Krill - The little amp that might...

Status
Not open for further replies.
Jump to Latest
#1,303
Hi Scott,
You're going to drive people mad with that stuff. Just to add to the bedlam, the airplane does take off normally. :devilr:

I guess we should try to stay on topic now. Since there are some Krill amplifiers nearing completion, it shouldn't be long before we are getting real, hard data. I can see where most instruments are not capable of that low residual noise, and even if they were, the connections might prove to be a problem.

-Chris
 
#1,304
anatech said:
Hi Scott,
You're going to drive people mad with that stuff. Just to add to the bedlam, the airplane does take off normally. :devilr:

I guess we should try to stay on topic now. Since there are some Krill amplifiers nearing completion, it shouldn't be long before we are getting real, hard data. I can see where most instruments are not capable of that low residual noise, and even if they were, the connections might prove to be a problem.
Click to expand...

Can't wait for "real, hard data"! Although I'm afraid we'll debate soon what exactly means "real", "hard" and "data". Certainly, the models we are using for these concepts are wildly different, so the process may not converge at all, because the system non-singular matrix transformation into a normalized Jordan tri-diagonal format failed.
 
#1,306
traderbam,
Hi. That seems a very sensible suggestion on the face of it. Thought experiment. Suspend disbelief and assume that this makes the performance of the OPS worse. Even though the drivers are now operating in a more linear region. What else is going on? What could the trade-off be?
Click to expand...
In the deep sense of it. Although you`re a mathematician, you should consider linearity as the main feature in audio, not losing sight of it for a second. Any switching action causes the worst kind of nonlinearity, measured subjectively. If you`d with some method try to accurately unite split signals, you probably would soon find yourself in trouble, noticing that they won`t allow you to do that.
 
#1,307
REVIEW: driving the Krill OS?

Just want to review, what did we say was required to drive the Krill OS?

I seem to remember someone saying something about the impedance looking in, or looking out at... can we reduce this to a current & voltage (power) requirement effectively?

Also, have we decided what the bandwidth and/or slew rate ought to be for the VAS that will drive the OS (outpoot stage)??
(assuming no global loop feedback for the moment)

I'm thinking that these will be useful data points for those who are building the OS only over in the Construction thread...

_-_-bear
 
#1,308
By bear-have we decided what the bandwidth and/or slew rate ought to be for the VAS that will drive the OS
Click to expand...

When I had my prototype going I did some crazy things...
first I ran the output of a lowly 741 through it.. not loud enough!
(sounded just like the 741 through headphones) , then I
ran the output of one of my blameless's into the krill..better.

The krill's OPS just copies anything "put to it". It slews very
fast (30+ V/us - depending on devices) so the advantages
(or disadvantages) of various VGS topologies would be heard
regardless. I'm looking into TI's OPA line,
but some spec out
kind of bad (10v/uS).. sloooow.
OS
 
#1,310
Lumba Ogir said:
traderbam,

In the deep sense of it. Although you`re a mathematician, you should consider linearity as the main feature in audio, not losing sight of it for a second. Any switching action causes the worst kind of nonlinearity, measured subjectively. If you`d with some method try to accurately unite split signals, you probably would soon find yourself in trouble, noticing that they won`t allow you to do that.
Click to expand...

I know where you are coming from. If you are right, what is your explanation for SD not running his drivers in class A?
 
#1,311
Steve Dunlap said:
I spent some time today simulating my circuit with some of the suggested improvements made on this thread. First, I simulated my circuit as I designed it. I am not ashamed of the simulated THD. It seems to agree quite closely with my measured results.

Some obvious observations:

The posted VGS is far better than some have claimed.

None of the recommended improvements showed an improvement in THD.

If you can't duplicate my results, that is not my problem.

I will not post my models. I bought them, but according to the licensing agreement I can not distribute them.
Click to expand...


Hi Steve,

I'm not sure you saw my post above, but I'm interested in knowing whose transistor models you are using for your simulations. I realize that models are sometimes proprietary, but often others here will have access to those models.

Thanks,
Bob
 
#1,312
Yes, it is self evident that the OS is unity gain.

My questions remain as before.

Do we need any current/power to control the input of the Krill OS?
What bandwidth is optimal?
What slew rate is best?

Not which circuit to use to do the job.
That's a separate question - this is a question of specification.



_-_-bear
 
#1,313
My simulation indicates the input Z is hundreds of kohms even up to 100kHz. Being a non-GNFB design, the input stage needs to roll off above 20kHz and have a slew rate as high as the highest input signal slew rate x gain.
John Curl will insist that the slew rate needs to be several millions of times higher than this 😉 but he is really observing that it is usually the case that a circuit that can slew at the minimum requirement, with extremely low distortion, will also slew at a much higher rate.
 
#1,314
By bear -Do we need any current/power to control the input of the Krill OS?
Click to expand...

Hfe of the KSA/C's , which pass 10mA (.01A) / 300 (high gain group KSA/C)= 33uA (my models show about 50uA reflecting
200 Hfe) So, no power/current to speak of..(you could put
a wire on the input and pick up AM radio. 🙂 )

As far as slew/bandwidth , my guess would for it to be as good as
possible. (20v+ /uS / flat to 50k+) Any decent discrete VGS
should do that or better. Most chipamps fall short (LM3886),
but people still like them. Same with tubes.

OS
 
Status
Not open for further replies.
Top Bottom