DIYHiFi.org

For the sake of audio
It is currently Fri Jul 21, 2017 6:47 pm

All times are UTC




Post new topic Reply to topic  [ 34 posts ]  Go to page Previous  1, 2, 3  Next
Author Message
PostPosted: Sun Jul 16, 2017 11:08 pm 
Offline
Cow

Joined: Mon Nov 28, 2005 1:10 pm
Posts: 479
OK. Like I said BC327 works fine in that position, even with a big C2. No blowing up.
Of course I don´t think I will be using C2 anymore.


Top
 Profile  
Reply with quote  
PostPosted: Sun Jul 16, 2017 11:19 pm 
Offline
Benjamin
User avatar

Joined: Mon Dec 13, 2004 3:29 pm
Posts: 4319
Location: Somewhere other than here.
Jocko Homo wrote:
Maybe I will measure a generic Toshiba, from my junque boit.

Or not.


Don't know why I waste my time, since there are so many parasite's that will use what I share, then bad mouth me, when they think I am not looking.

Well, anyway...............for the rest of you good folks..................

OK, took one of those NMOS depletion-mode CCS thingies that I have been known to use, and drove several parts with them. Measured the noise, of those parts, through the old 60 dB cannibalized old phono stage, and shoved that into our ancient HP FFT.

I did not normalize out the pick-up hum, and the preamps intrinsic noise, because, well, I didn't feel like, and screw you guys anyway. So, the numbers are not exact, but close enough for the parasites to steal and parrot.

So, first up, we have the good ol' green LED.

Please note the scale is nV/rt-Hz, on all of the plots, except the last one. On that one, it is uV/rt-Hz. Not my fault the s/w I used, to get the data off of the FFT is not perfect, and that one plot is missing that key detail.

(Also, I put a © notice, on all of the plots, since I know some horse's patoots will copy it, and put it on their crappy forums. Screw you guys!)

Anyway, back to the point at hand........................green LED:

Image

So, its noise floor is a bit over 1 nV/rt-Hz. Not bad.

But, that is only good for 1.8 V or so.

So, I then put in a 1K resistor, which should gives us close to the same voltage:

Image

So, in terms of voltage, it is around 10x more.

Well, what happens if we want something more than 10 V or so? Try a 10K resistor:

Image

Well, the noise is obviously more, but worse than that.......................more hum pickup. Which one would expect.

So, having gone through all of that..................

What happens when we put in a 2SA970, that is reverse-biased? (Remember, we are now at uV/rt-Hz):

Image

There you go: 2.5 uV/rt-Hz, through most of the audio band.

Voltage was around 13 V, in case anyone cares. I don't. But, you "asked", and I delivered.

Screw you guys, I'm going to eat dinner.


Top
 Profile  
Reply with quote  
PostPosted: Sun Jul 16, 2017 11:46 pm 
Offline
Benjamin
User avatar

Joined: Mon Dec 13, 2004 3:29 pm
Posts: 4319
Location: Somewhere other than here.
Oh, you know what..................

Those 10K numbers might be wrong. The voltage, on the CCS gizmo, may not have been high enough, to leave enough headroom, for a 10K load.

I'll re-run that one, tomorrow, and update it, if needed.

So, for now................kinda sorta ignore that one.

The bottom line is: avalanche noisy!


Top
 Profile  
Reply with quote  
PostPosted: Mon Jul 17, 2017 2:41 pm 
Offline
Benjamin
User avatar

Joined: Mon Dec 13, 2004 3:29 pm
Posts: 4319
Location: Somewhere other than here.
OK, I am down at the shop................and no one is here!

I am going to run those tests, again, with the preamp's hum and noise cancelled out, and with a high enough rail supply, so the 10K one is not crunched.

Stay tuned.

Or not.


Top
 Profile  
Reply with quote  
PostPosted: Mon Jul 17, 2017 3:10 pm 
Offline
Cow

Joined: Mon Nov 28, 2005 1:10 pm
Posts: 479
Do a zener measurement if you have it at hand! 6V8?
I´m reading up on noise and addition of noise sources.
http://cas.ee.ic.ac.uk/people/dario/fil ... -noise.pdf
(A string of leds is looking good to me. I reckon that is what you use Jocko.)


Top
 Profile  
Reply with quote  
PostPosted: Mon Jul 17, 2017 4:01 pm 
Offline
Benjamin
User avatar

Joined: Mon Dec 13, 2004 3:29 pm
Posts: 4319
Location: Somewhere other than here.
OK, green LED:

Image

Looks better, no hum, and noise floor (<1 nV/rt-Hz) is better!

Next...................1K:

Image

A bit lower, this way: 16 nV/rt-Hz

And the 10K, with lots of headroom:

Image

Not only looks better, but more believable.

OK................next time........................normalize (Not that you guys are worth the effort.)

And now, the 2SA970, which, as one would suspect, looks just as bad as it did before:

Image

Which, it should, since its noise swamps the preamp's noise.

(Remember, still uV/rt-Hz, on this one. Yuk.)

So, what have we learned?

1.) Normalize
2.) Avalanche noisy

Or, "Bleeding head, good. Healed head, bad."

Or not.

Screw you guys..................


Top
 Profile  
Reply with quote  
PostPosted: Mon Jul 17, 2017 4:03 pm 
Offline
Benjamin
User avatar

Joined: Mon Dec 13, 2004 3:29 pm
Posts: 4319
Location: Somewhere other than here.
alexandre wrote:
Do a zener measurement if you have it at hand! 6V8?


Do you really need to ask?

"No."

OK, then don't!

Quote:
I´m reading up on noise and addition of noise sources.
http://cas.ee.ic.ac.uk/people/dario/fil ... -noise.pdf
(A string of leds is looking good to me. I reckon that is what you use Jocko.)


"Yeah, but LEDs are noisy, because of light and stuff."

(Only if you reverse-bias them.)


Top
 Profile  
Reply with quote  
PostPosted: Tue Jul 18, 2017 2:23 am 
Offline
Muriel
User avatar

Joined: Thu Jul 26, 2007 5:45 pm
Posts: 1096
Hi,

alexandre wrote:
A brief note might help grasp this circuit.


IIRC, 12V Zenners had the sweet spot for noise, a reverse biased transistor as 6.8V was very noisy.

LED's generally are pretty quiet, but use the older "not high brightness" ones. I seem to remember green being better than Red and Yellow, but that may have been the specific ones I used. Infrared are supposedly low noise too. Important in all cases is to go mick Jagger on the LED's (paint it black).

Some recent voltage references have pretty low noise too.

With a BJT (and it's tempco) this shunt is never going to be a marvel of DC accuracy and low drift (though this can be compensated if needed). AC behaviour can be pretty good though.

Depending upon load and other requirements a low noise, high speed op-amp can be a great choice for a regulator. I tend to make an AC loop directly from the Op-Amp's outputs and a DC loop from after the build-out inductor. If you take care, you can get an excellent flat handover from the Op-Amp (plus inductor DCR) to the bypass capacitors after the inductor and excellent DC precision. In a pinch even a 5534 can work well enough, though better options exist these days.

Ciao T

_________________
"For a successful technology, reality must take precedence over public relations, for nature cannot be fooled." Richard Feynman


Top
 Profile  
Reply with quote  
PostPosted: Tue Jul 18, 2017 4:25 am 
Offline
Benjamin
User avatar

Joined: Mon Dec 13, 2004 3:29 pm
Posts: 4319
Location: Somewhere other than here.
OK, one thing needs to be mentioned, about stacking LEDs.

What happens when you stack 2?

Does the noise go up 6 dB?

Or 3 dB?

Most folks will say 6 dB, because that there are 2 of them, and the voltages add, so that is 6 dB.

Well...............no.

Noise is not correlated, so it is noise powers, and not noise voltages, that add up.

So, stack 2, and the noise goes up 3 dB. Up that to 4, and that means 6 dB more than a single unit.

Every doubling adds 3 dB of noise.

Etc., etc., and etc.

So, if you stack 7, to get somewhere around 12 V, it is 8.5 dB more than a single.

Find a zener that is that quiet....................


Top
 Profile  
Reply with quote  
PostPosted: Wed Jul 19, 2017 11:15 am 
Offline
Muriel
User avatar

Joined: Thu Jul 26, 2007 5:45 pm
Posts: 1096
Hi,

Jocko Homo wrote:
Does the noise go up 6 dB?

Or 3 dB?


Neither, in reality it's more like +4dB.

Rather than stacking References or LED's, we can simply make the AC gain unity and the DC an arbitrary value to give us the required DC output.

This produces lower overall noise, except at very low frequencies (depending on size of bypass cap) where stacked references may be a better choice.

That said I know VERY few applications that need both very low extremely low frequency noise AND relatively high voltages, but no doubt some exist. Again, it's a question of appropriate technology.

Ciao T

_________________
"For a successful technology, reality must take precedence over public relations, for nature cannot be fooled." Richard Feynman


Top
 Profile  
Reply with quote  
PostPosted: Wed Jul 19, 2017 6:09 pm 
Offline
Benjamin
User avatar

Joined: Mon Dec 13, 2004 3:29 pm
Posts: 4319
Location: Somewhere other than here.
Kuei Yang Wang wrote:
Rather than stacking References or LED's, we can simply make the AC gain unity and the DC an arbitrary value to give us the required DC output.

This produces lower overall noise, except at very low frequencies (depending on size of bypass cap) where stacked references may be a better choice.


And there's the rub.........................

Most of the folks I do work for prefer the sound, of the overall design, when the bypass cap is not used. And since it is their product(s), and their money (paying the bill), the bypass cap is usually not used. (And low voltage stuff, so not a lot of extra noise accumulated.)

I say "usually" because I can think of one recent case where sticking a really big one was deemed acceptable. From a guy that I thought would never use one.

Now that he has been declared schizophrenic, not sure how to approach the next job................!


Top
 Profile  
Reply with quote  
PostPosted: Wed Jul 19, 2017 6:24 pm 
Offline
Benjamin
User avatar

Joined: Mon Dec 13, 2004 3:29 pm
Posts: 4319
Location: Somewhere other than here.
Which reminds me of a story............................

Sometime, back in the 80s, we loaned our phono preamp to a guy, for his CES exhibit. He loved it, and demanded that we build one just like it, for him.

We did.

Except that he hated it, and went ballistic.

"What the **** did you do to this thing? It sounds like ****. I told you to make one just like it, and you sent us this ****."

"The only thing that could possibly be different is the bypass cap, in the pre-reg. Both units were made with whatever we found, at the surplus store, because we felt as far upstream as it was, it wasn't that critical. I can't imagine the one in ours being all that great."

(It really was some no-name brand, we found at the surplus store. Next one used a different no-name brand, that they had at that time. So, from that day on, we settled on Panasonics, as everyone seemed to be ok, with them.)

(Then they stopped making HF, and went to HFQ. Then to HFS, etc. We got tired of having to buy new caps, all the time, and long story short, we started to thinking about "What if we leave this damn cap out?")

(And speaking of Crapasonic..................they stopped making the snap-lok caps, used as the main PSU caps. Had to use different brands, and no one liked them. So, went to using their regular 'lyitcs, and folks were happy again. Until the day I stuck one of the "Con Brothers" brand in a proto.

Yeah, I caught hell for that. Back to crappy Crapasonics.)

Moral of the story: open-loop designs are really sensitive to the caps.

I hate caps.

Don't ask me about the 'lytics, in power amps. Unless you want me to never talk to you again.

Hint: now you know why I have not built a power amp for around 20 years.


Top
 Profile  
Reply with quote  
PostPosted: Wed Jul 19, 2017 9:23 pm 
Offline
Sheep

Joined: Thu Jan 13, 2005 10:34 am
Posts: 91
Location: Bath, UK
Kuei Yang Wang wrote:
Hi,
Rather than stacking References or LED's, we can simply make the AC gain unity and the DC an arbitrary value to give us the required DC output.

This produces lower overall noise, except at very low frequencies (depending on size of bypass cap) where stacked references may be a better choice.


I've played with this for a decade or more (er, more actually) in small regs and higher -voltage regs and surprised myself when I firmly came down on the side of running a higher voltage reference, and flat & low gain across the spectrum.

It made no sense first time given where the AC LPF was set - well below 10Hz - but to me, and for the things I use it for - no contest. The eye/earopener for me was wanting about +/-21v or so and comparing LV references and the approach you outline, to simply stacking say two LM329s fed by a good CCS, filtering that output - and using 1.5x constant gain in the Vreg error amp. it's not even that quiet an error amp! Note - this with a hybrid-based SMT discrete opamp of claimed stupid PSRR & high OL gain , and v low gain after feedback (no, none of the usual suspects, a custom jobbie) - not an open-loop stage.

Made a far, far bigger difference than messing about with board -level decoupling, anyway; and expanded my maths and analysis as a result, no bad thing.

(I also learned a few things about what LM329s do and do not do & do not like that way...)


Top
 Profile  
Reply with quote  
PostPosted: Thu Jul 20, 2017 1:28 am 
Offline
Muriel
User avatar

Joined: Thu Jul 26, 2007 5:45 pm
Posts: 1096
Hi,

Jocko Homo wrote:
Moral of the story: open-loop designs are really sensitive to the caps.


Well, hate to break it to you, so are closed loop designs...

Jocko Homo wrote:
I hate caps.


Yeah, but what else would you use?

Jocko Homo wrote:
Don't ask me about the 'lytics, in power amps. Unless you want me to never talk to you again.

Hint: now you know why I have not built a power amp for around 20 years.


I did a poweramp recently, strictly as experiment. Grounded Bridge design using current boosted LM3875 "Gainclone" Chips (discontinued) with some really beefy RET (also discontinued) current dumper transistors, with regulated floating 84V supply (so the bleed'n thing can swing 160V PP into a 4 Ohm load - yup, that's nearly 800W). Yes, it uses loop feedback and gainclone chips, the result sounded better at small signal levels (where the LM3875 do all the heavy lifting) than a Gainclone full of Elna Silmic Capacitors - probably because the PSU electrolytics no longer form part of the equation. At high signal levels, lets just say it's a beast.

Anyway, no-one wants boring, good sounding Class AB amp's when they can buy the latest totally over hyped ex Class D amplifier. So I'll not publish this one and not put it into production either, for now.

Ciao T

_________________
"For a successful technology, reality must take precedence over public relations, for nature cannot be fooled." Richard Feynman


Top
 Profile  
Reply with quote  
PostPosted: Thu Jul 20, 2017 4:44 am 
Offline
Cow

Joined: Mon Nov 28, 2005 1:10 pm
Posts: 479
This sort of amp works very well IMO. But mine is low powered - a composite amp (nested feedback) with THS4131 and two LM3875 (bridged).


Top
 Profile  
Reply with quote  
Display posts from previous:  Sort by  
Post new topic Reply to topic  [ 34 posts ]  Go to page Previous  1, 2, 3  Next

All times are UTC


Who is online

Users browsing this forum: No registered users and 1 guest


You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot post attachments in this forum

Search for:
Jump to:  
cron
Powered by phpBB® Forum Software © phpBB Group