Why does Leestereo insist on increasing power supply/decoupling capacitors?

[Leestereo]

I often get asked why should one consider judicious increases to the capacity of a power supply. I typically respond as such: It can be argued that the power supply is in the signal path, it's essentially in series with the output stage and hence does impact the sound of an amplifier. Nelson Pass uses the analogy of a water reservoir as the power supply and the output stage as the valve that controls the flow of the water: https://www.passdiy.com/project/articles/power-supplies. Hence, if you have a poor and/or inadequate supply, the quality and/or the quantity of water is affected. For an amplifier, a poor/failing power supply capacitor (e.g., low capacity, high ESR) may not be able to supply the voltage/current that is necessary for the output stage to perform as designed. IMO, this is primarily why improving the power supply (e.g., increasing capacity, lowering impedance/ESR) often does impact/improve the sound of an amplifier.

But didn't the original engineers of these vintage amplifiers know how to properly design a power supply? Well, yes, they very likely did, but were constrained by the capacitor technology of the day (e.g., capacitor size, ESR, ESL) and cost considerations. For example, in this excellent video demonstration from xraytonyb, increasing the power supply capacity from the stock 1000uF to 6700uF in a modest receiver significantly improved its performance wrt clipping behavior (~15 minute mark in video):

 

[Overundr1]

There is an interesting side note to your cap upgrades sir. I do not recall ever seeing a recommendation for a value increase that would exceed the inrush capability of the surrounding components. Couple that with lower esr values and a higher reserve for those ultra low frequency draw demands of modern recordings and the whole thing becomes a win-win in my opinion.
The flip side is it is becoming more of a challenge to find dimension-ally correct drop in replacements these days without going down the not so neat appearing rabbit hole.

 

[Scott Muni]

I did a quick experiment with my 331 receiver's power supply. It had a single 2200uF, swapped in a 4700uf.
Nothing else was changed, checked power on scope it was 11.2 watts. Afterwards , power increased to 14 watts. These are rated 12 watts @ 8 ohm. AC ripple was also reduced.
Sound had a slight increase in low frequency response.

I plan to replace the 10d-1 diodes with faster/slow recovery diode (SBYV27-200-E3/54) wonder how that will effect the power supply.

 

[stereofun]

Could not agree more. In my 999 I almost tripled the reservoirs from a combined 8,800 to 24,000. It absolutely gives the amp more audible breathing room.

To stay with Leestereos water analogy...... Many moons back, our family of four, moved into a new house. All was well, except for the puny 30 gallon capacity gas powered water heater.

During morning "peak demand" = four showers, we'd always run out of hot water, otherwise the 30 gal capacity was fine for 90% of the time.

So I installed a 50 gal heater, and never had cold showers again. When I think about it, that was - sort of - swapping a bigger reservoir cap into an existing "power supply" of 0.35 psi flow from a 3/4 gas pipe..

.....at least that how I have thought of supply caps since then.

 

[Hyperion]

Power supply decoupling capacitors are meant to 'decouple', in other words make the circuitry that has been 'decoupled' behave differently (better) from its source. The values specified and installed by manufacturers were working to a 'price point', but also to the 'size point' of capacitors produced ~40 years ago. In many cases a significant improvement can be seen by a decoupling capacitor value increase thus increasing the current reserve available to smooth out the fluctuations caused by music peaks which draw more current. This smoothing out of the supply is visible using an oscilloscope to observe the power supply and often audible as improved accuracy and clarity. This is a direct result of not having a power supply that's bouncing around all the time (because its been decoupled more efficiently), and one of the reasons for 'regulated' power supplies in the first place.

There is an added bonus, that increasing the value of a replacement capacitor very often makes the new capacitor 'match' the PCB lead spacing of the original (older, larger) capacitor. Thus reducing or eliminating the strain on the capacitor wire seals which is important to reduce premature failure of the capacitor, i.e. physical leakage.

Just a word of warning, it is not beneficial to increase the value of some, or all, capacitors - the opposite in fact. The values of capacitors have been chosen carefully by manufacturers, but potentially with other considerations in mind originally. Please only increase values if advised and recommended to do so by those who understand all the ramifications of doing so.

 

[PE9ZZ]

This is a fun illustration of what can go wrong with a cap refurb. These were even replaced by the original values. That said, this amp is a beast.

On a side note, swapping out diodes for faster ones will not change anything. The frequency is only 50 (or 60) Hz. Lower forward voltage drop will increase headroom by a tiny amount.

 

[specialcase]

I often get asked why should one consider judicious increases to the capacity of a power supply. I typically respond as such: It can be argued that the power supply is in the signal path, it's essentially in series with the output stage and hence does impact the sound of an amplifier. Nelson Pass uses the analogy of a water reservoir as the power supply and the output stage as the valve that controls the flow of the water: https://www.passdiy.com/project/articles/power-supplies. Hence, if you have a poor and/or inadequate supply, the quality and/or the quantity of water is affected. For an amplifier, a poor/failing power supply capacitor (e.g., low capacity, high ESR) may not be able to supply the voltage/current that is necessary for the output stage to perform as designed. IMO, this is primarily why improving the power supply (e.g., increasing capacity, lowering impedance/ESR) often does impact/improve the sound of an amplifier.

But didn't the original engineers of these vintage amplifiers know how to properly design a power supply? Well, yes, they very likely did, but were constrained by the capacitor technology of the day (e.g., capacitor size, ESR, ESL) and cost considerations. For example, in this excellent video demonstration from xraytonyb, increasing the power supply capacity from the stock 1000uF to 6700uF in a modest receiver significantly improved its performance wrt clipping behavior (~15 minute mark in video):

lee, i’ve seen you recommend this a number of times and i’m curious, what is your rule of thumb for determining the values of increased capacitance? is it case by case, dependent on the amp/circuit on your workbench, or can you generalize across most amp power supplies?

 

[Leestereo]

...what is your rule of thumb for determining the values of increased capacitance? is it case by case, dependent on the amp/circuit on your workbench, or can you generalize across most amp power supplies?

Increasing the capacitance of power supply/decoupling capacitors (whose sole function is DC filtering), is indeed determined case by case, and takes into consideration the circuit, power transformer and the physical size constraints. IMO, for decoupling capacitors it's not uncommon to increase them 2-3x if the stock values are <1000uF. For the main power supply capacitors, a more conservative approach is preferred, with increases often limited to 20%-50% unless the stock capacitors are very much undersized (as seen the video link above). FWIW, I do remember coming across the following rule of thumb many years ago: "For every 1A of power consumption, 3000uF is recommended".

 

[Hipocrates]

One thing to mention is the power supply rejection ratio of the amplifier, it seems that a good design do not need big chunky filter caps (maybe that's one of the reasons that the 9090db small caps , other than Sansui being sansui and cheaping out). Off course we enthusiast love the more the better, but I've tested some amp boards with crappy PS (one diode and a 470uf cap) with high ripple and they do ok.
Another rule of thumb from Randy Sloane is 1000uf for every 10watts, but we can see manufacturers using double than that (pioneer/marantz) and others way less than that (Sansui).
For me it seems more interesting investigate the PSRR mentioned by D.Self, but who doesn't like a nice beefy filter cap? Points extra if have cool colors. (black and gold or blue)
I would just add that IMHO if the PS is more than 18000uf per cap, a soft start circuit is a nice addition.

 

[leesonic]

I see a lot of "restorations" where original main filter caps are left in place. Reasons vary from "they tested good" to "they never go bad". If the things are 40 years old, how much more life do they have left in them? I always change these out except when I positively cannot get replacements. Case in point would be my own NAD 2700 power amp. The 10000uF 80V ones are easy to find, but the 10000uF 120V ones are impossible. 100V ones are available, but how close to the 95V supply rail can we go? United Chemi Con make ones that are rated at 100V but will take 125V surge.

Lee.

BTW @specialcase I don't think @Leestereo 's name is actually Lee.

 

[petes'picks]

Most receivers/amps built after 1977 or so have adequate filter sizes and will not see as drastic benefits from increasing filter cap sizes.. Its easy to see that a 1000uf filter cap for a 30 watt amp is undersized. I just had a Kenwood KR-100 on my bench and it had one 2200uf main filter cap, no doubt things would improve if you increase that cap to 4700uf or 6800uf. Move along to 1975 and you would likely see two 3300uf caps for main filter caps in a 30 watt amp, by late 70s you will likely see two 4700uf or 6800uf caps in a 30 watt receiver. You will not see much benefits increasing 6800uf to 10000uf in a late 70s or early 80s amp.

 

[PE9ZZ]

The power supply capacitors are not DC filters but reservoirs to store energy in during the zero crossing of the mains input voltage. Increasing them will put more strain on the rectifier diodes as the peak current recharging the power supply caps has to take place in a shorter time because the initial voltage is higher. The transformer usually doesn't mind but the resistance of its copper windings may become a factor too, reducing the DC output voltage. As always, YMMV.

As for rush-in current reduction I like to use those NTC surge limiters/thermistors (TDK/Epcos) that are also used in computer power supplies. 10R (R(25)) is a nice value to start with in case of a 2x 100 W amp.

 

[stereofun]

True they don't filter "away" DC, like coupling caps, but they smooth dirty DC coming in 120hz pulses from the rectifier. I for sure used to be confused by the many names: smoothing - filter - reservoir - supply caps

 

[KeithD]

For example, in this excellent video demonstration from xraytonyb, increasing the power supply capacity from the stock 1000uF to 6700uF in a modest receiver significantly improved its performance wrt clipping behavior (~15 minute mark in video):

Interesting to watch, but I noticed he was always looking at the "max output" of the receiver. So my question is, if one is no where near the power rating of the amp, is there any need to increase PS capacitance?

Out of interest, I keep an eye on the power meters on my QRX-8001 and 9090 when I turn the volume up a bit. Granted, these are not calibrated, but I find I don't even get to 1 Watt power output on the meters, louder would just hurt my ears. It does leave me wondering who's pushing their amps to the limit at 40, 50, 100 watts! Maybe it's due to my inexperience or lack of understanding, but I can't imagine getting anywhere close to that power output. So, if I'm staying at say 1-10% (at most) of the rated amp output, does increasing capacitance in the PS make any difference?

 

[specialcase]

BTW @specialcase I don't think @Leestereo 's name is actually Lee.

thanks lee. my apologies, @Leestereo.

 

[Hyperion]

So my question is, if one is nowhere near the power rating of the amp, is there any need to increase PS capacitance?

Maybe not the main PSU capacitors - i.e. the 'can type' big ones, but more the capacitors that decouple/filter/smooth the supply to the earlier stages of the amplifier, and I don't mean only the pre-amp.
Many amplifiers use the high current, high voltage supply to the output transistors to power the whole power amp stage, the OP transistors modulate this supply with their current draw. The Main PSU capacitors are sized to take care of most of this, however there are capacitors, usually after a resistor in that output stage supply, that filter the modulation so it doesn't affect earlier amplification stages including for example the 'differential pair', and the bias circuitry.

 

[damacman]

Ironically, the largest such experiment was a car audio amplifier - having 225,000 Joules of energy storage.

 

[GrisPato]

Being uneducated in electronics this is a really interesting topic.
I do think that the video in the first post is an extreme example of just enough.
For a Sansui example: AU-777, 2 X 25 Watt, has 4000uF, already more margin. Is it useful to increase this value a lot. I think for fit, yes, also considering that at that time capacitors were often rated -10/+50% it has a good chance it was actually closer to 5000uF in most instances. And lets not forget this is single rail!
Then another's comparison ( I mostly use Sansui designs to see what they did) between a 85 Watt/channel amplifier, AU-717 and a 300 Watt/channel G-33000.
Both have 2 X 15000uF/channel.
Of course space is a limiting factor in the G-33000. But did Sansui handicap the G-33000 or is the AU-717 overspecced?
Then something that I can not find an answer to. Actually two matters.
First: in the patent for the x-balanced design from Sansui, they bring up as one of the benefits, that you can just use one capacitor between positive and negative rail because the transformer secondary doesn't have a center tap (0volt, ground) and that the two capacitors used with center tap are for the circuit in series effectively halving there capacitance ( 2 X 15000uF becomes 7500uF?) I do not know if that is correct but you can see in alpha designs that they use really low capacitance compared to what they did before.
Example: AU-D11 compared with AU-D907X.
D11: 66000uF total for 2 X 120 Watt
AU-D907X: 33000uF total for 2 X 160 Watt.
second issue is the value of the capacitors in the decoupling of the inputstage of the power amp ( LTP and VAS ). With the TOTL amplifiers from Sansui they often have a separate secundary rectifiers and filtercapacitors and in the power amp board some extra decoupling. The confusing thing is that Sansui rarely goes above 470 uF in this functionality. And when they do it it's always in the lesser model of the AU serie. What Sansui does in some occasions is put more 470uF's parallel, like in the B2201 and B2301. So no space restriction because 4x 470uF can be replaced with 2x 1000uF with even less space (this is X-balanced, so just 50v capacitor needed).
So what I am questioning is this, are there other capacitor parameters that made Sansui use this smaller capacitance values in this funtion.
I know from Ishinolab that Sansui engineers found that less etched foil in the capacitors sounded/ funtioned better. Can it be that the extremely etched foil to increase capacitance for size, and extreme low ESR needed in Computer power supplies and Switch mode powersupplies is not the best for decoupling low current stages in audio amplifiers. I did read once about ions being slower in energy transfer in deeply edged foil in electrolytic capacitors.
So that are my thoughts about increase of capacitance, totally non scientific! So would love to get some technical educated responses to straighten my thinking.

 

[Hyperion]

First: in the patent for the x-balanced design from Sansui, they bring up as one of the benefits, that you can just use one capacitor between positive and negative rail because the transformer secondary doesn't have a center tap (0volt, ground) and that the two capacitors used with center tap are for the circuit in series effectively halving there capacitance ( 2 X 15000uF becomes 7500uF?) I do not know if that is correct but you can see in alpha designs that they use really low capacitance compared to what they did before.

I don't think this is entirely correct, because they DO create a virtual ground connection at the junction of the two capacitors, so I don't think the series capacitance effect comes into play here. And regarding low capacitance in the Alpha designs, I had noticed this, but was under the impression that they simply used more low value capacitors, say four or more? where previously they might have used two, and maybe distributed some of the filtering closer to the circuitry being filtered? This tendency may be to use more smaller capacitors may be for parallel ESR benefits, cheapness, or the need to better utilise smaller case sizes thus making more efficient use of case volume?

Can it be that the extremely etched foil to increase capacitance for size, and extreme low ESR needed in Computer power supplies and Switch mode powersupplies is not the best for decoupling low current stages in audio amplifiers.

Could be, which is why I believe there is still a place for 'Audio Grade' capacitors in many signal path circuit locations.
For some time I have been an advocate of low ESR capacitors for filtering and as Main PSU capacitors. However I am puzzled by the wonderful results achieved by using Audio Grade capacitors in key PSU positions as @Leestereo has done. I have followed this advice in several of my restores now, and have also achieved these really excellent results. Admittedly the values of the replacements have been very significantly bumped up, but I am still wrestling with this rather surprising outcome and result for which I have no explanation. Unless it is all down to the value increase and there is no mystery?

 

[Bob Chieldon]

An interesting read.

I've found that capacitor coupled amplifiers and receivers benefit hugely from vastly increased PSU capacitor capacity and voltage increases.

I recently restored a Trio KA2000A which puts out an earth shattering 12.5W at clipping and increased the main filter cap from 1500uF 50V to 6800uF 80V and the output caps from 1000uF 35V to 2200uf 50V and compared the square wave response before and after as well as listening.

It made a huge difference to the sound, better bass but also a feeling of effortlessness across the board that belied the 12.5W rating, especially impressive given the speakers are 6 ohm B&Ws which aren't particularly easy to drive. The square wave response also shows improvement across the board.

With more modern direct coupled amps the difference between standard and high performance caps seems to be much more subtle to me, unless the caps are underspecified from factory or bad it seems to tighten up the bass but to my ears at least, not as much as the cap coupled designs.

I confess to not being able to hear the difference between standard Panasonic SU series bipolar caps and premium Nichicon Muses' in power amplifier signal path decoupling positions, for example, having tried in things ranging from Sansui 9090s, Yamaha CR2020, various big Pioneers, etc. and so maybe it's just my ears.