Simple Symetrical Amplifier

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
In your grounding case with 100R, my question was, will there any chance that shorting the resistor will be much better when there is no earth connection to the chassis? This is not my environment/situation of course.

Not sure about the answer to your hypothetical question - LC would probably know as he mentions this scenario above when he is in testing phase before sinks are earthed.

However, I have two almost identical amps - one with this grounding scheme and the other with the 100R shorted. There is one other variable I have to eliminate but so far the amp with the 100R sounds considerably better.

I'm pretty busy right now but in a few days I'll do some experimentation to establish exactly what is causing what. Also I will add this 100R to every element in my system - I think only then will the full benefit of this mod be realised. As you say, we have to consider earthing in terms of the whole system.

I'll report my findings when I have them

mike
 
It can be solved only with some thermal independent Vref like TL431 etc. That's why mine CCS are like sch on the pic. ;)
We always start with good resolutions. "Simple is beautiful". Then we always end with the same kind of solutions, a little bit more complicated. ;-)

I have general questions about PSUs.
Why shunt regulation is supposed to sound better. Noise ? Regulation response time ?
Why battery powered system are supposed to sound better ? PSU impedance ? Current Noise in the grounds wires between equipments ?
(I have my own partial opinion , but i'll like to know yours)
 
Last edited:
Most of the measurements are noise measurements of PSUs made "as-it", rail side. If we cannot hear noise in our equipment, i believe the problem is not mainly here .

The problem behave, on my opinion, on two sides. The first one (main on the musical perception) is how react PSUs during transients, and the second how ground currents pollute the signals. More precisely what kind of IM is produced by this pollution. This, of course, for analog.
We all know what kind of errors noise can produce in DA converters.

I believe a (so called) fast amp is less sensitive to PSU or ground problems, because it produce less TIM and less IM.
I believe we can improve things on 3 ways.
First, boards side, using two layers, one dedicated to ground connections (star or plate ?).
Second, on power supplies (low impedance, less noise, less HF currents in the ground wires): C trasfos, or shielded toroidal etc).
Third, on the wirings (Symmetrical lines etc).

The only thing i know for sure is, designing audio, we have DEFINITIVELY to forget to think "voltage" and concentrate our minds on "CURRENTS". And keep thinking each wire or track is a F... coil !

I have no idea of how experiment all that together in order to make our mind and find general "Laws" instead of a compilation of cook recipes. If we believe LC on the musical improvement brought by a good CS on only one stage, we can have an idea of the improvement that can be done if such a work is done everywhere !!!!
 
Last edited:
However, I have two almost identical amps - one with this grounding scheme and the other with the 100R shorted. There is one other variable I have to eliminate but so far the amp with the 100R sounds considerably better.

I'm pretty busy right now but in a few days I'll do some experimentation to establish exactly what is causing what.

Thanks. I mentioned about environment variations (when you find no improvement, where for example you use power conditioner, it doesn't mean that there is no improvement with different environment), also my difficulty with solving interference or coupling issues. Those have inter-relationship.

You may believe that an amp will sound good without casing, and may sound terrible when put in enclosure. Then you may not believe that it is not a simple grounding problem. Theoretically we don't want a ground loop so theoretical (but often doesn't work in practice) approach is to ground at one point. But metal casing is not a perfect inductor. The casing also functions as "ground" to shield sensitive circuits. But they also seem to function as antenna, that reflect and radiate interference or so (especially in "dirty" environments).

In old stuffs, any metal even if small are grounded (tape machine is a classic example). The casing of potentiometers must also be grounded. The classics worst case is the use of metal casing of bad quality EI transformer.

It might also not be a secret that when you open the top metal casing, the sound will be improved than when the casing was in position. Also when power supply (aka transformer) is separated in different (or far away) enclosure.

The latest find that I couldn't believe was the effect of nearby switching power supply. I didn't understand how it could manage to interfere with my system (which I believe has no grounding issue).

The fact is that ground (planes) are often used (intentionally or not) as a shield. But the quality of the shield unfortunately depends on... the quality of ground? :D
 
The only thing i know for sure is, designing audio, we have DEFINITIVELY to forget to think "voltage" and concentrate our minds on "CURRENTS". And keep thinking each wire or track is a F... coil !

Exactly, voltage is just a potential, causing potential, unrealistic problems, but currents when started to flow, they do all the messy stuff in circuits: inductions, interferences, voltage drops, heat, noise, ...
If we concentrate more to correct current "regulation", we'll always gain in quality. Similar like the rivers, who cares about the hight or the nature of their source, local current nature is all what it matters. Is it wild? Smooth? Fast? That's what interests us, when we look at certain point of the river, asking ourselves how to cross it, make a dam, ... so path where the water flows matters. Similar in electric circuits, we have to ensure fast paths to the electrons, not causing to much negative effect to the neighburing parts, circuits.
Regarding amp's PSU, low source impedance and charge [A/s] is all what it matters. Filtrations, referencing, decoupling etc, can all be done locally, suited to each part of the circuit individually. Same as the upper case of CCS, where +/- 15 V potential quality is really not critical at all. :yes:
 
Not sure about the answer to your hypothetical question - LC would probably know as he mentions this scenario above when he is in testing phase before sinks are earthed.

Of course I know, from my experiments experiences I learned that direct grounding (zero ohm) is not always good, since it means that you practically connected one plate of the "capacitor" directly to GND, causing new problems that didn't exist before. Direct capacitor to GND coupling in other words means opening the gate to all kinds of AC currents to flow where they should not. To some currents at resonance frequencies we even enables oscillations etc. So to certain metal parts it is much better to connect them to GND via resistor, thus controling maximum current level and resonance. One example is VAS cascode BJT heatsink connection to GND via 47 ohm resistor. If not, that's same as you would connect 1 nF cap from collector directly to GND. One can just imagine what that causes to VAS's high impedance to GND. :eek:
 
Of course I know, from my experiments experiences I learned that direct grounding (zero ohm) is not always good, since it means that you practically connected one plate of the "capacitor" directly to GND, causing new problems that didn't exist before. Direct capacitor to GND coupling in other words means opening the gate to all kinds of AC currents to flow where they should not. To some currents at resonance frequencies we even enables oscillations etc.

This is one of the reasons to have some impedance in the earth lines going back towards the transformer.

Doesn't have to be a choke Hugh Dean puts little resistors in all legs going back to the supply.

If it is a choke it certainly doesn't have to be anywhere near a big as the chokes I am currently using.

But some impedance in all the power lines makes a big difference to the sound.
 
Similar like the rivers
Funny, when i work with electrons, i always have some sort of hydraulic equivalent in my mind. Where caps are tanks, resistances are tubes of different diameters, voltages are pressure (or level of the water) etc... I have a representation of each active device that way.
It helps to get the feeling.
And you get a very affraying idea of the signal you are supposed to treat, from wavelets to tsunamis.
...suited to each part of the circuit individually...
Funny too, thinking to PSUs, and unable to find a satisfying solution in my mind, i just began to design a power fet at the current source of the power transistor after the decoupling cap...
Looking to the article pointed by milkem, i had this strange thought too: "Why the hell did i always, by reflex, put caps after PSU regulation transistors ?
 
Last edited:
Most of the measurements are noise measurements of PSUs made "as-it", rail side. If we cannot hear noise in our equipment, i believe the problem is not mainly here.

Hehe - I think we have almost completely opposite view on this topic. I spent last 18 months systematically going through every aspect of my amps design with the top priority of reducing noise in every area I can possibly think of and am I am very happy with the results this process brought.

But when I say noise I am absolutely not referring to mains hum - rather the big culprit for me is HF noise which amps somehow manage to demodulated into a kind of audio hash that may be very difficult to measure but can really ruin the sound of an amp. I suspect this demodulated noise raises the high order HD noise floor let's say H6 - H20 There are very many potential causes of this noise.

A classic example of increased H6 - H20 HD ruining the sound of an amp ( though not caused by HF noise ) is adding a poorly designed servo to an amp. In spice we can see the higher harmonic noise floor rise by as much as 25dB and subjectively the nice sound of the amp is suddenly lost. I'm sure anyone who has experimented with servos knows what I mean.

Also when I refer to noise I'm not just referring to hiss levels because I have noticed that amps with similar hiss levels can sound very nice or quite horrible.

I have no idea of how experiment all that together in order to make our mind and find general "Laws" instead of a compilation of cook recipes. If we believe LC on the musical improvement brought by a good CS on only one stage, we can have an idea of the improvement that can be done if such a work is done everywhere !!!!

I think if we consider the o/p of an amp, everything that is present that was not present at the i/p signal can be grouped in 3 main categories.

1) Harmonic Distortion
2) Phase distortion
3) Noise

If there are other categories I am happy to hear about them but I don't see what else there can be.

One way we could talk about this forever and never get anywhere is by not defining our terms so I propose the following

Harmonic Distortion = H2 - H5

For me regardless of it's cause Harmonic Distortion H6 - H20 is undesirable and I regard it as noise ( even though I realise in reality measured H6 - H20 is going to be a mixture of HD, resonances not related to the signal & completely random noise )

Phase distortion = A delay in time of voltage or current or combinations of both

Noise = everything else that is not directly related to the signal

This can be a chaotic mix from very many sources. Sources of noise I am aware of are and have tried to address are:

mains born noise ( Caused by many things including poorly designed power supplies ! )
noise from switching diodes or transistors
HF ringing in secondaries caused by combination of inductance & switching diodes
noise on the domestic earth lead ( Also caused by many things including poorly designed power supplies ! )
noise from micro resonances caused by combination lead / track inductance & capacitance and excited by the signal - often caused by bypass caps.
noise from micro resonance in mosfet input leads
noise from micro resonances caused by combination lead inductance & capacitance and excited by other noise !
noise from slight instabilities caused by poor feedback design
noise from semiconductors.
noise from industry standard resistors.
noise from electrolytic capacitors.
RF noise picked up by speaker leads & casing and injected into the amp via the o/p or the earth line.

there may be other noise sources I have missed !


Given that I am not that concerned about low order HD all that is left according to my definitions above is phase distortion & noise.

I think phase distortion is important that why I like fast CFB amps like SSA that can produce beautifully clear transients

But even with the most wonderful amplifier topology in the world - which some version of SSA may well be - if the above issues are not addressed correctly it can still sound very bad and if they are addressed correctly it can sound very nice indeed

You guys may think I've got noise on the brain and you might be right but having just spent 18 months systematically working on all of the above issues I can testify to what a huge difference it can make when ALL of these issue are addressed.

Anyway - that's my rant for tonight - I hope some of it makes sense !

cheers

mike

p.s. I forgot noise from poor connections :)
 
Last edited:
Hehe - I think we have almost completely opposite view on this topic.
Like L.C:i will argue to morrow. But i'm sure we will agree together at the end with a nice cup of tea.. Once we will tune together our definitions of Harmonic distortion and Inter-Modulation distortions.

Just little words about. Let suppose there is no intermodulation distortions ? Who care about high order harmonic distortions that appears outside of our earing frequency window ? We do not ear them. It is the intermodulation product between those upper harmonics distortions and the instruments in the range -they are IN the window, - that we listen and hate.
Harmonic distortion themselves are not so disagreeable. They just change more or less the tonal texture of instruments. Loudspeakers produce lot of them, and we can enjoy them. Led Zepplin produce lot of them, i love Led Zepplin :)
Lot of audiophile loves old analog records. They are full of noise (tape hiss, vinyl cracks;) it do not bother.
Intermodulation (both static and dynamic) product artefacts, confuse the message and the separation between instruments and paint everything with mud.

If you love SSA sound, it is not because it produce little harmonic distortion (some tune their sample to add some), it is because it is fast, and so, produce very little IM. and TIM.
 
Who care about high order harmonic distortions that appears outside of our earing frequency window ?

I only complain about H6 - H20 for that very reason - I include up to H20 for our younger friends for me let's say H6 - H15 ;)

It is the intermodulation product between those upper harmonics distortions and the instruments in the range -they are IN the window, - that we listen and hate . . . Intermodulation (both static and dynamic) product artefacts, confuse the message and the separation between instruments and paint everything with mud.

Agreed - I do hate that sound - could you define static & dynamic IMD please

Harmonic distortion themselves are not so disagreeable.

Agreed

Lot of audiophile loves old analog records. They are full of noise (tape hiss, vinyl cracks;) it do not bother.

mmmm - not so fond of those things

If you love SSA sound, it is not because it produce little harmonic distortion (some tune their sample to add some), it is because it is fast, and so, produce very little IM. and TIM.

Let's say I love CFB sound if implemented well - didn't get to symmetrical design yet. My development has been with SE CFB design but never the less I absolutely agree about ( low order ) HD, I stated that already.

So I think we agree completely about HD ( not so bad ) & IMD ( very bad ) and I think can put those things to one side ?

But . . . my finished design isn't significantly faster that the design I started with 18 months ago - in fact I slowed it down a little bit by increasing FB impedance from 1000R / 50R to 2200R / 100R ( this gave less ringing on square waves with a capacitive load and sounded better )

But my finished design sounds much better now then my initial design did with the basic circuit topology hardly changed. The main difference is that I worked on the different types of noise that I list in my post above - so according to my definition noise is very important indeed.

I think there are three things we need to clarify

1) Do we believe / accept the HF noise ( let's say 20Khz - 20Mhz ) from whatever source, somehow get's demodulated by the amp and ends up as grundge in the audible H6 - H20 noise floor ?

2) Do we need to revise my definitions to make sure we're arguing about the same thing.

3) Can we define static & dynamic IMD

cheers

mike
 
Last edited:
Well, before going for the standard SSA CCS made of SSM, TL and jFETs I was thinking of a last take on a CCS commonly used in many amplifiers. I wonder why I didn't think of it before as I have many many times used it in my previous projects with good results(although not in a similar place as it is now).

I have simulated the CCS with the SSA and this one really shows some hope. I think everybody here understands this CCS more than me and so I am not gonna "describe" it. I will use trimmers there to control the current and measure it around the 100ohm load resistors.

It looks like this CCS is fairly tolerant of temperature variations. Heating the BD bjts to different temperatures seems to affect the offset to +-10mV at most. Also the current looks super stable with signal at 7uA CCS current per 10mA feedback current(1:1400) or per 1000mV of feedback voltage(1:140000), all AC. This is only what I see through the monitor but don't know how to derive the dB value of the CCS variation current. Please educate me. I attached some traces. I am going to build it and will update once tested. I hope this one will be less noisier than the Wilson.
 

Attachments

  • SimplerCCS.png
    SimplerCCS.png
    11.2 KB · Views: 441
  • SimpleCCS.PNG
    SimpleCCS.PNG
    86.6 KB · Views: 431
This will do...

Test done. Results very positive.

1. Very stable current supply, offset drift from cold to warm is less than 15mV.

2. Almost zero noise from speaker(input short to ground).

3. Turn on thump gone(only a little cone excursion). Also both + and - PS capacitors drain equally at turn off.

4. I was worrying as the CCS doesn't use the ground potential, what if it doesn't start. But it did. Something new for me, hasn't played with symmetrical CCS before.

5. Music come alive. Mission accomplished! Clean and Clear. Going nod-nod and tap-tap... WOOHOO!
 

Attachments

  • DSC07651.JPG
    DSC07651.JPG
    129 KB · Views: 395
  • DSC07652.JPG
    DSC07652.JPG
    141.5 KB · Views: 357
  • DSC07653.JPG
    DSC07653.JPG
    110.9 KB · Views: 344
  • DSC07654.JPG
    DSC07654.JPG
    140.7 KB · Views: 155
  • DSC07658.JPG
    DSC07658.JPG
    160.9 KB · Views: 181
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.