[ TNT | Who we are | Listening tests | HiFi topics | HiFi Shows | Tweakings | Inter.Views ]

Solid State Power Amplifier Supply

Part 3

[Italian version]
[Power Supply Exposed]


Large reservoir capacitors notwithstanding, for really high quality power supply we need additional filtering. As seen on the Diagrams, small value capacitors, in the 100-220nF region are used to filter out the very high frequencies, where large capacitor efficiency has already dropped to unacceptable levels. This is no big deal, you can't really go wrong here whatever you do - the only way to go wrong is not to do it, and even that may not be critical, depending on actual circuit layout (if your large filter caps are next to the output transistors, for example, the effects will be smaller, but they will still be there).

Off hand, use 100nF for transformer filtering and 150nF for large capacitor bypassing. Do not be stingy, lay out some good money and go for polypropylene capacitors in the 100V or more range.

If you are using single capacitors for every supply line, I would strongly suggest you also use the RC filter as shown on Diagram 4. In fact, I would strongly recommend you use it in any case, as it will ALWAYS improve your sound - how much or little depends on many factors, and obviously, the better components you use downstream, the less it will be needed, and vice versa.
But you don't need to take my word for it - try it and see, sorry, hear for yourself. Play some music without, then with the filters and you will have improvement in your treble range, from small to immediately noticeable.

Filters have an added complication at no extra cost, that of placement. It's not quite the same thing where you put them, just after the caps, after the supply lines on the PC board, next to the output transistors, etc. My suggestion would be to place the small value caps C7 and C8, as well as the RC filter after them on the PC board. That way, they also filter out whatever may have been picked up by the wiring if you have it.

How I'd do it

Looking over some notes from the last time I went for a 50W/8 Ohm amplifier, I notice I used a circuit as shown on Diagram 4. It follows the general outlines given above.

But, what if your requirements are different? What if you want to run a pure class A amplifier, or you want a 150W/8 Ohms amp using higher than usual bias levels? In short, theory is nice, but nothing beats hard practice.

If you're into pure class A amplification, you're going to have to pay some serious money. Class A draws constant currents, and they are not small, though of course, they depend on the required power output. For example, 50W/8 Ohms will require under 4 Amperes, while 100W/4 Ohms will require more than 7 Amperes - and this are CONSTANTLY drawn currents, not peaks! This means that you cannot count on your capacitors for surges, but must concentrate on their filtering abilities. Drawing large constant currents means you need much, MUCH bigger filtering capacities. Practically, this means at least double of what you'd need for a class AB design, and more if at all possible. For a 50W/8 Ohms class A amp, I'd say 20,000uF per supply line, assuming a dual mono construction - a total of 80,000uF.

The last obvious question is why should you be interested in this when you already have an amplifier? Well, you can always make a new power supply for it and bypass its own, if you feel it's inadequate (and on commercial units, it almost always is). To do that, you'll need to do some disconnecting - not a job I'd recommend to novices. You will need to disconnect the lines feeding the power amplifier stages, which means you need to know which lines are they, how much current is needed, what voltage is required (from off load, higher voltage to full load on, lower voltage).
You will also need to install at least one IEC socket on the back of your unit which is to receive power from an outboard power supply. In short, a project which generally yields good results, but is not very easy to do, as it includes both electrical and mechanical work.

Remember, power kills - these are voltages quite capable of killing the unsuspecting!!!
If you're not absolutely sure, refer to qualified personnel.

[Power Supply Schematic 4]

This brings us to diagram 4. It shows one channel only, but it can be used for the entire amplifier if necessary, possibly with some value changes. As shown, it uses separate bridge rectifiers for each rail, a philosophy I deeply agree with, my ears especially.
The rectified voltage is filtered by two large filter capacitors, mounted in parallel - this doubles their capacity and current handling ability, while halving their residual output impedance. Voltage is taken from +Vcc and -Vcc and taken to the power amplifier.
There, before reaching the output devices, no matter what they may be (bipolar, MOSFET, IGBT), a smaller value cap of 100uF in parallel with a 100nF capacitor filter out any RF signals or other disturbances which may have been collected en route.

Finally, an RC network, 1 Ohm/17W in series with 680nF, gets rid of any residual capacitor inductance. As far as I know, this was first used in 1972 by Prof. Matti Otala, in his legendary IEEE text on Transient Intermodulation. From that time, I have tried it many times, with I must admit varying results - sometimes almost insignificant, sometimes audible.
If audible, it tends to produce somewhat clearer, more coherent high tones, with benefits for ambience as well. If inaudible, it will quite surely make any amplifier more stable - the poorer quality its capacitors, the more benefit you will receive. But even with very high quality capacitors, there's always some residual inductance left over, perhaps small, but it's there - and it shouldn't be.

Try if you can to have the power amp filtering section as near to the power stage as possible, not quite touching, but the first next thing. However, if you're building an outboard power supply, you will probably have to put it inside an add-on box - not ideal, but still much better than without it.
In that case, try also replacing the local filter capacitor(s) which must exist on the original power amp board with better quality polypropylene or polycarbonate units.

How to work out the values? Simple, really. First, make up your mind whether you're supplying both channels from the same source, or if you're going for dual mono, which means each channel will have its own power supply.
As a general rule, if you're using a single source, take one channel values and multiply them by 2, and then multiply the result by 1.2 - this will give you a safety margin of 20% in case both channel suddenly require much current.

Second, take into consideration the actual power you have at your disposal. There's no sense in wasting money, just as there's no sense in falling short quality-wise. Be RATIONAL! Look at your amp's specifications, see what its manufacturer says it will do, and use that as a rule of thumb. Next, take a look at your actual output stage.
Let's say, for the sake of argument, that you have a SEPP (Single Ended Push-Pull) output stage, which means one NPN and one PNP transistor per channel, which is what most commercial units have inside up to around 60-70W/8 Ohms.

Recently, manufacturers will use anything in such stages, from a pair of cheap 80W transistors, to better and more expensive 130W transistors. They will almost always be plastic case types, as this makes mounting easy and eliminates wiring, but also limits the available power.
Such devices are rated at 8-12A continuous, 12-20A peak, sometimes a little more. How much will they actually deliver depends on many factors, the design being the dominant, and the power supply usually being the subsidiary. In other words, usually the over current protection will trigger BEFORE you run out power supply juice - but by then, it's very likely your power supply will be in some heavy weather.
I do not advise changing protection circuit limits unless you have detailed schematics, know the devices, know a lot of electronics, have a test lab and are possibly willing to change the devices for better ones. Bear in mind that protection circuits can NEVER be simply calculated, they are always the result of initial calculation and much, much practical experimenting, and are thus absolutely never as simple as they may appear at first glance.

[Yamaha Power Supply]

By way of example, Yamaha uses two 80W pairs per channel in their AX592 integrated (nominally 100W/8 Ohms),

[Harmon Kardon Power Supply]

Harman/Kardon use one pair of 130W devices per channel in their HK3270 receiver and two pairs of 130W devices per channel in their HK680 integrated amplifier (85W/8 Ohms),

Marantz uses similar pairs in their 7000 integrated (95W/8 Ohms), as does NAD in their 317 integrated amp (85W/8 Ohms), etc.

Since those devices are capable of rather large impulse currents, obviously they need to be well fed with current and voltage. In some cases, including some mentioned above, this is not quite so, due to reasons of economy - the Yamaha in particular would benefit from a beefier power supply, not to mention changing the output devices.

So, assume short term peaks up to the absolute limit as set by the device manufacturer - you can't go wrong there. If that's more than it will actually ever deliver in a given design, then at least it will do so without any trouble from the power supply, but in reality, it will always be less, if for no other reason than because such ratings are given for case temperatures of 25 degrees centigrade, which they will never be under normal operating conditions, and will consequently deliver less than their maximum power. Look up the transistors in comparative tables (ask if you don't have them), and if they are rated at say 12A continuous, assume you need that much current. Not very scientific, but very, very safe.

Right - 12A of current means at least 10,000uF per supply line in mono and 20,000uF in a single source. Multiply that by square root of 2 (to obtain peak values), or by 1.41 and you have 14,100uF - a non-standard value, use 15,000uF as standard value, or better yet, use two 6,800 or 8,200 uF caps in parallel. If you were from Sony, you would use 10,000uF in parallel with 4,700uF; there is merit in this concept too, as the smaller cap is better at filtering out higher frequencies, but I must admit I have never used this configuration. Also, smaller caps charge and discharge faster than larger ones, so speed may also be an issue here. For single source, multiply all by 2, then by 1.2.

Charging those caps is no small task and is likely to really stress the diodes in the bridge rectifier. This means you will need some very powerful rectifiers. Some, notably the Scandinavians, say fast switching diodes give better sound - I not only disagree with that in case of high power amplifiers, but in fact believe the opposite is true.
Standard bridges are available in metal packages, which makes their cooling easy and efficient; they are cheap, which makes using more easy and inexpensive. By using two, we share the load between them, thus doubling their capacity, or halving their load. In both cases, they become more reliable and are quite capable of charging the capacitors which follow.

My personal favorite is KBPC04-25; "04" defines voltage, in this case 400V, and "25" defines their current capability, here 25 Amperes. I like them for two reasons - one, they have shown themselves to be extremely reliable, and two, they have most reasonable prices.
They are packaged in metal cans with a center mounting hole, and are thus easy to affix and cool down, using the case bottom plate as the cooler. Of course, there are more powerful models available.

Capacitors are a separate story. Many are on offer, practically all of them will get the job done, but there's no doubt that good capacitors will impart better sound to the amplifier than low cost, low quality ones, no matter what else we may do. So, no money saving there. What to go for? I can only speak of those I have practical experience with, and these are:

Elna.   Well, that they are good is beyond doubt, but in my view, they are not half as good as some others, they are just more fashionable and have a much better marketing department.
This is not to say I think them poor - by no means, I'd stick them in any of my units without thinking twice - but only if I failed to find some of the competition. But you won't go wrong with Elna, series black and series blue, not to mention the Starget and Cerafine series.

Philips.  I have excellent experience with Philips caps, even those strictly commercial units. The new T-network capacitors are fantastic, but are very hard to come by and not at all cheap. Ideal for audio.

Siemens.   My favorite, however, only the very expensive, excellent Sikorel series. It can be shown that these are exceptionally fast, ultimate reliability units, but at no small price. Also, no snap-in caps in this series, so you need external fixing and wiring, which is a mixed bag. Any wiring is best avoided if possible, but on the other hand, these caps are affixed by rather heavy duty screws, and vibrate much less than the snap-in type. If you can, go for these, but be warned - they are very expensive.

Roederstein (Roe).    German made, these are good capacitors, but for smaller voltages only. A safe buy you won't regret, but not up to the highest standards.

Fischer & Tauche.   Another German brand, and one I often recommend because of its ideal price/performance ratio. However, two series are offered in same sizes and voltages; the first is purely commercial and is nothing special, it is the second you want. The second series is about 30% more expensive and is much better suited to audio. Also, these are bigger units. For example, the commercial series 10,000uF/63V cap is 35x70 mm (dia. x height) versus 50x80 mm for the audio series - it may not look like much, but that's 63% more volume, used for that much more foil inside.
They are about as fast as Elna series black (for Audio) caps, but cheaper, at least as far as I know. They use one large M8x12 or M12x16 screw affixed to the capacitor body; not ideal, but still better than the snap-in types.

Aerovox.   I tried only one pair of these, the T-series, and found them to produce a rather dull sound. Not for me, thank you, no matter what the paper press, notably British, says about them.

Whichever you go for, try putting them in parallel.

Now, for some easy-to-read tables.


Mono, or Per Channel for Dual Mono, Class AB
Amp Rating Transformer Bridge Rectifier Capacitors
20/40 Watts into 8/4 Ohms 80 VA 2 x B80C2200 4 x 3,300 uF
40/80 Watts into 8/4 Ohms 120 VA 2 x B80C3300 4 x 4,700 uF
60/120 Watts into 8/4 Ohms 170 - 200 VA 2 x KBPC04-25 4 x 6,800 uF
80/160 Watts into 8/4 Ohms 220 VA 2 x KBPC04-25 4 x 8,200 uF
100/200 Watts into 8/4 Ohms 300-300 VA 2 x KBPC04-25 4 x 10,000 uF
125/250 Watts into 8/4 Ohms 400 VA 2 x KBPC04-25 4 x 12,000 uF
150/300 Watts into 8/4 Ohms 500 VA 2 x KBPC04-35 4 x 15,000 uF
200/400 Watts into 8/4 Ohms 600 VA 2 x KBPC04-35 4 x 20,000 uF

Mono, Class A
Amp Rating Transformer Bridge Rectifier Capacitors
10/20 Watts into 8/4 Ohms 80 VA 2 x B80C3300 4 x 6,800 uF
20/40 Watts into 8/4 Ohms 100 VA 2 x KBPC04-25 4 x 8,200 uF
30/60 Watts into 8/4 Ohms 120 VA 2 x KBPC04-25 4 x 10,000 uF
40/80 Watts into 8/4 Ohms 150-200 VA 2 x KBPC04-25 6 x 8,200 uF
50/100 Watts into 8/4 Ohms 200 VA 2 x KBPC04-25 6 x 10,000 uF

High Load Tolerance Class AB, Mono or Per Channel Dual Mono
Amp Rating Transformer Bridge Rectifier Capacitors
30/60/120 Watts into 8/4/2 Ohms 150-200 VA 2 x B80C3300 4 x 4,700 uF
50/100/200 Watts into 8/4/2 Ohms 300 VA 2 x KBPC04-25 4 x 8,200 uF
75/150/300 Watts into 8/4/2 Ohms 450-500 VA 2 x KBPC04-25 4 x 10,000 uF
100/200/400 Watts into 8/4/2 Ohms 600 VA 2 x KBPC04-35 4 x 15,000 uF
150/300/600 Watts into 8/4/2 Ohms 800 VA 2 x KBPC04-35 6 x 10,000 uF

My favorite source for components is Munich, Germany. It's relatively near to where I live, the prices are reasonable and most items are on stock. Therefore, all info on availability and prices is based on my Munich sources; however, this should not be interpreted as any specific recommendation, you are free to find better sources as you will.

My prime source is a company called Buerklin (site: http://www.buerklin.de, e-mail: info@buerklin.de), so to help along, all prices and item numbers quoted refer to their '99 catalogue. Over the years, I have found that their prices change very little and are a good indication of general market prices; as usual, some items are cheaper and some dearer than elsewhere.


B125C3700/2200, 125V, 2.2A free, 3.7A on cooler, item 55 A 860, price DEM 3,20

B125C5000/3300, 125V, 3,3A free, 5 A on cooler, item 55 A 908, price DEM 3,80

PSB 35/02, 200V, 35 A on cooler, item 57 A 260, price DEM 29,40


Siemens Sikorel, 10,000uF/63V, 36x108mm, item 16 D 670, price DEM 47,80

Fischer & Tausche, 10,000uF/63V, 50x80mm, item 16 D 944, price DEM 20,10

Fischer & Tausche, 15.000uF/63V, 50x100mm, item 16 D 946, price DEM 26,70

Toroidal transformers

Well, get them where you can. I have a local supplier who is very good and has reasonable sizes, and is also willing to make whatever I want at no surcharge over the basic transformer size price. I generally use his 800VA units, which weigh in at 6.9 kilos and consume merely 14 mA of quiescent current, a sure indication that he uses good quality cores.
I pay DEM 140 per piece, which is quite reasonable, given that I have special requests. Also, proper shielding is available at a small surcharge, again something I just love to use. But transporting them across the continent is neither cheap nor practical, Anyway, you might be using his products without even knowing it, since about 99% of his production is exported and labeled differently, sometimes with rather well known labels (I promised not to tell on anyone, so I'll let you wonder, especially if you live in Britain and Scandinavia).

I hope all this has cleared up some points and perhaps helped you learn a few interesting details. That's what we strive for on TNT-Audio.

© Copyright 2000 Dejan Veselinovic - http://www.tnt-audio.com
HTML Editing by Scott Faller

Go back to Part 2
Go back to Part 1

How to print this article

[ TNT | Who we are | Listening tests | HiFi topics | HiFi Shows | Tweakings | Inter.Views ]