Discussion on the design of the T2 & other circuitry

[blubliss]

macbois don't do history

[Linear]

It would be pretty cool to take the digital output from an iphone and feed it into a balanced dac and then into my balanced SRM-Xh...

I have a Wadia 170i that takes the digital output of my ipad/iphone and digitally feeds it into my Wadia 861. The 861 has a totally balanced set of four 24-bit Burr Brown PCM-1704 DACs (2 per channel) that then feed into my "swift current" zero-feedback current conveyor I-V converter (US pat.4,983,930). However, I can't use the balanced output of the 861, because I then go into my Stax SRA14S driver (no balanced input ). If (or more hopefully when) I build a T2, I can then go balanced right to my headphones!

[jcx]

I think the tl431 would be best avoided - as an amplified band gap reference working down to low currents it is inherently noisy - and as a feedback loop amplifier it is poorly characterized

band gap reference noise can be OK for audio when you insert a lowpass filter after but the required low corner frequency complicates startup dynamics

http://www.analog.com/library/analogDialogue/archives/32-1/xfet.html - precision instrumentation designers often look at the .1-10 Hz noise number - audio really doesn't care given low frequency insensitivity of hearing

http://cds.linear.com/docs/LT%20Magazine/LTMag-V19N1-03-References-BrendanWhelan.pdf

buried Zener LM329 is still available and is probably the cheapest of the buried Zener devices

with the heavy output C the feedback amplifier inside the regulator probably ends up with the output Cload determining the dominant pole compensation

for T2 use "reasonable" 1/f, drift, tempco could be adequate - I'd look for the 1 KHz S/N - don't forget to divide by the Vref value

[luvdunhill]

yeah, the LM329 is used in the Maida. I'm not sure why you guys don't consider this design, it's simple, cheap, hard to destroy, and performs pretty well.

[kevin gilmore]

A maida circuit with a tl783 is possible, ray manages to do the same

thing at 600 volts without the front end voltage reducing transistors.

I'll try it. But the idea is to use an unregulated voltage of about 625

volts to make sure that there is enough to do the 580v bias. Will order

the parts and see what the real performance is. The problem will be

voltage drift over temperature as there is no extrememly stable reference.

[spritzer]

I have a Wadia 170i that takes the digital output of my ipad/iphone and digitally feeds it into my Wadia 861. The 861 has a totally balanced set of four 24-bit Burr Brown PCM-1704 DACs (2 per channel) that then feed into my "swift current" zero-feedback current conveyor I-V converter (US pat.4,983,930). However, I can't use the balanced output of the 861, because I then go into my Stax SRA14S driver (no balanced input ). If (or more hopefully when) I build a T2, I can then go balanced right to my headphones!

Ahh but the SRA-14S can be modified to take a balanced input. You'd have to rewire it completely and do some light mods to the PCB but the amplifier unit is basically a SRM-1 Mk2 which is fully balanced.

Kevin sent me your patent a while back, very cool stuff.

Btw. Here is the amp I was referring too, fully balanced and about the size of the old double CD cases...

[kevin gilmore]

So i've been playing with the miada thing at 250 volts due

to lack of some other parts. Its R1 and D1 that are the

problem, the spike noise from the power supply diodes rips

right thru and you end up with about 180 microvolts of spike

noise. Still way good enough for lots of stuff, but probably

noticeable in the ultimate S/N of an electrostatic amplifier.

[aerius]

What about reducing the noise at the source? From what I've heard TV damper diode tubes such as the 6CM3 aren't nearly as noisy as silicone diodes while still being able to pass a lot more current with lower voltage drop than traditional tube rectifiers.

[kevin gilmore]

The problem with tubes is ... well tubes

For a kgsshv supply you would need 4 of them.

For a T2 supply you would need 8 of them and

still need some solid state stuff.

As well as lots of extra floating filament

windings to run the heaters.

[luvdunhill]

So i've been playing with the miada thing at 250 volts due

to lack of some other parts. Its R1 and D1 that are the

problem, the spike noise from the power supply diodes rips

right thru and you end up with about 180 microvolts of spike

noise. Still way good enough for lots of stuff, but probably

noticeable in the ultimate S/N of an electrostatic amplifier.

nice! 180uV sounds about right, but I have other numbers I can track down.

I would think with PSRR of a good CCS, and the nature of an electrostatic headphone load (static), that this would be good enough, but I defer.

[Linear]

the tl431's works and is about 35 microvolts peak to peak noise.

So i've been playing with the miada thing at 250 volts due

to lack of some other parts. Its R1 and D1 that are the

problem, the spike noise from the power supply diodes rips

right thru and you end up with about 180 microvolts of spike

noise. Still way good enough for lots of stuff, but probably

noticeable in the ultimate S/N of an electrostatic amplifier.

nice! 180uV sounds about right, but I have other numbers I can track down.

I would think with PSRR of a good CCS, and the nature of an electrostatic headphone load (static), that this would be good enough, but I defer.

Huh??

It looks like my TL431 MOSFET Cascode idea works way better (35uV noise versus 180 uV noise) and has way fewer parts. Why would anyone even consider Miada?? What am I missing??

[luvdunhill]

Huh??

It looks like my TL431 MOSFET Cascode idea works way better (35uV noise versus 180 uV noise) and has way fewer parts. Why would anyone even consider Miada?? What am I missing??

What post is the schematic in again?

[Linear]

What post is the schematic in again?

Post #20

[luvdunhill]

Post #20

I think it still has a few more parts. I use a MOSFET, not the BJT Darlington configuration. I also drop the series resistor before the regulator and it seems to come out a few parts ahead. My noise measurement is right at 100uV.

What about a difference in short circuit protection? Also, what about a difference in supporting higher voltages?

Edited March 31, 2011 by luvdunhill

[kevin gilmore]

I always thought the idea was to make absolutely the best performing thing possible,

then once that is accomplished, find a way to reduce the parts count and price.

Reality has no place here. We are nuts by definition. If you are not nuts, then

please leave. John Curl class obsession is absolutely required.

For the high voltage stuff, everything i'm doing is either 450 volts or 500 volts.

Not really interested in anything else. OK, well 250 volts for possibly the

next generation T2 in some even bigger chassis.

The low voltage stuff is also however important and has different design issues, but

really much of the same stuff. For example dynahi/dynafet/B22 class things with

+/-30 volt power supplys, and even +/-40 volts for those that desire to some

quality to go with their 160 volts peak to peak into lcd2/he6.

The most successful stuff so far (i.e. the stuff with the lowest noise, and best regulation)

uses depletion mode mosfets as the current sources, and then either series pass

or shunt mode pass devices.

Until someone can show me some clever way to shut up the power supply diodes.

Maybe the SiC things, but i don't have any of those to test.

Compared to the unregulated supplies of the past with a few hundred millivolts of ripple,

any of this stuff is light years better.

Edited April 1, 2011 by kevin gilmore

[Linear]

I think it still has a few more parts. I use a MOSFET, not the BJT Darlington configuration. I also drop the series resistor before the regulator and it seems to come out a few parts ahead. My noise measurement is right at 100uV.

What about a difference in short circuit protection? Also, what about a difference in supporting higher voltages?

Well.....OK, let's just look at just the part that controls the series element (FET of bipolar):

Maida: LM 317H, zener, diode, 4 resistors, 2 caps.... Noise= 180uV

TL431 MOSFET Cascode: TL431, zener (not shown), 3 resistors, 1 cap for stability (not shown)....Noise= 35uV

Remenber, we are not including the components that implement the hybrid function. (That oscillated with the TL431.) Maida hadn't even THOUGHT of a hybid PS, and as far a I'm aware that is an original idea of KG's.

BUT, we are working on ultra low noise PS. We need to close of discussion on 180 uV noise power supplies for now! Maybe later, we can see how cheap and dirty we can get!

As I finish writing this, I see that KG has beat me to the punch (again!) and basically posted the same sentiments in the previous post. So YES, let's move on and discuss ultra low noise stuff!

Linear

[Linear]

Ahh but the SRA-14S can be modified to take a balanced input. You'd have to rewire it completely and do some light mods to the PCB but the amplifier unit is basically a SRM-1 Mk2 which is fully balanced.

Kevin sent me your patent a while back, very cool stuff.

Btw. Here is the amp I was referring too, fully balanced and about the size of the old double CD cases...

Thanks for your comments, spritzer.

Yes, I had already thought about doing a mod on the SRA-14s to give it a balanced input. I even went to a lot of trouble to get complete schematics and a service manual for it (from Japan). Then, when it came time to "put it under the knife", I just couldn't do it! I just couldn't cut the chassis to pieces on a relatively rare bit of Stax history.

So, for balanced, I'll have to build a T2!

P.S. Actually, the 14s schematic is significantly different than the SRM1 Mk2. For one, it runs on higher voltage rails.......I'll send you a PM with more details.

Linear

[Pars]

Have you ever looked at John Swenson's snubber for the transformer

My link

Not sure if this helps or not, different from the normal snubbers used on the diodes.

I'll go back to sleep now

[luvdunhill]

Ok.

What do you think of this design?

http://www.basaudio.net/datasheets/hvreg.pdf

[kevin gilmore]

I have tried every combination of capacitor of all kinds, and resistor and capacitor of all kinds

in parallel with the power diodes with no success. Now if i go back to really lousy diodes

and EI transformers, then the caps actually do stuff. But with the fast recovery diodes the

last remaining bit of noise is very hard to get rid of.

I really like the basaudio thing. Sure wish i had seen it earlier. The T1 current source

feeding the gate of T2 makes all the difference. But it also needs the low voltage supply

for the opamp, which i did on the T2 with a current source and zener.

Will study more tomorrow. Thanks marc.

Still i like the hybrid approach that shuts down the current source in the absense

of a load. And adding that to the basaudio thing probably turns it into an oscillator

like the problems i was having.

More to do

plenty of extra time now due to this...

http://gilmore.chem.northwestern.edu/dsc_1813.jpg

Edited April 1, 2011 by kevin gilmore

[luvdunhill]

Credit for that design goes to Emile Sprenger and the late Allen Wright. There might be some later iterations of it, sold through Vacuum State Audio.

Edited April 1, 2011 by luvdunhill

[Pars]

Swenson's snubber is not in parallel with the rectifier diodes, but in parallel with the transformer secondary(ies) and is intended to damp transformer ringing.

[kevin gilmore]

Swenson's snubber is not in parallel with the rectifier diodes, but in parallel with the transformer secondary(ies) and is intended to damp transformer ringing.

Yes i have tried a few versions of that too. That was the one time i got

my ass kicked, first time in a while. For some reason getting hit with

voltages like that does not have the same effect it used to...

I'm going to have to try again.

Edited April 1, 2011 by kevin gilmore

[luvdunhill]

One other thing I've found, when playing with this on LV supplies. Use carbon composition resistors. They perform better in these applications.

[kevin gilmore]

This gets into the discussion of which values of film resistors

(varies by manufacturer) is the least inductive. If you are

careful, you definitely can get metal film resistors that

match the carbon comps. Or buy those non-inductive bulk carbon

film on ceramic things which are very temperature stable.