John Curl's Blowtorch preamplifier part II

Status
Not open for further replies.
diyAudio Member RIP
Joined 2005
Kindhorman, Brad, are you guys aware (probably yes but I will still mention it ;-) of what Klippel is doing with his pre-correction of drivers? As an example, current versus excursion transfer function for most drivers is different when moving forwards or moving backwards, and is quite consistent from driver to driver. So he does characterize that and 'pre-corrects' the drive signal through DSP. Avoiding the obvious issues with trying to put a mech system within a feedback loop.

Jan

Yes, somewhat.
 
Jan,
I can't wait for the day when I can purchase one of Kippel's laser systems, it will make some things so much clearer when you can actually see some of the physical functions and not just measure the response. That an an upgrade to my Clio software and I will be a happy camper.

I wonder how the Kippel dsp software adjusts for different excursion levels with regard to spl? What the results are telling me is that he has found that the typical motor assembly is very asymmetric in magnetic energy field strength. This is one of the reasons that I designed my motor assembly the way that I have. Yes I also understand that the restorative forces of the surround and somewhat the spider are also not linear or symmetric about the X axis, that is an area that requires some serious study and changes in what is the norm. Not many would even consider some of these things in a design, commodity parts truly do drive this industry.
 
I'M glad Richard asked basically the same question I also asked twice - when do you think THD becomes audible?

I aksed this because experience has taught me that THD specs have very little to do with actual sound. For example, my old Sansui AU-X701, chock full of X-balanced circuitry, Diaond this and Diamond that, has a nominal THD of 0.004% at its rated 100W/8 Ohms. It has an "impedance adjustment" switch at the back, offering me th choice of high voltage or high current. On my regular speakers, there's no discernible difference in sound, but they are exceptionally well bhaved, while using it with my AR94 speakers, not nearly as well behanved but not catastrophic, the "high czrrent" setting is preferred. Ovearll, the amp presents a very detailed, clean and coherent sound stage, with a lot of detail, but unfortunately in a very 2D setting, everybody seems to be placed in a straight flat line, no depth to speak of. And, while not wildly expensive, it was one of the costlier of its kind in its day (made in 1986). It is, in a sense, lifeless.

In stark comparison with it is my old (1993) H/K 6550. It's rathed at 50W/8 Ohms, and has a very low NFB topology, global NFB 17 dB. It has a simple SEPP output, based on a pair of Toshiba's 2SC1382/2SA1302. However, it's sound borders on incredible for the money with only one change compared to a factory unit, it's volume pot has been changed from factory to ALPS Blue, which I regard as my standard, it's a reasonable quality pot at an equally reasonable price. It has more life in it, and certainly more depth, than the Sansui by a long shot. It's rated THD is 0.08/0.3% into 8/4 Ohms. The sheer force and appearent speed and force of bass and mid response is a joy to listen to. In short, it's hard to remain unimpressed with it, especially in terms of its price, which while not dirt cheap, was still quite reasonable and relatively quite a bit lower than Sansui's.

So, what's going on there? The tecnically measurement wise Sansui does not sound as good as the technically inferior H/K, even if it has all the bells and whistles, like FET input (H/K bipolar), X-balanced topology (H/K single input, but complementary VAS), high slew rate of 200 V/uS (H/K 80 V/uS) and whatnot. Measurably worse model sounds better than the measurably superior model - so, it's not how it measures, is it? And, truth be told, not all H/K models designed along the same lines, turned out so well, some are easily forgotten, so it's not their general concept either.

Where does one look?
 
AX tech editor
Joined 2002
Paid Member
Jan,
I can't wait for the day when I can purchase one of Kippel's laser systems, it will make some things so much clearer when you can actually see some of the physical functions and not just measure the response. That an an upgrade to my Clio software and I will be a happy camper.

I wonder how the Kippel dsp software adjusts for different excursion levels with regard to spl? What the results are telling me is that he has found that the typical motor assembly is very asymmetric in magnetic energy field strength. This is one of the reasons that I designed my motor assembly the way that I have. Yes I also understand that the restorative forces of the surround and somewhat the spider are also not linear or symmetric about the X axis, that is an area that requires some serious study and changes in what is the norm. Not many would even consider some of these things in a design, commodity parts truly do drive this industry.

It's been a few years since I attended an evening long of his presentations for the NL AES section, but yes correcting for asymmetry was one part. One other was to correct for compression by expansion through DSP right up to the point where the driver would bottom out and then gracefully 'soft clip'.

TBH, the most spectacular results were obtained with small drivers in small enclosures; but obviously that's where the money is these days.

Jan
 
So, what's going on there? The tecnically measurement wise Sansui does not sound as good as the technically inferior H/K, even if it has all the bells and whistles, like FET input (H/K bipolar), X-balanced topology (H/K single input, but complementary VAS), high slew rate of 200 V/uS (H/K 80 V/uS) and whatnot. Measurably worse model sounds better than the measurably superior model - so, it's not how it measures, is it? And, truth be told, not all H/K models designed along the same lines, turned out so well, some are easily forgotten, so it's not their general concept either.

Where does one look?
My immediate thought is that the PS of the HK is better suited to its amplifying circuitry; and that of the Sansui is much less at ease with the current demands. If I had the latter unit I would immediately increase the stiffness of the supply, especially at the higher frequencies. Would aim for, say, as a very bare minimum improving supply impedance by at least 10 times, where it counted.
 
In stark comparison with it is my old (1993) H/K 6550.

Wow , I did not know :eek: - that unit has a LTP driving a symmetrical super-pair.
.... and a CFP output stage.
H/K sure has quite a few different designs. :eek:
Apex's AX-14 is close to this amp ....
edit - the sansui has slightly (bigger) supplies. The H/K topology is what "shines" (SQ).
sansui has some sort of proprietary daughter card /IPS. H/K is a typical single PCB.
OS
 
Last edited:
Kindhorman, Brad, are you guys aware (probably yes but I will still mention it ;-) of what Klippel is doing with his pre-correction of drivers? As an example, current versus excursion transfer function for most drivers is different when moving forwards or moving backwards, and is quite consistent from driver to driver. So he does characterize that and 'pre-corrects' the drive signal through DSP. Avoiding the obvious issues with trying to put a mech system within a feedback loop.

Jan

Jan, you were ahead of me.

In order to make further progress in loudspeaker development, it is necessary to see each driver and its amplification as a single unit. The reason is that this allows the kind of pre-distortion that is required to do one of two things: 1) improve linear distortion, like Bose did far ahead of its time with the 801/901, or what the Linkwitz transform can do on the low end of closed subwoofers or 2) improve non-linear distortion. As to the latter, current drive does improve some drivers (I posted some measurements a while ago), but the even more targetted approach enable by Klippel measuring gear is of course even better.

As to the idea to control drivers behaviour by including them in the feedback loop of the amplifier, unfortunately this does not fly for the upper 8 octaves or so. The reason is that higher up in the frequency band, the errors to be corrected will be faster than the system that can correct them; drivers are very much bandwith limited; this also applies to the speed with which they can respond to error correction. Only at the deep end it can work.
 
My immediate thought is that the PS of the HK is better suited to its amplifying circuitry; and that of the Sansui is much less at ease with the current demands. If I had the latter unit I would immediately increase the stiffness of the supply, especially at the higher frequencies. Would aim for, say, as a very bare minimum improving supply impedance by at least 10 times, where it counted.

And I would immediately agree with you, Jan. The Snasui uses two 10,000uF lectrolytics with a floating ground, while the H/K use two classic 15,000 uF caps. Lower power, but bigger caps.
 
What is the level of distortion that you, the individual, can just detect? It can be thd or individual harmonics.
"Very good question. Thanks to have asked this. Next question ?"
©General de Gaule, during a press conference.

More seriously, i'm perfectly unable to answer. I made a lot of tries in studio to figure out this question. It depends on a lot of things. First, it is easier to discriminate an harmonic when its frequency is far from the fundamental.
Easier to figure out when the distortion is not correlated (IM), IE, when the added frequency is not an harmonic of the fundamental.
It depends of the harmonic content of the original source (the instrument you play).
It depends if the distortion is level dependent or not.
It depends of the listening level.
The only thing i can tell is i cannot hear anything under ~0.01%.
 
Last edited:
Wow , I did not know :eek: - that unit has a LTP driving a symmetrical super-pair.
.... and a CFP output stage.
H/K sure has quite a few different designs. :eek:
Apex's AX-14 is close to this amp ....
edit - the sansui has slightly (bigger) supplies. The H/K topology is what "shines" (SQ).
sansui has some sort of proprietary daughter card /IPS. H/K is a typical single PCB.
OS

Actually, OS, the H/K contains a grand total of 4 PC boards, one of which is dedicated to their Phono eq/amp (and in my view, it shines), one for some added resistors to the volume pot, and two smaller boards related to switching and their "phase orrect loudness" circuit. But like 90% of it is all on one board. The Sansui is by contrast split into a multitude of smaller cards.

That topology lived on until 1999, its pinnacle being the H/K PA 2400, but with some differences, mostly in the sense of evolution. It is fully complementary from input to output, using cascoded JFETs (2SK170/2SJ74 BL) and 2SC2240/2SQ970 for cascodes, a composite VAS and 4 pairs of Toshiba 2SC1381/2SA1302 - something has to deliver 170W/8 Ohms and short term currents of 100 A (a figure which I seriously doubt for anything above 1 uS, but it will deliver impulse power bursts of 620W/2 Ohms. Visually, it's the usual dull looking H/K black box, with visual memory fall time of < 1 uS. It's as if they are racing with NAD in who's going to produce a duller looking piece of gear.

It's greatest redeeming virtue is the ease with which it works in general. It's a dual mono unit, with one bigga toroidal transformer, with separate secondary windings for each channel, separate 25A rectifiers and very modest 12,000 uF caps (2 per channel)for PSU lines at +/- 59.4V. It also has a high voltage/ high current switch at the back, but I could not sense any differences in operation in either mode both with my highly tolerant main speakers, or with the more misbehaved ARs. Probably swamped by an abundance of power. Global NFB factor is just 12 dB. Most important, it doesn't change tonality when pushed hard.
 
Last edited:
What a shame that the two 10.000 uF cans of the Sansui contain a lot more energy than the ones of the Harman Kardon

(2 times 10.000uF at 64V makes 41 joules)

Even a bigger shame that all that energy has gone amiss, not to say waste.

Speaking of energy, the H/K 6550, despite its relatively low power rating, is fed by unusally high lines, at +/- 51V.

Two times 15,000 uF at 51V equals 19.5 Joules.

Remember that the Sansui has a single PSU, one for both channels.
 
Last edited:
I find these sonic differences between two 'well designed' amps to be the real question. Why? Of course, I have my own ideas, why, but I just can't convey them without a lot of static. '-)
The Sansui example is a perfect one. They apparently through everything that they could at that amp, BUT? It didn't quite make muster. For me, 10K uF is the lowest value I can possibly tolerate in any amplifier, 15K or more is better.
 
You know, Richard, it is almost impossible to put a number on distortion detection without knowing the properties of the distortion. Of course, you know that, but in this general discussion, it seems to get overlooked.
For example: Analog tape has about 1% almost pure 3'rd harmonic distortion at 0Vu. In fact, even as they have improved the tape itself, they moved the recording level to make 0Vu stay near 1% in order to improve the noise floor. Now, most of us who recorded with analog tape recorders found that putting the average value near 0Vu was usually the best place to record, So 1% 3rd harmonic distortion is pretty normal, and 10% would be there on 10dB peaks. Yet, I made hundreds of recordings that sounded pretty darn good, even recorded at these levels, and many 1000's of successful recordings have been released to the public to great acclaim. How is this possible if 3'rd harmonic distortion level is very important to what we hear?
We are told that 2'nd harmonic distortion is roughly twice as easy to listen to, so how can we use that as a reference?
Over the decades, working with analog recording, I found that HIGHER ORDER distortions, created often by Xover distortion in preamps and amps, and clipping the record amp were truly audible.
 
Ok, so what amp will sound better?

They are both CFB with same OPS but first have single ended gain stage, and second have complementary gain stage. They both have 40dB of global feedback and fundamental F is 1KHz +27dB.
 

Attachments

  • SE-GainStage.gif
    SE-GainStage.gif
    35.4 KB · Views: 202
  • Complementary-GainStage.gif
    Complementary-GainStage.gif
    34.5 KB · Views: 202
I find these sonic differences between two 'well designed' amps to be the real question. Why? Of course, I have my own ideas, why, but I just can't convey them without a lot of static. '-)
The Sansui example is a perfect one. They apparently through everything that they could at that amp, BUT? It didn't quite make muster. For me, 10K uF is the lowest value I can possibly tolerate in any amplifier, 15K or more is better.

As you probably know, I too am a big cap fan. However, I must note that practice shows that it's not all in the caps, and amps having seemingly modest caps still manage to pump out enormous currents, albeit in short term bursts only. This begs the question where does the big current live, in the big caps or in the concept and execution?

A good example would be my wife's H/K 680 integrated amp. It is nominally rated at 85/130W into 8/4 Ohms, it is also a dual mono design, one trafo but separate windings for each channel, separate bridge rectifiers and two 8,200 uF caps per channel. True, this is capaity wise inside your above specs, had it ben a single source PSU it would have had 2*8,200 uF per PSU line, or 16,400 uF per PSU line. Still, by modern standards I'd call that a modest PSU, yet the amp doesn't seem to lack oomph even into 2 Ohms, into which it will deliver just over 500W in peaks (t=20 mS, IEC stadards).

If the caps had been bigger, say 12,000 uF each, it would probably do even better, but I think you'd find that despite a formal +46.3% larger caps, the actual net gain into say 2 Ohms would have been more like 5-15%, or something like that. The gain would have been bigger in quality rather than quantity. Unfortunately, it's anything but a linear funtion. In my view, it's much more about the capacbilities of the electronics rather than simple cap size, even if bigger certainly can't hurt.
 
diyAudio Member RIP
Joined 2005
...

As to the idea to control drivers behaviour by including them in the feedback loop of the amplifier, unfortunately this does not fly for the upper 8 octaves or so. The reason is that higher up in the frequency band, the errors to be corrected will be faster than the system that can correct them; drivers are very much bandwith limited; this also applies to the speed with which they can respond to error correction. Only at the deep end it can work.

Of course one does not attempt to correct higher frequencies directly and "instantaneously". If you know where the cone is you can determine what the sensitivity of the driver is at a given high frequency and adjust the gain as a function of cone position. There will be a transition region in frequency where both effects are in play.

A sensor approach to determining cone position at low frequencies is one ploy. Klippel's is more of a "plant model" I believe, which can then deduce the cone position and process signals accordingly.

An anecdote about the early Klippel work, his backing out nonlinearities via his "mirror filter": it was demoed at Harman in the early days of the company's foray into desktop computer audio. It certainly worked of course, but the boss auditioning it determined that the bass sounded louder without it. Since bass excess is usually preferred by the average listener, the reduction in distortion was less useful for promoting such low-end systems.

Brad
 
I find these sonic differences between two 'well designed' amps to be the real question. Why? Of course, I have my own ideas, why, but I just can't convey them without a lot of static. '-)
The Sansui example is a perfect one. They apparently through everything that they could at that amp, BUT? It didn't quite make muster. For me, 10K uF is the lowest value I can possibly tolerate in any amplifier, 15K or more is better.

Back to the design issues, the ones left for the designer to "voice" the amp. If memory serves, it was you and me who advocated this many thousands of messages ago.

In lab measurement, the Sansui does just great, with noinal THD ate rated power into 8 hms, 20-20,000 Hz, at 0.008% or less. And it is less, Sansui remained Old School to their last day, it never exceeded 0.005% under nominal conditions unless you did something wild, like attaching a 1uF cap in parallel with 8 Ohms. Technically very well done, neat layout, and in fact the only thing working against it is the fact that the whole consists of some 5 or 6 boards, with attendent wiring in between. This a concept I try to avoid at all costs, but their technical solutions necessitated that if they were to keep the case size down to a still acceptable level, it's very big as it is.
 
Status
Not open for further replies.