest :
- input filter around Q1 feeds IC2 and Q2
- IC2 feeds error opamp and CCS chain to voltage reference.

Sketch is proposal to replace IC1 with a cascode ccs to suit this, because it needs less voltage overhead (12v in for ~6v max reference output is fine) and performs better. It would ensure IC2 'sees' more than minimum loading. It also contributes more glow!

 

[Bemused]

John many thanks for the benefit of your experience. I don't think its at all possible to go OTT with Buffalo regulators, we have proven this so many times that darned Buffalo just keeps giving more and more.

As always it will take ma a little time to digest your post but I see the filter idea.

Just looked at Martins sketch and it and he has gone and done it again, ill get the schematic redrawn soon.

 

[Yank]

1: If you put a bunch of these in one box (like a SuperCap), would you have a Flea Circus?

B: How much current can this supply provide?

III: Will there be a negative version as well?

 

[martin clark]

Well as it is constituted this really only suits low voltage, 1-5v supplies; plenty of current not a problem depending in how much waste heat you can tolerate.

I've another bodger's +/- V supply discrete reg idea for other uses. Perhaps a separate future thread

 

[Bemused]

OK a new schematic, a little scruffy sorry.

I added Martins new CCS which looks neat and also moved its feed point to after the 7812.
Sure to be something I have forgotten already so just remind me please, remember I'm no so fast as some of you guys.