Bruno says there's no such thing as too much feedback, I say there's no such thing as too many supply caps - I'll bet he's got a lot more feedback than decoupling farads.
On UcD I think he was using toroids for output chokes which I doubt very much would be ferrites, more likely powdered iron dust. Haven't inspected closely the Mola-Mola but I do recall a nice-looking pot core on the PCB, perhaps that's his ferrite output choke?
Just reading some general comments about the Mola Mola - it sounds very familiar to me: the amplifier is intrinsically extremely revealing, which means that any irritating audible defect, anywhere, is very obvious, it doesn't have a layer of Vaseline smeared over such, to take the audible edge off. People expecting an premium component to automatically make a system sound really, really good are always missing the point - it has the potentional to sound really, really good, but "hard work" still has to be done to eliminate all the stupid little things that drag the sound quality down - and the latter can be anywhere in the whole system. A good component won't "fix" a poor system!!!
Some amps that sound good are marginally unstable . Reading other peoples accounts makes me feel they are actually better . The instability we measure is not using music .
The reason I posted the class G stuff is it is not outside of the skills of anyone who understands a little to build them . Class D not so .
Bob Carver says it well . I will paraphrase . The whole industry be it high end or budget is limited to 20 amps per device . If at 10 V that might be possible ( it isn't if sine waves , more like 10 amps ) . If 90 V that's 2.2 amps . Thus an amp sitting at 90 V is very limited in it's SOA . Bob says his PSU delivers 6 V more than the transistors need at any time ( not headroom , actual voltage , if music at 10 V peak PSU at 16 V ). That's how he does it . Bet you anything it is more complex than class D ? I think I read Hypex class D current limit is 22 amps transient . The amps gives 6.7 amps 4 R continuous . Some say they can be banked like op amps with a small output resistor each ( 0R22 ) .
The PA industry says class G ( H ) never caught on in audio as no source of sound has the dynamics to need it . PA has those sources . What I like is an amp running 10 W PP class A and then class C up to 400 W 4R . The heat output no worse than an AB . 10 W is as far as I would go in A . Dvv , I forgot . How much class A power do you have ?
A good component will not fix a bad system . Also a bad component can be made to work if the design is strong . Remember my amp challenge 2N3055 and NE5534/2 , I have every reason to think amplifiers that really work can be made with the better examples of these . I especially dislike the NE 5532 . I always use it as a reference in my work . The worse I can say of it is I think it hides the reality of the music ( holding back and featureless ) . It never sounds less than " OK " in the " now what do I hear " head over heart debate . The main reason I dislike 5532 is the amps I found it in . 90% didn't justify a LM741 . I do suspect most op amps have hidden stability problems which even music finds . Dvv , your Philips discrete op amp should be one . It is ridiculously complex for the job it has to do . Did the designer loose sight of the music it should amplify . Problem is the objective observation of hearing stopped in 1947 . Now we have a headless chicken industry that has a world of it's own .
Everyone has a laugh at 1000 W 1 R . I have a high regard for it . NAD 3020 transient 1 R = 192 W on old Hi Fi Choice test ( stated 20 W , 40 actual 8R ) . Sony 170 W 8 R amps 2 W on the same test . I have heard the Sony . 2 watts actual seems about right 8 R . Made a NE 5532 sound like a goddess of a device . The Sony was there upper middle range product at about 7 times the price of the NAD . Sony the Police should get you for that . It is a crime to deceive people . I have heard NAD 3020 stand in for high end gear . Usually it is a giant killer . If it sounded rubbish when you heard one I suspect it was the company it was keeping . It loved the Linn Sara speaker more than Linn or Naim amps of any price . I bet it would love Gale 401 , a notorious load . Also Dalquist DQ10 .
Last edited:
Nige, I understand what you are saying, but for the record, that idea is hardly mine, actually I got it from the Otala & Lohstroh amp. They plainly suggested an amp should be in pure class to about 17 dB below its nominal output, if memory serves.
That seemed very logical to me, given that most of us spend most of our listening time below or at about 1W of average dissipated power, so in effect, we would spend that time listening to pure class A sound, but without the inefficiency of class A on a large scale, where it becomes a penalty.
They got their amp in pure class up to 2.9W/4 Ohms (bias at 600 mA), reasonable since at the time and in Philips, DIN 45000 standard were God.
In my view, the problem of op amps somehow generally settles aroud their poor output current capability, no matter what the data sheet says.
I have yet to hear an op amp, by anybody, driving a 20k load at 3V, thus theoretically needing only 0.15 mA of current, which did not do the same but better with an added pair of current boost transistors. Theoretically, there is no problem, but realistically, there seems to one nevertheless.
Eaxctly what it is I don't really care because I cannot open up an op amp and fix it. With discrete, I can avoid it altogether from even happening, althought of course I am likely to run into a whole set of new problems, like transistor matching for one.
Nige, technically that was a Philips circuit at the start, but no more. Philips has no copyright on it, similar circuits have been used by other before and after them, and it has changed by more tha 33%, which according to international copyright laws, voids copyrights, it is now considered to be something else.
I have yet to hear an op amp, by anybody, driving a 20k load at 3V, thus theoretically needing only 0.15 mA of current, which did not do the same but better with an added pair of current boost transistors. Theoretically, there is no problem, but realistically, there seems to one nevertheless.
Eaxctly what it is I don't really care because I cannot open up an op amp and fix it. With discrete, I can avoid it altogether from even happening, althought of course I am likely to run into a whole set of new problems, like transistor matching for one.
Nige, technically that was a Philips circuit at the start, but no more. Philips has no copyright on it, similar circuits have been used by other before and after them, and it has changed by more tha 33%, which according to international copyright laws, voids copyrights, it is now considered to be something else.
My gripe with a circuit like that is being large it doesn't have to have the fancy stuff . A few extra PSU sections instead ( same goes for power amps ) . KISS is the engineers watchword , Keep It Simple Stupid .
My output transformer idea might have one extra advantage . That is to run more in class A whilst at a good SOA compromise . Give it enough output devices to cope when the transformer has gain . In fact a NAD 3020 is about the right place to start . If someone says transformers are expensive I would say look at the market prices of some of this stuff . $300 should be enough .I started to think how the volume might work . A small computer always adjust the loop gain as the next step is chosen . Thus the volume seamlessly increases . This might keep the steps to 6 . Feedback loop volume usually doesn't work because the amplifier changes character too much with gain ( that is if distortion stays OK ) . Make it over a small range should change that . An amp I designed years ago used a 2 stage volume . First one conventional and the upper section relaxed the feedback loop . Just when the amp needed it most life gets easier . This idea I think is mine alone ? It is that most amps have ideal gain . However in reality it is two ideal gains . One for serious listening and the other for fun . I gave my friend an amp of 150 W . She complained it had hiss with horn speakers intended for 8 W . When I said what planet was that idea from she said . The high end guys insist we use speakers like this . The gain of this amp was 100 and still very low distortion . In the end she used some Swiss PA speakers that I thought were out of this world good . Not unlike ATC .
The bigger advantage of low voltage rails is the number of devices to choose from , admittedly a shrinking collection .
I dare say a bridge is possible . +/- 35 V rails gives a very wide choice of high current caps .
I would do the Dvv route as it is as good as class G for this application and is so simple
Did you ever try running your amps at 50% rails ? That would be to set up the PSU in four sections . If so a very nice amount of class A power could be had . 2 x 10 000 uF becomes 1 x 5000 uF in series and 1 x 20 000 uF in lower voltage mode . These truths are universal so shows the potential . You could use an octal valve base with different plugs to effect the changeover . I usually have 10 K across each cap to slowly discharge them . That is 8 minutes discharge . That would be a nice USP . How many people buy what they don't need ? How nice to find the better amp is the one you own ?
My output transformer idea might have one extra advantage . That is to run more in class A whilst at a good SOA compromise . Give it enough output devices to cope when the transformer has gain . In fact a NAD 3020 is about the right place to start . If someone says transformers are expensive I would say look at the market prices of some of this stuff . $300 should be enough .I started to think how the volume might work . A small computer always adjust the loop gain as the next step is chosen . Thus the volume seamlessly increases . This might keep the steps to 6 . Feedback loop volume usually doesn't work because the amplifier changes character too much with gain ( that is if distortion stays OK ) . Make it over a small range should change that . An amp I designed years ago used a 2 stage volume . First one conventional and the upper section relaxed the feedback loop . Just when the amp needed it most life gets easier . This idea I think is mine alone ? It is that most amps have ideal gain . However in reality it is two ideal gains . One for serious listening and the other for fun . I gave my friend an amp of 150 W . She complained it had hiss with horn speakers intended for 8 W . When I said what planet was that idea from she said . The high end guys insist we use speakers like this . The gain of this amp was 100 and still very low distortion . In the end she used some Swiss PA speakers that I thought were out of this world good . Not unlike ATC .
The bigger advantage of low voltage rails is the number of devices to choose from , admittedly a shrinking collection .
I dare say a bridge is possible . +/- 35 V rails gives a very wide choice of high current caps .
I would do the Dvv route as it is as good as class G for this application and is so simple
Did you ever try running your amps at 50% rails ? That would be to set up the PSU in four sections . If so a very nice amount of class A power could be had . 2 x 10 000 uF becomes 1 x 5000 uF in series and 1 x 20 000 uF in lower voltage mode . These truths are universal so shows the potential . You could use an octal valve base with different plugs to effect the changeover . I usually have 10 K across each cap to slowly discharge them . That is 8 minutes discharge . That would be a nice USP . How many people buy what they don't need ? How nice to find the better amp is the one you own ?
6 w/8 ohm ....
Power numbers are vastly over rated vs stability and guts, what gives an amplifier life ? The BAT VK500 was (250w/8ohm) as dead as nails @ 4 ohms or on ML/ESL's (embarrassingly bettered by an adcom on these) vs Levinson 27 or Krell ksa-80 , yet both had less than half the 8 ohm power and yes Nige it was bettered somewhat by the Sunfire amp.
Spectral has a lot of life too and very meager (comparatively) in the output stage and SOA , yet music ..
NP over builds and i have never had one issue , nada, no failures on any Threshold or his PASS stuff, that's over 25 yrs. Spectral focus is wide bandwidth , the others to some extent , but with brute force PSU, is there a magic formula..?
I can only suggest what works after trying most of the farm, even if only doing 25w/8R..
Spectral .....
Attachments
Last edited:
I never liked class H. Switching the power supply voltage in the middle of the signal?
It's a make-do to get high power with small heatsinks, that's all.
jan
That's what I tried to tell you, Wayne, obviously with little success.
There are many ways how to use output devces, some are less, and others are more EFFICIENT. In my view, most are not effcient enough.
For example, I have found far too many cases when an amp's output turned out to be limited by the drivers, rather than the output devices. This is of a more recent date. Beogre, output devices had much smaller beta factors, so like it or not, you HAD to use more powerful drivers or you wouldn't make the nomial output limit. The new, much faster power devices, also have much bigger and more even across the band beta factor, so designers became lazy and started using puny drivers.
H/K was a shining counter example. They not only avoided falling into that trap of basically complacency, but in fact developed and used extremely efficient NPN/PNP and PNP/NPN two transistor circuits, with local feedback, and with a tremendous composite beta factor.
Lately, they went back to the more conventional 3 stage output, meaning predriver, driver and output section. But a Locanthi circuit is a much better solution than the usual 2 way, driver and output, solution, for a number of reasons. It offloads the VAS, which is now much less influenced by what's going on after it, the overall beta factor is much larger and is programmable, the bandwidth is improved and the output impedance is reduced before any global feedback is applied.
All you need to do is to use robust output devices which you know inside out, so you KNOW, not guess, how far you can push it. This is precsiely the reason why I stick to Motorola/ON Semi and Toshiba output devices - I know them and have seen them do quite a bit more than I would have expected from them, time and time again.
Somebody else, say Milan Karan, will know someone else's devices, specifically Karan swears by Sanken output devices.
Just as an example, old TI BD 249/250 C trannies would blow you over in well applied devices. Nominally rated at 25 A is already way over what you'd ever expect from a dinky little 125W trannie, but even if you did, it will shock you further by delivering all of 40 A in peaks.
Imagine, you are listening to a nominally 75W/8 Ohm amp, and you decide to gun it a little. Instead of melting out, it rocks those speakers like they may have never been rocked before. It seems that they will do anything into anything you decide to throw at them, as if they were invincinble. Obviously, I am assuming a solid PSU behind them.
And such products were made once, by people you'd never ever expect it from, the ultra popular makes like Grundig. Their V5000 integrated amp was such a product. Uses two pairs of BD 249/250 C per channel, rated at 70/120W into 8/4 Ohms, it drives Apogees much bigger and incomparably more expensive amps struggle with. A one-off product, unfortunately, produced 1980-1983, thereafter Philips bought them out and their amps became sexed up Marantz amps in drag.
I have always believed that a surprisingly large number of German made amps somehow always sounded better than much more expensive Japanese offering, but in fact, the Japanese amps were only more sexed up from the outside. If the impedance drops below 6 Ohms, they struggle, but the Germans were designed to drive 2 pairs of speaker in parallel, and German DIN 45000 standards stipulated that speakers shall be with a 4 Ohm impedance. In other words, they were designed to interface with 2 Ohms straight off. The trade-off was in their relatively small 8 Ohm power ratings.
There are many ways how to use output devces, some are less, and others are more EFFICIENT. In my view, most are not effcient enough.
For example, I have found far too many cases when an amp's output turned out to be limited by the drivers, rather than the output devices. This is of a more recent date. Beogre, output devices had much smaller beta factors, so like it or not, you HAD to use more powerful drivers or you wouldn't make the nomial output limit. The new, much faster power devices, also have much bigger and more even across the band beta factor, so designers became lazy and started using puny drivers.
H/K was a shining counter example. They not only avoided falling into that trap of basically complacency, but in fact developed and used extremely efficient NPN/PNP and PNP/NPN two transistor circuits, with local feedback, and with a tremendous composite beta factor.
Lately, they went back to the more conventional 3 stage output, meaning predriver, driver and output section. But a Locanthi circuit is a much better solution than the usual 2 way, driver and output, solution, for a number of reasons. It offloads the VAS, which is now much less influenced by what's going on after it, the overall beta factor is much larger and is programmable, the bandwidth is improved and the output impedance is reduced before any global feedback is applied.
All you need to do is to use robust output devices which you know inside out, so you KNOW, not guess, how far you can push it. This is precsiely the reason why I stick to Motorola/ON Semi and Toshiba output devices - I know them and have seen them do quite a bit more than I would have expected from them, time and time again.
Somebody else, say Milan Karan, will know someone else's devices, specifically Karan swears by Sanken output devices.
Just as an example, old TI BD 249/250 C trannies would blow you over in well applied devices. Nominally rated at 25 A is already way over what you'd ever expect from a dinky little 125W trannie, but even if you did, it will shock you further by delivering all of 40 A in peaks.
Imagine, you are listening to a nominally 75W/8 Ohm amp, and you decide to gun it a little. Instead of melting out, it rocks those speakers like they may have never been rocked before. It seems that they will do anything into anything you decide to throw at them, as if they were invincinble. Obviously, I am assuming a solid PSU behind them.
And such products were made once, by people you'd never ever expect it from, the ultra popular makes like Grundig. Their V5000 integrated amp was such a product. Uses two pairs of BD 249/250 C per channel, rated at 70/120W into 8/4 Ohms, it drives Apogees much bigger and incomparably more expensive amps struggle with. A one-off product, unfortunately, produced 1980-1983, thereafter Philips bought them out and their amps became sexed up Marantz amps in drag.
I have always believed that a surprisingly large number of German made amps somehow always sounded better than much more expensive Japanese offering, but in fact, the Japanese amps were only more sexed up from the outside. If the impedance drops below 6 Ohms, they struggle, but the Germans were designed to drive 2 pairs of speaker in parallel, and German DIN 45000 standards stipulated that speakers shall be with a 4 Ohm impedance. In other words, they were designed to interface with 2 Ohms straight off. The trade-off was in their relatively small 8 Ohm power ratings.
Dvv . At No 12561 you questioned the slew rate or slew limiting of my primitive amp . It is 1mA/60pf ( 47 + Cob ) or better . That relates to about 17 V/uS ( Jung method ) . Not a bad slew rate if 80 kHz bandwidth and 80 W rms 8 R . My 7 transistor 100 W amp is 36 V/uS symmetrical slewing double PNP + NPN VAS . I have reverted to tail resistor for it . Bias set 100 mA .
The primitive amp was the antithesis of an amplifier suffering TID . Very sweet and light footed . Far better and similar to a Mission Cyrus amp .
The 1973 Bob Stewart amp followed practice of the day . 0.7 V at 4K3 = 0.16mA to drive the VAS . 100 pF and 12 pF internal Cob = 1.5 V/us ( 80 watts also ) . That is barely enough for classical music . I suspect these arrangements were to force true trans-conductance ? In my opinion this is the so called transistor sound . TID in HC Lin amps . Stupid as it is the one type of amp that should avoid it as the complexity is so low . Dogma I suspect . VAS Z in I guess 400 R trying to be fed by 4K3 .
I feel my design is free of TID because it has a 47 R emitter resistor to the VAS . Nowadays I would do it to the input pair in preference . You can choose this VAS resistor by ear . As it increases the sound becomes sweeter . Conversely the slam of the sound seems to decrease . All of this is logical . If you want to be good engineer correct the resistor feeding the VAS for least second harmonic . I could have offered a slam control on it as the LTP will servo the adjustment . Bass plus second harmonic is rather nice . The treble would be more incisive although metallic . Thus we can have dial a sound .
TID is a real issue in PA . Microphones and instruments can have large HF output .
At home it is RFI that matters .
Class H is useless for distorted and compressed signals . Either class AB or D will suit better . H will be fine , it's complexity will be wasted on such signals .
The primitive amp was the antithesis of an amplifier suffering TID . Very sweet and light footed . Far better and similar to a Mission Cyrus amp .
The 1973 Bob Stewart amp followed practice of the day . 0.7 V at 4K3 = 0.16mA to drive the VAS . 100 pF and 12 pF internal Cob = 1.5 V/us ( 80 watts also ) . That is barely enough for classical music . I suspect these arrangements were to force true trans-conductance ? In my opinion this is the so called transistor sound . TID in HC Lin amps . Stupid as it is the one type of amp that should avoid it as the complexity is so low . Dogma I suspect . VAS Z in I guess 400 R trying to be fed by 4K3 .
I feel my design is free of TID because it has a 47 R emitter resistor to the VAS . Nowadays I would do it to the input pair in preference . You can choose this VAS resistor by ear . As it increases the sound becomes sweeter . Conversely the slam of the sound seems to decrease . All of this is logical . If you want to be good engineer correct the resistor feeding the VAS for least second harmonic . I could have offered a slam control on it as the LTP will servo the adjustment . Bass plus second harmonic is rather nice . The treble would be more incisive although metallic . Thus we can have dial a sound .
TID is a real issue in PA . Microphones and instruments can have large HF output .
At home it is RFI that matters .
Class H is useless for distorted and compressed signals . Either class AB or D will suit better . H will be fine , it's complexity will be wasted on such signals .
Blomley amp and MOSFET's
I have just looked up the original Hitachi MOSFET curves in class B ( not AB ) . These include a comparison with bipolar transistors . The MOSFET's looking identical to the Blomley curves he shows , thus far better than standard bipolar curves . And guess what . The Blomley amp allows better feedback margins just like MOSFET's . The difference is the MOSFET's need one single resistor and the Blomley circuit a bag full of components that need selecting .
All I need for my headphone amp is some small L MOS ( not D MOS ) . I don't think they exist .
Dvv . Did you ever try MOS FET's to drive your outputs . The maths I have just done says they will work and support your argument . Exicon 10P/N20 . You might even steal some output current from them ( 2 amps ? ) .
I have just looked up the original Hitachi MOSFET curves in class B ( not AB ) . These include a comparison with bipolar transistors . The MOSFET's looking identical to the Blomley curves he shows , thus far better than standard bipolar curves . And guess what . The Blomley amp allows better feedback margins just like MOSFET's . The difference is the MOSFET's need one single resistor and the Blomley circuit a bag full of components that need selecting .
All I need for my headphone amp is some small L MOS ( not D MOS ) . I don't think they exist .
Dvv . Did you ever try MOS FET's to drive your outputs . The maths I have just done says they will work and support your argument . Exicon 10P/N20 . You might even steal some output current from them ( 2 amps ? ) .
Last edited:
D,
Telling me,
Obvious you're not getting me and To be polite , HK is paper tiger and not at the level being discussed...
Paul McCartney has one ( HK ) . There again maybe he is a Paper Tiger ? Garrard Lab 80 also and JBL's .
Just want to correct my Bob Stewart amp slewing rates . It is bound to be higher than I say as the VAS input impedance dominates . Sadly no reference points were given . Bob only showed it as typical . I was fooled by the shorted other side to get balance . This is another good reason to float the VAS . The maths is easier . Who knows in reality that amp might have been OK .
- Status
- Not open for further replies.