advertisement


New Regulator

More voltage overhead and no benefit here - the sensitive bits (Vref, error amp already have a pre-filtered then pre-regulated supply before their own massive inherent rejection (whereas on the ALWSR the pre-reg helps the current source driving the output no end)

Rejection of the output emitter-follower itself will be pretty good as it is, esp if you add the cap I suggested up thw thread a bit, item 0.
 
How is this

NewRegulatorr04_zps7b03b97c.png
 
For lily gilding, you could replace R3 by 0.5ma current source, like a J502 or J503. This would give about another 20dB of rejection at that point.
 
Er, CCS there is exactly what I did on the ones in my CD2...replace R3, R4, C10 with a 0.6mA.

I didnt suggest it here because it's silly %)

(that means - I'd design the board for R3, R4, C10; CRDs can be dropped in later)
 
Er, CCS there is exactly what I did on the ones in my CD2...replace R3, R4, C10 with a 0.6mA.

I didnt suggest it here because it's silly %)

(that means - I'd design the board for R3, R4, C10; CRDs can be dropped in later)

How about cascoding that CCS too. I'm willing to bet the effect of that would be completely inaudible :)
 
'you're very clever young man, it's cascodes all the way down'

A vaguely serious point I was going to make earlier - one can endlesly elaborate something, but it doesn't necessarily mean that you'll detect a useful difference, nevermind improvement - or even like the result.
 
Well gentlemen less than 50 posts in and were already settled.
Have I missed anything?

Do we have everyones input.

Martin, that was a quick edit :D
 
I don't think R22 does very much - it is effectively in parallel with the trimmer.. You can remove and save board space.

I think a lower value of R11 might be better - as it is, Q3 will run at about 0.7mA, plus base current of the BD139. The peak values of FT and HFE occur for currents occur round 10mA. The base bias current is going up at this point, but it is still well within the range of the AD797.

I would use say 220 Ohms for this.

There is no current limiting or protection at all - this will make the design fragile, but simplifies things.

If the load is 0.5 amp, the BD139 will get very hot - it is dropping about 12V, so you will get 6W of dissipation. This is inside the SOA allowed region, but means you must keep the case temperature below about 60C. This means you need a heatsink rated at beter than 10C/W, so allow enough space on the board for mounting this.

The input gyrator sees only a volt or two, so only needs modest heatsinking.
 
PD thanks for looking in.

A couple of things no obviously clear from my crappy schematic, it includes for multiple options.

R6 the trimmer would not normally be fitted, its there as a jail breaker should the Vref be a little out, I currently use it occasionally on the Flea when things don't pan out to plan.

R20, 21, 22 and C20 would not normally be on circuit. R23 would be linked.
This was John Luckins idea to change the mode of the op-amp and Vref for use when low down at 1v2 where the standard Flea implementation of the 797 went a little astray.

We have not really discussed actual BOM values, values on the schematic are my rather uneducated stabs at getting in the right ball park.

Note sure if the Darlington would be normal BOM, just an idea to get some lift with the 797 at 1v2.

I sort of expected normal 3-5 volt BOM would be just the BD139, R11 is no fit.

Difficult for me to try and make sense being unskilled.

Tony
 
Martin I have had a little play with your CCS and it is indeed very nice.
No detectable noise with anything I have at hand although thats not saying much as I have no expertise in noise measurement other than my HiFi :D
So long as it has more than around 5v5 its stable up to my 30 odd volt test capability and completely shrugs off line variation.
I have 3v38 rock steady with a nice handy 220R E12 for Rset.
Its temperature stability is nice and has a safe negative tempco.
I could only shift it down 140mV even attacking it from 1 inch with a hair dryer :D

Think I will modify my current Avcc Fleas with this CCS.
 
Have been playing catch up at Chez Luckins

A couple of fleas were delivered mid week and I have built one of them up to provide the housekeeping supplies to my ALWSR regs in Avcc duties on my Buffalo II. Even in basic setup, as per Rays supplied components, feeding the ALWSR opamp, ALWSR voltage reference via FET CCS and the ALWSR output stage CCS it was a substantial improvement over the ALWSR and its TPR reg feeding these loads for itself.

Over the last two days I have been tweaking same Flea to improve its performance and introducing a number of Tony's design improvements. The Avcc regulator is such a sensitive and responsive load that even changing R1 and C2 from 10k/3u3 to 1k/47u on the Flea was audible. Fitting a single fet current source in place of the 7812 feeding the LED's was an improvement, though not as great as I expected, so this evening, if I can keep off the Vino, I'll be knocking up my own cascode current source for these ref LED's. This afternoon I changed the 7812 for a LM1086 regulator as on paper at least this has better line and load regulation. This was worth doing and may be something to consider in place of the 7812 on the final circuit. an extra 3 components though, two resistors and a Tantalum cap. The pin outs of the 78XX series and the LM317/LM1086's are different so doing a layout suitable for either will be a little tricky, but I think worthwhile.

While I was at it I made some other changes to things that had disturbed me about the Flea for a while. More lily gliding perhaps, but don't we want to leave no stone unturned here?

Neither the 78XX series nor the LM317/LM1086's like to have a purely capacitive load. The 317/1086's want to see at least a series 2.7 ohm in the load and yet the standard Flea circuit only gives 0.5 ohms here. I put a further 2.2 ohms in series with R2 and R8 (R7 on Tony's design) to correct this. A nice little increase in dynamics and apparent volume resulted (normally a sign that a resonance has been removed IME). There must have been something untoward going on that this extra damping stopped.

I don't like putting two low ESR caps in parallel with each other without additional damping between them, yet this is how the two caps at the AD797 supply pins are connected. To sort this I added 0.5 ohms in series with the 4.7uF cap to damp any possible resonance between stray inductance in that cap and the 100nF ceramic across it. The result was a very slight increase in apparent dynamics, only discernible after extended listening to familiar music. Another way round this would have been to use a high ESR Electrolytic or Tantalum bead as the 4u7 cap, but at least 0.5 ohms of ESR is needed and the ESR of Electrolytics reduces with temperature.

I might try the two BJT current source as well now, but I think further improvement is limited to improved layout after that. I will also be getting the scope out to sniff around for anything nasty.

In the midst of this process I realised that I was getting a better sound now than I really have ever expected possible from my PMC FB1 speakers. There's a depth and realism to the bottom end that I really hadn't expected of them, no matter what amp was driving them. It is quite spectacular at times. It was the switch to using the Flea for housekeeping duties on the ALWSR that really made that significant step up in quality. Two weeks ago I sold my Naim CDS3 after 9 years of ownership due to lack of use and I have no regrets or doubts. The modified Buffalo was already better in all departments IMHO and more so now I have a modified Flea in service.

John
 
:)

The 317/1086's want to see at least a series 2.7 ohm in the load

Not quite, depends on cap size: http://www.acoustica.org.uk/t/3pin_reg_notes3.html

And that's quite why the orig flea has one small bypass cap for the opamp fed via a small resistor off the reg, and the other over a decade smaller across the load (for which the opamps R+larger C acts as a damper on the 78xx output = good snubbing practice). Also why the orig manual specs Tant caps... esp for things like the input to the 78xx reg*.

Yes - subtleties count!

* transients on the output of a reg get reflected on their input; maybe only 20dB down according to some e.g. Errol Dietz paper on the LM317 c/o National Semiconductor.
 
Nice post John I thought I was barking with the idea of a Flea and an ALWSR, good to know it works well.

Just speed read sofar and will read in fine detail later in the day.

You have not yet tried just the Flea on the Avcc terminals?
 
Yes had a read a few times now and a very informative post.

Some nooby questions for you John so I am clear what you did, if you please.

A couple of fleas were delivered mid week and I have built one of them up to provide the housekeeping supplies to my ALWSR regs in Avcc duties on my Buffalo II. Even in basic setup, as per Rays supplied components, feeding the ALWSR opamp, ALWSR voltage reference via FET CCS and the ALWSR output stage CCS it was a substantial improvement over the ALWSR and its TPR reg feeding these loads for itself.

Just to be sure, you used the Flea to feed just the AD797 supply terminals of the ALW?

Fitting a single fet current source in place of the 7812 feeding the LED's was an improvement, though not as great as I expected, so this evening, if I can keep off the Vino, I'll be knocking up my own cascode current source for these ref LED's.
John

This was just a jfet with gate and source connected to deliver Idss?

This afternoon I changed the 7812 for a LM1086 regulator as on paper at least this has better line and load regulation. This was worth doing and may be something to consider in place of the 7812 on the final circuit. an extra 3 components though, two resistors and a Tantalum cap. The pin outs of the 78XX series and the LM317/LM1086's are different so doing a layout suitable for either will be a little tricky, but I think worthwhile.

Well if it improves why bother having an option, I am quite sure you have it plugged into a position that puts the spotlight on performance, on the other hand re-jigging for 1086 with a single 7812 option slot is no hardship.
The hardship would be writing a stuffing guide for the amount of options we now have.

Thoughts of others?

While I was at it I made some other changes to things that had disturbed me about the Flea for a while. More lily gliding perhaps, but don't we want to leave no stone unturned here?

Neither the 78XX series nor the LM317/LM1086's like to have a purely capacitive load. The 317/1086's want to see at least a series 2.7 ohm in the load and yet the standard Flea circuit only gives 0.5 ohms here. I put a further 2.2 ohms in series with R2 and R8 (R7 on Tony's design) to correct this. A nice little increase in dynamics and apparent volume resulted (normally a sign that a resonance has been removed IME). There must have been something untoward going on that this extra damping stopped.

I don't like putting two low ESR caps in parallel with each other without additional damping between them, yet this is how the two caps at the AD797 supply pins are connected. To sort this I added 0.5 ohms in series with the 4.7uF cap to damp any possible resonance between stray inductance in that cap and the 100nF ceramic across it. The result was a very slight increase in apparent dynamics, only discernible after extended listening to familiar music. Another way round this would have been to use a high ESR Electrolytic or Tantalum bead as the 4u7 cap, but at least 0.5 ohms of ESR is needed and the ESR of Electrolytics reduces with temperature.

Interesting and something I completely missed. Indeed my first Fleas were stuffed with 4u7 Tants as the 797 bypass, I changed to the films in later versions as Rays BOM specified Films and thought the change was due to improvement. The penny never dropped that I had contradicted what I learned in my TPR developments with Martins help. Mainly a TPR from Acoustica's 3 pin reg notes, boy that was a nice sounding regulator when feeding my SBTouch perhaps I should try it in other positions now my SBT is dead.

I might try the two BJT current source as well now, but I think further improvement is limited to improved layout after that. I will also be getting the scope out to sniff around for anything nasty.
John

I ordered some parts and have a little strip layout for Martins CCS of this thread and may get around to grafting them to the pair off Fleas on the Avcc terminals.
Please post a sketch of your CCS when your sorted.

In the midst of this process I realised that I was getting a better sound now than I really have ever expected possible from my PMC FB1 speakers. There's a depth and realism to the bottom end that I really hadn't expected of them, no matter what amp was driving them. It is quite spectacular at times. It was the switch to using the Flea for housekeeping duties on the ALWSR that really made that significant step up in quality. Two weeks ago I sold my Naim CDS3 after 9 years of ownership due to lack of use and I have no regrets or doubts. The modified Buffalo was already better in all departments IMHO and more so now I have a modified Flea in service.

Yep my CD5 got taken out of the system, it was tweaked a little but absolutely LoFi in comparison to the tricked out Buffaloes.
 
So I built the first of Martins new CCS on a little board ready to plug into the Avcc Flea.

I ordered some new little diodes RS 826 521 and had a differing voltage so ended up with Rset as 120R giving 3v42 and the whole CCS circuit taking 7.9mA.

Need to build its twin and then swap out the CCS on the Fleas.

IMG_1891_zpsa38641be.jpg


IMG_1889_zps14d27f77.jpg
 
Got twins

IMG_1894_zps480e510d.jpg


Now mounted on the Fleas

IMG_1895_zps3640d541.jpg


And installed on the Buffaloes

IMG_1899_zps309ab204.jpg


Blimey, done it again :D
Don't have to worry about A / B comparisons this is another no brainer.
Happy bunnies and all things nice.

Thanks very much Martin
 
Oh heavens! So there're a couple of Buffalos in there somewhere?

Great stuff Tony - yours must be the trickiest Buffalo-based DAC ever! :eek:
 


advertisement


Back
Top