| akunec |
Is it true that using more capacitance on the supply rails can negatively affect high frequency performance? I find it really hard to believe but I heard it a few times on this forum. Can someone offer a logical explanation?
Thanks
AK |
|
|
| I_Forgot |
There are some who think that small capacitance on the power supply rails results in best sound from the chip amps. Keep reading these forums and you will see many, even stranger things...
Large electrolytic capacitors (thousands of microfarads) have relatively low Q at high frequencies. This means the ESR is relatively high compared to the capacitive reactance. The normal procedure is to connect small capacitor(s) (a few 10s of microfarads, and even 1 or 2 uF) with high Q in parallel with the large capacitor with low Q. This makes the overall Q higher and allows the power supply filter to perform better at high frequencies.
Using only small capacitors in the power supply filter/storage will allow the voltage across the amplifier to drop during loud signals (typically low frequencies) that require maximum output from the amplifier. The reduced voltage powering the amp will limit the power available to the load (your speaker). This will result in clipping or other behavior which ultimately results in distortion.
Some people seem to like that distortion and use only small capacitors in the power supply.
I_F |
|
|
| Upupa Epops |
| " ... some people like this distortion ... " - many people can't live without distortion :D ( sometimes they are very well known :D ). |
|
|
| jackinnj |
| quote: | Originally posted by akunec
Is it true that using more capacitance on the supply rails can negatively affect high frequency performance? I find it really hard to believe but I heard it a few times on this forum. Can someone offer a logical explanation?
Thanks
AK |
i think that someone pulled that thesis out of thin air -- which seems to be standard operating procedure (SOP) for some DIYrs'
the chipamps are generally run at high gain (A=20) and are quite sucsceptible to small hum (60, 120, 180 240Hz at US line frequencies) making its way into the circuitry -- this shows up as increased distortion at the fundamental and harmonics of the line frequency -- 10,000uF as recommended by National Semi is fine -- careful attention to layout is also important.
i can demonstrate the way in which ground loops, inadvertent power cable placement etc. measurable affect distortion. |
|
|
| analog_sa |
| quote: | | Is it true that using more capacitance on the supply rails can negatively affect high frequency performance? I find it really hard to believe but I heard it |
The effect is only one of perception and certainly not measurable. Not because it does not exist, but because we haven't a faintest idea how subjective perception works and what needs to be measured.
Of course, all the subjective benefits of better midrange and high frequency clarity with low capacitance are only possible with sympathetic speakers. I used to be happy with 1000uF driving Lowthers but none of my current speakers are listenable at all with such PS. Clarity? Yes, but with an unhealthy dose of anaemia, especially in the bass.
Has not this horse been beaten to death already? There are no simple explanations and practically no measurements to support half of what we hear. Is this a huge problem? |
|
|
| GregGC |
Last night I was playing with the GC (lm3875 NITGC).
First changed the PS from dual (220VA) transformer and 2 Bridges per channel to Single transformer (220VA center tap) and a single Bridge (400V/40A). I didn't hear any difference, so I left it that way. Now I have a transformer and extra bridge for another one. Then I decided to double the 1000uF PS caps. The base increased substantially and subjectively the mids and highs became more recessed. It changed the whole picture. It has more base but it doesn't sound as lively. I know that it's because of lack of low frequencies. Because the PS caps are in ser of the speaker I decided to calculate the F-3dB for 1000uf and 4 Ohm load (my speakers are rated at 6 Ohms). Well it's 40 Hz. No wonder it has less base with 1000uF. I think that the same effect could be achieved by decreasing the input cap and keeping the 2000uF in the PS. So, to make sure 20 Hz are not attenuated at all by the GC you need to have at least 2000uF per rail in the PS, just because all speakers dip below 8 ohms in the LF area.
I hope I haven't missed something.
/Greg |
|
|
| thomas997 |
| quote: | Originally posted by Upupa Epops
" ... some people like this distortion ... " - many people can't live without distortion :D ( sometimes they are very well known :D ). |
nice :)
| quote: | Originally posted by akunec
I find it really hard to believe but I heard it a few times on this forum. |
You will find quite a bit of stuff that is hard to believe on this forum.. not necessarily a bad thing, it does make you think and discuss :scratch2: (this picture is the best). |
|
|
| janneman |
| quote: | Originally posted by GregGC
[snip] The base increased substantially and subjectively the mids and highs became more recessed. It changed the whole picture. It has more base but it doesn't sound as lively. [snip]/Greg |
Greg,
This is quite normal. Whenever you get more bass, it is perceived as if there is less mid/high. It is all relative. Conversely, if you change to a more efficient tweeter, you perceive as if you have less bass.
This is perception-related, well documented, no secret.
So, if you decrease you ps capacitance, you may perceive relatively higher levels of mid/high because the bass level relative to the mid/high falls. But saying that less ps cap improves mid/high is just deceiving yourself. You get the same effect with a resistor in series with your woofer, or by using a cheaper woofer, and I don't think you would call that an improvement (unless you have unbalanced speakers to begin with).
In fact, playing with the ps cap in this way seems to me like a rather awkward way of re-introducing tone controls...;)
Jan Didden |
|
|
| janneman |
| quote: | Originally posted by analog_sa
The effect is only one of perception and certainly not measurable. Not because it does not exist, but because we haven't a faintest idea how subjective perception works and what needs to be measured.[snip] |
... but we DO know how it works in these circumstances. See my previous post. Not all people do know, surely, but a little research will show that the knowledge, experimentattion and etc IS available.
Jan Didden |
|
|
| GregGC |
| quote: | Originally posted by janneman
Greg,
This is quite normal. Whenever you get more bass, it is perceived as if there is less mid/high. It is all relative. Conversely, if you change to a more efficient tweeter, you perceive as if you have less bass.
This is perception-related, well documented, no secret.
So, if you decrease you ps capacitance, you may perceive relatively higher levels of mid/high because the bass level relative to the mid/high falls. But saying that less ps cap improves mid/high is just deceiving yourself. You get the same effect with a resistor in series with your woofer, or by using a cheaper woofer, and I don't think you would call that an improvement (unless you have unbalanced speakers to begin with).
In fact, playing with the ps cap in this way seems to me like a rather awkward way of re-introducing tone controls...;)
Jan Didden |
| quote: | Originally posted by janneman
... but we DO know how it works in these circumstances. See my previous post. Not all people do know, surely, but a little research will show that the knowledge, experimentattion and etc IS available.
Jan Didden |
Thanks Jan.
That's exactly how I see it too.
/Greg |
|
|
| janneman |
| quote: | Originally posted by I_Forgot
[snip]Using only small capacitors in the power supply filter/storage will allow the voltage across the amplifier to drop during loud signals (typically low frequencies) that require maximum output from the amplifier. The reduced voltage powering the amp will limit the power available to the load (your speaker). This will result in clipping or other behavior which ultimately results in distortion.[snip]I_F |
Not only that, but because there is riple on the supply (quite a lot with low cap values), with high output levels the amp output voltage will be modulated by double the mains frequency (assuming two-phase rectification), which IS audible and measurable. Again, all well established, no secret. Now, if your speaker lacks robust bass, this may actually give the perception of more bass. Something like turning your amp into a perceptual coder and saving on caps at the same time...;) . What was it again that the "fi" in "hi-fi" stood for??
Jan Didden |
|
|
| GregGC |
Talking of lowering PS voltage ripple:
Forgot to mention that by doubling the PS cap size the minuscule amount of hum I had in the speakers (audible only if your ear is in front of the driver) disappeared totally too.
/Greg |
|
|
| analog_sa |
| quote: | | Whenever you get more bass, it is perceived as if there is less mid/high. |
Maybe true but only audio-virgins wouldn't recognise the difference. You might try reducing a coupling cap and the sound just gets thinner and thinner without any surprising increase in clarity. |
|
|
| Joseph K |
If it would be true that the power supply ripple will modulate the output, then this would show up in the simple distorsion measurements. Why it is that I did not see anyhting of it, with 2200 uF/rail, 5 ohm? What I have seen was diminishingly low distorsion right up to the clipping. Also, I would challenge anybody here, to demonstrate the direct bass - enhancing effect of a bigger cap. Also that should have shown up in my measurements, but I have only seen the high pass effect of the input cap, with the 20 kohm input resistance, and it was ~ 3-5 Hz, now I don't remeber that well.
If we increase the power supply impedance, we only trigger the amp to correct it's own dynamic impedance. Until it runs out of headroom, it will do it. So, with the lower power supply caps, what is really lost is the dynamic range.
george |
|
|
| GregGC |
| quote: | Originally posted by analog_sa
Maybe true but only audio-virgins wouldn't recognise the difference. You might try reducing a coupling cap and the sound just gets thinner and thinner without any surprising increase in clarity. |
It may have something to do with also introducing harmonics that change the overall picture. Reducing the PS caps is only part of reducing the LF. The transformer is in par of the caps, so it's not as strait forward.
Greg |
|
|
| SY |
| George, it might be interesting to look at clipping and recovery at, say, 20Hz. |
|
|
| GregGC |
George,
2000uf/rail is fine. Do the same measurements with 1000uf/rail. It made a big dif. in my setup going from 1000 to 2000uf/rail. I don't know if adding more capacitance to the PS would make any dif. in my GC. Have to try that.
/Greg |
|
|
| carlosfm |
| quote: | Originally posted by Joseph K
So, with the lower power supply caps, what is really lost is the dynamic range.
george |
Certainly.
Dynamic range can't be the same.
In my case, with my Epos 11 speakers and 1000uf capacitance I didn't have lack of bass at all.
What I did have was untight bass, undefined, it was like listening to a bad cd player with a strong bass recording.
Untight, unprecise, slow bass always kills midband and treble and this was no exception.
Although very detailed with simple acoustic music, as soon as more instruments came in some sounds got masked behind the mix, unperceptible.
Summarizing: the amp didn't drive the speakers properly.
The results with the low capacitance PSU are very different from speaker to speaker.
With high sensitive (and easy load) speakers the results may naturally be quite different.
But they are never correct.
Welcome back, George.:cool: |
|
|
| GregGC |
On my speakers with 1000uf/rail GC sounded very clear and dynamic/fast. After doubling the PS caps it's like there is a curtain of base in front of the other instruments. In my case it was more like turning up the LF. My speakers are 6 Ohms 90dB/m/W though.
/Greg |
|
|
| I_Forgot |
| quote: | Originally posted by carlosfm
The results with the low capacitance PSU are very different from speaker to speaker. |
The ultimate performance of the amp chip is defined by the designers who assume and use stable, well bypassed, DC supplies. All bets are off if you don't give the chip a stable, well bypassed DC supply. Small power supply capacitance leads to high supply impedance, which leads to high output impedance. The results are unpredictable.
Would you put diesel fuel in your car's gasoline engine, then complain that the performance sucks?
I_F |
|
|
| GregGC |
| quote: | Originally posted by I_Forgot
The ultimate performance of the amp chip is defined by the designers who assume and use stable, well bypassed, DC supplies. All bets are off if you don't give the chip a stable, well bypassed DC supply. Small power supply capacitance leads to high supply impedance, which leads to high output impedance. The results are unpredictable.
Would you put diesel fuel in your car's gasoline engine, then complain that the performance sucks?
I_F |
Some speakers, depending on how they are designed may have increase of the LF (relative to the Mid and HF) because with increasing of the Rout of the amp the Q of the speaker increases and depending on the volume of the box you may get (relative to the mids and highs) more base. Speakers design for tube amps may benefit from this kind of high Rout amps. In my case the effect was the opposite.
/Greg |
|
|
| Joseph K |
Sy, I have to admit, I don't know [did not think about ] about the effect of power supply impedance on the recovery mechanisms. And yes, at 20 Hz, but even above, you will see the [serious] difference in the clipping levels. This is what I told, as well. The recovery from clipping is quite rude whit this chip, anyway.
What I wanted to emphasise, is that it is NOT a tone control. At least not in a simply measurable way.
And Greg, You are right, there is a difference between 2200 uF /1000 uF. Although less difference, if you take into account the BrianGt kit value [1500 uF]. But, in that measurement, I already had something like >10 V power supply ripple at max output & still very low distortion [in fact, it is the least just before clipping]. Changing the cap to 1000 uF will only cause further increase in ripple, and that would only, say, double the 0.007 % distortion? I think it's not what Jan meant?
What I wanted to say there, is that I also thought before this test, that, at least, some 50Hz induced spuries will be seen.. but not. Nothing. [Naturally, this happened only after that i cleaned up all ground loops, & resolved all shielding problems] In the same time we were looking at a push-pull EL34 tube amp, which hummed like crazy, intermodulated all the test signals with all the power supply spuries, [wanted to demonstrate Jan's claims] had shown a bit of ringing, severe bandwidth limiting, slew rate problems, and huge amounts of harmonic distortion [which did the intermodulating job with the power supply hum..]
This tube amp in the same time had blown off piste my GC, in the listening tests..
George |
|
|
| GregGC |
| quote: | Originally posted by Joseph K
Sy, I have to admit, I don't know [did not think about ] about the effect of power supply impedance on the recovery mechanisms. And yes, at 20 Hz, but even above, you will see the [serious] difference in the clipping levels. This is what I told, as well. The recovery from clipping is quite rude whit this chip, anyway.
What I wanted to emphasise, is that it is NOT a tone control. At least not in a simply measurable way.
And Greg, You are right, there is a difference between 2200 uF /1000 uF. Although less difference, if you take into account the BrianGt kit value [1500 uF]. But, in that measurement, I already had something like >10 V power supply ripple at max output & still very low distortion [in fact, it is the least just before clipping]. Changing the cap to 1000 uF will only cause further increase in ripple, and that would only, say, double the 0.007 % distortion? I think it's not what Jan meant?
What I wanted to say there, is that I also thought before this test, that, at least, some 50Hz induced spuries will be seen.. but not. Nothing. [Naturally, this happened only after that i cleaned up all ground loops, & resolved all shielding problems] In the same time we were looking at a push-pull EL34 tube amp, which hummed like crazy, intermodulated all the test signals with all the power supply spuries, [wanted to demonstrate Jan's claims] had shown a bit of ringing, severe bandwidth limiting, slew rate problems, and huge amounts of harmonic distortion [which did the intermodulating job with the power supply hum..]
This tube amp in the same time had blown off piste my GC, in the listening tests..
George |
What kind of speakers were used during the listening test, George? |
|
|
| Joseph K |
In that test we used my friends setup, that tube amp, AN CD1 as a mechanics / also CD player, North Star Dac 192, speakers Sonus Faber Concerto / Grand Piano.
They are not very effective, though not bad. Interestingly, my GC seems to sound better with my speakers, which are DIY Visaton AL170/ Vifa, low efficiency. [maybe also not so low] |
|
|
| Joseph K |
I would like to add, that tube amp measures rather well, for being a tube amp, it seemed only be ridiculous in a direct comparison with the GC. The hum was only measurable, not appearent during listening.
So, you are right, Greg, maybe those Sonuses were really made for tube amps. Also we concluded, that the ~ 4ohm output impedance of that tube amp might be a major culprit for it's beautiful sound...
And I would like to say it again, higher power supply impedance does not mean automatically higher output impedance! Only if the amp runs out of control, like when in clipping, or at high frequencies, where it's OLG has diminished already.
Thanks, Carlos! It is interesting, that your experience is somewhat different from mine. When I did try to raise the power supply cap value, [inspired by You] then what we experienced was properly the opposit - more bloomy, much less controlled bass, and somehow compressed, somebody sat up on it soundstage... [It was just adding more capacitance, not the proper setup suggested by You]
Ciao, George |
|
|
| GregGC |
| quote: | Originally posted by Joseph K
In that test we used my friends setup, that tube amp, AN CD1 as a mechanics / also CD player, North Star Dac 192, speakers Sonus Faber Concerto / Grand Piano.
They are not very effective, though not bad. Interestingly, my GC seems to sound better with my speakers, which are DIY Visaton AL170/ Vifa, low efficiency. [maybe also not so low] |
Did you find that Sonus Faber Concerto / Grand Piano sounded less clear on the GC, compare to the tube amp? I wonder if that was the same thing I experienced when increased the PS caps. I bet you were using more than 1500uF/per rail at that time. With a 1000uf you'd have liked it better on those speakers if that's the case.
Greg |
|
|
| carlosfm |
| quote: | Originally posted by I_Forgot
The ultimate performance of the amp chip is defined by the designers who assume and use stable, well bypassed, DC supplies. |
Sure.
"Stable" PSU with 10,000uf unregulated PSU, as they say, right?:dodgy:
| quote: | Originally posted by I_Forgot
All bets are off if you don't give the chip a stable, well bypassed DC supply. |
I did.
I bypassed as they say on the datasheet, and there's more.
I used big caps, small caps, 10,000uf, 4,700uf, 1,000uf...
I tested all kinds of bypass you can think of, and also paralleling 1000uf caps.
Above 2,000uf the mids/treble lost "air", definition.
It's not a question of less, it's not a question of "tone controls".
I had (and have) very good results with regulated PSUs.
Unregulated, things only started to work with high capacitance when I started testing snubbers.
| quote: | Originally posted by I_Forgot
Small power supply capacitance leads to high supply impedance, which leads to high output impedance. The results are unpredictable. |
As you can see from my previous post, I don't defend small capacitance.
And these days I don't regret that I don't have ultra-sensitive and easy to drive speakers on my main system.
If I designed an amp for those kind of speakers it would sound like junk everywhere else.:clown: |
|
|
| GregGC |
| quote: | Originally posted by Joseph K
And I would like to say it again, higher power supply impedance does not mean automatically higher output impedance! Only if the amp runs out of control, like when in clipping, or at high frequencies, where it's OLG has diminished already.
George |
You are right that it's not that strait forward, though to some degree, because the PS caps are in series with the speakers and the transformer's Zout in par of the caps and the efforths of the OPAMP to compensate for the Voltage changes, if I'm not mistaken. It's quite complex. So for LF and small PS caps the Rout probably increases to some degree together with other factors that kick in.
Greg |
|
|
| carlosfm |
| quote: | Originally posted by Joseph K
Thanks, Carlos! It is interesting, that your experience is somewhat different from mine. When I did try to raise the power supply cap value, [inspired by You] then what we experienced was properly the opposit - more bloomy, much less controlled bass, and somehow compressed, somebody sat up on it soundstage... [It was just adding more capacitance, not the proper setup suggested by You]
Ciao, George |
No, my experience is exactly the same, with high capacitance.
That's it!:angel:
Only my experience with low capacitance was not the same as yours.
But high capacitance mucks up the mids/highs and the bass doesn't get faster.
Only the snubber DID the job.
EDIT: one night, years ago, I tested my low capacitance (1,000uf) GC with more than 15 pairs of commercial speakers.
The amp couldn't drive anything properly.:bawling: |
|
|
| Joseph K |
Greg, yes, they are in series, and while correcting, the amp should work harder. But I feel it is a make or brake relation - until it is in control, you should not see " somewhat raised impedance".
But I only stick to this - otherways i admit, that there are a full bunch of effects which suddenly get in the picture!
And yes, there is also the transformer and the diodes, as well. And the time spent in direct conduction is not negligible, it starts with ~ 30 percent at low volume, and raises from there.
The way the GC failed in that test was rather somthing that Carlos would call "incapability of driving the speaker", which surprised me a lot, because that speaker is a first order crossover, rather high impedance speaker..
And it is not only the caps that count. For me, changing the diodes caused a much more dramatic change, than a couple of uF-s. Also the transformers are never the same..
Ciao, george |
|
|
| carlosfm |
| quote: | Originally posted by Joseph K
The way the GC failed in that test was rather somthing that Carlos would call "incapability of driving the speaker", which surprised me a lot, because that speaker is a first order crossover, rather high impedance speaker.. |
Well, if you don't know the Epos speakers, the woofer is direct and the tweeter is first order (a cap in series and a resistor in parallel, to attenuate the output).;)
They are 8 ohms, 86~87db and hard to drive by many commercial amps.
On my tests with commercial speakers and low capacitance PSU, 4 ohms was out of the question, and multi-driver speakers (like two woofers, or 3-way, etc.) just sounded plain horrible.
It is not a question of "claimed" average sensitivity, it's just that most bass-reflex speakers these days (even 8 Ohm) have impedance dips in the bass to below 4 ohms.
To sound decent and go with the low capacitance PSU the speaker must have a very simple crossover AND a very benign load.
A very difficult thing to find these days...:xeye:
Why bother, when you can make a decent amp?:D |
|
|
| Joseph K |
Carlos,
Yes, I know the Epos series. Mmmmm, nice one.
It should be noted, that I have far far less experience in tests, than You, or for that matter, a lot of people here. So I don't argue with You on this. Even more, as my far less experience tells me, it is reasonable what You say here, and I concur.
Ciao, george |
|
|
| Joseph K |
Now, I would like to get back to the original question.
Personally, I find it interesting, or, say challenging. Same like the snubber puzzle, which I still don't get, I have to admit. But, thinking about the snubber, enlightened me about my false ideas, which were so nicely set before.. [Thanks, Carlos] First of all, I realized what this thread is about, that is, we don't even know, why the different size caps sound different.
One thing that i think of is that the recharge current pulses would play a major effect. With such high value strong pulses sharing the same path with the output signal, accross such a non-linear something like the cap's ESR, we SHOULD experience some strong effects!
Now, if we try to look at these pulses, what are the controlling factors? Here is where I had my nicely set, pre-biased false ideas. I always supposed, that the bigger the filter [reservoir] cap value gets, the higher amplitude these pulses are taking. [higher capacity, lower ESR, etc.]
While I tried to simulate the transformer - rectifier - filter cap supply model, I realized that the sims didn't give the same results, that were visible on the scope. The sims gave a sharp front edge, narrow high value current pulse starting at the moment when diode anode tension exceeds that of the cathode.
But in reality it is a round, ~ gaussian - looking medium amplitude pulse, which rarely gets much narrower than 1.5 -3 mSecs. So something other is limiting than the transformer DC R, or cap's ESR.
Here again, like in the case of non-snubbed diode ringing, the LC tank formed by the transformer leakage inductance and the reservoir capacitor gets in the picture. Only that now the formed tank circuit has a much lower resonance frequency. It's in the hundreds Hz region, maybe kHz.
So the current surge in the diodes excite this slow LC tank, and this is the limiting factor, which defines the pulse shape & rise time. Now the simulation started to look like real!
This tank circuit is actually getting SLOWER with filter cap value rising! That is, the current pulses recharging the filter cap, at a given output current load, are getting LOWER amplitude & broader, while we are going from 1000 uF to 20000 uF ! The other dominant factor here is the DC R value of the secondary, which though does not tell us the amplitude, but actually damps this LC tank formed.
Then, what is the dynamic behaviour of such a system? As told, it is a quite well damped, low Q LC parallel resonant circuit. What if we now, all of a sudden, change the load current to a much higher value, let's say, for 100 mSec? [a drum beat?]
In a balanced status, for a given setup & load current [say, 0,5A], we see a current pulse train, each 10 mSecs a ~3mSec wide, 1-2 amper amplitude pulse.
____________________________
I I I I I I I I I I I I
I I I I I I I I I I I I
I I I I I I I I I I I I
In case of 1000 uF & a 30 msec sudden load current change, we will see:
____________________________
I I I I I I I I I I I I
I I I I I I I I I I I I
I I I I I I I I I I I I
I I I I I
I I I I
I I I I
I I I I
I I I I
I I I I
I I I I
After the three, 10 msec distanced high [load] current excited high current pulses the cap gets back to it's almost original state in the fourth pulse.
With 10000 uF and the same 30 mSec load current change:
__________________________________
II II II I I I II II II II II II II II
II II II I I I II II II II II II II II
I I I I I I I I I
I I I I I I I I
I I I I I I I
I I I I I I
I I I I I
I I I I
So, what I have seen is that the bigger cap is SLOWER to get back into it's original balanced status!
It "remembers" of the load change that happened earlier! In the case of the 1000 uF, it does not "remember"! It just gets recharged in the next pulse, following the falling edge of the load pulse.
But, the bigger cap sees smoother, smaller amplitude pulses, while the smaller cap experiences heavier, shorter pulses. [these can be 6-10 amps, or more..]
I don't know, what does it mean with reference to the sound. Maybe not a big deal. But I see it as a demonstrable, existing difference, and which somehow reflects the "big cap is slower" or "muddy", impression.
Here I was talking about actual measurements and sims, that I made, but at this moment I was unprepared to nicely serve them, so comes this ugly illustration, pardon me.
Ciao, George |
|
|
| Joseph K |
I've found some files from the past measurements.
This one here would be the 1000 uF, the recharge current pulses measured using ~.05 ohm series resistance, the load current draw is ~0.6 - 0.7 A; the recharge current pulses are ~ 2 A; the load peak is ~3.5 A, the pulses are 8 A.
Previously I forgot to mention, an another limiting factor in the developing current is the core saturation. |
|
|
| Joseph K |
This another pic would be the previous setup, but with 20000 uF.
Also, here I forgot to turn off AC coupling, for this reason the baseline shifts. |
|
|
| Joseph K |
| here is the 1000 uF, again, but zoomed in: |
|
|
| Joseph K |
| With some luck, it is even possible to notice some ringing in the process [here there was 2200 uF in the setup] |
|
|
| GregGC |
George. that's great work.
Now, what happens if the load current is 20 Hz, 50 Hz, 100Hz. For frequencies close to the 60/120Hz (for you guys in Europe 50/100Hz) things will get complicated. The charging current of the PS caps will get mixed with the discharging current going to the amp/load. I wont be surprised if some harmonics show up, though the PSRR is quite high. I think the distortion figures you got were at static state charge/discharge of the PSU caps, I mean constant level of the input signal. Did you try frequencies closer to the mains frequencies (50/100hz). What if it's a burst 30, 50, 100 .... Hz signal. How can you measure if it gets amplified without any distortions of the original signal and in which case the attack of the signal gets represented more correctly (1000uF or 10000uf). George, if you have the simulation model of the amp with the caps, did you run a frequency response to see if anything changes in the LF area with 1000 and 10000uF (most likelly nothing will, but...).
I'll try tomorrow to install the snubbers and see if I'll hear any difference in my setup. My ears are not as sensitive to small changes, though the diff. doubling the PS capacitance was really big (A lot more base, 100hz and below, soft and controlled and extends much deeper than before).
Greg |
|
|
| akunec |
| My next speaker system will be a 3 way fully active system with an LCY ribbon tweeter, PHL 1120 mid (350-3000hz) and lambda TD12S for 30-350Hz. Would it be best to use lots of capacitance (30000uf) on the TD12S, and then not so much (2200uf) on the mids and the tweeter? |
|
|
| peranders |
| quote: | Originally posted by I_Forgot
There are some who think that small capacitance on the power supply rails results in best sound from the chip amps. |
This has been a fashion for some time but is to going away(?) now when a high capacitance power supply is being marketed.
I think when we are taking about taste, the sound will be different but the question is if you like it. One thing for sure is that high ripple voltage sneaks into the amp, mixes with the audio signal and eventually some part of it comes out as distortion products. Noone can deny it. |
|
|
| analog_sa |
| quote: | | Would it be best to use lots of capacitance (30000uf) on the TD12S, and then not so much (2200uf) on the mids and the tweeter? |
This is an important decision in the final 'voicing' of your speakers. It is not something to take lightly and especially not to delegate to others - you'll have to use your own ears. |
|
|
| peranders |
| quote: | Originally posted by akunec
Is it true that using more capacitance on the supply rails can negatively affect high frequency performance? I find it really hard to believe but I heard it a few times on this forum. Can someone offer a logical explanation? |
My vision (not answer really) is that there is a range where you have an optimum. Too little, not good, too much, not good either. This is a theory at the moment and I have no proofs. |
|
|
| peranders |
| quote: | Originally posted by jackinnj
i can demonstrate the way in which ground loops, inadvertent power cable placement etc. measurable affect distortion. |
I think we can assume that everything has been taking care of, in other words, a perfect environment for the lM3886 but with different cap values. |
|
|
| Upupa Epops |
| P - A, it is about psychology ( in this case ). Man tends to take newest things as better, 'cos they are not still in his brain " wearing out " ;) . |
|
|
| janneman |
| quote: | Originally posted by Joseph K
[snip]What I wanted to emphasise, is that it is NOT a tone control. At least not in a simply measurable way.
[snip]I already had something like >10 V power supply ripple at max output & still very low distortion [in fact, it is the least just before clipping]. Changing the cap to 1000 uF will only cause further increase in ripple, and that would only, say, double the 0.007 % distortion? I think it's not what Jan meant?[snip]
George |
Hi George,
Indeed, it is not really a tone control, I said that tongue-in-cheek, I thought that was clear. But it CAN be used to modify the speaker tonal balance.
What I meant on the ripple modulation (I think that is what you refer to) is related to large output levels. If you indeed have up to 10V ripple, with large output signals, the output level varies with the actual supply voltage which varies in a 100Hz (or 120Hz for those 115V mains voltage challenged countries:D ) rithm. That is plain signal modulation, aka as an AM signal. Just the frequencies are different from a standard AM radio station. Put a lf detector and headphone on the amp output and you can listen to the ripple! But you know all that.
In another thread people worried about minuscule AM modulation resulting from PIM and TIM conversion. It seemed that it really ruined their enjoyment of music. Here you got many percentage points, maybe even tens of it, AM modulation depth, and people like it! Someone above, I think it was analog_sa, coined the term "audio virgins". Its a good term, although it probably wasn't what he meant.
Jan Didden |
|
|
| janneman |
To Joseph K:
Joseph, I have just read your posts on this on the previous pages. Good posts. What I see in the graphs is that what you call "slowness" of the cap is of course a slow variation of the charging current pulses. It is wrong to relate that to "slowness" in sound reproduction. In the limiting case, if you would have a perfectly regulated DC power supply, this is infinitely slow, there is NO variation of the voltage on the cap. But surely an amp with such supply isn't "infinitely slow" or being able only to reproduce DC!
I maintain that the "slower" in this respect a power supply is, the better. Amplifiers are designed with the implicit assumption that the supply is DC, it is almost always also acting as a ground point for the signal. Any ripple or signal residue on the supply line messes up the signal fidelity to some degree. It may be a little in competently designed amps, but it is there.
Jan Didden |
|
|
| carlosfm |
| quote: | Originally posted by Upupa Epops
P - A, it is about psychology ( in this case ). Man tends to take newest things as better, 'cos they are not still in his brain " wearing out " ;) . |
This just shows total ignorance.
The "snubber" is used on audio amplifiers for some 35~40 years.
I just picked it up, tested it and optimized it for power op-amps, or chipamps if you like.
Because I knew it could be the answer to the problem of the high(er) inductance of the bigger caps.
Why would I be biased to listen for ghosts just because I wanted it to work?:clown:
Belive me, if it didn't work I would forget it and I would not even have opened the thread and reported it.
But you tend to call everyone stupid, as a dictator calls his own people stupid.
You are THE voice, you are THE reason, you know it all, you are GOD.
Amen.:angel:
PS: while you fail to understand that the PSU is also part of the signal, that everything is on the signal path, you fail to make a decent amp.
You blindly believe that the best you can do with a chip is what they say on the datasheet.:clown: |
|
|
| peranders |
| No matter what Pavel thinks, have you tested to remove the resonace peaks which are present at the supply pin of LM3886, Carlos? Have you any comment about my "enhanced" snubberization? |
|
|
| carlosfm |
| quote: | Originally posted by peranders
No matter what Pavel thinks, have you tested to remove the resonace peaks which are present at the supply pin of LM3886, Carlos? Have you any comment about my "enhanced" snubberization? |
We agree, no matter what Pavel says.:D
The resonance peaks are much attenuated by the snubbers, and the bypass caps serve, as you should know, to provide a low impedance path to ground at high frequencies. |
|
|
| Upupa Epops |
| Carlos, I'm so tired by your snubber. You are permanent claiming, that for good sound is this circuit mandatory, my meaning is different. Let your opinion and I let mine. All is crazy stupidity and I can't talk about this, good amp make something quite different. I'm not God, I know, that I know nothing, but don't make my teacher, it's funny for me. |
|
|
| akunec |
| I would really like to see what a sine wave at various frequencies looks like with different supply capacitance. |
|
|
| carlosfm |
| quote: | Originally posted by Upupa Epops
Carlos, I'm so tired by your snubber. |
I don't need to talk about it, everyone talks, specially those who tried it.
And notice: I'm not selling or trying to sell anything.
| quote: | Originally posted by Upupa Epops
All is crazy stupidity and I can't talk about this... |
You still don't understand what's this thing about, so yes, for you it may be stupidity.
Anyway, besides this, what I don't like to see is your way of posting:
| quote: | Originally posted by Upupa Epops
P - A, it is about psychology ( in this case ). Man tends to take newest things as better, 'cos they are not still in his brain " wearing out " ;) . |
You can all see that without opening my mouth this guy posts like this.
You keep ridicularizing and calling psychos to those who TRIED and claim improvements, when you don't even have the guts to try it.
And if you do, you will not be honest enough to admit improvements.
It's just too late for you.:dodgy: |
|
|
| carlosfm |
Oh, and PS:
Those who understand it, encourage it's use.
Like here, check Phred's last paragraphs, before the links:
http://www.diyhifi.org/forums/viewtopic.php?p=3780#3780
Quote, Fred Dieckmann:
"Oh yeah........ before I forget, gratuitous snubberizing is welcome and encouraged, but not three weeks discussion about it."
He's just a PSU man...:D |
|
|
| GregGC |
Last night decided to double the 1000uF caps and the result was more solid base. The presentation though became somehow unclear. I thought the base was too much and masked the rest of the spectrum.
So I decided to try the "controversial snubbers" and to be honest I approached the modification with a dose of skepticism. Hey what if... I had my first lesson when I decided to change the opamp in my CD player. I didn't believe that it may sound different/better (I am an EE and we are talking 20kHz after all). Well, most of you who tried it know the answer.
The setup:
NITGC,
2x1000uF/rail/channel at the chip (lm3875),
(I don't need to go to more than 2000uF/rail/channel. There is plenty of base and loudness)
No 100n bypass caps on the rails.
One PS for both channels with a single bridge rectifier (600V/40A and 220VA CT transformer).
Snubbers 100n/1R.
I installed the snubbers as close as possible to the chip.
There is definitely a big improvement in the HF spectrum. The easiest way to explain it is to say that it removed a layer of HF grain. It makes the highs polished and more defined, more sparkly.
The improvement is quite obvious and not at all subtle.
It makes the whole picture more defined. It's like looking through a clean window rather than a foggy one.
For testing I used mainly Destiny's Child/Survivor. A lot is happening on that track. It tests the capability of the audio system quite well.
Mind you my ears are not the most sensitive. I really admire people that can catch some small differences in the sound. I think it takes a lot of time of listening to music to develop that kind of sensitivity.
So, the snubbers are staying for good. Thanks, Carlos! I'm a happier man now.
I didn't try just a 100n bypass cap instead the snubbers, but that'll be another day. May have a positive effect too. I don't know yet.
/Greg |
|
|
| analog_sa |
| quote: | | I would really like to see what a sine wave at various frequencies looks like with different supply capacitance. |
The same? Unless you go really overboard with reducing either the capacitance or load. Nothing can be less revealing than this test. |
|
|
| Dr. Photon |
I don't see how you can get better high-end response from smaller capacitors (or that large capacitors have "worse" trebble). I do understand that undersized filter capacitors can hurt you in the bass region and a power supply input for the chip that is jumping around would make the op-amp's feedback network do some serious work to maintain the output. I can also see how the chip can maintain decent sound quality even with a very poorly regulated power supply (untill the rails sag dangerously close to the output) due to the action of negative feedback.
I think that for the maximum sound quality, the power supply should be as stiff as possible at all frequencies. I have typically used large main reservoir capacitors (several 1000's of uF, such as 4700 or more per rail), mounted whever makes the most sense in the mechanical layout. These are supported by capacitors of a few 100's of uF mounted on the same board as thr chip(s), as close to the chips(s) as possible, perhaps one set per chip in a stereo or bridged unit that shares the power supply. And finally I would bypass the rails with a 0.1 uF capacitor RIGHT AT THE CHIP. these capacitors (I've used both ceramic and the green "chicklet" film capacitors) should be mounted as close to the chip as practically possible. Either right in front of the chip in a PCB layout, or perhaps under the board, hanging off of the chip's pins in a DIY protoboard layout. Sometimes I'd also use a small capacitor at the chip between the positive and negative supply rails. I always make sure to bypass my chips at the chip with a .1 capacitor or so, since I have had some bad experiences (with a PICmicro and some op-amps) by leaving it out.
The only bownside to a large capacitor (besides size/cost and inrush current) would be that a single large capacitor might have a somewhat higher internal resistance and inductance (taken care of by the smaller board mounted capacitor). Annother solution to this problem would be to use several smaller capacitors. For example 3 2200 uF units in parallel rather than one 6800 uF cap. |
|
|
| peranders |
| quote: | Originally posted by carlosfm
We agree, no matter what Pavel says.:D
The resonance peaks are much attenuated by the snubbers, and the bypass caps serve, as you should know, to provide a low impedance path to ground at high frequencies. |
Note that I'm not arguing how much good the snubber does just what it does electrically.
The resonace peak is dampend at the left part of the PS but, Carlos, what do you say about the local decoupling very closely to the supply pin where you have an additional 100 uF// 100nF? There you must have some resonance too and what about it? Isn't that resonance even more important?
My point is that you have two resonance peaks but treat only one of them which has the least influence and the most important peak is untreated. |
|
|
| peranders |
| quote: | Originally posted by Upupa Epops
Carlos, I'm so tired by your snubber. You are permanent claiming, that for good sound is this circuit mandatory, my meaning is different. Let your opinion and I let mine. All is crazy stupidity and I can't talk about this, good amp make something quite different. I'm not God, I know, that I know nothing, but don't make my teacher, it's funny for me. |
Pavel, skip the emotions and just think technically. There are resonance peaks, how to remove them?
Is this important? It's an another question. I can't pick any huge difference but that's me. |
|
|
| Upupa Epops |
| P - A, how is value ( amplitude ) of this oscilations ? 10 mV ? 100 mV ? 1 V ? What is ratio of this amplitude related to DC ? 1 % ? Or less ? How is theoretical mechanism of penetration to signal path ? And PSRR ? In every my amps I'm using bypasing rail to ground, but also rail to rail - these oscilations have theoretical oposite phase, will they in this case by bypasing shorted ? Etc... - many questions are here, but any answer, only with postcoital smile declared claims. :xeye: |
|
|
| peranders |
The peak is reallly, really small..... and not to many dB either 3-6 dB! in the lower MHz region.
I'm trying to sell in that the first and only snubber doesn't do much good at the supply pins, still the effect is big or even huge!
My "product" here is to snubberize also the supply pins. How much this peak has any real influence, I have no idea becasue I can't detect it and I don't want to hear any degrading remarks about that over at "the other" forum, please. |
|
|
| rha61 |
Hi JosephK
Again , thank you for your very analytic work
i have another aspect to relate about big capacitance :
I was very happy with a buffer stage (ad826 in class A driving Hypex ucd module ) powered by a CRCRC+snubber psu
I have heard about memory back up capacitors ( 1F/5v5) that some use in their audio psu
i put two in // with my second C , my psu became CRCC'RC+snubber
it takes some minutes to load but the sound is really impressive ! and it approaches the battery//lowESRcap final solution
we all know that a good psu is primordial , and unfortunately , filtering is a very complexscience !
alain |
|
|
| GregGC |
| quote: | Originally posted by analog_sa
"I would really like to see what a sine wave at various frequencies looks like with different supply capacitance. "
The same? Unless you go really overboard with reducing either the capacitance or load. Nothing can be less revealing than this test. |
I don't think you'd see any difference. I remember when I was swapping opamps in my CD player and finding such a huge diff. in the way they sounded I wanted to see it on the scope. Nothing. 20hz-20khz @ +/-0.2db. No change at all.
I suspect (in the case of GC) the diff. to be in the fact that the NFB/opamp has to work a lot harder to compensate for all the anomalies and imperfections of in the PS parameters, thus effecting the sound. Just my speculation, but I'm 99.9% sure you wont see it on the scope. I wish it was that easy to catch the differences in the sound.
/Greg |
|
|
| carlosfm |
| quote: | Originally posted by peranders
Note that I'm not arguing how much good the snubber does just what it does electrically.
The resonace peak is dampend at the left part of the PS but, Carlos, what do you say about the local decoupling very closely to the supply pin where you have an additional 100 uF// 100nF? There you must have some resonance too and what about it? Isn't that resonance even more important?
My point is that you have two resonance peaks but treat only one of them which has the least influence and the most important peak is untreated. |
Amazingly, anyone here has yet realized that snubbing the main filter cap(s) is also snubbing the resonant circuit formed by the wiring from the raw supply, PCB traces and the local capacitor decoupling at the IC. |
|
|
| peranders |
| Carlos, simulations and real measurements show that your circuit doesn't remove the peaks. One single resistor won't kill the resonance as you have done. The local decoupling creates also a peak which you leave untreaten. |
|
|
| carlosfm |
| quote: | Originally posted by peranders
The local decoupling creates also a peak which you leave untreaten. |
It is taken care of, don't worry.:D
Feel free to leave the 100nf local decoupling cap on the chip if you feel better.
But try the snubber and report your results.:devily: |
|
|
| yopi3622 |
| quote: | Originally posted by janneman
... I maintain that the "slower" in this respect a power supply is, the better. Amplifiers are designed with the implicit assumption that the supply is DC ... |
This assumption seems right to me too.
I'm surprised nobody tried a Gyrator circuit for the PSU to slow down charge pulses... :angel:
Take a look at this page to find a "Gyrator PSU" example :
http://www.itechu.co.uk/007/circuit...r%20Circuit.htm
And ... no more expensive big caps with this solution :D
Yopi |
|
|
| peranders |
| I'll guess you don't want to comment my "enhanced" power supply, can be seen here, so I'll claim it as a real enhancement unless someboby has anything to object. And yes, pavel, I don't believe it's particulary necessary to remove small unmeasurable in the MHz region. It's more like a cool thing, like when I made a hole in my groundplane at the input transistors in my QRO amp long ting ago. I could trim the slew rate behaviour but it hardly had any impact on the sound. More info here. See especially page 11. If you reduce the stray capacitance under the emitters of the input transistors you will improve the step response under slew rate limiting conditions. |
|
|
| Mr Evil |
| quote: | Originally posted by yopi3622
This assumption seems right to me too.
I'm surprised nobody tried a Gyrator circuit for the PSU to slow down charge pulses... :angel:
Take a look at this page to find a "Gyrator PSU" example :
http://www.itechu.co.uk/007/circuit...r%20Circuit.htm
And ... no more expensive big caps with this solution :D
Yopi |
That's a capacitance multiplier, and plenty of people have used them before, including myself. They don't reduce charging pulses in the way that an inductor does, since they can't store energy. |
|
|
| carlosfm |
| quote: | Originally posted by peranders
I'll guess you don't want to comment my "enhanced" power supply, can be seen here, so I'll claim it as a real enhancement unless someboby has anything to object. |
YOUR "enhanced" PSU? :eek:
You mean that you are converted to the snubbericerad PSU?:clown:
Take notice: if I didn't reported anything you would not even know it.
| quote: | Originally posted by peranders
...so I'll claim it as a real enhancement unless someboby has anything to object. |
I don't see the point of using a second snubber, and even worse, the last thing before the chip's PSU pins.
Also, that way the 100uf caps will not be as near to the pins as they could.
You are walking around in circles, and I don't see your point now. |
|
|
| jackinnj |
| quote: | Originally posted by yopi3622
This assumption seems right to me too.
I'm surprised nobody tried a Gyrator circuit for the PSU to slow down charge pulses |
maybe an inductor is easier. |
|
|
| Joseph K |
Carlos!
| quote: | | Amazingly, anyone here has yet realized that snubbing the main filter cap(s) is also snubbing the resonant circuit formed by the wiring from the raw supply, PCB traces and the local capacitor decoupling at the IC. |
Hey, this time you are picking at something interesting! Though may I disagree with the first statement of Yours? [I know that You were not aiming at me, but I feel like being included in that "anyone"]
What could have been in my mind in this moment? post52
Only at that moment, after having lost my post, had to start for our ski trip..
But I have to admit, that your above statement is still a tad confusing for me, could You clarify it a bit further? Which Pcb traces? Supply board or amp board? Which raw wiring? to the rectifiers from transformer or from rectifiers/supply board to amp board? Given that the word "snubbing" is not properly used, what do you mean by that? I am poking you, because you made me curious about your view of it!
Ciao, George |
|
|
| GregGC |
If the snubber compensate for the parasitic inductances of the caps and the traces it would make sense if they are connected as close as possible to the PS pins of the opamp (if I'm not mistaken).
Carlos,
Did you try the effect of snubbing at diff. locations (PSU side vs. Chip side). I tried it on the chip side and it improved the sound quite a lot.
http://www.diyaudio.com/forums/show...4466#post604466
Thanks
Greg |
|
|
| carlosfm |
| quote: | Originally posted by Joseph K
Hey, this time you are picking at something interesting! Though may I disagree with the first statement of Yours? |
You can disagree, as I was not including you.
Actually you were and are the one here who maby better understands what's this all about.
Just by being so curious, you already show that.
And thanks for your measurements on the other thread.
| quote: | Originally posted by Joseph K
But I have to admit, that your above statement is still a tad confusing for me, could You clarify it a bit further? Which Pcb traces? Supply board or amp board? |
Everything after the snubbers.
If they are located at an unreg. PSU on a separate board, everything after them.
Including the bypass I use on the big caps.
| quote: | Originally posted by GregGC
If the snubber compensate for the parasitic inductances of the caps and the traces it would make sense if they are connected as close as possible to the PS pins of the opamp (if I'm not mistaken). |
It does make sense to snubberize at the chip's board, close to the chip, but not the last thing.
Remember, it's very very important to have the small caps as near as possible to the PSU pins.
So, I would use the snubber before the 100uf caps, if I was to use it on the chip's board.
Improvement or not, that depends on your layout, cabling, etc.
It may be a good improvement on your case, maby unnoticeable on other cases.
What's very important here is to use small caps (no more than 100uf) on the chip.
And thick wire (or traces) for PSU, very thick for power ground.
And finaly Greg, I love to see open-minded people, thinking with their heads and accepting the challenge to understand, try to improve, experiment.
Instead of sticking to what says on a schematic, or on a datasheet, or on the books, whatever.
Congratulations.:cool: |
|
|
| Joseph K |
Carlos, though I have probably different opinion impedance -compensation wise [see my following post to Jan], on this last point I agree with you.
In a "classic snubbered" config, you tie down, "anchor" the HF impedance of the circuit at two, physically separated points: at the supply pins, the 100 uF+ 100nF applied locally effectively bypasses the reservoir cap ESL plus [much greater] stray L of the wires; this bypass config, though not perfectly balanced for max. smooth phase & impedance, is peaking much less then if it were with a greater cap value [ the 100 uF cap's ESL is less than that of the 1 - 10 kuF]; the 100 uF is needed to "swamp out" the stray inductance presented [mainly by] the wires.
Then it is true, that at the other point instead [the diodes / filtercap node], if you don't apply anything, HF-RF wise you leave it floating.
If you apply the film cap, you create a resonant parallel LC circuit, where L is formed by the filter cap ESL + plus the stray L presented by the wires at this side. The Q of this tank circuit will be high, and a "snubber" here will dampen it, in this way "anchoring" this floating node. A low impedance node at this point will help to block the RF noise [common & diff] creeping in from the Line; the RF noise generated by the diodes; the harmonics in the half wave-rectified audio signal.
Because the L part is formed only partially by the filter cap, a "snubber" here will have similar effect even without the cap.
So, it is HF filtering, not impedance correcting. [at the power pins]
This far, this is how I can imagine the things. But, somehow I never felt satisfied about this all, this is why I don't stop.. |
|
|
| Joseph K |
Jan,
| quote: | | What I see in the graphs is that what you call "slowness" of the cap is of course a slow variation of the charging current pulses. It is wrong to relate that to "slowness" in sound reproduction. |
You are right, I was not clear enough in my conclusion. I think I picked up the wrong expression. It is quite hard to describe the effects experienced while adding - subtracting capacitance to the amp's power supply. Though this all is very subjective, there seems to be an agreement that bigger caps will bring [seemingly] more bass, and have an influence on sundstage / air /mid articulation. I experienced something like a compression, less ability to separate things.
But, I would not like to divulge into it, - this is not my field..
The fact is, that something is changed. Like when you change the type of diodes, transformer, Dac bypass, interconnects, digital cable, transport - all things have their influence. Only that at most of the time it is really hopeless to find a plausible explanation, or I should say, the explanations are really above my head..
In this special case, I hoped to be able to get somewhere, and again, it was only false hopes..
Anyway, while thinking about it, my first instinct told me that the difference is maybe in the caps electrical parameters, in it's frequency dependent impedance. But, after having a look at it, the reality is that today's big electrolytic caps are so well constructed, that their ESL is comparable with the stray inductance of that length of wire, which would replace them in the circuit. In some cases, [like the Jamicon which I used] they can actually have less ESL then the 1000 uF "classic GC" cap! [for example, my Elna Cerafine]
Anyway, we are talking about a dozen nH, and Fc [3db] at about a couple of hundred kHz, with respect to their 20 - 50 mohm ESR, and Fc [3db] of a couple of MHz, if we refer to 1 ohm.
So they are lightning fast in the audio range, though their ESL can be a problem RF stability wise.
But, even more to it, the caps inherent ESL is just nothing, when compared to the stray wiring inductances present in most of the configs here.
So, I started to look out for other possible causes, and I found, amongst many other things, this "memory" effect present in the recharge current pulse train. Now I really don't know the weight of it. Nor the exact mechanism, in which way it can result in distorsion. The signal current & and charge pulses are sharing the same, surely non-linear ESR path. These pulses are strong, might trigger electromechanic responses in the cap.. [it's not my idea]
But maybe their effect is really small, and I'm on the wrong path.
From one hand, there is a listening experience, the other hand, this nice little mechanism, [amongst many others, true]. So I felt like it might worth to call you all's attention to it.
Ciao, George |
|
|
| carlosfm |
George!
You are getting there.:angel:
Let me quote Phred:
Quote Fred Dieckmann:
"You have just scatched the surface of a most interesting and very important aspect of decoupling power supplies for high bandwidth amplifier circuits and most of these guys just don't get it. I do get it." |
|
|
| Joseph K |
Carlos,
Now it's your take on bullying me! In fact, it would be most interesting to get grilled by Fred.. :clown:
Ciao, George |
|
|
| GregGC |
Carlos, thanks for the good words.
George, quite interesting...
Keep up the good work guys.
OK, just tried 100n caps instead of the snubbers (just shorted the 1R). The sound is not good again. Fuzzy and not as clear and defined HF picture. Same as before installing the snubbers. So, installing small caps in par of the big ones doesn't always solve the problems.
I put the snubbers back on and all is good again!
Snubburize!
Greg |
|
|
| Vix |
Now one stupid question, discussed probably 1000XXX times:
1. Assume that there are 2200uF caps per rail, bypassed by 100nf ones. If I insert 1 ohm resistors in series with existing 100nf caps, will that qualify?
2. Or, there need to be other 100nf +1 ohm in series?
{so, the final will be 2200uf+100nf+(100nf+1 ohm)}
I think that the first option should work....
Sorry for repeating the basics.....
Vix |
|
|
| carlosfm |
| quote: | Originally posted by GregGC
OK, just tried 100n caps instead of the snubbers (just shorted the 1R). The sound is not good again. Fuzzy and not as clear and defined HF picture. Same as before installing the snubbers. So, installing small caps in par of the big ones doesn't always solve the problems.
I put the snubbers back on and all is good again!
Snubburize!
Greg |
I have been telling this for years...
You can do what you want with bypassing, it won't make a bad amp into a good one.
From around 2,000uf capacitance and up, the sound really mucks up.
Only the snubber works.:cool: |
|
|
| GregGC |
| quote: | Originally posted by Vix
Now one stupid question, discussed probably 1000XXX times:
1. Assume that there are 2200uF caps per rail, bypassed by 100nf ones. If I insert 1 ohm resistors in series with existing 100nf caps, will that qualify?
2. Or, there need to be other 100nf +1 ohm in series?
{so, the final will be 2200uf+100nf+(100nf+1 ohm)}
I think that the first option should work....
Sorry for repeating the basics.....
Vix |
Yes, you can insert 1R resistor in ser. with the existing 100n cap and it qualifies as a snubber. And as Carlos said, snubbers are more important in this case than just bypassing with 100n. That's exactly what I proved to myself by trying it on my GC last night.
http://www.diyaudio.com/forums/show...5208#post605208
You can leave the 100n in and add the snubbers too (that may be your best option because you cover all the angles). That's what I ended up with, just in case. Adding 100n in par of the 2000uF and the snubbers did not change anything in my setup. So to finalize it: >2000uF || 100n+1R is a must.
Greg |
|
|
| carlosfm |
Greg, you must have really easy to drive speakers.
What I would consider a minimum is around 4,700uf per rail. |
|
|
| GregGC |
Hi Carlos,
http://www.yamaha.ca/av/speakers/PDMSeries/NS200.asp
Those are the ones I have. 2000uFrail/channel gives me an excellent base. And the MF and HF are amazing too. When I drive them with the Yamaha receiver the HF are not as open an free compare to the GC. The base is the same as the GC. The receiver has 2x4700uF in the PS, so it's the same. I should try one day 4700uf/rail/channel too. It should be better now that the snubbers are in.
Thanks for not giving up under the pressure of the "snubbers non-believers" that never tried it. I have to admit that I questioned the rational of it too, but had to try it first before I have any say on it. Just theory is not enough. It's much more complex and there are a lot of angles that can be missed. ONLY AFTER IMPLEMENTING a mod. I can say what the results have been on MY set up. I don' think personal ego should stay in the way of understanding, learning and building better GCs:) .
Greg |
|
|
| carlosfm |
Greg, if you want, and if it fits, you could try changing the 2x1,000uf caps per rail with 2x2,200 or 2x3,300uf.
But if you have good results as it is now, let it be.
| quote: | Originally posted by GregGC
Thanks for not giving up under the pressure of the "snubbers non-believers" that never tried it. |
Some more white hair on my head, no problem.:D
Some centuries ago a man insisted the the Earth is not flat, and that all planets turn around the sun.
He was almost burned alive.:xeye:
And the Church only recognized this in 1956.:clown:
Not wanting to compare anything but deam, why does a cap and a resistor creates such a spanking? |
|
|
| GregGC |
| quote: | Originally posted by carlosfm
Greg, if you want, and if it fits, you could try changing the 2x1,000uf caps per rail with 2x2,200 or 2x3,300uf.
But if you have good results as it is now, let it be.
|
Not to much space inside the amp, but I'll give it a try if I find small enough caps.
| quote: | Originally posted by carlosfm
Not wanting to compare anything but deam, why does a cap and a resistor creates such a spanking? |
And it's not that hard to see if it makes any diff. or not. It's not like getting on a ship and traveling around the world for a year to check if the Earth is round or not.
I think in this case the only people that lose are the ones that never tried it. They are the ones that potentially can have better sounding amps. Some people rather have their ego in good shape than enjoy nice sounding equipment. If that's what makes them happy, who am I to argue...
Greg |
|
|
| rdf |
| quote: | Originally posted by Upupa Epops
P - A, how is value ( amplitude ) of this oscilations ? ... And PSRR ? ... |
Hi Epops, in Spice simulations on the multi-segment RC supply for my SET amp the answer was, surprisingly, a PSRR after the secondary winding of ~ 0dB at 10 MHz. Better PS supply components - in the sense of lower ESR and inductance, etc. - cause higher Q resonances and are trickier to correct. Keep in mind that the PS caps in this amp are 40 uf oils with self-resonances around 100 kHz. The caps typically used in a GC become inductive at much, much lower frequencies.
And yes, as a result I "snubberized" a 3 watt EL84 single-ended tube amp to good effect. ;) |
|
|
| Upupa Epops |
| To rdf : Is funny to see here neverending story about bypasing of PS and about quality of components, while on 95 % photos of assebled amps showed here I see uncorrect internal wiring. Isn't problem here ? All are calling " simply, simply .. " but nobody " shortly, shortly ... " ;) |
|
|
| carlosfm |
| quote: | Originally posted by rdf
And yes, as a result I "snubberized" a 3 watt EL84 single-ended tube amp to good effect. ;) |
Yes, snubberizing is a good practice and not exclusive of GCs.;)
I also had reports of big improvements with class-D amps.
| quote: | Originally posted by Upupa Epops
To rdf : Is funny to see here neverending story about bypasing of PS and about quality of components, while on 95 % photos of assebled amps showed here I see uncorrect internal wiring. Isn't problem here ? All are calling " simply, simply .. " but nobody " shortly, shortly ... " ;) |
Well, the snubber vastly improves the performance of P-A's "industrial SMD GC", so I suspect it may be a problem with the wiring. :clown: |
|
|
| Upupa Epops |
| Carlos, don't worry, but your photos show me the same wire salat which I'm talking about ( 95 % ones ). |
|
|
| cjd |
| quote: | Originally posted by carlosfm
Some more white hair on my head, no problem.:D |
At least you get to keep the hair! :cool:
C |
|
|
| rdf |
| quote: | Originally posted by Upupa Epops
To rdf : Is funny to see here neverending story about bypasing of PS and about quality of components, while on 95 % photos of assebled amps showed here I see uncorrect internal wiring. Isn't problem here ? |
Can't disagree, but I went ground plane point-to-point. Does that admit me to the other 5%? :p
http://www.diyaudio.com/forums/atta...tamp=1104397052 |
|
|
| Upupa Epops |
| To rdf : Show me still top side :cool: . |
|
|
| carlosfm |
| quote: | Originally posted by Upupa Epops
Carlos, don't worry, but your photos show me the same wire salat which I'm talking about ( 95 % ones ). |
No, I don't worry.
You will | | | |
