Discussion on the design of the T2 & other circuitry

[livewire]

Here is the first cut of the hybrid shunt regulator version of the kgsshv power supply.

http://gilmore.chem.northwestern.edu/shuntregver2.pdf

Unlike some other shunt regulators however, the current sources

shut down if the shunt element pulls too much current, so its

safe to power up with no load.

edit: silly bench mistake fixed. Actual noise under load about 3.5 microvolts peak to peak.

So what is the benefit of that design over another one of your earlier

kgsshv power supply designs? (see attached file below)

I am currently using the kgsshvps2 design because it works. Is it safe to power up w/o a load?

I am striving to learn the mechanics of the hybrid shunt regulator as well.

(the one in the kgsshvps3 psu)

Heck, if you and Linear come up with something better, I may use it.

Link: http://gilmore.chem.northwestern.edu/kgsshvps2.pdf

[kevin gilmore]

The current kgsshv power supply is a series pass design.

Works just fine from 0 to full load.

Alex Cavalli is now famous for making shunt regulators

that blow up real good when the load disappears for

any reason.

In theory shunt regulators can be a bit quieter

than series regulators. In practice, once you get

into the hundreds of microvolts of noise or less

its the power transformer and diodes that are most

of the problem, and something not easy to fix.

Compared to the hundreds of millivolts of ripple

on most power supplies for tube use, we are well

into the major tweek realm.

I think that added zener diodes across the tl431's and tlv431's

are probably necessary for startup and shutdown conditions.

It sure would be nice to have 100% isolated packages for

the entire designs.

Edited March 29, 2011 by kevin gilmore

[Linear]

With the TL431 MOSFET Cascode scheme that I posted a few messages back, one can, of course, make a T2 type 740 V active battery. Please see the attached drawing. I don't think you can get much simpler than that, and with a 431A (1%) and 1% resistors, we can hit damn close to 740 V without any adjustment.

The accepted Head-Case naming convention for circuits appeals to my ego, so I'm calling this the "DWAB1". (I guess that the 431 MOSFET Cascode shunt regulator is "DWSR1").

Now, there is only one problem with the DWAB1. It won't work as an active battery in the T2! I'm almost sure that it's noise level (220 nV/rt Hz) is way too high. By comparison, the original JFET scheme is about 25 nV/rt Hz.

These noise levels come from theory, so I'm going to measure them experimently. (Forget Spice simulation - the macro model for the TL431 almost certainly doesn't have accurate noise parameters. You would have to model the 431's complete schematic at a transistor level.)

Let's see: I've got a breadboard of the original T2 active battery, got a breadboard of my DWAB1, got a distortion / noise analyzer, so just hook them up to a power supply.......wait a minute, that has to be an ultra low noise PS, at 740 V! I've got it! I'll use eighty-two 9V batteries is series.....which would cost well over $100...

If I don't electrocute myself with a 9V battery (actually, 82 of them), or blow up my distortion analyzer, I'll report back. Stay tuned.

Linear

[livewire]

@KG - thanks for the reply, time will tell if the second iteration of the kggsshv supply works for me.

In the meantime I'm all ears.

@ Linear - 82 batteries, what a novel idea. That should erase any noise caused by an input trafo & bridge.

Dont need them. As I stated before, I'm all ears.

[Linear]

.........eighty-two 9V batteries is series.....which would cost well over $100...

I've just been getting together a bunch (engineering term: "shit-load") of 9V batteries.......they clip together in series so nicely.......idea........what's the date? March 29? OK, it's 3 or 4 days early, but here goes anyway......

*****************************************************************

BATTERY POWER YOUR T2, KGSSHV, OR KGBH AMPS!!!!!!

Don't be tied to that wall plug and that heavy power supply!

Get Mobile! Go jogging with your T2 and Stax headphones!

Specifications:

500 V supply: 52 batteries (9V)

250 V supply: 26 batteries (9V)

T2 users: 6 V car battery (for valve/tube heaters)

Battery life @ 50 mA: 7.5 hr

PS Regulation: -21% @ 7.5 hr (494V > 390V, 247V > 195V)

Price: 500V supply = $208 (duracell batteries)*

250 V supply = $104 (duracell batteries)*

* Car battery not included

******************************************************************

April Fool!

[luvdunhill]

oh, I'd be interested to know what the noise is of that many batteries in series.

[livewire]

QUOTE: Don't be tied to that wall plug and that heavy power supply!

Get Mobile! Go jogging with your T2 and Stax headphones!

Ha!

I think that KG beat you to the draw on that one.

Although he used the Stax SRM-2XX with his Omegas.

Somewhere around here is a pic of a travel bag outfitted

with what looks to be a motorcycle battery (gel cell?),

the amp, a Sony Diskman CD player source and Stax.

Got a good laugh when I saw it.

Edited March 29, 2011 by livewire

[spritzer]

I did something similar with the ESX. I needed a simple C- supply so four 9V batteries off the B- for a -460V supply...

[Linear]

oh, I'd be interested to know what the noise is of that many batteries in series.

Good point again, Luvdunhill. I found the following on the web. The pink trace is ONE 9V alkaline battery. (The blue trace is a LM337 PS the guy was boasting about.)

Anyway, the battery is about -110 dBV @ 1kHz. Now, 100 batteries in series - 40 dB worse? = -70 dBV? Maybe the noise adds rms, and is not quite that bad, but still, it's not an ultra low noise PS! I've cancelled my bulk order for 500 durcells! Low noise is hard work, but then you all knew that.

On the site where I fould the battery noise measurements, it states "TL431's are prone to wild oscillations". So, we will have to see how this 431 mania pans out, once we build these circuits. Maybe my cascode MOSFET will isolate the 431 from the load to some degree.

But the MOST inportant question of all. Where did you get those massive vertical fins heatsinks, Luvdunhill???

Linear

[luvdunhill]

I'm assuming you mean the really long heat sinks I posted in the original T2 thread?

Well, if a LM317 is so good (and assuming we trust those guys), what about a Maida regualtor:

http://www.national.com/ms/LB/LB-47.pdf

The above is a layout I did.

[kevin gilmore]

OK, so linear's first circuit with the ixys current source devices and the tl431

with a 10uf cap across it actually tests at 20 microvolts peak to peak noise.

(10k load) about 4 to 5 times worse than the T2 battery version.

With anything less than 10uf it has very nasty startup oscillations.

Obviously with the cap across it, it is nowhere near as fast as the T2

battery.

Absolutely needs a 24 volt protection diode across it.

Going to try the tl431 as the current source next, but

anything i have tried like this previously has a lot more

noise from the switching of the diodes because of the

resistor to the gate.

Edited March 29, 2011 by kevin gilmore

[kevin gilmore]

QUOTE: Don't be tied to that wall plug and that heavy power supply!

Get Mobile! Go jogging with your T2 and Stax headphones!

Ha!

I think that KG beat you to the draw on that one.

Although he used the Stax SRM-2XX with his Omegas.

Somewhere around here is a pic of a travel bag outfitted

with what looks to be a motorcycle battery (gel cell?),

the amp, a Sony Diskman CD player source and Stax.

Got a good laugh when I saw it.

Hey, i used that for years and years. But after 9/11 there

was no way in hell they were going to let me on an airplane

with that. It was a 12 volt 7 AH gel cell.

Edited March 29, 2011 by kevin gilmore

[Linear]

OK, so linear's first circuit......about 4 to 5 times worse than the T2 battery version.

With anything less than 10uf it has very nasty startup oscillations.

Absolutely needs a 24 volt protection diode across it.

tl431 as the current source.......has a lot more

noise from the switching of the diodes because of the

resistor to the gate.

I'm working on a version of the T2 battery that is looking really good. It's essentially the same as Dr. Hayashi's circuit, but turned upside down so that it only needs 1 HV device (FQPF8N80C) and the 2sc3381 are gone! It should drop right in to the kgsshvpower3 schematic and give the better performance KG quotes above. But maybe the TL431 approach would be OK as a low cost, easy-build tradeoff, versus the "flat-out absolute best PS possible" approach.

I hope the TL431 current sources work out! Right now, I can't see any other option except the ixsy devices. Maybe the NPN could be used if the TLV431 acts up. We would have a Tc on the current, which would lower the current with increasing temp(~0.5%/degree C). Maybe that would be a good thing (to prevent thermal runaway). By the way, the T2 active loads (with the LED biasing) seem to have this kind of Tc. Has anybody meaured the "standing current" of the T2 over temp?

I'll try to get to testing the '431 current sources tonight, but KG will probably beat me with some results!

Luvdunhill - Yes, "the really long heat sinks I posted in the original T2 thread". Where did you get them? It wouldn't have been from the land down under, would it? Eh, mate?

[livewire]

Hey, i used that for years and years. But after 9/11 there

was no way in hell they were going to let me on an airplane

with that. It was a 12 volt 7 AH gel cell.

Same battery I have in my Yamaha 1000.

When I saw your portable rig I was was impressed by the concept, albeit somewhat bulky.

Realy dig the dated "old school" CD transport.

[kevin gilmore]

media players did not exist back then

http://gilmore.chem.northwestern.edu/porta.gif

date... 2/20/2000

besides which, compressed to hell music from one

of the early mp3 players on stax headphones... be serious.

to bad sony never actually sold their portable sacd player.

I would have bought it. Now back to our regularly scheduled

program, food, a bit of rest, now lets see if i can't send

some parts flying across the room.

Edited March 29, 2011 by kevin gilmore

[luvdunhill]

Linear:

That profile is sold by a number of different people, including Austerlitz, Fischer and Seifert. The ones I have are from Fischer.

[kevin gilmore]

The hybrid shutdown part combined with the required caps across the tl431's cause

massive oscillations. hundreds of volts. with the shutdown part removed, it works

and is about 35 microvolts peak to peak noise.

The tl431's really like to oscillate, and once you add enough cap to shut them up,

everything slows down.

[grawk]

oh come on, the ipod did lossless in 2000

[kevin gilmore]

well for one thing, i had already been using that setup for

some time, not sure how long. But the 16 bit dacs in the

sony (and later the d-ej01 or some such) beat the crap

out of the less than 14 bit ipod at the time.

Too bad the stax eats so much power, a new version of

that setup is something on my todo list.

edit: d-ej01 corrected. I still have 2 of them.

Edited March 30, 2011 by kevin gilmore

[justin]

oh come on, the ipod did lossless in 2000

first ipod wasn't released until 10/23/01, and only interfaced with Macs. 2nd version had Windows compatibility but that wasn't until mid-2002.

[Linear]

OK, I woke up this morning and I've recovered from "431 fever"! This thread is what I hoped it would be - raw design in real time!

Therefore, please don't go and order parts, ask when PCBs will be available, or otherwise committ to something you see here just minutes after it has been posted. We are just throwing ideas around, and there will be a lot of churn before we get to the perfect solution. Correct that, we will NEVER get to the perfect solution! The best we can hope for is to find a a combination of trade-offs that work to our satisfaction.

So, where are we on the path to designing a hybrid power supply? Well, we are trying to combine some of KG's and some of my ideas. Hopefully, the resutl will be a PS that can be a shunt type, a series pass type, or a "hybrid" of the two. Switching between the modes will be as easy as changing two or three resistor values. This is what KG proposed with: http://gilmore.chem....huntregver2.pdf

What I hope to add to this, is simpler series and shunt elements that only use isolated package devices, thereby resulting in a very flexible PS, that will be easy to build and inexpensive. We already have a bunch of building blocks, and I think that we can reach our goal by putting them together in the right combination.

The '431 building block looks like it is too unstable (too much gain with too many transistors, including SLOW lateral PNPs) to be usable with FAST PS designs, especially those with multiple feedback loops like the Hybrid PS. It might work in a simple pure shunt, or pure series pass topology that would have higher noise. A noisy, but low cost solution? Do we really NEED < 5uV of noise? Anyway, at least while I was under to influance of "431 fever, I was able to demonstrate that it can be used in HV PS design, with a MOSFET Cascode arrangement.

Where do we go from here? I'm going to send KG an email and discuss all the building blocks that we have. We will see if we can come up with a working KGDWHPS ( .........Hybrid Power Supply )!

Linear

P.S. As I'm finishing this post, I see that we have an on-going discussion on the DESIGN thread about ipod release dates!

[justin]

how much will PCBs cost and can you send me a BOM for Mouser/Digikey?

[Linear]

how much will PCBs cost and can you send me a BOM for Mouser/Digikey?

:angry: :angry: :angry: :angry: :angry:

[kevin gilmore]

The universal series/shunt/hybrid regulator board will be posted soon

Seriously, some things work out great, some things not so great.

The tl431 has too much gain to make it useful. There may be other

similar devices out there that work better, but it is unlikely.

The idea of an amplifier with all isolated packages sure would

make things easier. Don't know of any isolated versions of the

2sj79/2sk216 either.

Any current source that references the unregulated input voltage

(like the cavalli thing) absolutely guarantees that the spike

noise from the diodes rips right thru the current source and

the rest of the circuitry. This is why the 10m90s part works

so well.

[spritzer]

well for one thing, i had already been using that setup for

some time, not sure how long. But the 16 bit dacs in the

sony (and later the d-ej01 or some such) beat the crap

out of the less than 14 bit ipod at the time.

Too bad the stax eats so much power, a new version of

that setup is something on my todo list.

edit: d-ej01 corrected. I still have 2 of them.

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...