"weak" is not a technical quality of a power supply. An overdamped or underdamped power supply distorts the signal.
Useless personal subjective rant: I get subjectivly "best" dynamics (is that what you call "swing"??) with a power supply where the last decoupling cell (in my case mostly LC) has a Q of 0.7. I haven't heard a ringing power supply that anyone preferred over a proper one.
Useless personal subjective rant: I get subjectivly "best" dynamics (is that what you call "swing"??) with a power supply where the last decoupling cell (in my case mostly LC) has a Q of 0.7. I haven't heard a ringing power supply that anyone preferred over a proper one.
Hi thanks this is difficult for me Honestly if i had the space i would set up a little testing bench
As i said before stressing an amp driving a robust dummy load and placing a voltmeter on the voltage rails directly on the output devices pins will tell a lot These voltages should not sag even during the musical peaks
If they sag it must mean that the output devices are starving of energy
I know it is very trivial but it is the only thing that i could possibly do i.e. read and look at a voltmeter needle
Peeping inside amps i see how amps should be layed out Big fast caps close to the outputs The pic below would be my reference as best implementation at least for a 50Watter
or this one as well
however it could be that some additional uF (i cannot do the calculation but i could measure at a bench if i had it) soldered directly on output device pins could limit the sagging
I really do not know how much Maybe 3300-4700uF ? one for each output device Maybe the offsett will vary
this could be done quite easily in some cases where there is space
Ok i take back what I said because I can't understand the theory And this is my biggest frustration
But the second, even more serious because it is fundamentally an economic and not an intellectual problem, is that I cannot find a comfortable corner to dedicate to the measurements on devices (i am seriously looking for it by the way)
There is little reason to invest in scopes signal generators without the space to carry out the testing
The measurements, if done well, ultimately confirm the theory and are much easier for me to understand
Between calculating the noise of a power supply and measuring it, it is a thousand times easier to measure it
Again my brain has limits Often when i try to understand the only result is an headache For sure this does not make me happy
However this summer i would try to do what i have said above on a little sacrifical amp i have at hand
In case the voltages sag i will try the caps mod with some 4700uF pieces
But the second, even more serious because it is fundamentally an economic and not an intellectual problem, is that I cannot find a comfortable corner to dedicate to the measurements on devices (i am seriously looking for it by the way)
There is little reason to invest in scopes signal generators without the space to carry out the testing
The measurements, if done well, ultimately confirm the theory and are much easier for me to understand
Between calculating the noise of a power supply and measuring it, it is a thousand times easier to measure it
Again my brain has limits Often when i try to understand the only result is an headache For sure this does not make me happy
However this summer i would try to do what i have said above on a little sacrifical amp i have at hand
In case the voltages sag i will try the caps mod with some 4700uF pieces
I would do it by ear, because a simple voltmeter won't show anything in the lower power range, and high power is rare and is usually noisy anyway because the speakers and room don't cooperate. In addition, the optical measured value is not a statement regarding sound or audio measurement.
There are at least two things to try with capacitors: different types and different capacitances. I would simply buy 3 different capacitors of one type - even one lower than the original - and > 5 different types of the same capacitance. I would always remove the small bridging capacitors, whether on electrolytic capacitors or diodes, first. Then you get a first access to a comparison. And here it is first and foremost about WHETHER a difference can be heard - I am addressing all the audio doubters and audio beginners;-)
Offset and bias should and will not vary.
There are at least two things to try with capacitors: different types and different capacitances. I would simply buy 3 different capacitors of one type - even one lower than the original - and > 5 different types of the same capacitance. I would always remove the small bridging capacitors, whether on electrolytic capacitors or diodes, first. Then you get a first access to a comparison. And here it is first and foremost about WHETHER a difference can be heard - I am addressing all the audio doubters and audio beginners;-)
Offset and bias should and will not vary.
Last edited:
actually i was thinking about using a 5ohm/200W dummy load that i should have somewhere
i ran 3 amps in the past through it to test by hand their current delivery on 5 ohm load
One amp heatsink got crazy hot and the dummy load cold Another one got hot and so the dummy load
The third one the one clearly better built and most expensive got warm and the dummy load very hot
It is qualitative i know but it can tell something nevertheless
absolutely you are very right It would be just a qualitative test My next buy and i will ask about this in the relevant section will be a decent signal generator with not to much distortion I am decided about this
For the ADC part i could try to use at first my sound card But i prefer to have a standalone signal generator
An audio analyzer could be justified only if this thing will get me hooked
I understand the entry price is around 800 euro ... not little money even if the unit is very worth its price
this is very interesting indeed and thank you for the precious advice
At least one magazine did a test similar to this And there was a winner The caps were judged mainly on the basis of the sound delivered by the same integrated
Years ago i bought 6 huge Siemens Sikorel said to be very very good They should be still in my cellar somewhere
I had to have them after reading how fabulous they were Never used Just admired and put away for the future
Then i tried some big Mallory and another pair of industrial grade big caps on a Rotel integrated then after i fried
In this particular last case the result was not exciting Only later i knew that big caps need time to reach their potential
Ignorance is a very bad thing I am sure they were quite good
Unison Reasearch used them They were red from Itelcond
As a note i think that class A sounds often nice because the current draw is constant at least within the limits of Class A operation
AB designs sound good if the power supply is well designed and able to cope with the changing demand of energy from the output devices
Just to add one thing The amp that got very hot with the dummy load staying cold had no bass to speak of at listening
Like it had a 100Hz high pass filter inserted I was very surprised because it had positive reviews
I am about to buy a cheap signal generator to start a more scientific approach And also to save my fingers
Like it had a 100Hz high pass filter inserted I was very surprised because it had positive reviews
I am about to buy a cheap signal generator to start a more scientific approach And also to save my fingers
The fact that the load became cold and the amplifier hot can also be due to the fact that it started to "oscillate". For example.
The fact that an amplifier audibly produces no bass is often due to the parts, and also to the construction. I have many transistors that simply don't make any bass, it's just a milky soup, or the bass is time delayed, as example. This cannot be determined by means of line-view measurement, but only by ear;-)
The fact that an amplifier audibly produces no bass is often due to the parts, and also to the construction. I have many transistors that simply don't make any bass, it's just a milky soup, or the bass is time delayed, as example. This cannot be determined by means of line-view measurement, but only by ear;-)
Hi thanks a lot I thought about that But also on a pure resistive 5 ohm load ? the funny thing is that it was a pro amp intended for a quite heavy use Maybe i did some mods first to be fair
I always wanted to take out any current limiting circuit I prefer a simple good quality fuse instead
i remember that the output devices were 150W big Toshiba Fakes maybe ?
The transformer was a 400VA thing This one here to be more precise
Well, getting to the bottom of the sound is pointless with complex push-pull amplifiers. And these are at the end of the line in terms of sound very early on anyway: the concept of half-wave separated amplification, which is practically implemented using components with different sounds. It's simply not worth investigating any bad habits - for example, a Cyrus 2 had fried my Manger drivers - but a Cyrus 2 is not an audio amplifier either;-)
And transformer power is not a statement regarding sound. The differences in cable quality alone, in the copper range alone. For example, I have transformers with aluminum winding. They sound, for example, consistently more colorless and uncontoured, "weaker". Most electrolytic capacitors sound gray. These "ceramic" power resistors in the output stage sound gray, sandy and grainy. And so on;-)
And transformer power is not a statement regarding sound. The differences in cable quality alone, in the copper range alone. For example, I have transformers with aluminum winding. They sound, for example, consistently more colorless and uncontoured, "weaker". Most electrolytic capacitors sound gray. These "ceramic" power resistors in the output stage sound gray, sandy and grainy. And so on;-)
I was thinking similar thoughts. I designed a preamp that was highly respected at all the audio publications that used half wave rectification. No complaints about poor bass. For the most part components do not define the operation with a few exceptions- X7R ceramic caps should only be used to stabilize DC supplies. Carbon resistors should only be used where their peak power capability is useful, etc. These are not audiophile fetish, they are industry accepted understanding.
That's the case even for a two-component circuit. Take an emitter follower for example. That'll add distortion. Not much, but some. So a clean tone going in doesn't come out as clean.
Or look at a resistive divider. Due to the nonlinearities of the two resistors you will have some amount of distortion. If you choose the ratio between the resistors correctly you can get the even order harmonics to cancel, but you still end up with the odd-order distortion products.
And 52 components for an opamp sounds low, actually. At least for an IC opamp that's expected to have rail-to-rail I/O and work well across a wide range of temperatures and voltages. The opamp itself might only be 10-20 devices, but will require many more parts for biasing and such.
In IC design, devices are practically free. Some devices, like capacitors, are expensive because they require a lot of area, but most others are free. So if adding another device improves performance, another device is added no questions asked. Not so in the discrete world where you have to pay for the device and then pay someone to populate it on a circuit board.
I spent the first three years of my career designing opamps for National Semiconductor. I've tinkered with some discrete circuits in the simulator as well. I even built some of them.
My general experience has always been that I could get vastly better performance on an IC than I could even in the simulator with a discrete circuit. And reality has a nasty way of being slightly worse than the simulated results, even with good simulation models. I think it comes down to the smaller devices and better ability to get high-impedance nodes on an IC. One can do wonders by shifting currents around when the leakage is low. That's just my hunch.But getting to the question about optimization: Most optimizations are multivariate ones. It's sorta like pushing on a balloon. You push in one spot and something else pops up. So the trick is to push a little in many spots. And you may have to accept that there is no global optimum. Maybe you can find a local optimum.
Every now and again I go through an exercise where I take a look to see if I can further optimize the Modulus-86. I designed the circuit some ten years ago and I do get new ideas every now and then. Most recently I explored various frequency compensation techniques. I tried 4-5 different ideas and some did result in lower distortion, which is my main goal. But it came at the expense of worse performance near clipping or it pushed the amplifier closer to instability than I was comfortable with. So I basically spent two weeks figuring out that what I already had was as close to optimal as I can get at this time.
I did find some opportunities for reducing cost, though. For example, I used both 180 pF and 220 pF in the circuit but using 180 pF for both made no impact on performance, but opens up the possibility for a lower cost as I'd get a bulk discount on many 180 pF instead of no discount when I ordered both values. Same performance, lower cost adds value.
Tom
Spot on. I find simulation to be a great tool for determining if a circuit will even have a chance in the lab.
I like low THD designs because they're a fun technical challenge both to design, implement, and measure. I also believe that what goes into an amp should come out just amplified. Not have a bunch of stuff added. So I design amps that add very, very little. Getting to that point requires a pretty strong knowledge of how devices work and how the PCB layout impacts the circuit performance. All the stuff that isn't covered by the simulator ... or at least not covered without deliberately adding in those effects.
Tom