I think 70dB CMRR for a main amp is already impressive and as you say it can be increased. You will need a 10 turn pot and a ceramic trimmer to take it to a higher level like 90dB by trimming out all kind of side parasites. Look at expensive amps from Nelson Pass or Mark Levinson, although having balanced inputs, nobody specifies CMRR.
You haven't yet figured out what you really want, a Current Amp or a Voltage Amp with a series resistor, more like a Tube Amp or some Hybrid. Not yet having made up your mind is not a problem at all, but commenting at the start of this process on things like CMRR, protection for Amp and Drivers and protection against EMI only retard in finding the solutions you are looking for.
Then another such thing that you mentioned, too much input noise. At this point not yet an important issue. But looking again at these very expensive amps, they all seem to hover between 100uV and 200uV output noise between 20Hz and 20Khz.. The circuit in #53 equipped with the OPA1652 for U1,U4 and U5 and with the given resistor values, has 49.6uV output noise over a 8 Ohm load for this same FR range. So what are we talking about, this is a non issue.
Coming back to your question to control U3 with the input signal. I do not see what benefit this could bring over the given proposal. What you really want is to have equal but opposite voltages from U2 and U3 to get the most Power out of this bridged Amp. When offering a different signal to U3, there is no longer an even distribution in output voltage between U2 and U3 resulting in a lesser maximum Power where either U2 or U3 reaches it limits before the other one does.
Have you thought about placing a resistor in parallel to the Driver. With a resistor having 10 times the value of the driver, you will have 90% current drive and 10% voltage drive in your wording. Maximum power to the driver will drop to 81%, but at the same time you have solved the high output voltage problem with no Driver connected or a driver with a cap in series. For an 8 Ohm speaker, output impedance will drop to 80 Ohm.
Hans
Not sure who you were asking, but I vaguely recall something around 200 ohms. I need to rebuild it and make some measurements.
A line level Linkwitz transform can easily resolve the overly feared and hyped low frequency resonance. Cascaded linear filters work.
Ok, I have two possibly interesting solutions for regulating current drive level and frequency response:
1: DSP
A processor dedicated for high speed, I.E. a video processor or an FPGA, and set it up to automatically respond to increased gain. If the gain is high in a certain range, it can mix this area with more voltage control using filters.
2: Analog
A sensing circuit has a "setup" mode. In this mode the voltage gain limit is lowered by 6dB. The circuit moves the amplifier in the direction of voltage drive in steps until the voltage gain is managable. When "setup mode" is turned off, this circuit is a safety circuit. We should also have the ability to override and set the current drive level manually.
1: DSP
A processor dedicated for high speed, I.E. a video processor or an FPGA, and set it up to automatically respond to increased gain. If the gain is high in a certain range, it can mix this area with more voltage control using filters.
2: Analog
A sensing circuit has a "setup" mode. In this mode the voltage gain limit is lowered by 6dB. The circuit moves the amplifier in the direction of voltage drive in steps until the voltage gain is managable. When "setup mode" is turned off, this circuit is a safety circuit. We should also have the ability to override and set the current drive level manually.
With all respect, what is the problem that those solutions are going to solve.
It’s better to first make a clear definition of the problem before thinking of possible solutions.
You seem to be unhappy with a pure current amp because of 1), 2), 3) etc.
And you prefer some automated set up process because of 1), 2), 3) etc.
Hans
It’s better to first make a clear definition of the problem before thinking of possible solutions.
You seem to be unhappy with a pure current amp because of 1), 2), 3) etc.
And you prefer some automated set up process because of 1), 2), 3) etc.
Hans
I thought that was fairly well covered already. But let me clearify:
The problem:
When you run current drive, the gain is constant for a certain resistance. A change in impedance of 10/1 will cause a change in voltage gain of 10/1 as well, and that is not uncommon. If you have a typical series capacitor, the impedance will rise far more than that, resulting in a increaseing voltage gain of maybe 40-80dB, or even more. That is an extremely significant reduction in headroom, and the amplifier will go straight into clipping even at moderate signal levels.
Practical approach:
In some situations, the amplifier will be connected to a driver where the impedance is smooth, and the filtering has been done properly. In these cases, a 100% current drive is no problem.
In other situations, the driver has huge impedance variations. In other situations, the filtering has been put in series with the driver, while the driver is being connected and signal is added. This will then lead to a huge increase in voltage gain at certain frequencies. Many drivers are brittle and rather expensive, and the user runs a huge risk of damage if the signal is suddenly out of control.
In a test setup, this is trivial and managable. But when things are left to others, things are a bit more out of control.
The solution:
So, if we make an amplifier that effectively has infinite output impedance, the gain will increase according to the load impedance. If the output impedance is set to twice the nominal load, the voltage gain will increase with maximum around 6dB. If the situation is something in between, the output impedance could be set to for example 4, 8 or 16 times the nominal load impedance, leading to varying sensitivity to loads.
The safety circuit could monitor the input signal and compare it to the output voltage. If the voltage gain is too high, it is a sure sign that the impedance is above the safe limit. With the input signal and output voltage the system can easily determine the safe setting. It could also be set to ignore signals below a certain treshold. That way, the system will also be able to handle situations where the signal is cut digitally prior to the amplifier. But I still think it would be very useful to determine errors in passive filtering. If the system is set to high current drive, and the filtering is not done properly, the right thing to do would be to decrease the output impedance to a level where the amplifier does not sustain the current through the driver outside the passband of the filter, and the voltage gain of the amplifier is kept from rising above a certain level.
As a product:
So the amplifier will step down its output impedance when needed. This is actually a useful tool for the user to determine if the setup has been designed correctly for current drive. If it is not done correctly, the amp will respond by stepping down its output impedance to a more managable level.
If the driver is disconnected, the amplifier will immediately step down is output impedance so the output voltage is kept under control. If the output impedance is set to "fixed", the amplifier will regain its output impedance as soon as the conditions are back to normal again.
The problem:
When you run current drive, the gain is constant for a certain resistance. A change in impedance of 10/1 will cause a change in voltage gain of 10/1 as well, and that is not uncommon. If you have a typical series capacitor, the impedance will rise far more than that, resulting in a increaseing voltage gain of maybe 40-80dB, or even more. That is an extremely significant reduction in headroom, and the amplifier will go straight into clipping even at moderate signal levels.
Practical approach:
In some situations, the amplifier will be connected to a driver where the impedance is smooth, and the filtering has been done properly. In these cases, a 100% current drive is no problem.
In other situations, the driver has huge impedance variations. In other situations, the filtering has been put in series with the driver, while the driver is being connected and signal is added. This will then lead to a huge increase in voltage gain at certain frequencies. Many drivers are brittle and rather expensive, and the user runs a huge risk of damage if the signal is suddenly out of control.
In a test setup, this is trivial and managable. But when things are left to others, things are a bit more out of control.
The solution:
So, if we make an amplifier that effectively has infinite output impedance, the gain will increase according to the load impedance. If the output impedance is set to twice the nominal load, the voltage gain will increase with maximum around 6dB. If the situation is something in between, the output impedance could be set to for example 4, 8 or 16 times the nominal load impedance, leading to varying sensitivity to loads.
The safety circuit could monitor the input signal and compare it to the output voltage. If the voltage gain is too high, it is a sure sign that the impedance is above the safe limit. With the input signal and output voltage the system can easily determine the safe setting. It could also be set to ignore signals below a certain treshold. That way, the system will also be able to handle situations where the signal is cut digitally prior to the amplifier. But I still think it would be very useful to determine errors in passive filtering. If the system is set to high current drive, and the filtering is not done properly, the right thing to do would be to decrease the output impedance to a level where the amplifier does not sustain the current through the driver outside the passband of the filter, and the voltage gain of the amplifier is kept from rising above a certain level.
As a product:
So the amplifier will step down its output impedance when needed. This is actually a useful tool for the user to determine if the setup has been designed correctly for current drive. If it is not done correctly, the amp will respond by stepping down its output impedance to a more managable level.
If the driver is disconnected, the amplifier will immediately step down is output impedance so the output voltage is kept under control. If the output impedance is set to "fixed", the amplifier will regain its output impedance as soon as the conditions are back to normal again.
Last edited:
Hi Snickers,
There is very good reason why voltage amps are so popular. When properly designed, you can connect almost anything.
But as you stated in the beginning, driving a speaker with current is more correct, since current is what makes the speaker move.
Doing so extends the FR range and avoids several types of anomalies.
BUT, current drive has also it's problems.
It will never be as universal applicable as voltage drive, some of the problems depicted above.
That's where I would like to respond. It seems that you are looking for a sort of Swiss Army Knife of current amp, a jack of all trades.
Well that will be very hard to achieve, and the question is, who will need such an amp loaded with features, taking care of any possible sort of load or the lack of it.
To me it sounds like a fata morgana unless you keep sticking to a voltage amp with a series resistance of 8 Ohm !
So my advise would be make one step back and follow a more practical approach and define the borders of the application.
Reading Esa's book "Current-Driving of Speakers" he proposes a Current Drive Index or CDI defined as |Zt|/|Zl.
Zt being the output impedance of the amp and Zl the speaker impedance.
A CDI of 10 is supposed to be as optimum, 5 very satisfactory and even 1 a step forward, possibly explaining why Tube amps with their high output impedance are sounding so nice and warm.
If we take this as guideline, a par. load to the driver ten time its value, may already be enough, giving a CDI close to 10 for most of the FR range when driven above Fres.
Having no speaker connected prevents at the same time the amp from breaking the feedback loop.
And all the frequency shaping and crossover filtering can best be done electronically between preamp and main amp.
The only thing that could be added is some speaker protection system, although this is also absent in speaker systems that are voltage driven.
But with a current driven amp this is easier to implement since nothing damaging happens when the protection system presents a short circuit in protection mode.
Hans
There is very good reason why voltage amps are so popular. When properly designed, you can connect almost anything.
But as you stated in the beginning, driving a speaker with current is more correct, since current is what makes the speaker move.
Doing so extends the FR range and avoids several types of anomalies.
BUT, current drive has also it's problems.
It will never be as universal applicable as voltage drive, some of the problems depicted above.
That's where I would like to respond. It seems that you are looking for a sort of Swiss Army Knife of current amp, a jack of all trades.
Well that will be very hard to achieve, and the question is, who will need such an amp loaded with features, taking care of any possible sort of load or the lack of it.
To me it sounds like a fata morgana unless you keep sticking to a voltage amp with a series resistance of 8 Ohm !
So my advise would be make one step back and follow a more practical approach and define the borders of the application.
Reading Esa's book "Current-Driving of Speakers" he proposes a Current Drive Index or CDI defined as |Zt|/|Zl.
Zt being the output impedance of the amp and Zl the speaker impedance.
A CDI of 10 is supposed to be as optimum, 5 very satisfactory and even 1 a step forward, possibly explaining why Tube amps with their high output impedance are sounding so nice and warm.
If we take this as guideline, a par. load to the driver ten time its value, may already be enough, giving a CDI close to 10 for most of the FR range when driven above Fres.
Having no speaker connected prevents at the same time the amp from breaking the feedback loop.
And all the frequency shaping and crossover filtering can best be done electronically between preamp and main amp.
The only thing that could be added is some speaker protection system, although this is also absent in speaker systems that are voltage driven.
But with a current driven amp this is easier to implement since nothing damaging happens when the protection system presents a short circuit in protection mode.
Hans
What are you worried about Hans? Why can't the amp be a bit more advanced and still be an attractive amp? I think that makes it more attractive... I also think it is cool to actually offer something really close to pure current drive. There are already lots of amplifiers that offer some output impedance. I want to turn it to 11.
I'm thinking the design criteria aren't really clear yet.
Is the design you're talking about a general amplifier design for all speakers (passive speakers), or is it specifically designed to be used as the amp that drives 1 driver only (active approach).
If it's passive then the whole talk about capacitors that protect expensive compression drivers can be thrown out of the window, if it's active then one could always place a big inductor parallel to the output/driver which would only work well when there has been a stage where active filtering took place, right?
Looking at it from a different angle:
This problem with high load impedances combined with unknown speaker sensitivity and unknown demands of the volume that needs to be reached by the customer (that's it in a nutshell isn't it), is easily solvable if opamps were not the basic ingredient of the output stage of the amplifier. If you replace those with basic solid state you could start designing and only care e.g. about what range of, or adjustability of output impedance you would need for the range of speakers it should be able to drive and be very happy with modifying the impedance to get sonically satisfiable results.
In short: use opamps where you need them(filtering and difference amps), solid state to not have those headroom problems that make the current amp a great sounding solution but for only a handfull of speakers.
Btw, iirc, "the Book" talks about modifying the physical damping of the bass driver as sometimes needed. This makes a general amplifier approach also provlematic without alteration of the driven speaker. I think the idea in general is that current drive should be seen as a total system approach. Damping of the bass driver, changes in x-over are all needed to get the really stunning results. That would mean building an active speaker, not a jack of all trades.
If you want to make a jack of all trades, just take a high power amp and put a variable resistor in series to the speaker. Any L-pad would just work fine there (pins 2&3, about 0-40 ohms). Just for testing what kind of maximum output impedance you could tolerate for your own speakers and what level of current you have to deliver. See if an opamp can take you there.
P.S. I drive 15" and 18" woofers with big horns and they do great with an amplifier impedance of a few ohms. The bass doesn't suffer when properly designed and the integration of horn and woofer is much easier than with a voltage amp. The whole systems gets a lot more flow, transparency etc. It's hard to say what part is caused by the current approach and which by changes in level. I don't know the output impedance exactly, but it doesn't allow for e.g. Tannoy dual concentrics, just to name a few. So already with a few ohms the amp gets "niche status" very quickly. In this case the main problem is found in the mid frequencies where the system just gets too loud and current drive isn't really current drive anymore when you add parallel resistors to mitigate, right? But it almost makes it a jack of all trades;-)
Is the design you're talking about a general amplifier design for all speakers (passive speakers), or is it specifically designed to be used as the amp that drives 1 driver only (active approach).
If it's passive then the whole talk about capacitors that protect expensive compression drivers can be thrown out of the window, if it's active then one could always place a big inductor parallel to the output/driver which would only work well when there has been a stage where active filtering took place, right?
Looking at it from a different angle:
This problem with high load impedances combined with unknown speaker sensitivity and unknown demands of the volume that needs to be reached by the customer (that's it in a nutshell isn't it), is easily solvable if opamps were not the basic ingredient of the output stage of the amplifier. If you replace those with basic solid state you could start designing and only care e.g. about what range of, or adjustability of output impedance you would need for the range of speakers it should be able to drive and be very happy with modifying the impedance to get sonically satisfiable results.
In short: use opamps where you need them(filtering and difference amps), solid state to not have those headroom problems that make the current amp a great sounding solution but for only a handfull of speakers.
Btw, iirc, "the Book" talks about modifying the physical damping of the bass driver as sometimes needed. This makes a general amplifier approach also provlematic without alteration of the driven speaker. I think the idea in general is that current drive should be seen as a total system approach. Damping of the bass driver, changes in x-over are all needed to get the really stunning results. That would mean building an active speaker, not a jack of all trades.
If you want to make a jack of all trades, just take a high power amp and put a variable resistor in series to the speaker. Any L-pad would just work fine there (pins 2&3, about 0-40 ohms). Just for testing what kind of maximum output impedance you could tolerate for your own speakers and what level of current you have to deliver. See if an opamp can take you there.
P.S. I drive 15" and 18" woofers with big horns and they do great with an amplifier impedance of a few ohms. The bass doesn't suffer when properly designed and the integration of horn and woofer is much easier than with a voltage amp. The whole systems gets a lot more flow, transparency etc. It's hard to say what part is caused by the current approach and which by changes in level. I don't know the output impedance exactly, but it doesn't allow for e.g. Tannoy dual concentrics, just to name a few. So already with a few ohms the amp gets "niche status" very quickly. In this case the main problem is found in the mid frequencies where the system just gets too loud and current drive isn't really current drive anymore when you add parallel resistors to mitigate, right? But it almost makes it a jack of all trades;-)
check out post #615/616 under
https://www.diyaudio.com/forums/solid-state/250272-current-drive-loudspeakers-62.html#post6707142
https://www.diyaudio.com/forums/solid-state/250272-current-drive-loudspeakers-62.html#post6707142
Current drive is good for protecting speaker drivers. Heat is generated by current not by voltage. So controling current is a nice thing even if voltage swing is high. I never allow to have clipping for any type of drivers. So adding clip protection for my current drive amps. So i can’t play clipped audio even if i want. Clipping is a fault condition and can’t be allowed in any circumstances if we want pure sound
Heat is generated by both, not one or the other. P=U×I. Only when R=0 your statement holds true ;-)
And regarding clipping and playing clipped audio, how would you repair music files, or is it just the amplifier is prohibited from clipping(different things)?
Most speakers are electrodynamic, a coil in a magnetic field. That means that cone movement produces a voltage, ie "back emf". The current determines the force causing the cone movement, but the impedance of the that movement is not constant so a constant current results in a frequency response analogous to the impedance, which is nothing like flat. Some speaker resonances are "parallel" impedance peaks, and some are "series" impedance dips. These are best driven by voltage for impedance peaks and current for impedance dips. I general, woofers are best controlled with a constant voltage and rolled off below the impedance dips. Mids and tweeters are best driven with a compromise, usually around the driver rated impedance.
It is not difficult to implement whatever impedance you like by combining negative and positive feedback and a current sensing resistance. But since you usually need to attenuate the mids and highs anyway, it's best to put impedance increasing resistance from the attenuators in the speaker cabinet.
Sure once made a small PA system that used a current sensing transformer to increase the amplifier impedance at high frequencies. But users would mix the amplifier with other speakers that did not match, resulting in poor results. In such cases, I removed the current feedback, many years ago.
There is a feedback circuit that produces an output impedance that is defined at Nx the build-out resistor, which was very useful for modems and telephony. I would post it hear but it was years ago that I used it and I don't want to get it wrong. I'll be looking for it in my archives and perhaps post it later.
It is not difficult to implement whatever impedance you like by combining negative and positive feedback and a current sensing resistance. But since you usually need to attenuate the mids and highs anyway, it's best to put impedance increasing resistance from the attenuators in the speaker cabinet.
Sure once made a small PA system that used a current sensing transformer to increase the amplifier impedance at high frequencies. But users would mix the amplifier with other speakers that did not match, resulting in poor results. In such cases, I removed the current feedback, many years ago.
There is a feedback circuit that produces an output impedance that is defined at Nx the build-out resistor, which was very useful for modems and telephony. I would post it hear but it was years ago that I used it and I don't want to get it wrong. I'll be looking for it in my archives and perhaps post it later.
Here's a paper from TI which covers synthesizing the output impedance of an ADSL driver, aka 'active termination' : https://www.ti.com/lit/an/sloa100/sloa100.pdf
P=UxI corect. Just my thinking aproach is diferent. I think in P=IxIxR. Where R is resistive element of the coil and where the heat is produced. R is roughly constant where impedance is not. So I is good aproximation of heat.
Except I is a function or impedance, making P for the same speaker and SPL in the R-part of the impedance bigger outside the R-part.
It doesn't matter, it's just P=I×I×R doesn't do what your thinking approach assumes it means.
Yes the talk is about nothing i know🙂 But if you now voltage and driver resistance you don’t know how much heat it produces. You have to measure or calculate current. The heat kills the driver. And heat is proportional to current. With voltage you have to consider the impedance. Current is more direct indicator