AKSA 55, 100 - Listening impressions

[Tube_Dude]

Originally posted by Christer


With they type of "music" most people listen to nowadays, I wish they did listen to null tests.

That is the crux!!

When the residual of the null test is only thermal noise , you don't listen to the amp anymore , you listen only to the music that pass trough the amp without alteration. The amp come out of the equation.

[Christer]

Originally posted by Tube_Dude



I was talking about, the performance of the addiction of a resistor to a CCS , in the tail of a LTP in an amplifier and not of the analysis of a perfomance of a CCS per si.

And in a amp, the rail is the greatest source of noise , with many harmonics , from the swithing of the output devices and from the rectifiers .

Yes, I realized that after your previous post. However, as I said the improvements are small, and there are other ways get much better performance. On the other hand, the resistor won't do much harm either, except for adding a bit of thermal noise, which probably doesn't matter for line level amps.

But in the simulations the rails are pure DC.

Only if you forget to add a signal source to them to use for PSRR simulations.

[Christer]

Originally posted by Tube_Dude


That is the crux!!

When the residual of the null test is only thermal noise , you don't listen to the amp anymore , you listen only to the music that pass trough the amp without alteration. The amp come out of the equation.

Yes, but both I and mastertech (I think) was jokingly referring to listening to the actual result of the null test.

[Tube_Dude]

Originally posted by Christer
Only if you forget to add a signal source to them to use for PSRR simulations.

And how you simulate a signal with spikes and wideband spectral noise , as we find in a rail of a power amplifier?

[Christer]

Originally posted by Tube_Dude


And how you simulate a signal with spikes and wideband spectral noise , as we find in a rail of a power amplifier?

AC sweep and trusting Monsieur Fourier.

Of course, simulations have their limitations, but sweeping the frequency gives you PSRR vs. frequency graph. You can also do transient analysis using pulses or other curve forms, and you can of course add several signal sources if you want to study IM effects etc.

As usual, no simulation is better than its models. Measurements can tell you some things the simulator can't tell you, but the simulator can also tell you some things you would fail to see if trying to measure them, at least without fancy state-of-the art test equipment.

[Christer]

Originally posted by lineup
A resistor from differential pair emitters to set current
is in fact not a constant current source.

Of course not, but resistors have been used as approximations of current sources since the early days of electronics. The lower the AC to DC ratio, the better the approximation.

LTPs originally used a resistor only, and many textbooks present them that way. The use of a CCS was a later invention to improve the CMRR and with todays cheap transistors, is probably the most common method.

Some seem, however, to prefer the sound of using a resistor. Somebody recently mentioned in some other thread that there was an amp (McIntosh, HK, ...?) that had an extra winding on the transformer just to provide a -100 V supply for the LTP tail resistors.

[hifryer]

If you want to do null tests, here is a way:

http://sound.westhost.com/sim.htm

http://sound.westhost.com/project57.htm


How revealing this is will depend on what tortue you feed it.

It may reveal speaker interactions as well.

Good for gross distortions but how will it reveal problems in ambience, imaging, decay etc and how will we interpret these?

Useful but no cigar I think. Our main problem is that way have no scale to evaluate the importance of differences when we find them. In isolation they can be seen in magnitude but not importance.

Our ears listening to actual presented music still remain the most sensitive and reliable instruments we have in evaluating the qualities of this music.

Yes, we can be fooled but if we are, over time we either fall out of love or remain happily fooled ;-) Tubes continue to happily fool me!



cheers

[Tube_Dude]

Originally posted by hifryer
If you want to do null tests, here is a way:

Like Frank Sinatra , I prefer do it , "My way ".

See post # 8 at

Null Difference Testing

Cheers

[lineup]

Originally posted by Christer
, but resistors have been used as approximations of current sources since the early days of electronics

Christer
I think you miss what I am trying to say with my post.

CCS always makes other transistor decrease its current
with same amount as input transistor increase.

While using a resistor it may happen
that
both transistors increase +0.1 mA at input of +1 Volt
this is because the increase across emitter resistor
is 0.2 mA

And to me this is a clear difference, when using constant current source versus resistor.
And CMRR, whatever this is - most of us dont know and use these words, was not my issue.

---------------------
sometimes I dont understand how people read:
they miss the main message
and add a sidenote comment on some remote detail you used to give your message
or they even 'reply' by start talking of something else

if someone quote your post
you would expect a response to your main message
Wouldnt you?


another example:
when you want to explore an amplifier, subjectively and obejctively
people can be allwoed to sabotage a whole page (10 posts)
by compete who can make most funny post about beer

this is clearly not what we want
Do you want such response?

my advice:
those who wants to discuss beer in diyaudio.com
should start threads in Off Topic
instead of polluting other peoples discussions.
This is what will benefit diyaudio.com
and keep this a respectable website for exploring DIY audio

[lumanauw]

I experimented that in average DIY amps (not all), resistor do sounds better as CCS. The question is why, how come?
Lineup gives an explenation that R is not working so good as CCS (for differential pair). But what makes it "seems" to sound better?

I think the answer is not as simple as "it gives harmonic distortion bloom". It has to be a technical explenation why. (I'm not EE here )

The schematic that Christer mentioned maybe this one http://www.diyaudio.com/forums/showt...768#post752768
Why that amp makes the difficult+expensive way of making additional high-voltage rail just to feed plain R as CCS? It is more easier for them to make 2 transistor based CCS with the same rail.

I ask a question in this thread Base Stoppers
The answer seems lies in the transistor itself. Transistor is not all "real number" device. It has capacitance, so its properties and behavior has to be expressed in complex number (i+jw).
But since the capacitance properties is so small, this (i+jw) behavior appears at frequencies sweep above 100khz-Mhz region. In this high frequency sweep, this (i+jw) does makes a difference.
(There are 2 very good papers in this thread, one pointed by Dimitri and one pointed by JCX)

The question is : Audio is below 20khz, so what is the importance of Mhz behavior there? I have tried transistors with higher fT, usually they do sounds better. Again, what is the relation of Mhz behavior to below 20khz reproduction? It seems intermodulation is the answer
What is wrong?


But I have an opinion (for myself) that transistor based CCS sounds different than plain R because of this Mhz behavior, which is more important than the drawback of using R as CCS (like Lineup pointed).
Input-output properties of transistor based CCS (like impedance, etc) in 1khz will not be the same with those properties in 1Mhz or 10Mhz.
In the other side, Resistor is the same at 1khz or 1Mhz or 10Mhz, it only follows Ohm law V=IxR, no (i+jw) there.

This maybe still the missing link. What is the corelation of Mhz behavior to below 20khz reproduction properties?