Hi all. My first post here is prompted by a thought about a way to reduce the dissipation associated with Class A amplification. I have built several small ZEN-like amplifiers (another debt of gratitude owed to Nelson Pass) but I always seem to fall short of a ending up with a "daily driver" for my listening habits. What I mean by this is that in one way or another the excess power dissipation becomes a problem - I think most people know what I mean by this so I won't elaborate.
Anyhow, after searching here and there I can find nothing that directly addresses the concept of using a switch-mode constant current source in place of the linear variations (resistor/transistor/inductor/light-bulb!). On offer seems to be a way to almost halve the usual dissipation - a very significant reduction indeed!
This can't really be called Class D, although granted there may be similar issues. It's not to easy to search for either (so I apologise if I have missed it - searching for A/D isn't much help!).
Anyway, as an experiment I substituted a 2Amp switching current source into the original ZEN constant current source design and sure enough the results seem promising. The switching noise (around 500KHz in my case) would seem to be fairly simple to remove using an inductor between the current source and the lower transistor (a much smaller and more practical inductor to the sort used as the sole current source) but no doubt this is where people will be focussing their criticism of the idea.
I'm interested to hear the criticism as I can't measure (or hear!) any downside to this strategy - although my measuring (and hearing) equipment will probably not be up to the standards around here. I'm almost certainly missing something - but what is it most likely to be?
Anyhow, after searching here and there I can find nothing that directly addresses the concept of using a switch-mode constant current source in place of the linear variations (resistor/transistor/inductor/light-bulb!). On offer seems to be a way to almost halve the usual dissipation - a very significant reduction indeed!
This can't really be called Class D, although granted there may be similar issues. It's not to easy to search for either (so I apologise if I have missed it - searching for A/D isn't much help!).
Anyway, as an experiment I substituted a 2Amp switching current source into the original ZEN constant current source design and sure enough the results seem promising. The switching noise (around 500KHz in my case) would seem to be fairly simple to remove using an inductor between the current source and the lower transistor (a much smaller and more practical inductor to the sort used as the sole current source) but no doubt this is where people will be focussing their criticism of the idea.
I'm interested to hear the criticism as I can't measure (or hear!) any downside to this strategy - although my measuring (and hearing) equipment will probably not be up to the standards around here. I'm almost certainly missing something - but what is it most likely to be?
Still thinking about this.... Here's a photo of a breadboarded "proof of concept". The switched current source is a separate module (bottom left). Its built around a mosfet & driver with a 0.1 Ohm current sense resistor feeding a diff amp & comparator. The comparator trips as the voltage on the 0.1 Ohm ramps up and down by 10mv or so. Choosing the inductor value vs. current required I can keep it switching in the 500K~1Mhz range.
Above the current source is a separate module with a mosfet follower biased to mid rail. That's fed from a "Bride of Zen" preamp above. A 10uH inductor in series with the output sees to removing most of the HF ripple from the loudspeaker signal. Whatever remains that the loudspeaker sees - I can't hear. It sounds no different to a linear equivalent current source - but dissipates only a watt or so in the mosfet/diode/inductor.
Above the current source is a separate module with a mosfet follower biased to mid rail. That's fed from a "Bride of Zen" preamp above. A 10uH inductor in series with the output sees to removing most of the HF ripple from the loudspeaker signal. Whatever remains that the loudspeaker sees - I can't hear. It sounds no different to a linear equivalent current source - but dissipates only a watt or so in the mosfet/diode/inductor.
What I wanted to say is you cannot just swap-in any switching circuit in place of a negative voltage-referenced constant current source. Just to remind bascis.
You must aim to have zero avarage voltage drop on it, so you need to use both supplies for that. This makes the constant current source design as complicated as a class D amplifier with output impedance forced to be high by clever feedback. The schematic/photo/idea picture/whatever is still missing...
You must aim to have zero avarage voltage drop on it, so you need to use both supplies for that. This makes the constant current source design as complicated as a class D amplifier with output impedance forced to be high by clever feedback. The schematic/photo/idea picture/whatever is still missing...
Oops - thanks stinius!
2 Amps is all my test PSU can deliver - 4A should be a breeze with the same mosfet & diode. The efficiency is somewhere around 90% so nearly all the dissipation takes place in the follower - and this should average out to (supply/2)*I bias.
I sort of trust myself building a switching current source alot more than going the whole hog with class D (getting linear PWM ramps etc.) The goal here is *constant* current and the self-oscillating section is really simple (both to make and keep an eye on!).
An externally hosted image should be here but it was not working when we last tested it.
2 Amps is all my test PSU can deliver - 4A should be a breeze with the same mosfet & diode. The efficiency is somewhere around 90% so nearly all the dissipation takes place in the follower - and this should average out to (supply/2)*I bias.
I sort of trust myself building a switching current source alot more than going the whole hog with class D (getting linear PWM ramps etc.) The goal here is *constant* current and the self-oscillating section is really simple (both to make and keep an eye on!).
darkfenriz said:
Hello darkfenriz! Sure, but only if you're using non-reactive components (e.g. resistors, transistors) but the circuit I'm referring to uses an inductor and transistor to connect the inductor between the source of the follower stage and -v (where a constant current circuit goes in a conventional single-ended Class A topology).
The inductor takes time for the current to build when it is "in circuit" (storing power as a magnetic field). The transistor switch keeps on connecting and disconnecting the inductor to maintain a controlled current about the set-point. The inductor returns "unused power" via diode at the open switch to the +v rail. Using this arrangement there is (theoretically) no power loss in either the inductor or switching transistor, yet there is a DC current flowing across a DC voltage!
This is the basic principle behind switch mode power supplies (and Class D amplifier topologies).
I've just sketched it out roughly. I'm so lazy, I should have started with a diagram!
(EDIT: Due to being under moderation as a new member these latest posts of mine were delayed in appearing. They should still address the points raised subsequently though - thanks everyone for the feedback so far!)
An externally hosted image should be here but it was not working when we last tested it.
(EDIT: Due to being under moderation as a new member these latest posts of mine were delayed in appearing. They should still address the points raised subsequently though - thanks everyone for the feedback so far!)
Hi,
is the switched current sink ON for half the time?
Does the inductor smooth the switched current to effectively half the peak current?
Does the amplifying transistor see an effective constant current load ~=average smoothed current through the inductor?
Does any of that mean that the average bias has been reduced to ~50% of the peak bias?
Is it simpler to just turn down the bias to half? i.e. quarter (-6dB) of the maximum ClassA output power.
is the switched current sink ON for half the time?
Does the inductor smooth the switched current to effectively half the peak current?
Does the amplifying transistor see an effective constant current load ~=average smoothed current through the inductor?
Does any of that mean that the average bias has been reduced to ~50% of the peak bias?
Is it simpler to just turn down the bias to half? i.e. quarter (-6dB) of the maximum ClassA output power.
kenpeter said:
Not sure what you mean here (but it sounds intriguing)... equal impedance to what?
I (simply) see everything below the junction between the follower source and load as a "black box" constant current sink. Its apparent DC impedance varies as a direct function of the follower voltage divided by the set current.
On the subject of "free power" there can, of course, be no such thing! As you know we're only re-distributing it here. The 2 Amps on the meter above is what's being drawn from the PSU. The current in the sense resistor (i.e. the actual bias current in the Class A stage is 1.9 x as much (1 + the efficiency factor of the switcher) thanks to the diode pumping the unwanted potential back into the reservoir capacitors of the PSU (albeit that it must be re-introduced carefully to avoid noise issues).
So on my breadboard setup shown above there's actually a fairly repectable 3.8 Amps bias flowing through the follower stage.
AndrewT said:
No, hopefully my previous post will make it clearer how the efficiency is gained. What's happening in broad terms is that the unwanted energy that is normally lost in a a resistive element entirely as heat is temporarily accumulated in the inductor and then released back into the supply reservoir capacitor(s).
I have to say here and now that the principles are absolutely sound ~ it's not a perpetual motion device 
The fundamental principles are the same as those used in Class D amplification but their application is different. I really like single ended Class A. I like the simplicity of a ZEN style two-transistor amp (voltage gain in pre-amp, current gain in power follower). But the heat kills it for me.
Class D amps are complicated beasts that I doubt could be assembled on a breadboard - but keeping the switching out of the signal path and using it to manage the unwanted energy in the current sink is proving to be reasonably straight forward and is delivering a massive saving in energy.
The fundamental principles are the same as those used in Class D amplification but their application is different. I really like single ended Class A. I like the simplicity of a ZEN style two-transistor amp (voltage gain in pre-amp, current gain in power follower). But the heat kills it for me.
Class D amps are complicated beasts that I doubt could be assembled on a breadboard - but keeping the switching out of the signal path and using it to manage the unwanted energy in the current sink is proving to be reasonably straight forward and is delivering a massive saving in energy.
Wavebourn said:
Yes, although there aren't many (at the moment) that deliver more than a couple of Amps. Also they tend to operate at a fixed frequency ~500kHz with variable mark-space. This increases the peak-to-peak ripple voltage on the inductor which the self-oscillating design above keeps symmetric. Instead, the frequency shifts around with the audio frequency of the follower.
Fortunately the switching current is relatively low in the world of switch-mode designs so a small-ish transistor with low gate charge can be selected and the switching frequency can be kept high. The inductor I've used is also a lot higher in value (220uH) than most switchers so the ripple is kept small. I've gone for a toroidal with powdered core to store as much energy as possible while keeping stray field to a minimum. I think this is the best strategy.
The differential amp has to be high speed and have an excellent CMRR to handle the rail-to-rail swing of the sense voltage. I used an old OP37 from the junk box with a gain of 10 and it works fine.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.