Nige, we all have our own little beliefs, some which may be right, and others which may be wrong. That's what makes each one of us who we are.
I am well aware that I have negative preconceptions about the NE 55XX, possibly because it was almost idolized by the industry and escpecially the UK audio industry, and I tend to react emotionally to idolization. Like any op amp, it can surely be used to come on song, just as it can be misused.
I also have my preferred devices, beside the transistors - I love using the old OP37 op amp, the old LF356 and 353 dual, and of course, the Analog Devices boys, like AD826, AD829, AD818, AD828.
The fact that op amps sound lacklustre in many cheap commerical devices does not mean that with proper pplication and some add-ons the same op amps could - and do! - come on song. They have their issues, but so does each and every electronic device.
I intensly dislike the sound of Burr-Brown's 213X series of FET input op amps, so popular today, I can actually hear them inside a device most of the time. Others trip out with it, never mind for which reason.
And yes, I have heard several very good power amps based on 2N3055/2955 of the older generation, as well as with 2N3055/BD18, as used by Bob Stuart in Lecson amps, for example. Technically, it should have sounded somehow poor, depleted of life and ambience, yet they did exactly the opposite, they were bubbling with life and had reasonable ambience. A particularly good implementation was on German made Wega receivers from 1976 and 1977, i.e. before Wega was taken over by Sony and started peddling repackaged Sony products.
It's not so much the devices as such, but much more the designers and implementations.
I am well aware that I have negative preconceptions about the NE 55XX, possibly because it was almost idolized by the industry and escpecially the UK audio industry, and I tend to react emotionally to idolization. Like any op amp, it can surely be used to come on song, just as it can be misused.
I also have my preferred devices, beside the transistors - I love using the old OP37 op amp, the old LF356 and 353 dual, and of course, the Analog Devices boys, like AD826, AD829, AD818, AD828.
The fact that op amps sound lacklustre in many cheap commerical devices does not mean that with proper pplication and some add-ons the same op amps could - and do! - come on song. They have their issues, but so does each and every electronic device.
I intensly dislike the sound of Burr-Brown's 213X series of FET input op amps, so popular today, I can actually hear them inside a device most of the time. Others trip out with it, never mind for which reason.
And yes, I have heard several very good power amps based on 2N3055/2955 of the older generation, as well as with 2N3055/BD18, as used by Bob Stuart in Lecson amps, for example. Technically, it should have sounded somehow poor, depleted of life and ambience, yet they did exactly the opposite, they were bubbling with life and had reasonable ambience. A particularly good implementation was on German made Wega receivers from 1976 and 1977, i.e. before Wega was taken over by Sony and started peddling repackaged Sony products.
It's not so much the devices as such, but much more the designers and implementations.
I would say a 5534 will tell you 90% of what you need to know . One chip holder and 1 minutes work to insert another will say if a different grape would be better . Being that 5534 is tricky suggests others are bound to work well .
Two of my favourite general purpose op amps would be TL074 and MC33079 . Sadly the range of alternatives is not vast . Often I start a project with 5532 or TL072 as upgrades are easier to try .
Two of my favourite general purpose op amps would be TL074 and MC33079 . Sadly the range of alternatives is not vast . Often I start a project with 5532 or TL072 as upgrades are easier to try .
DVV . Ages ago you and I discussed the linearity claims of Motorola . They claimed remarkable linearity from their devices ? I usually dismiss linearity in current amplifier stages . However it must be better . Here is a conjecture . If 1 MHz or better then linearity matters more ? I have never thought of this before in these terms , sorry if very wide of the mark on this . My assumption is the linearity they talk about is zero feedback as a current amp ? My second assumption is the current curves are closer to a horizontal straight line especailly at high current ? Sorry to say I forget which Motorola transistor made a point of mentioning it in the PDF . Douglas Self also made a vague reference to the same . My very vague memory say less than 0.8% distortion without feedback typical , less if a matched pair ? My experiences with complimentary feedback pairs says distortion is almost zero without any great effort , they have local feedback which I see as being almost a zero feedback concept as it is not loop feedback . Forgive the vague way I put this . It is not how these things are discussed usually . It might be important to get someone who does know to say something ?
Well Nige, I would say that at all times, in all places within a circuit, linearity is of paramount importance.
In output stages, it's just hardest to obtain. Consider what you are aksing those devices to do, work into a shifting load which changes its characteristics as it heats up, and which is anything but linear. We all know impedance is anything but a straight line.
Also, we need to examine output devices very carefully. Case in point - the Euro version of TI's TIP 35/36 C, the BD 249/250 C. Initially, I wondered why was it more expensive than the TIP pair, and why did it have a habit of sound so good in almost anything you drop it in. Until I purchased TI's massive "Power Semiconductor Data Book".
From it, I learnt a few very interesting things.
These transistors, despite their modest overall specs, are nevertheless capable of 25A of constant current, 40 A (!!!) peak. So how will do your bass line, with voltage only?
Secondly, their response is uncommonly linear, despite a declared Ft of "> 3 MHz". It doesn't go far up there, but where it does go, it is rather linear, and in those days, it was very linear.
Gain is so-so, nothing to write home about, but not bad.
And lastly, the Ton, Tstore and Toff times. That's the catch. In those days, which is late 70ies, for power devices typical Ton was like 1,500 uS; they do it in 350 uS. Tstore was also something awful, and the do it in about a third of that time. Toff was very often 1,500...2,000 uS, yet the do it in 450 uS.
Now, connect this with the talk we had here about symmetry of square wave rise and fall times. As I see it, these devices were made to be as symmetrical as possible in those days, and did it better than three times the usual values of the day. Percentages can make a difference, and 3 TIMES (sic!) cannot help but making a darn big difference.
These shortened response times are the key difference between TIP 35/36 and BD 249/250 - the TIP types are also faster than the norm of the day, but only by about two times, i.e. the BD types were faster still.
Now add it all up. They are more linear than most, they can do currents which very few others dare even think about, and they have exceptionally fast switching even today, after decades, and after others have narrowed that gap, with only MOSFETs and IGBTs being faster still, but falling short elsewhere.
Motorola took its sweet time, but to their credit, they never gave up. The old 56xx and 60xx complementary power trannie devices were replaced by the 15 series, first by 15003/15004, then by evolution all the way to present day 1501x and ultiately to MJ 21193/21194, which are the end of that line, I think.
Add it all up and you come to the conclusion that it's not about any one particular thing, but about the overall compromise they have to make, how well they married individual aspects into one coherent whole.
That's why I firmly believe that power transistors actually do have a sound of their own, and consequently, that it's not all the same which you use. Tis also means that the kind of sound you prefer will also influence your decision on which type to use in case of a DIYer; in case of production for sale on the open market, additional factors will influence, sometimes critically, your choice, such as availability, multitude or sigularity of sources and of course, prices.
In output stages, it's just hardest to obtain. Consider what you are aksing those devices to do, work into a shifting load which changes its characteristics as it heats up, and which is anything but linear. We all know impedance is anything but a straight line.
Also, we need to examine output devices very carefully. Case in point - the Euro version of TI's TIP 35/36 C, the BD 249/250 C. Initially, I wondered why was it more expensive than the TIP pair, and why did it have a habit of sound so good in almost anything you drop it in. Until I purchased TI's massive "Power Semiconductor Data Book".
From it, I learnt a few very interesting things.
These transistors, despite their modest overall specs, are nevertheless capable of 25A of constant current, 40 A (!!!) peak. So how will do your bass line, with voltage only?
Secondly, their response is uncommonly linear, despite a declared Ft of "> 3 MHz". It doesn't go far up there, but where it does go, it is rather linear, and in those days, it was very linear.
Gain is so-so, nothing to write home about, but not bad.
And lastly, the Ton, Tstore and Toff times. That's the catch. In those days, which is late 70ies, for power devices typical Ton was like 1,500 uS; they do it in 350 uS. Tstore was also something awful, and the do it in about a third of that time. Toff was very often 1,500...2,000 uS, yet the do it in 450 uS.
Now, connect this with the talk we had here about symmetry of square wave rise and fall times. As I see it, these devices were made to be as symmetrical as possible in those days, and did it better than three times the usual values of the day. Percentages can make a difference, and 3 TIMES (sic!) cannot help but making a darn big difference.
These shortened response times are the key difference between TIP 35/36 and BD 249/250 - the TIP types are also faster than the norm of the day, but only by about two times, i.e. the BD types were faster still.
Now add it all up. They are more linear than most, they can do currents which very few others dare even think about, and they have exceptionally fast switching even today, after decades, and after others have narrowed that gap, with only MOSFETs and IGBTs being faster still, but falling short elsewhere.
Motorola took its sweet time, but to their credit, they never gave up. The old 56xx and 60xx complementary power trannie devices were replaced by the 15 series, first by 15003/15004, then by evolution all the way to present day 1501x and ultiately to MJ 21193/21194, which are the end of that line, I think.
Add it all up and you come to the conclusion that it's not about any one particular thing, but about the overall compromise they have to make, how well they married individual aspects into one coherent whole.
That's why I firmly believe that power transistors actually do have a sound of their own, and consequently, that it's not all the same which you use. Tis also means that the kind of sound you prefer will also influence your decision on which type to use in case of a DIYer; in case of production for sale on the open market, additional factors will influence, sometimes critically, your choice, such as availability, multitude or sigularity of sources and of course, prices.
And yes, Motorola does indeed claim that their MJ 21193/21194 devices will produce 0.6% distortion up to around 10 kHz or so, or abut 0.8% at 20 MHz.
It's a graph on their standard Data Sheet.
Of course, paralleling pairs will reduce this to less stil, up to about 4 pairs in series/parallel for 8 and 4 Ohm loads. More brings you nothing in terms of reduced distortion, but obviously enables more power dissipation without significant degradation.
The rest is on the rest of the ciruit.
It's a graph on their standard Data Sheet.
Of course, paralleling pairs will reduce this to less stil, up to about 4 pairs in series/parallel for 8 and 4 Ohm loads. More brings you nothing in terms of reduced distortion, but obviously enables more power dissipation without significant degradation.
The rest is on the rest of the ciruit.
Last edited:
That's the ones . I hope someone comes in with more to say . It might be a mystery solved . Namely why do medium speed transistors sound best ( very crudely put ) . As I said before complimentary feedback pairs have almost no distortion worth talking about . I would be interested to know if they favour the " nicer " types of transistors . If my theory is correct they will be less fussy with the proviso that they are stable ? Anyone who hasn't tried feedback pairs without loop feedback be prepared for a surprise . Very sweet sounding . I built a small class A with OPA 604 that way .
Nige, you might try using a Burr-Brown 604 op amp, which is rated up to +/- 24V, if memory serves. Throw in a predriver, say MPSA 06/56, a driver pair, say MJE 15030/15031 and finish it with a pair of MJ 21193/21194.
Given the low voltage power lines, I imagine a single pair of MJs would be able to drive even less than 2 Ohms in a steady state mode. They probably wouldn't even notice any 8 Ohms speakers.
That could be a low power amp, I imagine limited to OPA 604's output swing, which would be something like say +/-20V, or around 25 Watts into 8 Ohms.
Or perhaps you should investigate a myriad of power amps which use an op amp input. Even Dan d'Agostino, "Mr Krell", designed such units for people like Aragon and B&W (subwoofer amp). My Karan integrated amp KA-i180, delivering over 180/250 Watts into 8/4 Ohms, uses a BB 2604 as its input stage, one half, and the other half acts as a DC Servo. VERY compact, VERY short signal lines, and it is still the best sounding integrated amp I have ever heard. Also darn expensive, if memory serves, around $6k in the US, and around Lstg 5K in the UK.
Given the low voltage power lines, I imagine a single pair of MJs would be able to drive even less than 2 Ohms in a steady state mode. They probably wouldn't even notice any 8 Ohms speakers.
That could be a low power amp, I imagine limited to OPA 604's output swing, which would be something like say +/-20V, or around 25 Watts into 8 Ohms.
Or perhaps you should investigate a myriad of power amps which use an op amp input. Even Dan d'Agostino, "Mr Krell", designed such units for people like Aragon and B&W (subwoofer amp). My Karan integrated amp KA-i180, delivering over 180/250 Watts into 8/4 Ohms, uses a BB 2604 as its input stage, one half, and the other half acts as a DC Servo. VERY compact, VERY short signal lines, and it is still the best sounding integrated amp I have ever heard. Also darn expensive, if memory serves, around $6k in the US, and around Lstg 5K in the UK.
I don't know about 30+ year old caps, Tom. I wouldn't touch them with a flagpole, too old and almost surely dried up for me.
Also, while I do enjoy vintage gear, I do so mostly for fun. To actually keep it, it has to be a Marantz made February 1978 to March 1980, and even so, only some of the models. Which is what I do have, some in several editions. I admit to being VERY particular.
ALL my Marantz units (2 integrateds 1152, 1 preamp 3250B and one power amp 170 DC) were all made in March 1978, according to internal date stamps from the factory.
That aside, I have done some refreshing work for strictly good friends, i.e. non-commercially. I have turned down so far each and every offer to do it commercially, that ain't no fun. The last one really made me jump the hoops, I had to have a new wooden case constructed for it, since the original was in my view beyond repair. But the expression on my friend's face when he saw it "after" is priceless; after living with it for a few weeks, he bought me a rich lunch, so I'm guessing he's happy with the way it sounds.
My personal list of wanna haves is very short now. I'd like to grab hold of a German made Grundig V5000 integrated amp, because I have heard it and it sound just right, plus it uses two pairs of my beloved BD 249/250C power devices per channel. And I'd just LOVE to grab another German made product, the ASC 5000 integrated amp, but of the last series, from 1982 - these eschewed DIN 5 pin plugs for standard RCA Cinch. They caught on kinda late.
And perhaps a Sansui AU-7900 integrated, but it would have to be in pristine condition and with an acceptable price. Just a "maybe".
Last edited:
I honestly don't know. I got that info from an interview with Dan d'Agostino, in the section mentioning his external consultant design work, and was obviously surprised enough to remember that bit of info.
And I got the interview as reprinted material from Krell Industries which I contacted some years ago, and they appearently mistook me for a customer and sent me their tutta forza promo pack.
And I'm not a Krell freak, so I read it, and then threw it away.
So sorry.
Nige, you might try using a Burr-Brown 604 op amp, which is rated up to +/- 24V, if memory serves. Throw in a predriver, say MPSA 06/56, a driver pair, say MJE 15030/15031 and finish it with a pair of MJ 21193/21194.]
It was something like that . They can be arranged in Darlington or Complimentary feedback pairs . I seem to remember that at higher voltages the output stops at the +/- 18 V level with identical distortion . I suspect it has some built in regulation ?
This is all I can find of that project . Hope it is of some interest ? Note that the feedback cap can be shorted because the input is JFET , 14 mV offset is fine . The best capacitor is no capacitor . Try it with to be certain , sometimes against logic it is better . 10 uF is a bit small . The circuit was for my notes and not for showing to the world .
If anyone tries it you may want to enhance the output stability . Also an unregulated supply to the transistors and simple regulated supply to the op-amp .
The output cap was to be part of a crossover . 10 uF feedback cap OK if so .
BTW . From memory this version will already show very low distortion if reconnecting the feedback to OPA 604 output . Doubtless on simulations which I never do you will see a difference ? On my spectrum analyzer I didn't if memory is correct ? I set the base line at - 70 dB 2W . The bias was work in progress . I think 2 x 1N4007 and 5R4 almost was good enough to replace the Vbe ? It drifted a bit , not enough to matter ( 0.3 A ? ) ! I call this wrongly class AB 2 . The class B more like class G except no extra devices . The very low Vbe of the bias shows how close to acting like a single transistor a feedback pair is . An obvious way to do it , seldom seen .
The bass was driven by a Tripath class D amp ( not my choice ) . Using class A above 150 Hz is a nice touch and allows a simplified PSU . In my own system I would not use an output cap . For my friend it was a precaution , he had the choice . In the end we did it with KT 88 valves .
Last edited:
Just going out now to try to fix my friend Renault heater !
Very likely some parts reversed labeled or the like in my dia . As said just my notes .
The 10 K Vbe current is low as it suited 2 x 1N4007 . Above that too much bias . I would try 4K7 or less if using this Vbe bias . This project progressed a bit from here , alas the notes are filed under some mysterious name I suspect . Good news is that this is a great launch point . Do try 2 x 1N4007 fixed directly to the outputs with epoxy . The resistor to set bias will be below 10 R from memory . The Vbe is mostly so as to have more current , measurements do not seem to indicate that it is required .
Very likely some parts reversed labeled or the like in my dia . As said just my notes .
The 10 K Vbe current is low as it suited 2 x 1N4007 . Above that too much bias . I would try 4K7 or less if using this Vbe bias . This project progressed a bit from here , alas the notes are filed under some mysterious name I suspect . Good news is that this is a great launch point . Do try 2 x 1N4007 fixed directly to the outputs with epoxy . The resistor to set bias will be below 10 R from memory . The Vbe is mostly so as to have more current , measurements do not seem to indicate that it is required .
I think that was one of the terms of trade. Aragon made no mention of it either.
I believe, but cannot prove, that it was d'Agostino who didn't want any mention made, lest some forelorn soul should deduce from it that Krell is not doing too well, so he has to moonshine work elsewhere.
....
Well, I wouldn't do it quite like that. Darlington output sections have a nasty way of oscillating, and they are not as fast as emitter follower.
So, as I said, I'd use an MPSA trannie to drive an MJE1503x trannie, to drive whatever is at the output.
Also, I'd get the supply lines to run at say +/- 25...30V, and then use zeners to cut that down to say +/- 22V for the opa. That way, we have no effective voltage drop across the output stage (well, we do, but it doesn't matter), which would allow us to use the opa's full voltage output.
BTW, what the hell is that capacitor doing in the output line? Wouldn't it be better to use the usual trim pot to null the circuit's offset?
Well, well, ... this looks very promising indeed.
http://products.semelab-tt.com/pdf/bipolar/MG6331 MG6331-R.pdf
http://products.semelab-tt.com/pdf/bipolar/MG6331 MG6331-R.pdf
That does look interesting. The lumpy Ft curve looks like it could be based on real measurements too!
Comparable bandwidth to the old Sanken devices, better SOA I think. And based in merrie olde England
Agreed, Brad.
I compared them to MJL 3281/1302 curves and they look very similar, including the Ft curve.
Based on the graphs and specs, this trannie definitely has a lower output capacitance, which is almost identical to the new generation of Toshiba's transistors which replaced the old 2SC3281/2SA1302, the 2SC5200/2SA1943.
The fact that it is British was a surprise, even if a very pleasant one - I don't know what took them so long, but better now than never. Despite the fact that my personal views differ significantly from the British norms, so to speak, it was about damn time they rejoined the fray.
All I need to do now is to find out what do they cost, and then get my hands on a few samples, they seem too good to pass up.
As I said the output cap was to reassure my friend and was part of his crossover . As to output offset . No one has tested if it might be beneficial . I doubt if it is large enough to be so in typical examples . The idea being either to slightly reposition a voice coil or just to have slightly lower distortion . Having the cone very slightly held away from centre might do good ?
We make our houses too clean and help create immunity problems . Could low offsets being doing the same ? My instinct tells me it will do some good .
The stability is important . My conjecture is we can make the feedback pair become a super transistor . It has gain and linearity to rival any single device . We can even change Cob a little to make it even more obliging . The instability of this set up I believe to by mythical . I never saw it do anything one couldn't cope with . Millions of amps used it when Quasi Complimentary was popular . The extend it to an all complementary arrangement of 4 devices is no great leap . It also does make the two halves virtually identical and cheaply .
Also with Vbe bias it is possible to change this to boosted op amp use . I guess about 1.1 V required ?
If you calculate my use of power the critical question is how high to place the DC . I reckon +/ 18 V op amp and +/- 20 V outputs . As said the OPA 604 hits a brick wall at +/- 18 V from memory . The heat difference is not subtle when running even a low 1 amp standing current .
My research showed 4.5 W to be a critical class A requirement . I would hasten to say at 2 R is not daft . I would then say to convert to class G with 5 Watts A and 100 W G is perfect .
BTW . Like those transistors .
- Status
- Not open for further replies.