Amplifier Design Configurations: So simple yet so powerful aspects of design.....

[Ravindra Desai]

Hello all,

Attached is an impedance graph of a speaker that we normally do not see, yet it is the type of driver we all would like to have in our systems.
Also take a look at the frequency response that extends so nicely to the full end of the audible spectrum.
The special thing about this driver is that the impedance remains almost constant about its nominal value which is not the case, normally.
This is the type of impedance curve that you should look for, for most amplifiers.
Needless to say drivers with such impedance curve do not come cheap.
For more info and examples, look for Scanspeak drivers.

Also, an amp is a voltage source, it will swamp any impedance variation of the load, it must do it, to do it, it has low impedance in its intended bandwidth.

Isn't that contradictory?
A voltage source will follow Ohms law.
For the same voltage, current will be proportional to impedance. So if impedance varies, the current will also vary!
A big thing to understand, always, is that it is the current flowing through the speaker that causes the movement of the cone, not voltage.

However, you helped me make my point that almost all amplifiers in the market are voltage amplifiers.

Thanks to you all for contributing to this thread and taking it forward.

Regards,

Ravindra.

 

[surfatwork]

Hello all,

Attached is an impedance graph of a speaker that we normally do not see, yet it is the type of driver we all would like to have in our systems.
Also take a look at the frequency response that extends so nicely to the full end of the audible spectrum.
The special thing about this driver is that the impedance remains almost constant about its nominal value which is not the case, normally.
This is the type of impedance curve that you should look for, for most amplifiers

Is this a real driver, or a theoretical simulation? If the former, please share which driver.

 

[diyaudio]

Isn't that contradictory?
A voltage source will follow Ohms law.
For the same voltage, current will be proportional to impedance. So if impedance varies, the current will also vary!
A big thing to understand, always, is that it is the current flowing through the speaker that causes the movement of the cone, not voltage.

No, its not contradictory. If impedance is higher at resonance then the driver needs very low current at resonance (which a voltage source will provide). The speaker simulation programs all assume a voltage source and not current source. And no, ohms law is not violated. Try to give a constant current at all frequencies and see the fun, near the resonance of the driver, the driver will easily exceed xmax

 

[Ravindra Desai]

No, its not contradictory. If impedance is higher at resonance then the driver needs very low current at resonance (which a voltage source will provide). The speaker simulation programs all assume a voltage source and not current source. And no, ohms law is not violated. Try to give a constant current at all frequencies and see the fun, near the resonance of the driver, the driver will easily exceed xmax
Some of your statements are misleading.

I know. I need to work on my English. Really.


Let us agree on a few things as we take this forward:
1. With most amplifiers in the market, if you connect any speaker in series, the power output from the amplifier reduces. i.e volume reduces. Right?
2. With most amplifiers in the market, if you connect any speaker in parallel, the power output from the amplifier increases. i.e volume increases. Right?

So can we conclude that amplifier power output is inversely proportional to the speaker impedance? I think we can.

Now, the point I am trying to make is that with 'most' amplifiers in the market and with a 'normal' speaker like Ahuja 12" full range, whose impedance increases as we go to the two extremes of the frequency range, the bass and treble response drops and this is a fact! More often than not, its the mid-range of this speaker that leaves most of us impressed.
However, the same speaker, in the same enclosure, produces wonderful sound when connected to a valve amplifier!
Can we attribute this to the tube signature and harmonics only. The tube still needs to produce the fundamental frequency around which the harmonics will be attached. Something the SS counterpart was not capable of doing.

Here is the fun part:
If we use a speaker whose impedance does not vary much, like the one I referenced in my post # 24, it will sound the same on every amp whether it is normal SS, special SS or tube. The reason being the speaker is now posing like almost a resistive load.

That is the reason why I say that it is the ability of an amplifier to handle variations in impedance that will make it stand apart.

Regards,

Ravindra.

 

[diyaudio]

I would not try to correct flaws of the driver with a special amp, i would use Eq for it.

Many drivers have rising impedance due to dominating inductance and the constant voltage does give constant sound output even with this rising impedance, the driver manufacturer does the matching of rising impedance/constant output as part of the design of the driver.

Also driver impedance flattening is not done for amplifier, it is done for the passive crossover. The crossover does not like variations in impedance. The amplifier is happy with even varying impedance

 

[sound_cycle]

the power output from the amplifier increases. i.e volume increases. Right?

2x power will give you only 1.23 times loudness (ie 3db) increase. 10X power is required for a doubling of perceived loudness (a 10 db change).

As to your other points I will only embarrass myself with a display of my utter ignorance but don't sensitivity, efficiency what the transducer is doing to phase as such like stuff matter, or the fact that amps could be designed to be current or voltage sources, or whether they depend on feedback and so on. Like Ygritte said "I know nothing" (to Jon Snow and not in an audio context, just a confession that I actually know nothing of electronics and am simply a remote button pusher )

ciao
gr

 

[diyaudio]

But, Ravindra, what is the point you are trying to make?
Go on, pls share how you have flattened the impedance

 

[Hari Iyer]

I know. I need to work on my English. Really.


Let us agree on a few things as we take this forward:
1. With most amplifiers in the market, if you connect any speaker in series, the power output from the amplifier reduces. i.e volume reduces. Right?
2. With most amplifiers in the market, if you connect any speaker in parallel, the power output from the amplifier increases. i.e volume increases. Right?

So can we conclude that amplifier power output is inversely proportional to the speaker impedance? I think we can.

Now, the point I am trying to make is that with 'most' amplifiers in the market and with a 'normal' speaker like Ahuja 12" full range, whose impedance increases as we go to the two extremes of the frequency range, the bass and treble response drops and this is a fact! More often than not, its the mid-range of this speaker that leaves most of us impressed.
However, the same speaker, in the same enclosure, produces wonderful sound when connected to a valve amplifier!
Can we attribute this to the tube signature and harmonics only. The tube still needs to produce the fundamental frequency around which the harmonics will be attached. Something the SS counterpart was not capable of doing.

Here is the fun part:
If we use a speaker whose impedance does not vary much, like the one I referenced in my post # 24, it will sound the same on every amp whether it is normal SS, special SS or tube. The reason being the speaker is now posing like almost a resistive load.

That is the reason why I say that it is the ability of an amplifier to handle variations in impedance that will make it stand apart.

Regards,

Ravindra.

You have completely ignored impedance phase in your analysis. The impedance can be inductive or capacitive or even resistive depending on the phase angle. If the capacitive phase angle is above 60 Deg and the impedance is around 2 ohms, it's as good as short ckt for the amplifier.

The high, low and mid frequency from the speaker will depend on what load ( inductive, capacitive or resistive) load is presented to the amp.

 

[Hari Iyer]

But, Ravindra, what is the point you are trying to make?
Go on, pls share how you have flattened the impedance

Rising impedance of the woofer can be flattened by a simple zobel network. Rising Q at resonance can be reduced by using a parallel resistor across speaker terminals to lower Q. Both methods can reduce some gain from your amplifier though as they behave like voltage divider.

 

[Kannan]

This is enlightening discussion from a layman's point of view with some basic understanding of concepts like impedance, inductance etc.

Let me try to understand how it works without going in too technically, and sorry if I make some glaring technical errors, and do please correct my understanding if so.

I guess an ideal situation is a full range speaker with an absolutely flat response with constant impedance of 8 ohms or above for the entire bandwidth.
The needed amplifier power should then be decided upon the needed spl and the speaker efficiency assuming all speakers are 8 ohms and above.

Then I think we can build that magic amplifier which will work with all speakers.

But sadly the speaker designs vary drastically, primarily because of constrains, not in design, but the materials used in making the speakers and external environmental factors like air that offers its own varying resistance and the acoustics of the listening environment.

I am yet to come across full range driver that is compact in size, can play all types of music and is not finicky about amplifier type.
A multi way speaker takes care of many short comings with a well designed crossover. So in effect the amplifier sees the crossover impedance and not the individual speakers, and this impedance tries to be as constant.
This is the biggest advantage of a multi way design.
Active crossovers have big time advantage, but are expensive to implement.

Now one may ask why not a crossover design for a full range to handle the impedance vagaries. Though I don't know why, I guess, it is possibly because you can't correct what a speaker is not physically capable of performance at some frequency ranges with a crossover. There are what are called zobel networks for drivers, but these are more like equing a vagary in a frequency.

In a multi way crossover, the unplayable bandwidth of the speakers is completely removed from the response and it is allowed to play within its comfort zone where it offers the least variable impedance.

Overall, i see crossover in a full range is like employing an equalizer, but in a multi way, keeping the speakers within their playable bandwidth without or with very minimal equalization.

I hope what I say makes sense.

Now I have no idea if an amplifier can be designed to play with even poorly designed crossovers where the impedance fluctuates and still make it sound good.

 

[surfatwork]

Methinks this thread has lost its way....

 

[Kannan]

Methinks this thread has lost its way....

Happens, but in a way good as well as it also gets steered back.

 

[Ravindra Desai]

Hello again all,

I went through your responses from post 29 to 37 and rather than replying to each of those, I'll come to the point of it all:

What do we do with the impedance curve?
We use it to generate an EQ curve. This EQ curve will adjust the output voltage of the amplifier in such a way that the POWER delivered (not voltage, current) into the load is in accordance with the input signal. Thus we eliminate every other variable that contributes to irregularity.

How do we do that?

The analog way:
The general idea is to sense the output voltage (Vsen, voltage across the driver), output current (Isen, voltage actoss a sense resistor resistor in series with the driver/ Rsen) to find the power delivered into the load. Use the power (Vsen x Isen) to correct the feedback loop. This method is generic, driver independent and has the fastest response time as it does the loop gain correction in analog domain and in real time.

The look-up table way:
This will need a signal processor and will be driver specific. Using the impedance curve, we generate an EQ curve for gain vs frequency so that the voltage developed at the amplifier output for the impedance at any given frequency produces constant power. In this method there is no sensing of signals as the characteristics of the driver are previously known and hence the response time is as good as the one in analog method.

Use smart amplifiers:
Look at SSM3525 here: http://www.analog.com/en/products/ssm3525.html
It needs a processor to use the sensed data and process it as per ones need.

Going into the design details is beyond the scope of this thread.
I used the analog method and will soon be trying the one using SSM3525.

The need for this discussion arose after I mentioned why an amplifier that can best handle variation in load impedance stands out most.

I will leave you with some food for thought:
What would be most correct? Correction based on constant POWER curve or flat response (constant amplitude) curve?

Regards,

Ravindra.

 

[IndianEars]

Aniket,

Your offer to start a design thread ... would be great for a few.

Most here, will probably not be comfortable with Ohms Law.

The few Tech boffins (or those that consider themselves so) will probably soon get into a slinging match given their own personal views and opinions....

Another Cable debate but at a higher technical level ?

But its great to see some knowledgeable folks here.... Cheers !

 

[reubensm]

A friend of my father who was into repairing amps had collected over 400 amps and various speakers over the years, but suddenly passed away one fine day when I had relocated from Chennai for work.
I came to know couple years later that the entire stuff was sold to a reseller for pittance by his family who did not have any idea about the value of the gears.

Sorry, little off topic here but I think we all should start discussing the context of an "Audio-Will"

Recently while speaking with my lawyer I brought up the subject and he had no clue about what I was talking about. In the western world it is quite common to get audio equipment professionally valued and included in inheritance.

A 1000 LPs at even Rs.200 a piece is Rs.2L. Turntables, amps, speakers, etc will add on.

 

[diyaudio]

Ravindra, do you have any driver in mind that you are going to try this on?
Constant electrical power may not get constant acoustic power, a lot will depend on the chosen driver.
Can the circuit do baffle correction?

 

[Ravindra Desai]

Ravindra, do you have any driver in mind that you are going to try this on?
Constant electrical power may not get constant acoustic power, a lot will depend on the chosen driver.

I have been using the impedance correction in the analog feedback loop since last 2.5 years and I can confirm that the hardware algorithm is mature.

It has worked well with everything that was thrown at it. My own fullrange speakers, Quad bookshelfs, Wharefdale towers, Paradigm speakers and many more. Sorry I don't remember the model # of any. My friend, who is a system integrator, gets a whole lot of different speakers and once in a while we do tests by switching the amps. My amp has thrown sonically good results every single time. I don't mean to brag, but its my experience so far.
I have done A/B comparison with Denon, Marantz, Atoll and few other amplifiers. It produces a noticeable difference, in a good way, with each one of them.

I have not done an A/B comparison with any valve amplifiers so far.

I have not done frequency response mapping with any of them except during prototype testing to ensure that the method works as expected and tracks well with theoretical test plans.

I will soon be carrying my amplifier and my processor to @tuff again.
The last time I had carried only 20Wrms amplifier and it was not able to drive his Magnepans sufficiently.
Even then, it did not trip and this was something he noted particularly since the speakers are known, world wide, to be fussy with what drives them

This time I will be carrying 160 + 160 Wrms stereo amplifier.
The reason I want to test it with Magnepans is because its not a conventional speaker and so piques me.
I know it will work but where does it stand against Crown XLS1502 that he owns is of interest to us.
What difference does my processor bring in against his Peachtree is also something that we will be looking into.
Also, where would he place my processor- amplifier combo on the quality ladder matters to me and interests me.

The above paragraphs don't mean a lot to most of you. I know that.
The real reason why I got into this is because I wanted to make an amplifier that would work wonders with cheap (cost wise) speakers. This was the only way I could keep the system cost down.

In then end, constant power EQ graph was my means of getting to a point where I can connect ANY speaker, turn the power switch ON, and adjust the volume to listen without making ANY adjustments.

I hope that you get the picture now.

Regards,

Ravindra.


P.S. To your question after the edit: There is only electrical feedback and no acoustic feedback so, no, it cannot do baffle correction.

 

[vktarun]

I have been using the impedance correction in the analog feedback loop since last 2.5 years and I can confirm that the hardware algorithm is mature.

It has worked well with everything that was thrown at it. My own fullrange speakers, Quad bookshelfs, Wharefdale towers, Paradigm speakers and many more. Sorry I don't remember the model # of any. My friend, who is a system integrator, gets a whole lot of different speakers and once in a while we do tests by switching the amps. My amp has thrown sonically good results every single time. I don't mean to brag, but its my experience so far.
I have done A/B comparison with Denon, Marantz, Atoll and few other amplifiers. It produces a noticeable difference, in a good way, with each one of them.

I have not done an A/B comparison with any valve amplifiers so far.

I have not done frequency response mapping with any of them except during prototype testing to ensure that the method works as expected and tracks well with theoretical test plans.

I will soon be carrying my amplifier and my processor to @tuff again.
The last time I had carried only 20Wrms amplifier and it was not able to drive his Magnepans sufficiently.
Even then, it did not trip and this was something he noted particularly since the speakers are known, world wide, to be fussy with what drives them

This time I will be carrying 160 + 160 Wrms stereo amplifier.
The reason I want to test it with Magnepans is because its not a conventional speaker and so piques me.
I know it will work but where does it stand against Crown XLS1502 that he owns is of interest to us.
What difference does my processor bring in against his Peachtree is also something that we will be looking into.
Also, where would he place my processor- amplifier combo on the quality ladder matters to me and interests me.

The above paragraphs don't mean a lot to most of you. I know that.
The real reason why I got into this is because I wanted to make an amplifier that would work wonders with cheap (cost wise) speakers. This was the only way I could keep the system cost down.

In then end, constant power EQ graph was my means of getting to a point where I can connect ANY speaker, turn the power switch ON, and adjust the volume to listen without making ANY adjustments.

I hope that you get the picture now.

Regards,

Ravindra.


P.S. To your question after the edit: There is only electrical feedback and no acoustic feedback so, no, it cannot do baffle correction.

Hi Ravindra,
Very innovative way of auto - correction, but is there a risk in this method that we end up with a sound signature which is altogether different from the recording ( What the recording engineer wants us to listen). The question is how much correction and how your system determines the same?

Regards,
Vipin

 

[vktarun]

Do you first measure the impedence curve of the speaker - amplifier combo, feed it into your system to generate the correction algorithm, is it like that?

This was in my mind in college days but never got a chance to experiment.

Regards,
Vipin

 

[Ravindra Desai]

Hello Vipin,

Thank you for your interest in this topic.

Hi Ravindra,
Very innovative way of auto - correction, but is there a risk in this method that we end up with a sound signature which is altogether different from the recording ( What the recording engineer wants us to listen). The question is how much correction and how your system determines the same?

Regards,
Vipin

The method itself is not new to the industry and so don't look at it as unique or something that I am proposing for the first time. Companies just don't talk about the approach they take to achieve certain quality improvements.
Generally electronics has frequency response specified as flat within Adb from XHz to YHz.
This in itself is meaningless if your load is not constant.
Additional measures are required to ensure flat power delivery into load from Pohms to Qohms.

If the power delivered depends on input signal and only input signal, the electronics will now deliver consistent quality and sound signature with a wide variety of speaker systems. This is not the case today.

Coming to your concern:
The correction method is not adding any unique signature not is altering the signal recorded. It is only ensuring that the irregularities in the loudspeaker impedance is taken care.
Practically speaking: Amitabh Bacchan will sound exactly like him and Lata Mangeshkar will sound exactly like her.

Do you first measure the impedence curve of the speaker - amplifier combo, feed it into your system to generate the correction algorithm, is it like that?

This was in my mind in college days but never got a chance to experiment.

Regards,
Vipin

The problem can be solved in many ways. Three of which are mentioned in my post # 38.
Going into design details is beyond the scope of the thread. I hope that you will understand.

Regards,

Ravindra.