Arturo,
Thank you for this observation. I will try this out.
I've also been looking at reducing noise as it is dominate at small power levels. The input stage degeneration resistors along with the transistors play a big part in the noise. Do you have any thoughts or observations on this?
Ken
About noise
Hi Ken,
yes one BIG source of noise are the input degeneration resistors, this noise is amplified by subsequent stages. I prefer as little degeneration as possible (100% margin when leaving oscillation condition), this must be evaluated with the amp as a whole at different reactive load conditions, high dv/dt input impulses, clipping ... Another factor is the noise injected by the NFB network, so low impedance NFB network helps, but changes DC offset, that's why (among other considerations) I prefer a complementary topology and adjust the DC offset with the current at one of the input current sources. The circuit showed uses low degeneration, stabilising VAS 'dumpers', and low NFB loop, temperature compensation diodes (the source of the numbers I posted previously).
Cheers
Arturo
Hi Ken,
yes one BIG source of noise are the input degeneration resistors, this noise is amplified by subsequent stages. I prefer as little degeneration as possible (100% margin when leaving oscillation condition), this must be evaluated with the amp as a whole at different reactive load conditions, high dv/dt input impulses, clipping ... Another factor is the noise injected by the NFB network, so low impedance NFB network helps, but changes DC offset, that's why (among other considerations) I prefer a complementary topology and adjust the DC offset with the current at one of the input current sources. The circuit showed uses low degeneration, stabilising VAS 'dumpers', and low NFB loop, temperature compensation diodes (the source of the numbers I posted previously).
Cheers
Arturo
Attachments
So, I reduced the degeneration resistors to just above where the high frequencies fall off and changed the input transistors to lower noise ones 1nv sqrt Hz. I made a few adjustments to the TMC values and basically cut the noise in half. The first image is the noise of the fig 3.10 front end with the allison CAS back end.
Attachments
Arturo,
Are you running this in Class A only? In comparison to your scheme, I found that with the configuration I have if I reduce the NFB much below this value, the THD begins to rise. It also doesn't move the noise much lower. Seems I've found a decent balance between noise and THD. I also checked the noise of the transistors with Bob Cordell's noise simulation test and found that his models have very similar noise profiles (though the manufactures specs are different). In the spice thread I think he mentioned that he is updating his models trying to get the noise values to work accurately - many thanks again to Bob!
Here is the latest version of my "low noise Allison with Fig 3.10 modified front end".
Ken
Attachments
Hi Ken,
This amp run with a current bias of 1.4A, That gives full Class A for output power below 15W (8OHM), which is reasonable low not to need such a bulky heatsink at living room listening level. Power above 15W to 60W enters in Class AB region.
About the NFB network there is a point where the noise reduction cannot be improved further and is dominated by the input devices+degeneration resistor noise, there is a trade off between noise and distortion, each amp has it's own optimum.
Cheers
Arturo
This amp run with a current bias of 1.4A, That gives full Class A for output power below 15W (8OHM), which is reasonable low not to need such a bulky heatsink at living room listening level. Power above 15W to 60W enters in Class AB region.
About the NFB network there is a point where the noise reduction cannot be improved further and is dominated by the input devices+degeneration resistor noise, there is a trade off between noise and distortion, each amp has it's own optimum.
Cheers
Arturo
CAS = Current Amplification Stage; these would be the pre drivers, drivers and output transistors, how they are arranged, and the componets that support them.
Class A avoids cross over distortion, if well implemented it should sound better than Class AB. The simulations that I have presented stay in Class A until about 16watts then it transistions to Class AB. You should order Bob Cordell's book if you want to understand this.
You can see from the prior graphs that the simulated THD is very very good. My physical measurements show very good results as well, but, not at the level of the simulations - I can measure at the simulation resolution, nor, would I expect the amp to match the simulations, it's more about relative performance compared to known baseline amps.
If you download Bob Cordell's simulations, they include the models and have preset plots as well, so, all you do is hit the run button. It's a good starting point. Also look in this forum for the Spice Simulation thread Software Tools - diyAudio.
Ken
Hi Arturo,
I with you on this... your sims with the BD139/140 did you tune the models? Or just use the manufacturers models?
Ken
I think the most problems with the CAS arises when it asked to deliver and thus is driven into switching...seems like this sends spikes back through the feedeback....The sonic qualities of this output stage is also due the the fact that it runs rather hot and because of the it rarely leaves class A operation...
came to think of the krill output stage that is none switching....vonder how that would compare sonically..??
came to think of the krill output stage that is none switching....vonder how that would compare sonically..??
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.