I have been doing some noise simulations on this design trying to reduce the noise and there is is something I just dont quite understand.
Usually resistor noise goes UP when you increase the resistor value and DOWN when you reduce the resistor value.
But R8 and R5 in my current mirror does something I can not explain.
22R degeneration resistor gives me 12nV/Hz from each of those 2 resistors. Increasing them to 62R gives me 9nV/Hz, further increase to 120R gives 7nV/Hz, 240R gives 5.1nV/Hz, further increases gives slight improvements but amp distortion starts to suffer.
What gives? What am I overlooking?
Usually resistor noise goes UP when you increase the resistor value and DOWN when you reduce the resistor value.
But R8 and R5 in my current mirror does something I can not explain.
22R degeneration resistor gives me 12nV/Hz from each of those 2 resistors. Increasing them to 62R gives me 9nV/Hz, further increase to 120R gives 7nV/Hz, 240R gives 5.1nV/Hz, further increases gives slight improvements but amp distortion starts to suffer.
What gives? What am I overlooking?
Attachments
Last edited:
I asked the same question a while back!
A quick search will turn up the thread and helpful replies.
There has also been a flurry of posts recently after some reconsideration.
They are worth a search too.
Best wishes
David
A quick search will turn up the thread and helpful replies.
There has also been a flurry of posts recently after some reconsideration.
They are worth a search too.
Best wishes
David
Last edited:
the mirror degen indirectly form a V amp with the diff pair gm
a goal is to have enough V gain that the amplified diff pair input noise is the only major contribution - typcially when the V gain is >3-4 the noise contribution of the mirror Q themselves is swamped
a goal is to have enough V gain that the amplified diff pair input noise is the only major contribution - typcially when the V gain is >3-4 the noise contribution of the mirror Q themselves is swamped
I read your thread on the issue.
It is interesting that douglas self uses 100R and he states you want around 60mV across the CM degeneration resistors in his books while Bob Cordell uses 470R which in the examples he shows gives around 235mV across the CM degeneration resistors so it looks like there is no "golden rule" when it comes to sizing the resistor. Rather it depends on the rest of the design.
I have 240R in the LTP emitters and from simulation it looks like 240R is also a good value for the CM emitters. Noise contribution is more than halved, distortion is increased slightly and the total amplifier noise as simulated from the output is also reduced by a fair amount. Total noise goes from 33nV/Hz to 28nV/Hz, replacing the opa132 input buffer with another lower noise opamp, like an opa827 reduces total noise to abou 18nV/Hz which equals -109.7 dBu. I think this is more than good enough, even if this is a design meant for headphones. 2Vrms source, times 2.5 amp gain equals 5Vrms and 100mW into 250Ohm(DT880 Pro), 5Vrms is 16.2dBu, theoretical SNR if referenced to 5Vrms max output is 125.9dBu.
It is interesting that douglas self uses 100R and he states you want around 60mV across the CM degeneration resistors in his books while Bob Cordell uses 470R which in the examples he shows gives around 235mV across the CM degeneration resistors so it looks like there is no "golden rule" when it comes to sizing the resistor. Rather it depends on the rest of the design.
I have 240R in the LTP emitters and from simulation it looks like 240R is also a good value for the CM emitters. Noise contribution is more than halved, distortion is increased slightly and the total amplifier noise as simulated from the output is also reduced by a fair amount. Total noise goes from 33nV/Hz to 28nV/Hz, replacing the opa132 input buffer with another lower noise opamp, like an opa827 reduces total noise to abou 18nV/Hz which equals -109.7 dBu. I think this is more than good enough, even if this is a design meant for headphones. 2Vrms source, times 2.5 amp gain equals 5Vrms and 100mW into 250Ohm(DT880 Pro), 5Vrms is 16.2dBu, theoretical SNR if referenced to 5Vrms max output is 125.9dBu.
Last edited:
If the resistor is working into a low impedance, like a BJT emitter, then the issue is not noise voltage but noise current. This increases with smaller resistance.
An alternative way of seeing this is to think of noise voltage but not from an ideal voltage source but from a source with internal impedance equal to the resistor value. Noise voltage rises as the square root of R, but increasing R causes more voltage drop in the potential divider formed by the resistor and whatever is loading it. Hence you get noise current which varies as the inverse of square root R.
An alternative way of seeing this is to think of noise voltage but not from an ideal voltage source but from a source with internal impedance equal to the resistor value. Noise voltage rises as the square root of R, but increasing R causes more voltage drop in the potential divider formed by the resistor and whatever is loading it. Hence you get noise current which varies as the inverse of square root R.
So basically, in current mirrors, low value emitter resistors will have higher noise than high value resistors due to increased current noise because the resistor is working into a low impedance BJT emitter?
DF96 has answered your question adequitely. Further if you want to keep distortion low you should try using low Rbb transistors in which case you could use lower values of degeneration. With the transistors you have use large degeneration. You should also try stay away from wilson mirrors for use in input LTP which have much higher noise than common mirror, the small THD improvement with the wilson is in most cases not worth it, check THD differences at 500hz and 20Khz.
The Wilson mirror is nothing final, I am just trying out various ideas.
The reason for the somewhat large LTP degeneration resistors is stability, when you have an amp with very low gain, increasing the LTP degen resistors is a must, oherwise you get an amp that has way too much open loop gain as well as loop gain. This is for driving headphones and I want it to be stable into 10nF capacitive loads, which it is. Reducing the LTP degeneration resistors gives way too low phase and gain margin for my liking and makes it unstable into just a few nF. I already have 40dB loop gain at 20kHz, I do not really need more. 🙂
Last edited:
Sorry I should have stated that my comment is all related to the mirror degeneration and transistors and not the LTP degeneration or their components.
Ok, but low value mirror degeneration resistors gives higher noise than higher value resistors due to increased current noise, as per DF96.
I think we have a breakdown of communication here. 🙂
This is what DF96 posted :
"If the resistor is working into a low impedance, like a BJT emitter, then the issue is not noise voltage but noise current. This increases with smaller resistance."
Lower value resistor = Lower degeneration.
Higher value resistor = Higher degeneration.
What he said is that noise current increases with lower resistance and lower resistances gives lower degeneration.
Which is why you want to increase the degeneration resistor since this gives you more degeneration and lower noise.
Resistor goes DOWN, the degeneration goes DOWN, the noise goes UP.
Resistor goes UP, the degeneration goes UP, the noise goes DOWN.
🙂
Last edited:
I usually use high degen resistance, to get lower Gm. The sound will become less mid bass, more clear.
R = 0.26/Iq.
Goldmund use very high value degen. 75R for JFET@1mA ~ 2mA, 75R for VAS Differential (5.5 ~ 6.5mA), 330R for current mirror.
R = 0.26/Iq.
Goldmund use very high value degen. 75R for JFET@1mA ~ 2mA, 75R for VAS Differential (5.5 ~ 6.5mA), 330R for current mirror.
Last edited:
Evenin' all.
Just to confirm that as regards current-mirror degeneration it is a matter of current noise, so higher resistors mean less noise. This was pointed out a while back by Sam Groner. There is a new section on this in the 6th edition of my power amp book.
However, while increasing the resistance lowers noise the increased voltage drop can cause biasing difficulties. Furthermore, I doubt if there is much to be gained by reducing the amplifier noise. With the standard 68R/100R resistors the noise output of the amp is increased by 3dB if you put a 5534 voltage-follower in front of it, and there isn't much that's quieter than that.
I'm away from my files at present, but if memory serves there is also some new info on this in the Third edition of the book Self On Audio will be out in a couple of weeks, and it contains new & extra information on a lot of my designs.
Just to confirm that as regards current-mirror degeneration it is a matter of current noise, so higher resistors mean less noise. This was pointed out a while back by Sam Groner. There is a new section on this in the 6th edition of my power amp book.
However, while increasing the resistance lowers noise the increased voltage drop can cause biasing difficulties. Furthermore, I doubt if there is much to be gained by reducing the amplifier noise. With the standard 68R/100R resistors the noise output of the amp is increased by 3dB if you put a 5534 voltage-follower in front of it, and there isn't much that's quieter than that.
I'm away from my files at present, but if memory serves there is also some new info on this in the Third edition of the book Self On Audio will be out in a couple of weeks, and it contains new & extra information on a lot of my designs.
- Status
- Not open for further replies.