advertisement


Salas Shunt Regulator Build

Moved on a little more now.
The Salas regulators now directly feed the Dual Tridents, I lifted the Vd supply from beneath the Trident and dropped the Salas feed in from above.
Getting a little messy as there are now twelve cables jointing off around or on the Trident.

IMG_1826_zpsd612659d.jpg


I used the old Placid to feed my 6.7mA into the Buffalo main Vd terminals, this load seems to be just I2C housekeeping. My thoughts were that the Arduino and LED screen probably puts noise on to the Vd 0v rail which was shared with the Avcc Trident.
I have now separated the I2C and Avcc via different regulators and transformers.

Prior to this latest mod the Salas fed the Buffalo main Vd terminals and the Avcc Trident picked up of this same supply. Whilst it did provide some improvement in SQ I felt that I had lost out in some other areas.

This latest bodge has brought the aspects of SQ that I lost back whilst keeping the original gain.
Recommended bodge.
 
Bemused said:
Crikey what a complete cock-up.

I thought to just try feeding the Salas straight into the Buffalo Avcc terminals, no Trident.
I was not expecting anything really other than a quick comparative test.

So I linked out one of my red Vref LEDs on the Salas and gave it a try, but the voltage was still to high. Actually I should have heard the gentle tinkle of the warning bells but oh no I had to start cobbling.

To this end I linked out my Vref LED series resistor leaving just the trimmer potentiometer in series. Be okay I thought so long as I don't do something stupid and turn it all the way to zero ohm, so I thought.

So Salas all wired up directly to the Avcc terminals, meter in one hand, screw diver in the other and I start winding down, and down, and down...

Click

End of trimmer travel click, Vref Led goes out, voltage sense feedback circuit no open circuit and the Salas does its job and winds the output voltage to the rails.

Eeek

18 volts Avcc, not funny at all.

Big lesson learned, I gotta slow down and think when things don't go to plan. The Salas was shouting 'FETS don't go so low' at me, but oh no I'm to smart to listen. Now whats that old saying about a little knowledge?

The only smart thing I did was have some 3v9 Zeners sitting on the DAC.

So next plan is to feed the Salas at 5v3 directly into the Dual Trident to which I have a question.
Can I just feed into one Trident Vin terminal or do I need to loop to both. They seem to be connected by a bottom board track. Latest Dual Trident by the way.
Click to expand...

Sounds like a close call Tony. The Salas Reg can be made to go down to 3.3 volts but you have to replace the shunt Fets to 15030/1 bjt transistors to keep shunt current flowing as low as 3.3 volts. The group buy has BJT kits for this purpose.

I don't think the Salas would have actually made any improvement as a replacement for the TP Avcc regulator. The Avcc needs a very low impedance regulator and the higher feedback opamp based circuit TP use here is better at this than the Salas and quite hard to improve upon methinks. I am certain however that a 5.3 volt feed from a Salas reg to the TP Avcc regs will give you a noticable improvement over any other TP regulator in that position. BTW I'm still trying to figure out other ways to improve the TP Avcc in situ.

I'm glad your TP Avcc Reg survived, thanks god for the zeners. I'd never have been that lucky.


John
 
john.luckins said:
I don't think the Salas would have actually made any improvement as a replacement for the TP Avcc regulator. The Avcc needs a very low impedance regulator and the higher feedback opamp based circuit TP use here is better at this than the Salas and quite hard to improve upon methinks.
Click to expand...

The output impedance of the Salas reg is actually extremely low, well under 1 mOhm, indicating high feedback, plus noise can be made extremely low too. I'd be surprised if the Salas shunt wasn't a better reg.
 
Sparts155 said:
The output impedance of the Salas reg is actually extremely low, well under 1 mOhm, indicating high feedback, plus noise can be made extremely low too. I'd be surprised if the Salas shunt wasn't a better reg.
Click to expand...

In this role it has some inherant disadvantages compared to the rather special opamps used in the TP Avcc Reg. For starters the Salas will inevitably be further away, and where milliohms are critical distance can make or break performance. Don't get me wrong, I like the Salas a lot (I've built more than 20 of them), but this is a very particular application for which I don't think it is the very best option. It does make an excellent pre reg for the TP Avvc reg though.

Check out the data sheet for the LMP7732 which was the opamp used in the first version of the TP Avcc reg. This is used as a voltage follower with a high speed small signal npn on its output. It has very good PSRR and CMRR in the audio range which is what counts in this application and high gain so very low zout in the audio range. Hard to beat when close up to the load. Russ White uses an even more appropriate opamp in the latest version.

I'd rather like it if I was wrong on this. My DIY bones don't like admitting defeat to an opamp based circuit.
 
I did a couple of calculations, rough estimates based on the data sheet for the LMP7732 and the TP circuit for the Avcc reg and we're looking at a Zout of circa 1 micro ohm (theoretical as the tracks and connections will dominate) at 100Hz rising to circa 1 milli ohm at 10 kHz. Pretty impressive. Noise figures look good too. Has a very limited voltage range though compared with the Salas.

J
 
Found some further data John.
Output impeadance for Salas BIB mosfet
10Hz > 2.09mOhm
100Hz > 1.98mOhm
1kHz > 1.75mOhm
10kHz > 3.13mOhm
100kHz > 26.07mOhm
 
The graphs published by Linear and your stated numbers seem to contradict each other by a factor of 10? I wonder which is correct?
Edit - Sorry looking at the wrong line - the line types are confusing even in colour........
 
Bemused said:
John have you seen this data.
Click to expand...

Yep, I saw that for the first time last week. I'm ordering the full article/volume for more detail. I'd say the TP Avcc reg probably fits in somewhere around the two Jung regulators on output impedance, and certainly better than the Salas. It has two major advantages to me, firstly it is already built and working reliably, secondly it is small and a designed fit in close proximity to the load. Bettering it, even with a Jung style reg will be a challenge for me at least. The current capability and versatility of the Salas are not valuable factors in this application.

It might be possible to mount two Salas or ALWSR Regs so that pins on the output tabs mate directly with the Avcc terminals on the Buffalo board and give it a try though if keen. ALWSR are smaller and I've got a couple built up from an old 24v Hicap clone, so it is just a matter of getting the voltage down to 3.4 volts.

John
 
themystical said:
The graphs published by Linear and your stated numbers seem to contradict each other by a factor of 10? I wonder which is correct?
Click to expand...

I wasn't expecting them to agree and they could both be correct though I wouldn't expect that as I don't have accurate information on the TP circuit (yet) mine are just ballpark.

I haven't quoted any numbers for the Salas Reg to compare with those in the Linear Audio article if that's what you mean? Perhaps you are refering to the Linear Technology data sheet for the LMP7732 when you say "graphs published by Linear"?

I've just examined what I believe to be the circuit for an early version of the TP Avcc Reg and the opamp it uses and approximated Zout figures from that. Bear in mind the circuit I have been using has an emitter follower inside the loop with the opamp so it will be different to the figures for the opamp alone (lower) and I would certainly expect it to be different to the figures of the unrelated Salas Reg. I haven't seen the full Linear Audio article but assume it didn't include the TP Avcc regs.

I wouldn't be surprised to find in the eventual analysis that the TP Avcc regs do offer a lower output impedance than the Salas Regs by a factor of 10 or more over much if not all of the audio bandwidth, just as the Jung regulators do.

Does that clear things up?

John
 
Bear in mind 1mOhm is just 12mm of 7/0.2 hookup wire; so unless you use (very) local sensing, you're unlikely to realise such small numbers at the load.
 
themystical said:
Edit - Sorry looking at the wrong line - the line types are confusing even in colour........
Click to expand...

That's easy done. It took me a while to spot how good the Jung regs are wrt output impedance. BTW I'm interested to see how the TPS7A4700 LDO, the DEXA and other devices in your other "what regulator thread" stack up with regards to this TP (Twisted Pear) Avcc reg replacement role as well as the PH regs. Once the best configuration and device have been confirmed it will be all about minaturisation, proximity to the load and load capacitance/interaction with the regulator, all of which can spoil a promising regulator candidate.

J
 
john.luckins said:
That's easy done. It took me a while to spot how good the Jung regs are wrt output impedance. BTW I'm interested to see how the TPS7A4700 LDO, the DEXA and other devices in your other "what regulator thread" stack up with regards to this TP (Twisted Pear) Avcc reg replacement role as well as the PH regs. Once the best configuration and device have been confirmed it will be all about minaturisation, proximity to the load and load capacitance/interaction with the regulator, all of which can spoil a promising regulator candidate.

J
Click to expand...

John
The only regulator I have of that list is the TPS7A4700 and that currently feeds the Avcc of my Buffalo24 through a LM49720 Op amp (to provide the low impedance output) and the circuit linked below to provide the Vref. The TPS7A4700 feeds both the Op amp and the Vref.
http://www.diyaudio.com/forums/twisted-pear/128137-buffalo-dac-ess-sabre-9008-a-108.html#post1695381
The Data Sheet for the TPS7A4700 does not publish the output impedance of the chip but in the role that I have it in, I would guess that ultra low impedance is not critical, that role being performed by LM49720.
When I added the TPS, the leap up in sound quality was not subtle and you could tell the improvement immediately on the first listen - no need for endless repeated listenings of the same tracks...:)
From the Linear Audio article, it does look like you can not get better performance for analogue duties than a Jung, it seems to have it all - ultralow output impedance, ultralow noise and extremely high PSRR and to top it all, it came at the top of their listening tests - so if any reg is going to be better for a complete replacement of the Twisted Pear arrangement for Avcc, it is going to be a Jung variety. (in theory......). I am waiting for the bare ALWSR board that I have bought from Tim to arrive, so I can see how it works in practise.:)
Like you, I would like to how well the other regulators discussed perform comparatively and try and identify the best performers for the Digital side as well. I am due a visit to Tony's (Bemused) and like you he has (almost) every type available installed in his Buffalo (bar Paul Hynes/DEXA) so I am hoping to get some insights....:)
 
john.luckins said:
In this role it has some inherant disadvantages compared to the rather special opamps used in the TP Avcc Reg. For starters the Salas will inevitably be further away, and where milliohms are critical distance can make or break performance. Don't get me wrong, I like the Salas a lot (I've built more than 20 of them), but this is a very particular application for which I don't think it is the very best option. It does make an excellent pre reg for the TP Avvc reg though.

Check out the data sheet for the LMP7732 which was the opamp used in the first version of the TP Avcc reg. This is used as a voltage follower with a high speed small signal npn on its output. It has very good PSRR and CMRR in the audio range which is what counts in this application and high gain so very low zout in the audio range. Hard to beat when close up to the load. Russ White uses an even more appropriate opamp in the latest version.

I'd rather like it if I was wrong on this. My DIY bones don't like admitting defeat to an opamp based circuit.
Click to expand...

You're right about the size of the Salas reg being an issue for installing close to the load, that's why it was developed to use sense lines where it's output impedance can be maintained.

Thanks for the info about the TP op-amp, it looks a good device with typical figures for a modern op-amp with regard to gain, PSRR etc. Obviously it's performance will fall away as frequency rises and gain drops, so I wonder how the actual circuit performs - I couldn't find any performance figures from TP.

Here's the impedance from a basic version of the Salas reg. Later versions were much better, though you have to wonder whether past a point any improvements in impedance will matter in the real world, which goes for any reg.

reg-salas-1a-zout.jpg


Having said all this, my feeling is that it's wise not to take any tech data as gospel, and feel free to try different approaches. The original demo board for the ESS9008 used a single AD797 to drive both channels, and achieved the desired spec (circa 130dB dynamic range). Yet people have developed many other solutions that sound better.

As for noise, I don't know what TP use for a reference, but with the Salas if you use a current source feeding a resistor/capacitor, it's not going to get any better. The original demo board and the Buffalo mk1 used a plain-Jane 3.3V LDO with a 150HZ low pass filter for a reference - not exactly low noise! Using a proper low noise reference certainly helped the sound greatly, which shows that noise is a parameter worth keeping in mind as well as output Z.
 
Well I've just got some more buffalo bits and was going to get placid HD like my other one but these are supposed to be better and should work out far cheaper. Also doing a phono build with NJ's motherboard and 323 cards so this should be good for that, low noise and low Z :)

I posted selfishly as they need 100 boards on the list before ordering and I only take it up to 23!!

To give a quick short cut to save reading the whole thread you get 2 positive and 1 negative versions on one 'break' board. Great if you are powering a dac and an IV stage. You'll then need one of the minikits to populate it, mosfet kits for 5V + and BJT for less

Stefan
 
I'd like to order some boards & bits but, heavens, it's all very confusing! I'd like to order a couple of complete kits but it appears there are two types of board?
 
There's a low voltage board (LV) and high voltage (HV). I think all we'd be interested in is the low voltage, although I don't know offhand what voltage/current it can supply up to.

Quite tempted as well as I could do with a couple of positive supplies for waveio
 
You must log in or register to reply here.


advertisement


Back
Top