???
A comp change the CF of a stage, not his slope. Adding a stage will add a 6dB/oct supplementary filter. Most of the time, they are all near 10 MHz.
1. See keantokens reply above - that's part of the reason.
2. Look up AFEC - that's the other part
I usually separate the front end and TIS from the main power section with a simple RC filter. On my e-Amp,I used cap multipliers though.
The reduction in THD is cos the amp is unstable with very specific loads.
I'm not advocating expensive caps. I'm advocating cheapo small electrolytics.
If you leave them out, you will have problems .. especially if you put in Golden Pinnae films. If you have my cheapo electrolytics at the right places, you can add any number of Golden Pinnae caps without evil effect. 😀
_____________
PSR is a different important animal. The amount of cheapo electrolytic needed to make the instability go away is very small .. too small to make a difference to PSR.
Even a physically small 1u electrolytic makes a huge difference to stability in the right place. These days, physically small 1mF can be fitted in.
Last edited:
The film's Q is very high, so a small series R works.... but so would a cheap-o electrolytic with relatively high esr. Either needs to be placed at the OPS devices. Most people prefer - via listening - the film with series R. It is also more stable itself.... as many parameters of the cheap electro vary with temp quite a lot.... such as, esr and DF and C. So, I wouldnt go too low on the C value and use the not AS cheap 105C versions. Cheap films are also suitable (with R).
THx-RNMarsh
THx-RNMarsh
Last edited:
I like Keantokens idea of injecting a signal into andunpowered circuit, sweeping and looking for resonances. a network analyzer would be great, biut most do not have that - but a scope and sig gen most people have.
Anyone else got any comments on this?
Anyone else got any comments on this?
I like this idea too. Keantoken mentioned it a while back in one of my threads. Like to use an additional RC snubber when paralleling films and electrolytics. It gives an additional degree of freedom when trying to damp the power rails.
Probably over kill but this is DIY.
You could sweep a frequency into a power rail but wouldn't a simple square wave do the same job? Looking for peaking/ringing.
I also like the idea. Any thoughts about the power required for such an injection? Probably some power amp between the generator and the tested circuit would make sense?
The problem with this is that this tells you NOTHING about how the rails interact with the power amp driving wonky loads.
If adding resistors and film caps to damp stuff you see on this test moves the important cheapo electrolytics further away from their important positions, you might get worse results in the final amp.
And where & how are you injecting your signal? Don't forget an important part of the problem is track inductance. So your exact injection point is critical.
_______________
BTW, Bonsai's
usually introduces sufficient damping (cos the series resistor) to make things OK.
Last edited:
I think these may be separate mechanisms Richard.
One would be about resonances in the rails excited by load induced transients, while the second would simply be the power supply performance when driving heavy loads.
I think keantoken is refering to the fitst case.
Of course, EF3 instability, cascode instability etc add more worms . . .
One would be about resonances in the rails excited by load induced transients, while the second would simply be the power supply performance when driving heavy loads.
I think keantoken is refering to the fitst case.
Of course, EF3 instability, cascode instability etc add more worms . . .
On a stereo amp you could:-
Bridge the two amp outputs with a load (8 ohm, reactive etc)
Feed no input signal into amp A (lets call it the 'ground amp'), and then drive amplifer b (lets call it the 'signal amp')
Then you can monitor the rails of the ground amp and/or inject signals into the rails to look for problems - via a suitable sized capacitor of course.
However, its very easy to get anal about PSRR . . . .
I dont think you can differentiate below about 0.1% of anything - PSRR and/or distortion. The nx-amp and sx-amps use simple rail filtering to get rid of HF hash (there are some pics in the build document) and the PSRR at LF is certianly not better than about 50 dB, whereas a VFA can easily be 90 dB or more. Yet, there is ample evidence that listeners are not rejecting low PSRR amplifers (any/all CFA without regulated PSU's or cap multipliers, almost all of NP's designs . . .) in favour of designs that have high PSRR.
This does not mean we can design sloppy power suppliers, layout etc, but have to recognize that the law of diminishing returns applies. Of course, I will try to design amplifers with 150 dB PSRR, 1 ppm distortion at 420 W (sims of course) simpley as an intellectual exercise, but I am not going to cripple practical designs with encumbrances that bring no aural benefit whatsoever.
🙂
Bridge the two amp outputs with a load (8 ohm, reactive etc)
Feed no input signal into amp A (lets call it the 'ground amp'), and then drive amplifer b (lets call it the 'signal amp')
Then you can monitor the rails of the ground amp and/or inject signals into the rails to look for problems - via a suitable sized capacitor of course.
However, its very easy to get anal about PSRR . . . .
I dont think you can differentiate below about 0.1% of anything - PSRR and/or distortion. The nx-amp and sx-amps use simple rail filtering to get rid of HF hash (there are some pics in the build document) and the PSRR at LF is certianly not better than about 50 dB, whereas a VFA can easily be 90 dB or more. Yet, there is ample evidence that listeners are not rejecting low PSRR amplifers (any/all CFA without regulated PSU's or cap multipliers, almost all of NP's designs . . .) in favour of designs that have high PSRR.
This does not mean we can design sloppy power suppliers, layout etc, but have to recognize that the law of diminishing returns applies. Of course, I will try to design amplifers with 150 dB PSRR, 1 ppm distortion at 420 W (sims of course) simpley as an intellectual exercise, but I am not going to cripple practical designs with encumbrances that bring no aural benefit whatsoever.
🙂
Last edited:
You guys should read Bruno Putzeys' article in Linear Audio Vol 5 (and build his preamp - the book comes with a free PCB!): 'The G-word - how to get your audio off the ground'.
Star grounding is a bandaid, and not a good one at that.
Jan
Here is a good article by Bruno explaining the basic principles (publicly available at Hypex web site):
Dealing with legacy pin 1 problems
Dealing with legacy pin 1 problems
They are indeed separate issues.
But the one which both kean & your good self have missed is to do with stability.
My contention is that stability is the important issue. It can't be investigated without the amp working and driving loads including wonky ones.
But in any case, the cure is simple. Have cheapo electrolytics near the VAS.
If you have your usual RC filtering of the VAS & earlier rails physically near the VAS, you are probably OK.
________________________
Mr. Marsh, I second your recommendation to use good 105 electrolytics vis a vis my caveat about electrolytics near hot outputs.
Last edited:
Jan, what the hell about grounding ?
Ground is a reference point. You know that well. Each time you get a wire or track with current in it, you will have a voltage between their two ends because this connection is not 0 Ohms. Star ground is only a way to get no common wire to ground between two circuits.
An other solution, with printed circuit, is to use a ground layer, to make ground connections not far from 0 impedance. But simple wiring is possible, when you don't mix a high current reference with a low one, everybody is aware about this.
Between different devices, other evils are carried by the common grounds, like AC leakages etc..
Of course, symmetrical cabling is a solution, by the suppression of the ground reference to transmit signals, used in professional studios because there is too much devices witch can be connected together... or not ;-)
*There is no universal solution*, and again things will never be perfect. The purpose is to keep evils far away enough. Thinking about *currents*, considering wires and tracks for what they are, and find the most simple solution in the circumstance is the only way.
I always treat loop stability and parasitic/ PSU induces instability as separate issues.
Certainly parasitic oscillation can trigger loop stability problems and vice versa. But, if one comps an amp from the theoretical perspective (sensible ULGF, no SID issues and so forth) and you still get problems, then of course the designer would be expected to go looking elsewhere.
I am not ware of decoupling issues affecting loop stability, other than in poorly designed PSUs or bad layout.
Certainly parasitic oscillation can trigger loop stability problems and vice versa. But, if one comps an amp from the theoretical perspective (sensible ULGF, no SID issues and so forth) and you still get problems, then of course the designer would be expected to go looking elsewhere.
I am not ware of decoupling issues affecting loop stability, other than in poorly designed PSUs or bad layout.
I don't think star grounding is a bandaid. It's been the defacto way to do things in electrical engineering for about 100 years.
It's just there are other techniques now, like Bruno's for example, that allow one to take noise reduction to new levels at REASONABLE COST
It's just there are other techniques now, like Bruno's for example, that allow one to take noise reduction to new levels at REASONABLE COST
I use one el-cap at each device, separate tracks to common inlet, and common gnd star point. I believe it's important to make a solid circle around the speaker drive, the caps are then only loop connected through the devices and the emitter resistors, and they share only the lowest impedance points, this way chances of psu resonances are diminished