looking for "ultrafast power amp"

[KEES48]

ultra fast witj ring emitters

Hi

I have a schematic from myself, I have design this 4 years ago, it is a currend feedback, with some of the alexander amp in it.

I have use this until now, sounds good with transitors, not with fets.

send me a email or give me a possebility to post the schematic here, let me now or you like the sound and also the not good messages.

greetz

 

[amplifierguru]

Lumanauw,

Have you found your fast amplifier ccts, yet? To build car amps? What speed do you want - 100K/Hr? Sound to be not left behind?

Seriously do you think it will make for a better sound?

I recall doing some comparative tests when I was enslaved to Madrigal - between a Spectral (renowned for it's wideband/high speed), a ML anything, and my low cost lashed up prototype with a moderate I think 15V/uS.

My lashup was a unanimous winner in the 'sounds good' stakes.

Good design is about many things.

Greg

 

[luclucmir]

hi speed amps

Hi,

I have interested in your conversation and I did have to consult website LC audio and Spectral amp.
I check the spec. of Goldmund amps. about of slew rate + risetime + settling time + wide band in the order of 1.5Mhz to much similar of Spectral that you mentioned.
I thinking about some spec. they come of the high speed power supply and very short signal path. But I have a doubt about that high speed spec. are to reached by the common, power bipolar transistor. maybe only characteristics of the mosfet audio by Hitachi or Toshiba.
Who model transistor work Spectral or LC Audio.?
LC audio have this hi speed spec.?
Some brand work with bipolar?

Luc

 

[Miles Prower]

Hugh, I have seen the schematics of your AKSA amp (got it from you) and I think you are doing the right thing to keep it non-public. If the amp had been really complicated like Alaskan Audio's monster, it would be very hard to copy so it would be easier just to buy the pcb from you.

Why does this not surprise me. That's been my experience: "fancy" and "elaborate" don't perform all that much better. Stick with something simple, optimize the hell out of it, and you'll do better than coming up with something "wierd". That all too often seems to become an end in and of itself.

 

[lumanauw]

send me a email or give me a possebility to post the schematic here, let me now or you like the sound and also the not good messages.

Could you post the schematic here or send to lumanauw@bdg.centrin.net.id ?

Have you found your fast amplifier ccts, yet?

No, I haven't seen the 2 works of Stochino until now. How do the schematics look like? Do you have them?

To build car amps? What speed do you want - 100K/Hr? Sound to be not left behind?

Seriously do you think it will make for a better sound?

Maybe yes, maybe no. You are right. The more I learned and experiment, good sounding is not coming from 1 factor only. Even poor measurement amp (low damping, low slewrate) can sound good. I will keep on learning

a ML anything

I've heard ML33. Your power amp must weigh more than this

I check the spec. of Goldmund amps. about of slew rate + risetime + settling time + wide band in the order of 1.5Mhz to much similar of Spectral that you mentioned.

Many "expensive" audio power amp stressed on how fast they are. Like FM-Accoustic, they claimed that they use "special order high speed super matching" transistors for their products. Have to wait 1 month for 1 preamp and 9 months for ordering power amp, because they have to match all transistor used.
I learned that a system with FM-Accoustic from phono, preamp, power amp, they seems revealing all the details, but I'm not sure it will fit everyone's ear. Seems lack of depth. Seems more pleasing the same sytem with Air Tight 200W power amp (changing the amp only), more relaxed to listen to, even not so fast as FM Accoustic. This shows the slower can sound better, not necessarily the faster will sound better.

Why does this not surprise me. That's been my experience: "fancy" and "elaborate" don't perform all that much better. Stick with something simple, optimize the hell out of it, and you'll do better than coming up with something "wierd". That all too often seems to become an end in and of itself.

There are many ways to design audio power amp. Epupa Epops and PMA using different approach than AKSA or NP.

 

[Pan]

Luc,

LC audio uses big Sanken BJT´s in their poweramps. One older design managed a 4Mh powerBW with those. I think the class A Zapsolutes have about 1Mhz.

There are small signal BJT´s with 1GHz BW as used in Zapfilter.

/Peter

 

[luclucmir]

Who are the best mosfet or transistor for Hi-definition audio amps.?

Hi Lumanauw,

Then the wide bandwidth 1 to 2 Mhz from Goldmund and Spectral grants clarity and sound detail. Who provide?

Is possible obtain the same speed and wide bandwidth with bipolar transistor. ?
I appreciate from Krell, Burmester and Levinson that doesn't have nice high frequency sound. in case of Krell I listen a grain sound and hard and cold impression. In Levinson and Burmester the middle band is dark and backward image.
This, don't happens with Goldmund amps.

Peter,

Then, who model transistor work FM Acoustic, Spectral and Ayre.?

Regrads
Luc

 

[amplifierguru]

Hi luclucmir,

"in case of Krell I listen a grain sound and hard and cold impression. "

My impression as well - maybe that explains why every time I hear them at a show they're set up playing dirges or Leonard Cohen!

"In Levinson and Burmester the middle band is dark and backward image."

Once again, agreed. Since they've had so many different designers, one has to assume it's their listening panels that are questionable?

Greg

 

[Pan]

Luc,

bjt is bipolar and in LC amps make the best HF reproduction I´ve heard.

/Peter

 

[lumanauw]

Hi, Luc,

Then the wide bandwidth 1 to 2 Mhz from Goldmund and Spectral grants clarity and sound detail. Who provide?

Personally I don't have evidence that Mhz specs gives more pleasant sound, everything else being equal . I can get good sounding amp with transistors like 2955 and 3055. Faster transistors tends to oscilate more easily, and when it happens you will already forget about good sounding, you only think how to make it stop oscilating . I think Mhz specs are produced by the factory for the brochure/catalog/magazine for the potential customer to see. We can only hear up to 20khz, whats the use of 1mhz signal in audio?

Is possible obtain the same speed and wide bandwidth with bipolar transistor. ?

I think Mosfets or bipolars can reproduce audio (20-20khz) perfectly. How do they sound depends on other things, not the type of transistor.

I appreciate from Krell, Burmester and Levinson that doesn't have nice high frequency sound. in case of Krell I listen a grain sound and hard and cold impression. In Levinson and Burmester the middle band is dark and backward image.

I think what you observed doesn't come from type of transistor, but from other things, like crossover distortion or type of low-order high-order harmonics contents. Thay both can make "grainy" or "harsh" perception.
Quite complex thing(s) that determine how an amp will sound, not just what transistor they use

We have Charles Hansen (who makes Ayre) here. Why don't you ask him what transistor he use

 

[ilimzn]

G. Stochino's amp is, if memory serves, a voltage feedback design - I have the schematic somewhere. It does make extensive use of various tricks to prevent stages from saturating. One thing i do not like about it, it uses IRF640/9640 MOSFETs in the output (2 pairs) which are not very complementary at all. And, when you are into designing the fastest amplifier, you really want better matched capacitances in your MOSFETs...

 

[tlf9999]

No transistors are idealy, perfectly 100% complementary. The key test is if we can hear those less than perfect complementary devices. Sometimes, our persue for perfection might just be in those areas that make zero sonic differences.

 

[Upupa Epops]

... we have Charles Hansen here ...

All " guru's " or manufacturers have their own " fan clubs " and their members don't belive to members of other fan clubs. As we can see, all gurus are very reserved in evaluation of other guru . Ask them for their opinion is funny in this case .

 

[ilimzn]

Although you never find perfectly complementary parts (I certainly wish there was such a thing!), IRF640 and 9640 are by design mismatched by at least 50%. One can certainly do better than that!
Regarding audibility, it will produce more even ordered distortion, which may or may not be good. Regarding specs, Stochino has lower negative slew than positive, more so than differences in front end PNP and NPN BJTs would alow. Granted, considering it slews at some 400V/us or so, a 10-15% difference hardly counts for much, it's juct that the point was to build an 'ultrafast amplifier' and this particular asymetry could have easily been avoided.
Regarding ruggedness, well, that's a whole different thing. IRF640 is rated for nearly twice the current of the 9640. Not something that one would find a problem during normal operation, but driving very difficult loads or, heaven forbid, a short circuit, well, that would make a difference. TO220 devices are not the best choice for such applications anyway...

Oh, and I am definitely NOT a guru, and heaven forbid I ever become one...

 

[Upupa Epops]

Rather nonperfectly complement than quasi complement .

 

[ilimzn]

Well, i can only agree with that.
However, since the P ch part is the limiting one, better complementarity could easily have been found in a IRF740, though even that is far from perfect.
Using IRFP9240 and IRFP340 solves two problems at the same time: the IRFPs are much more robust given a more massive package, the complementarity fit is within 10% (which is on the order of differences of parts of the same kind from the same batch), and the capacitances are the same.
Alternatively, a cut-down in power (80-ish W) version could be built with one pair of FETs, IRFP240 and IRFP9140 would be the choice here, for even better complementarity.
A friend of mine built the Stochino about 6 years ago but i have not had a chance to audition it with a source and speakers that I know. What I have heared gives a good impression, though.

 

[Upupa Epops]

Yes, it is thruth, it is good way. But I have build with these noncomplementary Mosfets ( 540/9540 ) an amp, which is very good and I belive, that if I make it with " not declared " complements, in the end sound will be the same - any ear don't hear difference between .002 and .001 % of distortion.

 

[tlf9999]

IRF640 and 9640 are by design mismatched by at least 50%.

I guess it depends on what your base the 50% on.

End of the day, no one has shown that one can hear the mismatch. If that's the case, the "50% mismatch" may be good enough for this purpose. Then why bother?

 

[jacco vermeulen]

I am definitely NOT a guru, and heaven forbid I ever become one...

Mind if i ask which irf(p)'s are the best complementary set ?

(you are the Mosfet Messiah, the thread that you go to is the thread that i go to )

 

[ilimzn]

I have build with these noncomplementary Mosfets ( 540/9540 ) an amp, which is very good and I belive, that if I make it with " not declared " complements, in the end sound will be the same - any ear don't hear difference between .002 and .001 % of distortion.

For small signal output (the first 1W) in a follower configuration, you may actually, depending on bias, get sonically more pleasing results from not so complementary transistors. Complementarity becomes less and less of a problem the higher the bias current (but keep in mind you need to view bias current in the context of rated device current - have a look at gm vs Id curves). As much shortcomings as MOSFET models may have, this is actually quite realistically seen in simulator.

I guess it depends on what your base the 50% on.
End of the day, no one has shown that one can hear the mismatch. If that's the case, the "50% mismatch" may be good enough for this purpose. Then why bother?

The mismatch is based on a number of things, the foremost being gm, followed by the various capacitances.
If your output has source resistors, the differences will obviously be smaller as these will linearise the gm somewhat, but at the expense of lowering it's already low value.
If you don't have any source resistors, unless you have a gross mismatch of Cgs and Cgd, in a follower configuration it will be least audible, UNLESS you are driving your FETs close to the limit. Then, you need to look at other parameters, such as rail loss (uneven clipping), maximum drain current (uneven current limiting), as well as different points of inflection on the gm vs Id curve, including how they change with temperature.
For configurations where the MOSFET has voltage gain, such as various CFP derivatives, mismatches can have intuitively subtle, but sonically quite audible conseqeuences, as people often think the intrinsic local feedback will cure all ailments. Keep in mind that with MOSFETs, gm is much lower than for BJTs, so you can much more easily get into a situation where differences in said gm matter.
Of course, this all implies that everything else in the given amplifier is 'perfect' - otherwise these diferences will be swamped out by other problems.

Mind if i ask which irf(p)'s are the best complementary set? (you are the Mosfet Messiah, the thread that you go to is the thread that i go to )

Well, if there is such a thing as a MOSFET messiah, it's definitely NOT me. i just learned a couple of hard lessons.
people often get hung up on this word, 'complementary'. While Hitachi and Toshiba laterals and pi-trench devices have as much as achievable, electrical complementarity, if you looked at their dies, you would see they look very different (electrically complementary BJT dies also look different but not as much as MOSFET dies).
IR hardly calls their complementaries electrical, an N channel with Idmax 21A and P channel with Idmax 11A could hardly be called an electrical complement. But, they are process complements, in the sense that identical masks (but slightly different proces parameters, save for theobvious reversal of N and P) are used to make them.
Finding proper complements is made harder by often different scales on the output and input characteristics, but it can be done.

So far I have identified 6 pairs of transistors from the IRF, IRFD and IRFP series that can be used as good complements. For the purposes discussed in this thread, I have found:

IRFP240 / IRFP9140 for rail voltages up to +-50V max (I would go with +-40V max for safety).
IRFP340 / IRFP9240 for rail vltages over +-50V, and up to +-100V max (I would use +-80 for safety).

Note that the Nch parts must NOT be the versions with N at the end!!!

I have to admit I have not used the others in a long time and will have to look them up in my notes... although I'm already airing my trade secrets as it is

Two notes:
1) IRFP9140/IRFP240 are the best complementary fiit.
2) IR spice models are actually quite accurate and it is possible to construct a simple SPICE circuit to display the gm of the P part in relation to the N part.