I tell you Frank, the Otala/Lohstroh paper and proposed amp quite literally changed my life.
And I blush at the sheer stupidity of selling my sample. Now I have to make another one.
The other big turning point in my life was meeting with that German made LAS Mega 1 power amp. Also full of really good thinking and logic. Just looking at the schematic, you wouldn't expect much of it, but switching it on changes literally everything.
Double idiot badge for me here, I sold my sample of that one as well.
There are only single V+ and V- leads, but the heatsinking tab is connected to V-, that's probably what you were thinking of. With connections, I trimmed every leg to the absolute minimum that could still be soldered to, and connected the power leads to the edges of the copper planes - the chip and copper planes formed a right angle, with the heatsink running vertically, chimney style, forming a T with the planes.
The funny thing is, that many people wouldn't "get" it when listening - correct sound can be, so ... unimpressive on first hearing; you wind up the volume and they still don't get it - until a sudden transient whacks them over the head ...
Agreed.
SOMETHING has changed over the years. It seems to me that from a quest to make audio as neutral and invisible on its own, so we hear only the musical content, things have degraded to customers wanting to be impressed by whatever. And on the first bat, too.
In my view, it is a sad state of affairs to discover that my 36 year old Marantz 170 DC power amp with its 3250B preamp ends up sounding better, and often way better, than their own higher market tier offerings today. And worse, that old combo sounds more neutral and way more believable than most of their competition does today.
And if I swop the 3250B preamp for a Luxman C-03 preamp (not yet refreshed), I end up with really serious sound I have not heard from quite a few modern High End products.
Something is not right in all that.
And the only thing I changed from the original when refreshing that Marantz amp was that I threw out the dual concentric 2x12,000 uF single cap and replaced it with two 22,000 uF BC Components caps. I practically doubled the capacitance.
MOS FET's
Maybe class D gives us an insight into why MOS FET class AB amps work as well as they do. What they lack in bipolar good manners they make up for in speed.
I think I have found an answer. High slew rates inside a bipolar amp make up for what isn't happening at the output. Try a bipolar in a class D to find out how it matters. The special FET's of class D are one step further down the road. The AB FET is moderately linear when feedback applied and moderately fast. Also the only way to deal with the minority carriers that clugg up the base emitter region of bipolar is an iron grip VAS cum driver. The FET isn't bothered by that and merely forms a pole in the response. As the output is fast a slightly slugged VAS is a minor problem.
Blomley designed an amplifier that divide the class B amplifier further back so as to have two more perfect halves. The curves he gave look remarkably like Hitachi curves given for MOS FET's ( cross conduction ). The suggestion being that the MOS FET is a more ideal class B device. Class D sort of says that would be right.
http://www.keith-snook.info/wireles.../New Approach to class B Amplifier Design.pdf
A place where MOS FET's are used which is daft is tube amps , high voltage would be a valid excuse. Most designers will not touch a bipolar current source of sink. My tests say that is bonkers. As close to a resistor as you might ever get. Humble MJE340/350 ideal as 6 MHz is fine for that. Equally a tube in cathode follower is bonkers and not very good sounding to my ears. Use a bipolar.
One thing I found from tube amp current sources is they aren't low distortion if being careless. I think it is Early effect. When wrapped in a feedback loop you will never see that. The distortion is of the nicest sort. Take a LM317 as a current sink for an output tube. Clone it with a standard current sink. You will find it has has more second harmonic distortion. Play about a bit and it will not. This usually means loosing 1 V. The LM317 is horribly good . I just can't bare the thought of doing that. The last amp I built had stage one pentode current source , G2 supply and EL 34 output current sink all on the same totem-pole quiet supply. There's economy for you.
The complimentary feedback ( npn/PNP pnp/NPN) stage I showed before as far as I can tell in class A has no sound. A MOSFET without help does. It is the MOSFET's acceptance of feedback that gives it this small fighting chance. It has no place in a class A design.
If wondering why I did a 1964 amp in MOS FET, it was to see if repairing old germanium amps would be worthwhile. Loose the drivers and put in FET's . Bias will be OK. I chose to do it the hard way and say what if ?
Maybe class D gives us an insight into why MOS FET class AB amps work as well as they do. What they lack in bipolar good manners they make up for in speed.
I think I have found an answer. High slew rates inside a bipolar amp make up for what isn't happening at the output. Try a bipolar in a class D to find out how it matters. The special FET's of class D are one step further down the road. The AB FET is moderately linear when feedback applied and moderately fast. Also the only way to deal with the minority carriers that clugg up the base emitter region of bipolar is an iron grip VAS cum driver. The FET isn't bothered by that and merely forms a pole in the response. As the output is fast a slightly slugged VAS is a minor problem.
Blomley designed an amplifier that divide the class B amplifier further back so as to have two more perfect halves. The curves he gave look remarkably like Hitachi curves given for MOS FET's ( cross conduction ). The suggestion being that the MOS FET is a more ideal class B device. Class D sort of says that would be right.
http://www.keith-snook.info/wireles.../New Approach to class B Amplifier Design.pdf
A place where MOS FET's are used which is daft is tube amps , high voltage would be a valid excuse. Most designers will not touch a bipolar current source of sink. My tests say that is bonkers. As close to a resistor as you might ever get. Humble MJE340/350 ideal as 6 MHz is fine for that. Equally a tube in cathode follower is bonkers and not very good sounding to my ears. Use a bipolar.
One thing I found from tube amp current sources is they aren't low distortion if being careless. I think it is Early effect. When wrapped in a feedback loop you will never see that. The distortion is of the nicest sort. Take a LM317 as a current sink for an output tube. Clone it with a standard current sink. You will find it has has more second harmonic distortion. Play about a bit and it will not. This usually means loosing 1 V. The LM317 is horribly good . I just can't bare the thought of doing that. The last amp I built had stage one pentode current source , G2 supply and EL 34 output current sink all on the same totem-pole quiet supply. There's economy for you.
The complimentary feedback ( npn/PNP pnp/NPN) stage I showed before as far as I can tell in class A has no sound. A MOSFET without help does. It is the MOSFET's acceptance of feedback that gives it this small fighting chance. It has no place in a class A design.
If wondering why I did a 1964 amp in MOS FET, it was to see if repairing old germanium amps would be worthwhile. Loose the drivers and put in FET's . Bias will be OK. I chose to do it the hard way and say what if ?
The funny thing is, that many people wouldn't "get" it when listening - correct sound can be, so ... unimpressive on first hearing; you wind up the volume and they still don't get it - until a sudden transient whacks them over the head ...
There is a film with plasticine puppets called " The wrong trousers". Here we have the wrong speakers. Many amps do Taming of the shrew as their major function. I have just built some shrews. Rotel keeps them tame. That is handy as that's what Colleen has.
While I am aware of the technical benefits of MOSFETs, somehow I find that their sound tends to lack that last bit of "oomph!" in the bass region.
They sound a bit too polite to me.
However, it could very well be that I just haven't heard the really good ones yet. On the other hand, I believe quite a few designers, including John Curl and Pavel Macura here, still also prefer straight BJT output devices, even if they do use MOSFETs to drive them. Why is that, I wonder?
They sound a bit too polite to me.
However, it could very well be that I just haven't heard the really good ones yet. On the other hand, I believe quite a few designers, including John Curl and Pavel Macura here, still also prefer straight BJT output devices, even if they do use MOSFETs to drive them. Why is that, I wonder?
A planar bus structure uses the same principles as power, ground, and signal planes in a multi-layer pcb. When conductive planes are stacked, closely spaced, their loop inductance is reduced, while their mutual inductance is increased.
The bus planes have a certain loop inductance per square, scale independent. So, a large planar bus will have a similar inductance as a small bus of the same shape.
When closely spaced, the loop inductance is proportional to (d x l / w), where d is the spacing, l is the length, and w is the width. The behavior of an actual planar bus will depend on the details of the implementation, but it will usually be a lot better than a bunch of wiring, or even large bus bars.
OK thanks for the more detailed explanation - I am already familiar with the concept of power planes for PCBs so this sounds like a physically scaled up version of the same. However your original comment was in reply to my pointing out the difficulties of whole amplifier power supplies, not just the interconnects between source and load. So far it seems your 'easily achieve low inductance' remark only applies to the interconnect and not to the two ends of it, where ISTM the greater problems lie in achieving low inductance.
Yes, the devices can be the limiting factor. In high frequency power devices, the physical construction must be optimized to minimize the parasitic inductance inside the device. Much work has been done on this is recent years.
Some devices even eliminate the bond wires. Semikron has pioneered in this, check out their power modules.
Here's a link to an overview of the laminated bus, showing examples as actually built for various applications. A number of companies specialize in manufacturing these.
http://www.efo-power.ru/BROSHURES_CATALOGS/MERSEN/FUSES/Bus_Bar_Solutions.pdf
Right on, Richard, thanks for the addition. Agreed.
Overall, they just tend to sound a little rounded off, as if the sharp edges have been filed off, so to speak.
Sometimes, that's a boon, but other times, when you want that Billy Cobham to sound wild and biting, it just kills the enjyoment.
Perhaps I should consider an A/B design, where A is a MOSFET output stage, and B is the BJT output stage?
Boy oh boy, now that would be a novelty!Frankly, Richard, to me, nothing beats a few pairs of MJ 21195/21196 (250V, 16A continuous, 30A impulse, 250W, TO-3 metal can) trannies. For reasons which escape me, those transistors are capable of exceptionally integrated and most believable sound I know of, despite their relative old age now. Even their brothers, much newer MJL 3281/1302, TOP-204, 200W, 15/30A continuous/impulse, but Ft of >30 MHz and much more linear and higher beta curve, can't match them on sheer sound quality.
In fact, their only weak side is their drooping beta curve, which drops down to something like 12:1 at high power levels, which means they need a hefty driver. This would be MJE 15030/15031 for power levels of up to say 150/300W into 8/4 Ohms, NOT the MJE 15032/15033, which are higher voltage, but lower current capability (6A vs. 8A for lower down the line models). Don't even dream about 300W/4 Ohms, when your drive transistor current will exceed 150 mA at say 49 Vpeak, not if you want to keep some reserve for those Wayne-like low impedances.
And another thing. I find that most newer very high Ft devices are MUCH more susceptible to misbehavior than those older MJs, which are rated at "4 MHz minimum", but are in fact more like 5-6 MHz in the range I would use them in. Lastly, Motorola/ON Semi seem to have really gone to town in their manufacture, because their data sheet states that even if you take two random trannies for an NPN/PNP complementary pair, your net THD will not exceed 0.8% THD, and matching will reduce this to 0.08% THD. By "net" I mean just the distortion produced by them alone.
Sonically, they are wonderfully even top to bottom, it all comes out as a coherent whole, and it's almost palpable. That good.
For those who dislike metal can TO-3 packaging, the TOP-204 package is available, rated at 200W, with an MJL prefix.
To be fair, I am also very pleased with what I get from Toshiba's 2SC5200/2SA1943 as well, they have also never let me down (Toshiba, I mean).
IGBT I haven't even heard, let alone tried myself. I seem to remember a text by John Linsley-Hood related to an amp with them he had made, stating that some problems came up with the NPN side in particular, which needed to be taken care of for best operation.
The output devices were by Toshiba, designated with something like 204 or some such.
The output devices were by Toshiba, designated with something like 204 or some such.
All the current work on power devices are on power mosfets. You can get devices with on current ratings of 300A and RDS on of less than 1 milliOhm. However switching is different from linear operation.
I'm still not convinced that there needs to be any difference or limitation from using FET or bipolar as an amp, especially if there is feedback involved. Bipolar seems to get more complex in order to manage protection, biasing and to have enough drive and bandwidth. I do not understand why a bipolar would have "more "oomph!" in the bass region." if both devices are working inside a functioning feedback loop, otherwise what is the point in having the feedback? Has anyone measured anything that correlates?
I'm still not convinced that there needs to be any difference or limitation from using FET or bipolar as an amp, especially if there is feedback involved. Bipolar seems to get more complex in order to manage protection, biasing and to have enough drive and bandwidth. I do not understand why a bipolar would have "more "oomph!" in the bass region." if both devices are working inside a functioning feedback loop, otherwise what is the point in having the feedback? Has anyone measured anything that correlates?
Here is some info on current IGBT's http://www.ixys.com/Documents/AppNotes/IXYS_IGBT_Basic_I.pdf IXYS Website > Product Portfolio > Power Devices
They are well suited to motor drivers and the like but less for audio. The complements are not complements. Vsat is either around 2V or around 6V depending. They are high current and rugged so good for traction motors in things like locomotives. Ixys also has mostfets that could work in poweramps: 100A @ 10V, which could fry any voice coil I know of: http://ixapps.ixys.com/DataSheet/DS100397A(IXTK-TX210P10T).pdf http://ixapps.ixys.com/DataSheet/99186.pdf Driving the 10000pF gates will be the hard part.
They are well suited to motor drivers and the like but less for audio. The complements are not complements. Vsat is either around 2V or around 6V depending. They are high current and rugged so good for traction motors in things like locomotives. Ixys also has mostfets that could work in poweramps: 100A @ 10V, which could fry any voice coil I know of: http://ixapps.ixys.com/DataSheet/DS100397A(IXTK-TX210P10T).pdf http://ixapps.ixys.com/DataSheet/99186.pdf Driving the 10000pF gates will be the hard part.
- Status
- Not open for further replies.