Bob Cordell Interview: Error Correction

[Wavebourn]

It is possible to have good sound with .8 ohm output resistance. In fact, the original AR-1 loudspeaker originally had a recommended damping factor of 1. I experimented with this over 40 years ago and found this to be true. Variable damping factor can be designed into power amps, but we usually default today to high damping factor just to give a consistant standard to work with by the loudspeaker manufacturers.
Personally, I have to add .3-.5 ohms in series with my amps in order to drive my WATT 1 loudspeakers. These speakers drop their impedance to about .5 ohm at 2KHz. This is a design compromise that I must live with. Adding the resistor does not effect the sound very much, thank goodness.

John, what do you think of a behaviour of a speaker on the zero crossing piont damped by non-linear small impedance against higher one, but significantly more linear?

 

[john curl]

The very low impedance is caused by a resonant network that I presume is fairly linear, so what should happen is that a minor notch might be generated at 2K. BUT, having the amp try to drive this very low impedance can cause even more trouble.
Bob, can you drive .5 ohms without problem, even for a short time?

 

[gerhard]

Re: Importance of Damping Factor

The real issue is that low damping factor implies relatively high output impedance. A classic vacuum tube amplifier, or even a solid state amplifier with no negative feedback, may only have a damping factor of 10. This corresponds to an output impedance of 0.8 ohms.

Now recognize that most loudspeakers have widely varying impedance as a function of frequency - not just in the bass region. It is not uncommon for a so-called 8-ohm speaker system to have a minimum impedance of 4 ohms and a maximum impedance of 40 ohms. Think about what this does to the final voltage frequency response as seen at the speaker terminals.

At places where the speaker impedance is actually the rated impedance of 8 ohms, we are down by about 1 dB as a result of the amplifier's 0.8 ohm output impedance against the 8 ohm load (relative to the amplifier's no-load output). At places where the speaker impedance is very high, we are down very little, so relatively speaking we are up 1 dB relative to the nominal. Where the speaker impedance dips to 4 0hms, we are down about a total of 2 dB, so we are down by 1 dB relative to the nominal. We thus have a frequency response swing of +/- 1 dB. This is definitely non-trivial.

All OK, but a damping factor of 10 is an extremum left for those who castrate themselves by declaring feedback evil. A factor of 100 should be easy enough to achieve and then we are somewhere near 0.1 dB. Let's pump 50 Watts into a bass voice coil for a minute. What is its temperature? What is the TC of the resistance of copper? That could easily make more than 0.1 dB. And tweeters have an even hotter life. I've seen voice coils that had pretty burnt Kapton isolation. So, our system is not even time-invariant.

I once have done the math, decided that the only sensible way would be active speakers -- and saw that this project would either never come to an end or that it would be a desaster for my professional work as a RF / microwave consultant.
So I choose to buy the B&Ws and a fat used Krell to keep me from going astray. (Something active like a BM40 is just too much.) Guess what? Now I have a new preamp in my computer trying to get out.


regards, Gerhard

 

[Bob Cordell]

Good point Bob to follow available power voltages!

Unfortunately I oversaw such an idea in my design, but I used a chip containing set of 5 transistors to thermostabilize the clipping point. Your idea is brilliant, may I steal it?

--Anatoliy

Thanks for the compliment. Of course you can use it. I came up with the design about 20 years ago when I was working on a 200 wpc successor to the MOSFET power amp. I called the circuit the Klever Klipper because of its power supply tracking feature.

Cheers,
Bob

 

[mikeks]

Look at my new prototype of amp with Hawksford / Cordell error corection... ultrashort signal paths, fast VAS devices, precision thermal coupling...

Where?

 

[mikeks]

Of course you can use it. I came up with the design about 20 years ago when I was working on a 200 wpc successor to the MOSFET power amp. I called the circuit the Klever Klipper because of its power supply tracking feature.

Cheers,
Bob

Where is this Klever Klipper schematic...and the folded emitter follower schematic?

 

[mikeks]

Or is it just moi?

 

[Bob Cordell]

The very low impedance is caused by a resonant network that I presume is fairly linear, so what should happen is that a minor notch might be generated at 2K. BUT, having the amp try to drive this very low impedance can cause even more trouble.
Bob, can you drive .5 ohms without problem, even for a short time?

John,
I agree with your point that an amp with 0.8 ohm output impedance can sound good. Of course, it will tend to sound a bit different when driving different speakers with different impedance characteristics. And of course, some people may prefer some of those combinations.

That's something about the impedance dip of the Watt 1 to about 0.5 ohm at about 2 kHz. I had no idea. I wonder what amp Wilson used to voice them???

My 50-watt MOSFET power amp was able to drive a 0.5 ohm (resistive) load for brief intervals. I showed some two-cycle tone bursts of it doing so in my AES paper on the amp, mainly to demonstrate the high-current capability of the MOSFETs. That prototype amplifier employed no protection. The test worked fine with a 1 kHz tone burst, but I must admit that I tried the same thing with a 20 Hz tone burst and blew the thing up.

Cheers,
Bob

 

[Bob Cordell]

Where is this Klever Klipper schematic...and the folded emitter follower schematic?

I'll try to get them up on my site this week

Bob

 

[john curl]

Good response, Bob. I didn't know for sure, before you told me. Sounds OK to me. As you know many amps fail that sort of test.

 

[Bob Cordell]

Correction

My 50-watt MOSFET power amp was able to drive a 0.5 ohm (resistive) load for brief intervals. I showed some two-cycle tone bursts of it doing so in my AES paper on the amp, mainly to demonstrate the high-current capability of the MOSFETs. That prototype amplifier employed no protection. The test worked fine with a 1 kHz tone burst, but I must admit that I tried the same thing with a 20 Hz tone burst and blew the thing up.

Cheers,
Bob

John,
I just checked my paper on my web site, and I found that the figure I showed in the paper was of the amplifier driving 1.0 ohm, not 0.5 ohm, to a peak current of 22 Amps. I'm pretty sure I tested it with 0.5 ohm as described, but that was 25 years ago, so I could be wrong. I do definitely remember blowing it up with a 20 Hz tone burst .

Bob

 

[Wavebourn]

Thanks for the compliment. Of course you can use it. I came up with the design about 20 years ago when I was working on a 200 wpc successor to the MOSFET power amp. I called the circuit the Klever Klipper because of its power supply tracking feature.

Cheers,
Bob

Thanks. It gives some extra headroom for free!

 

[Wavebourn]

The very low impedance is caused by a resonant network that I presume is fairly linear, so what should happen is that a minor notch might be generated at 2K. BUT, having the amp try to drive this very low impedance can cause even more trouble.
Bob, can you drive .5 ohms without problem, even for a short time?

Of course I meant non-linear output impedance of the amp when crossing zero!
Resistor in series decreases impact of this non-linearity.

 

[mikeks]

Many theory, too little practical experiences...

I wouldn't worry too much about your lack of experience, old chap.

As we've seen much too often on this forum, it's not so much time served as what is learnt from that experience that is important.

Way too many folk have a lot of ''experience'' without actually learning much that is true in fact.

 

[lumanauw]

Many theory, too little practical experiences... Look at my new prototype of amp with Hawksford / Cordell error corection... ultrashort signal paths, fast VAS devices, precision thermal coupling...

Hi, UpupaEpops,

Very beautiful. You and PMA make beautiful boards
Is it a board of power supply? The transistors mounted on the heatsink only have 2 legs, are they diodes? Where are the output transistors?

 

[vuki]

Is it a board of power supply? The transistors mounted on the heatsink only have 2 legs, are they diodes? Where are the output transistors?

Metal bars are "power bars" so the output transistors with their metal surfaces conected to drain/collector are in direct contact with power supply. Better thermal coupling & large conducting surface.

 

[Upupa Epops]

Prism bellow " diodes " is " rail " with voltage and is insulated from heatsink...This solutiom have two advantages : precision temperature coupling of all devices and great rail's voltage distribution....

 

[PMA]

Many theory, too little practical experiences... Look at my new prototype of amp with Hawksford / Cordell error corection... ultrashort signal paths, fast VAS devices, precision thermal coupling...

. As you know, both is important. Without theory hints of error correction by B.C. you would have no DPA designed .

Nice board, as always.

Cheers,
Pavel

 

[traderbam]

Why don't we get objective?

Just how distorting is a "typical" loudspeaker? What is the distortion content of the output air pressure? Consider these "theoretical" inputs:
1) A perfect voltage waveform (zero distortion, from a zero output Z source)
2) A perfect current waveform (zero distortion, from an infinite output Z source)
Thinking about the likely distortion content for these inputs will help inform us about the relative importance of amplifier output Z to other amplifier distortion mechanisms.

 

[PMA]

Distortions superimpose. Distortion spectral content of speaker is far different from that of the amplifier.