My recent goal is to have flat distortion 20 - 20khz ... 
agreed - large bandwidth alone does not = good sound
I think it has to be combined with very low noise ( good PSRR ), good stabilty, reasonably low distortion and good dynamics ( micro and macro )
( ....and also, most important, a good front end ! - a transparent amp is is not much of an asset if the front end is sounding horrible. )
Normal bandwidth amplifiers have OLG that begins to roll off at quite low frequency. This then has to be rolled off even more so that in closed loop mode they are stable. This reduces the amount of feedback db in the high midrange and HF area
With wide band amp the roll off begins at higher frequency and the compensation for stability can operate at a frequency way above audible sound.
All of this means that there is are more db's of feedback available at high audible frequencies.
In my experience this gives a richer fine texture to the sound which I enjoy.
Also in feed back amps, PSRR, to some extent, relies on the amps feedback.
In my experience one very major factor in getting better sound from audio gear is reducing noise in general, but particularly the HF sonic & ultra sonic power supply noise.
So, if one has a broadband amp with a low pass filter at the input this can be a powerful means of preventing broadband PS noise from reaching the output.
My question was ( and still is ... ) to what extremes can one apply this principal and still notice an improvement both in reproduced HF texture, and the sweet smooth sound that comes with very low noise throughout the system.
I will continue to explore this area and let u now what I find
Happy New Year
cheers
mike
agreed - large bandwidth alone does not = good sound
I think it has to be combined with very low noise ( good PSRR ), good stabilty, reasonably low distortion and good dynamics ( micro and macro )
( ....and also, most important, a good front end ! - a transparent amp is is not much of an asset if the front end is sounding horrible. )
Normal bandwidth amplifiers have OLG that begins to roll off at quite low frequency. This then has to be rolled off even more so that in closed loop mode they are stable. This reduces the amount of feedback db in the high midrange and HF area
With wide band amp the roll off begins at higher frequency and the compensation for stability can operate at a frequency way above audible sound.
All of this means that there is are more db's of feedback available at high audible frequencies.
In my experience this gives a richer fine texture to the sound which I enjoy.
Also in feed back amps, PSRR, to some extent, relies on the amps feedback.
In my experience one very major factor in getting better sound from audio gear is reducing noise in general, but particularly the HF sonic & ultra sonic power supply noise.
So, if one has a broadband amp with a low pass filter at the input this can be a powerful means of preventing broadband PS noise from reaching the output.
My question was ( and still is ...
) to what extremes can one apply this principal and still notice an improvement both in reproduced HF texture, and the sweet smooth sound that comes with very low noise throughout the system.I will continue to explore this area and let u now what I find
Happy New Year
cheers
mike
make your bandwidth as wide as your design will tolerate. if your amp bandwidth is 1Mhz with no instability (and doesn't act as an AM receiver). then use it. you will never have a full power output at 20khz from most music sources anyway (that would also eat a lot of tweeters). but your distortion measurements at 20khz would probably be close to your measurements at 1khz. that's assuming of course your output devices are fast. bipolars begin going into common mode conduction around 100khz with most designs, which will eat fuses as well as output devices.
mikelm said:My recent goal is to have flat distortion 20 - 20khz ...
Not possible, different distortion mechanisms have different frequency dependence - Vbe modulation is flat over audio freqs, nonlinear Ccb causes increasing distortion with frequency, thermal modulation falls off quickly with frequency...
Not a useful way to look at it, most amps use single pole compensation, therefore larger GBW product gives larger open loop gain at all frequencies where the gain has a 1st order slope and is a quantity we should want to maximize - the limit is determined by the speed of the output devices (usually the slowest transistors in a amp), slow output devices add extra phase shift and cannot operate stably in single pole compensated amps at as high a overall amplifier GBW (even if we add more front end gain to make up for the low output Q GBW) as faster devices http://www.diyaudio.com/forums/showthread.php?postid=647995#post647995
The low/high frequency open loop gain terminology is misleading, for single pole open loop response there is a low frequency corner determined by the open loop DC gain intercept with the single pole slope, for the same GBW a high DC gain gives a low open loop corner frequency and low DC gain gain gives a high open loop corner frequency - but with less gain than the "low open loop corner frequency" at every frequency below the "high corner frequency"
http://www.diyaudio.com/forums/showthread.php?postid=489927#post489927
"A General Relationship Between Amplifier Parameters, And Its Application to PSRR Improvement"
http://citeseer.ist.psu.edu/cache/p...zcas91.pdf/a-general-relationship-between.pdf
This paper shows that high open loop gain is a necessary condition for high psrr, it also shows some topological ways around this limitation
filtering the input to the amp should have nothing to do with ps noise at the output, psrr is measured with zero input signal
Hi unclejed613,
Thanks for your comment - this is what I will now experiment with.
Hi jcx,
Thanks for your comment aswell - I think I will have to organise my ideas and choose my words a little more precisely in future - but despite my unprecise expression I hope u managed to get the general thrust of my reasoning.
I mention that I have i/p low pass filter because without it one might just be amplifying HF noise present at the i/p which is not that clever if the objective is to reduce HF noise at the o/p !
I will not comment much further on this now because I feel it's time to make a few more broadband amplifier designs, test what I have been talking about and find the answers to the questions I have been asking.
I have seen some designs in spice that do indeed have distortion charactoristics that are flat up to 20khz and now I will see if I can realise this in real life.
Spice, I know, does not show thermal distortions very easily but I usually try to design thermal distortions out - at least on the i/p stage one way or another.
cheers
mike
Thanks for your comment - this is what I will now experiment with.
Hi jcx,
Thanks for your comment aswell - I think I will have to organise my ideas and choose my words a little more precisely in future - but despite my unprecise expression I hope u managed to get the general thrust of my reasoning.
I mention that I have i/p low pass filter because without it one might just be amplifying HF noise present at the i/p which is not that clever if the objective is to reduce HF noise at the o/p !
I will not comment much further on this now because I feel it's time to make a few more broadband amplifier designs, test what I have been talking about and find the answers to the questions I have been asking.
I have seen some designs in spice that do indeed have distortion charactoristics that are flat up to 20khz and now I will see if I can realise this in real life.
Spice, I know, does not show thermal distortions very easily but I usually try to design thermal distortions out - at least on the i/p stage one way or another.
cheers
mike
lineup said:i don't think there is such a thing
as a general optimum -3dB upper bandwidth
it depends on the actual circuitry of this amplifier
it also depends on chosen transistors
very much, I think it depends on how fast the output stage transistors are
===========================================
There is no such general -3dB optimum, that would suit every, or even most, power amplifiers.
However for each amp + load setup, there is a maximal upper frequency, that is possible.
As many amplifiers use similar transistors ( monkey see - monkey do
and the usual topology ( Op-Amp style )
we end up in somewhat same area, for -3dB overall bandwidth:
.... something like 50-200 kHz.
The General Rule is to try to trim and make the amplifier
as fast as possible in conditions at hand
and still do a performance we can be satisfied with.
A too fast amplifier will have more distortion,
than if it is tamed a bit.
There is no good reason to slow down an amplifier,
just for the sake of making it slow.
lineup
unclejed613 said:
lineup
The General Rule is to try to trim and make the amplifier
as fast as possible in conditions at hand
and still do a performance we can be satisfied with.
A too fast amplifier will have more distortion,
than if it is tamed a bit.
There is no good reason to slow down an amplifier,
just for the sake of making it slow.
unclejed613
make your bandwidth as wide as your design will tolerate.
=====================================
Here is your answer, to the 'optimal -3dB'
for every amp / load ... there is a max upper limit
to get best performance within the audioband - your chosen criteria for 'good'
there are 5 factors that will effect what high freq you can use:
1. amplifier circuit type and topology
2. transistor type and individual devices data
3. PCB layout / practical set up of amplifier
4. what load or different loads you will use
5. what kind of music you will be playing/amplifying - the sound source material
regards
lineup
I have started looking at the closed loop bandwidth requirement in a different way and would appreciate folks comments.
Looking at the requirement from a NFB point of view, I care about how much NFB is available at the closed loop corner frequency. Based on Baxendall's work, we know that NFB can add distortion if there is not enough. While we have no audio signals in the upper part of the bandwidth, we do have noise from external interference and power supply. With low NFB, the noise distortions can "foldback" into the audible spectrum. Certainly good psrr helps here (JCX, thanks for the link to the psrr paper).
If the goal is a closed loop 200kHz bandwidth with a gain of 26dB, then I would look for around 66 dB or more of openloop gain at 200kHz (based on Baxandall's 40dB data). If this is not available, then I want to reduce the closed loop gain until it is. This seems to be compatible with what Lineup is telling us as well as JCX's thoughts.
Hope this is clear and not confusing.
Comments?
Bill
Looking at the requirement from a NFB point of view, I care about how much NFB is available at the closed loop corner frequency. Based on Baxendall's work, we know that NFB can add distortion if there is not enough. While we have no audio signals in the upper part of the bandwidth, we do have noise from external interference and power supply. With low NFB, the noise distortions can "foldback" into the audible spectrum. Certainly good psrr helps here (JCX, thanks for the link to the psrr paper).
If the goal is a closed loop 200kHz bandwidth with a gain of 26dB, then I would look for around 66 dB or more of openloop gain at 200kHz (based on Baxandall's 40dB data). If this is not available, then I want to reduce the closed loop gain until it is. This seems to be compatible with what Lineup is telling us as well as JCX's thoughts.
Hope this is clear and not confusing.
Comments?
Bill
Hi Wwood,
26db gain and 200kHz give a gain bandwidth product of 4MHz, certainly achievable.
But to get 60db of feedback at this same frequency is I suspect impossible with current technology and power amp devices.
Taking your first figures of 200kHz and 26db closed loop gain and assuming open loop GBW product of 20MHz results in 14db of feedback. This could be increased using Dr Cherry's nested feedback ideas and adopted (or preceeded) by some others in opamps and power amps.
But how far above that 14db is a reasonable target?
I hope my arithmetic is correct, I'm sure you'll tell me if it isn't.
26db gain and 200kHz give a gain bandwidth product of 4MHz, certainly achievable.
But to get 60db of feedback at this same frequency is I suspect impossible with current technology and power amp devices.
Taking your first figures of 200kHz and 26db closed loop gain and assuming open loop GBW product of 20MHz results in 14db of feedback. This could be increased using Dr Cherry's nested feedback ideas and adopted (or preceeded) by some others in opamps and power amps.
But how far above that 14db is a reasonable target?
I hope my arithmetic is correct, I'm sure you'll tell me if it isn't.
how easy - mikeks perfect little world
for dear mikeks,
there seems to exists only
yes
no
black
white
and
only one perfect -3dB figure:
200 kHz
==============================================
Hurray Everybody!!
if reality was as simple as for small children
where it is yes-no, good-bad and Nothing in between.
No shades of colours.
No exceptions.
Just two rigid sides to anything.
How simple and easy and good such reality would be for all of us.
But how interesting would this be?
How few different experiences and feelings and thing to explore.
There wouldn't be more discussion, than to decide yes-no
and no other category, colours, shade than black-white
to enjoy.
Thank God, our reality is complex
.. that there are a very few case
we can catch in one word
and still be in line with true reality.
And when we answer a categorical Yes or No
we have to add some reservations at hand.
Under this these conditions, but not when then .....
.... and with these important exceptions
How many times, at this www.diyaudio.com forum
haven't we heard wise elderly audio chaps
have to say to somebody
>>>> It is not quite as simple as that <<<<
We can try to Keep amplifiers Simple,
I found this is a good thing,
but EVEN THIS gives still a complexity
that will need modifications and adjustments.
===========================================
To tell what I think is the case, I repeat my answer to this topic
as our friend mikeks did not get my reasoning:
New Year 2007
lineup
mikeks said:
for dear mikeks,
there seems to exists only
yes
no
black
white
and
only one perfect -3dB figure:
200 kHz
==============================================
Hurray
Everybody!!if reality was as simple as for small children
where it is yes-no, good-bad and Nothing in between.
No shades of colours.
No exceptions.
Just two rigid sides to anything.
How simple and easy and good such reality would be for all of us.
But how interesting would this be?
How few different experiences and feelings and thing to explore.
There wouldn't be more discussion, than to decide yes-no
and no other category, colours, shade than black-white
to enjoy.
Thank God, our reality is complex
.. that there are a very few case
we can catch in one word
and still be in line with true reality.
And when we answer a categorical Yes or No
we have to add some reservations at hand.
Under this these conditions, but not when then .....
.... and with these important exceptions
How many times, at this www.diyaudio.com forum
haven't we heard wise elderly audio chaps
have to say to somebody
>>>> It is not quite as simple as that <<<<
We can try to Keep amplifiers Simple,
I found this is a good thing,
but EVEN THIS gives still a complexity
that will need modifications and adjustments.
===========================================
To tell what I think is the case, I repeat my answer to this topic
as our friend mikeks did not get my reasoning:
...... by lineup in post above:
as a general optimum -3dB upper bandwidth
it depends on the actual circuitry of this amplifier
i
===========================================
There is no such general -3dB optimum, that would suit every, or even most, power amplifiers.
However for each amp + load setup, there is a maximal upper frequency, that is possible.
As many amplifiers use similar transistors
( monkey see - monkey do )
and the usual topology ( Op-Amp style )
we end up in somewhat same area .. for bandwidth
The General Rule is to try to trim and make the amplifier
as fast as possible in conditions at hand
and still do a performance we can be satisfied with.
---------------------------
There is no good reason to slow down an amplifier, set some -3dB point,
just for the sake of making it slow.
New Year 2007
lineup
Hi Andrew T.
I was looking for feedback on the idea that Baxandall's work is viable/applicable to setting the closed loop corner frequency. If we have only 14 dB of feedback, then we are in the regime where Banxandall's work suggests that NFB can add distortion. If that happens, then it is possible that "garbage" information around 200kHz can be distorted by the NFB and may fold back into the audible spectrum.
My example was meant to show the open loop gain consequence of choosing 200Khz as the closed loop corner frequency. A conclusion some take away from Baxandall's work is that at least 40dB of feedback is required to be sure that you are reducing distortion and not increasing it. Thus if someone wants a 200kHz closed loop bandwidth and 26db gain, then they need 40dB + 26 dB of open loop gain at that frequency.
Anyway the specifics are not what is important, what is important is the idea that the closed loop corner frequency and the open loop gain at that frequency may have a relation driven by NFB theory.
Does this makes sense?
Bill
I was looking for feedback on the idea that Baxandall's work is viable/applicable to setting the closed loop corner frequency. If we have only 14 dB of feedback, then we are in the regime where Banxandall's work suggests that NFB can add distortion. If that happens, then it is possible that "garbage" information around 200kHz can be distorted by the NFB and may fold back into the audible spectrum.
My example was meant to show the open loop gain consequence of choosing 200Khz as the closed loop corner frequency. A conclusion some take away from Baxandall's work is that at least 40dB of feedback is required to be sure that you are reducing distortion and not increasing it. Thus if someone wants a 200kHz closed loop bandwidth and 26db gain, then they need 40dB + 26 dB of open loop gain at that frequency.
Anyway the specifics are not what is important, what is important is the idea that the closed loop corner frequency and the open loop gain at that frequency may have a relation driven by NFB theory.
Does this makes sense?
Bill
darkfenriz said:Actually in most hi-fi applictions cutting bass around 35Hz can be superior,
because most woofers are happier when not tortured with freqs they cannot cope with (coil heating etc..).
Absolutely no need to cut at 2Hz. And please do not say anything about phase of bass...
.
that is true, fenriz
... and your remark is well in line with my statement
...for every amp / load combination, there is an 'optimal' bandwidth
Actually, while we want each exemplar of amplifier to be
as fast as possible and still perform in a way we like
.... there are some exceptions to this, I can see:
1. Active amplifier systems
for example with: Low - mid - high. Separate amplifiers / woofer/ tweeter loads.
Bass woofer amplifier has other requirements than tweeter hi-freq amplifier.
2. Dual channel stereo amplifiers
Even if, due to transistor / components tolerance variations,
you can make one of these amplifiers with higher -3dB,
you will want BOTH channels to match.
So the -3dB is set to the one that is a bit slower,
when this 'slower one' is tweaked and adjusted for Maximal High -3dB
lineup
I have to add my $0.02 here:
Hearing is 0-15 KHz give or take.
A square wave requires infinate bandwidth.
Most signal sources don't preserve impact or rise time. They can't becuase of bandwidth allocated.
If you try to represent a square wave at 10KhZ you will need aboout 8 fourier terms (sine waves) to get a good representation.
Thus 80-100 KHz is a good number for a rolled off response.
RF filters required.
DC low-pass filters required.
Hearing is 0-15 KHz give or take.
A square wave requires infinate bandwidth.
Most signal sources don't preserve impact or rise time. They can't becuase of bandwidth allocated.
If you try to represent a square wave at 10KhZ you will need aboout 8 fourier terms (sine waves) to get a good representation.
Thus 80-100 KHz is a good number for a rolled off response.
RF filters required.
DC low-pass filters required.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.