You shouldnt use wilson current mirrors for input LTPs, they have two flaws and not much is gained compared to enhanced mirror. Wilson mirrors are about the noisiest mirror there is, if you going to use this then use even higher degeneration. The standard mirror rules here. Second flaw is that they are the most prone to peaking at high frequencies. All this to get 0.01 percent better accuracy than the enhanced ???
astx's tests show that the current mirror might be improved. I don't know if the dominant distortion is noise. If it is, then you're right. If not, then the Wilson mirror may be the way to go. Why be paranoid about noise if we don't know whether it is significant yet?
As far as performance gain, I am just following astx. He seems to think the previously demonstrated increases in performance are significant.
While designing an amplifier, it is important to think logarithmically rather than fractionally when improving distortion. However when you know you are not going to lower distortion much more, it is reasonable to look to fractional gains in performance.
As far as performance gain, I am just following astx. He seems to think the previously demonstrated increases in performance are significant.
While designing an amplifier, it is important to think logarithmically rather than fractionally when improving distortion. However when you know you are not going to lower distortion much more, it is reasonable to look to fractional gains in performance.
Last edited:
Sam Groner deals with this in detail in his critique of Self's book comments-audio-power-amplifier-design-handbook-douglas-self
Loadsa other good solid stuff on his website.
______________________
As the heretic among yus guys, when sim THD approaches 1ppm, I'm more likely to see if I can simplify the circuit further and retain the level of performance.
Once you get below about 0.01% THD, layout, decoupling, grounding & other stuff take on equal, if not MORE, importance. If I can remove another device from the forward path, I will have more stability in hand.
astx, what we need to know is whether increasing CM degeneration worsens or improves THD at this point. If it improves THD (assuming negligible noise), then the Wilson mirror is the way to go. If it worsens THD, a buffered CM should be best.
Wow, so many different factors affecting the use and effectiveness of current mirrors. I think I am beginning to get a handle on this.
Wow, so many different factors affecting the use and effectiveness of current mirrors. I think I am beginning to get a handle on this.
Dear keantoken, kgrlee and manso,
still waiting for distortion magnifier to be able to get more measurement headroom...
Maybe upcoming weekend we know more about the CM variants and it's effects for this amplifier topology.
Btw: noise simulation shows about 124nV/rthz with current gain settings at output. Don't know if this is good, bad or something else...
How to measure this quickly in real life to compare with simulation?
Thx, Toni
Here is a set of commands for outputting noise in a more scope-friendly format. I did not come up with this and I don't remember who did, but it is probably standard procedure.
.step dec param x 11 10meg 10
.noise V(Vop) VsineA oct 100 10 {x}
.measure ns INTEG V(onoise)
.measure snr param -20*log10(ns)
You will need to go to change the .noise output node and input source to match your circuit. After simulation finishes, go to the error log, Rclick->plot. Then Rclick->add trace and click SNR and ns, and okay. I believe the ns value is the RMS output noise at the BW of the frequency, and the SNR value is in decibels.
.step dec param x 11 10meg 10
.noise V(Vop) VsineA oct 100 10 {x}
.measure ns INTEG V(onoise)
.measure snr param -20*log10(ns)
You will need to go to change the .noise output node and input source to match your circuit. After simulation finishes, go to the error log, Rclick->plot. Then Rclick->add trace and click SNR and ns, and okay. I believe the ns value is the RMS output noise at the BW of the frequency, and the SNR value is in decibels.
Attachments
= 4nV/rtHz input referred. This is usually considered excellent.
Still not quite silent if bi-amped with compression drivers, damn.
Best wishes
David
Thanks a lot. Just tried it. Great - works out of the box. Simulated SNR seems to be realistic. Will make some real life measurements to compare with simulation.
BR, Toni
That is curious, the models are fairly similar.
Could you post the exact sim file that shows this difference?
I don't even care much about overload
But I share your enthusiasm for simplicity, once the distortion is undectable of course.
Best wishes
David.
Well, I have not built my really quiet amp yet. 4nV/rtHz probably inaudible at the sweet spot but the horn mouth is at ear level and I don't want to hear any hiss as I walk past. Less than 2nV/rtHz would be desirable and seems possible with extreme care.
Best wishes
David
I first saw this the model Toni's posted in #274. (see #333) But the huge difference in overload is obvious in any of his .ASCs
All you need to do is to run .TRANS with more than 3Vp @ 20kHz so no fancy THD meters or processing required.
Got any ideas on how to fudge the models?
I posted a query at finding-spice-models but only Harry Dymond replied
The other SPICE model gurus have left the planet.__________________
I share Bob Cordell's view that overload behaviour is one of the biggest audible differences between amplifiers.I don't even care much about overload
I lived with an amp about the same size as Toni's for a long time and confirm even something this size (with 90dB/W @ 1m speakers) overloads much more often then you'd imagine.
In fact, even a 1000W/channel amp will easily overload if you try to emulate a well played piano in your living room. This is 'real life' BTW. We used such a beast for our Blind Listening Tests on speakers for nearly 2 decades
_______________
My view of required amp performance is coloured by doing Double Blind Listening Tests bla bla for a long time. The quest for 1ppm THD20k is a fun exercise for me but for a 'real life' product, I'd sacrifice this without hesitation for some of the criteria which DBLT bla bla show to be audible to the man in the street.
'Overload' in its many varied forms on 'real life' signals, is among these.
Last edited:
On increasing the VAS resistor R22 to about 27R the overload behaviour betters significant - but also the THD especially at higher output levels is increasing.
So one can decide: better overload behavior or better THD at high power listening levels.
Btw: I am driving my speakers with 4 amplifier channels (bi-amping with active crossover) so a short overload spike at low frequency speaker would not be hearable at mid and high range speakers.
But a single amplifier which must play the complete audio range produces hearable distortion also before viewable clipping of the LF wave as mid- and high range signals are modulated on top of this LF-wave!
So one can decide: better overload behavior or better THD at high power listening levels.
Btw: I am driving my speakers with 4 amplifier channels (bi-amping with active crossover) so a short overload spike at low frequency speaker would not be hearable at mid and high range speakers.
But a single amplifier which must play the complete audio range produces hearable distortion also before viewable clipping of the LF wave as mid- and high range signals are modulated on top of this LF-wave!
Only Lot's of RAW Power for Life-Like ... Counts
@ kgrlee
Cannot more than 100% AGREE with You on this
I use very SIMILAR little monoblocks, & for the same amount of time as You, and only those little ones CAN do easily . . . .
all the rest little smokers are tired in less than 5 - 15 minutes . . . . at the end of the day only pure RAW Power counts,
almost all the rest these days are stories for the BrainWashed masses with the expansive Devil Code . . .
BRGDs
Andreas
In fact, even a 1000W/channel amp will easily overload if you try to emulate a well played piano in your living room. This is 'real life' BTW. We used such a beast for our Blind Listening Tests on speakers for nearly 2 decades
_______________
@ kgrlee
Cannot more than 100% AGREE with You on this
I use very SIMILAR little monoblocks, & for the same amount of time as You, and only those little ones CAN do easily . . . .
all the rest little smokers are tired in less than 5 - 15 minutes . . . . at the end of the day only pure RAW Power counts,
almost all the rest these days are stories for the BrainWashed masses with the expansive Devil Code . . .
BRGDs
Andreas
There are a few oddities in the models so, yes I have a little idea to work on.
Actually, I think that is all very sensible and true, even better that you have actual honest tests to back it up.
Just that my new amp is for 115 db sensitive compression drivers.
So a 100W amp is equivalent to about 30kW on your 90 dB example.
So I don't care much about overload
Best wishes
David
Last edited:
The wilson current mirror s output impedance is not much higher than a heavily degenerated common mirror (THD improvement will be minimal and likely to be higher when high frequencies like 20Khz are encountered), it is however much noisier. If one uses the wilson, high degeneration should be used to lower the noise; since degeneration doesnt effect the output impedance here it will not affect THD (probably a little worse off at higher frequencies. Also while the wilson seems to be better than the EF mirror it will probably have worse THD at high frequencies, higher than even the common mirror. This is all assuming that the transistors are matched otherwise other factors come into play.
A degenerated EF mirror offers about the best compromise between noise and lowest THD, that is 20Hz-20Khz.
You are aware that the lowest noise achievable is dependent on the input impedance, so if youre looking for the lowest noise youll have to custom design your input stage for your preamp or whatever source you using. The design will not necessarily be low noise with a different source.