"Power Tracking" in supply, ever heard, DIY ideas?

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hi!

I just saw this amplifier module
http://www.lautsprechershop.de/hifi/det500.htm]RCM DT500[/URL]
or
http://www.lautsprechershop.de/hifi/index.htm?http://www.lautsprechershop.de/hifi/bau_rave.htm#raveland]PDF about amp[/URL]

, and the site / shop stated that by using a "Power-Tracking" supply, this amp was capable of reaching Class D efficiency in standard class AB topology (and if I take a look at small size of the heatsink, and if the output power of 500W RMS is true, what surely must be the case).

I find that very interesting, does anybody know if the DIY community has something to offer in this regards?

Bye,

Arndt
 
Re: "Power Tracking" in supply, ever heard, DIY ideas?

Cradle22 said:
Hi!

I just saw this amplifier module
...URLs deleted...

, and the site / shop stated that by using a "Power-Tracking" supply, this amp was capable of reaching Class D efficiency in standard class AB topology (and if I take a look at small size of the heatsink, and if the output power of 500W RMS is true, what surely must be the case).

Arndt -

The idea behind this - and it is a theoretically sound one, if not practically so! - is that the voltage of the output stage rails is varied in proportion to the signal input to maintain the minimum amount of Vce across the output stage transistors (or Vds across the FETs...). One therefore gets the efficiency of Class D with the superior linearity of Class A (or, at least, this is the idea). The problem here is not unlike the most frustrating one facing Class D circuit designers: how to get the switchmode power supply to respond quick enough to make a difference! The pole-zero compensation of the SMPS' error amplifier is where most of the trouble can be found, and when one finally confronts all the variables that interact - switching frequency, output capacitor size, leakage inductance, output filter inductance vs. current, stabilization against load variations, etc... - most designers have the good sense to shelve the idea.

Gilmore - Bob Carver is the most prominent proponent (was?) of this particular idea, but the product you referred to wasn't on the market for very long, IIRC, so presuming that everyone should be familiar with it is a bit, well... presumptious!
 
"Bob Carver is the most prominent proponent (was?) of this particular idea, but the product you referred to wasn't on the market for very long, IIRC, so presuming that everyone should be familiar with it is a bit, well... presumptious!"

The M400 (Carver Cube) was made for about ten years. All the other models above 100 watts use the same idea. The M1.5 put out 1.2KW bridged, and twice that on peaks.

The only heatsink the M1.5 had was the bottom of the chassis that the output transistors bolted to, no fins, nothing. The PM1.5(pro version) added balanced inputs and a tiny 2" fan with what looked like a toy slot-car motor on it, the fan just sucked air in through the back and out a slot on the front panel.

Rail voltages on the M1.5 were ±37V, ±77V, ±125V, so in bridge mode the amp could swing 400V peak-to-peak on transients(2.4KW/8R).
 
Re: Re: "Power Tracking" in supply, ever heard, DIY ideas?

jeffreyj said:
The problem here is -snip- how to get the switchmode power supply to respond quick enough to make a difference!
Here's how I would do it if I were doing it, which I am not.
The bitstream that is coming out of your CD player and into the dac, have it first go into a cheap-and-nasty dac, the output of which controls the supply rail voltage with a fast attack slow decay setup. The bitstream then gets clocked through a delay line (say a static ram or whatever) and gets delayed for 1 second then finally gets to the proper dac that makes the music.

The point is, the supply rail voltage is given a 1second view into the future. It knows what is going to happen 1 second *before* the "proper" dac actually makes it happen. That means the power supply doesn't have to respond so terribly quickly. Heck, you could use triacs to switch taps on the power tranny secondary if you really wanted to go low tech. Lower noise than a switchmode supply. Easy for diy too. :cool:
 
Re: Re: Re: "Power Tracking" in supply, ever heard, DIY ide

Hi!

Circlotron said:

Here's how I would do it if I were doing it, which I am not.

Disappointing ;) . Actually when I posted this, I was kind of hoping that you would pick this up and tried implementing something... for the benefit of us all... :clown:


The bitstream that is coming out of your CD player and into the dac, have it first go into a cheap-and-nasty dac, the output of which controls the supply rail voltage with a fast attack slow decay setup. The bitstream then gets clocked through a delay line (say a static ram or whatever) and gets delayed for 1 second then finally gets to the proper dac that makes the music.

The point is, the supply rail voltage is given a 1second view into the future. It knows what is going to happen 1 second *before* the "proper" dac actually makes it happen. That means the power supply doesn't have to respond so terribly quickly. Heck, you could use triacs to switch taps on the power tranny secondary if you really wanted to go low tech. Lower noise than a switchmode supply. Easy for diy too. :cool:

Mhhhm... Since I use my Soundcard in a somewhat unusual way, as a freely programmable DSP, it would be no problem at all giving a signal with no delay on one output, and a signal with a delay though a delay tramline on another output...

But since I can only dupe designs already developed - I have very basic knowledge of electronics - that would not help me very much...

I remember from the TriPath chips that they used a somewhat similar approach on "monitoring" the actual output of the power transistors, and adjusting all parameters according to that output...

Sadly TriPath is very deep into debts, and development seems to have slowed down a lot, if not altogether stopped...

Bye,

Arndt
 
I too have been researching the net on how to implement the power tracking system that you are mentioning.QSC has schematics of its Class H amplifier online namely the RMX 2450.It uses a comparator to sense the output and raise the voltage when necessary,most of the time it uses the lower voltage thereby keeping output transistor dissipation low.I hope this helps.
 
kevin gilmore said:
Ever seen a carver cube?

Works the same way.

Actually, the Carver M100 ("Cube Amp") was a class G design that used (I think) 3 rail voltages. I have also heard that the power supply, at least on the original models, wasn't up to par.

Bob Carver developed his patented Tracking Downconverter for his Sunfire products, after he left Carver Audio. I don't know if he brought the technology back to Carver when he returned.

It is my understanding that a Tracking Downconverter is a Class H supply.

Regards,

Mark Broker
 
The weakness of all of these techniques seems to be that the switching activity itself introduces noise. However, in the last year or so, Doug Self published articlkes in Electronics World showing how, in at least one configuration, much of the noise can be reduced. One nice touch was that the inner (low power) outputs and rails can be operated as Class A without much penalty.

From a DIY standpoint, one of the more practical impediments is a tranformer with multiple voiltages. For instance to duplicate the rails (25V, 50V & 75V) of the M-400, one would need a transformer rated approximately 55-35-18-0-18-35-55. This is not something to be had "off the shelf". You could use thre separate transformers but that isn't going to be space efficient. I'm sure one could be custom made but I'll bet it would rndup 80%-90% of the total cost of the DIY project. There is a place where you buy the primary and a kit and instructions for winding your own secondaries. This might be useful for anyone who enjoys tedium.
 
Re: Re: Re: "Power Tracking" in supply, ever heard, DIY ide

Circlotron said:
The bitstream that is coming out of your CD player and into the dac, have it first go into a cheap-and-nasty dac, the output of which controls the supply rail voltage with a fast attack slow decay setup. The bitstream then gets clocked through a delay line (say a static ram or whatever) and gets delayed for 1 second then finally gets to the proper dac that makes the music.
....
That means the power supply doesn't have to respond so terribly quickly. Heck, you could use triacs to switch taps on the power tranny secondary if you really wanted to go low tech. Lower noise than a switchmode supply. Easy for diy too. :cool:

This is almost a very good idea... A seductively good one, I'd say, but it does not eliminate the problems incurred trying to yank the output voltage of a SMPS all over the place! To wit: you will need just as fast response time whether you delay the bitstream or not because you still have to be synchronized to said stream! So, whether the SMPS must vary its output now or 1 second from now is irrelevant, its the fact that it must vary it up to 20,000 times per second which is important!

There are many problems in doing this, and most of them are tradeoffs; e.g. - increasing the loop bandwidth demands that the switching frequency be proportionally increased.

The one thing you wouldn't have to worry about, though, is noise. A SMPS operating at the necessary 400kHz or higher fs will not affect audio circuits (unless there are those among you who claim they can hear up to 400kHz!) This is not to say that stray magnetic fields transformers or chokes could not cause problems! I am only speaking of the inevitably higher ripple SMPS exhibit compared to their linear counterparts.
 
A Schottky diode solves the noise problem, 200V types are common now.

Carver's Amazing Subwoofer uses a bridge amplifier with a tracking regulator for 2.7KW output with no line tansformer.

Here is a bridge amplifier combined with a tracking circuit to form a Class H amplifier, note that all you need is an ordinary 30-0-30 transformer to get the voltages required.

http://164.195.100.11/netacgi/nph-P...0&s1=6304138.WKU.&OS=PN/6304138&RS=PN/6304138

Go to figure 5 in 'Images', D1 is a power Schottky type.

Any amplifier may be substituted for the 3886, although with two pair of TIP35C/36C driven by the power supply pins and a suitable supply, 800W/4R is possible.
 
Great minds not thinking alike ;-)

jeffreyj said:
This is almost a very good idea... A seductively good one, I'd say, but it does not eliminate the problems incurred trying to yank the output voltage of a SMPS all over the place! To wit: you will need just as fast response time whether you delay the bitstream or not because you still have to be synchronized to said stream! So, whether the SMPS must vary its output now or 1 second from now is irrelevant, its the fact that it must vary it up to 20,000 times per second which is important!
What I was envisaging was a setup that simply makes the supply rails jump up to slightly more than what the peaks of the signal require, about 1 second before the output stage actually makes that signal. The supply rail is then allowed to reduce of it's own accord over several second perhaps, and tough luck about the excess dissipation for a few moments. Not quite as fancy as yours jeffreyj, but maybe a bit easier to do.
 
I am also one of those people that are intrigued by the Carver power supply design, but what to do about that custom transformer [multitap & a 60Hz-SMPS)? (Actually never seen the schematic of a m400).

One interesting solution was what the tda1562 did. The idea was to use a single supply voltage and then temporarily, during peaks, raise the voltage above the rail-voltage with a capacitor voltage reservoir. I am sure that inside that chip there are more transistors then you can shake a stick at to accomplish this task. But, it's one way to not have to use a multitap transformer.
 
"One interesting solution was what the tda1562 did. The idea was to use a single supply voltage and then temporarily, during peaks, raise the voltage above the rail-voltage with a capacitor voltage reservoir. I am sure that inside that chip there are more transistors then you can shake a stick at to accomplish this task. But, it's one way to not have to use a multitap transformer."

The TDA chip is a bridge amp with a high voltage tier. The high voltage tier comes from two bootstrap caps on either end of the speaker outputs. This is a non-inverter solution to getting high voltage for a car stereo amplifier.

Any regular center tapped transformer will work fine to provide the tiers for a bridge design.

You do not need a special transformer.
 
The newer designs by Bob Carver are indeed VOLTAGE-tracking downconverters (like the one used in the sunfire sub).

The old cube was somewhat different by using a circuit like it is used for dimmers (don't know the English expression, in German it is "Phasenanschnittsteuerung").

An alternative hybrid method would be the use of PARALLEL linear and switching power stages.


Regards

Charles
 
Re: Great minds not thinking alike ;-)

Circlotron said:

What I was envisaging was a setup that simply makes the supply rails jump up to slightly more than what the peaks of the signal require, about 1 second before the output stage actually makes that signal. The supply rail is then allowed to reduce of it's own accord over several second perhaps, and tough luck about the excess dissipation for a few moments. Not quite as fancy as yours jeffreyj, but maybe a bit easier to do.

Love that subject line, btw ;)

Let's see if I am now following your line of thinking, as it occurred to me that perhaps I wasn't.

Ok - because we aren't going to worry about decay time, we can dismiss the time constant of the output caps. What we want the SMPS' error amplifier to look at, then, is not the absolute output voltage, per se, but the voltage across any one output device. Monitoring this differential would be fairly easily as long as the common mode input limits of the differential (instrumentation?) amp were respected! :devily:

Integration of this voltage to limit its bandwidth will be necessary, of course, but I would estimate 2kHz to be a reasonable upper limit for the error amp's response. I propose this because with so little energy present in the upper octaves of any signal except white noise, there is clearly a diminishing return from making the SMPS slew faster!

All that remains, then, is whether a delay stage would be necessary. I still say that it wouldn't because the speed at which the SMPS must respond is dictated by the dv/dt of the signal and delaying the signal by any amount of time does not affect its instantaneous rate of change.

Getting closer?
 
Re: Re: Great minds not thinking alike ;-)

phase_accurate said:
The old cube was somewhat different by using a circuit like it is used for dimmers (don't know the English expression, in German it is "Phasenanschnittsteuerung").

"Phase control" is the literal english translation. Yes, this is how a typical light dimmer functions: ZVS


jeffreyj said:
Integration of this voltage to limit its bandwidth will be necessary, of course, but I would estimate 2kHz to be a reasonable upper limit for the error amp's response. I propose this because with so little energy present in the upper octaves of any signal except white noise, there is clearly a diminishing return from making the SMPS slew faster!

Yep :) As I recall from the whitepaper (that I can no longer find online :mad: ), the original Sunfire amp used a limit of 6 or 7kHz as the power supply response. I have no idea if it has been upgraded or not in the newer models.

I don't think I have ever seen the results of a listening comparison from a source I trust between any of the Sunfire products featuring the Tracking Downconverter and some other manufacturers' similarly priced models.

Regards,

Mark Broker
 
"Yep As I recall from the whitepaper (that I can no longer find online ), the original Sunfire amp used a limit of 6 or 7kHz as the power supply response. I have no idea if it has been upgraded or not in the newer models."

Doesn't really matter, the tracking supply is fed by the audio, the audio is then fed into an all-pass delay line before amplification.

The end result is the change in supply voltage is always ahead of amplifier demand.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.