MKP DC-LINK vs electrolytic power supply 4 tube amp

Hi all...

I am considering replacing a 15 years old power supply in a tube amp with MKP film. The variety of DC LINK caps and the price are compelling.

This amp uses 6 x 4700 @100 for the end result of +-800uf @600vdc for the B+ (480vdc), followed by a choke to the screen supply of 47uf, followed by the preamp (4 x 12ax7) supply at 47uf. The power supply is rectified by 8 IXYS hexfreds feeding the 6 caps as noted.

This one channel of a dual mono amp with six 6L6GC per channel in triode or pentode. Output impedance of the transformer is high for a tube amp but also follows the concept of maximum power transfer to the load. I bring this up as there is a relationship between this and the power supply. Comments are welcome.

Size makes the difference here, and will require an outboard box, or a second box under the amp to house the MKP caps. Price: well very good grade electrolytics x 6, is about the same as a CDE [email protected] or WIMA [email protected], about $100/$125 per channel.

The specs on these caps is very impressive. If I am going to go down this road,i have a choice of a large cap as mentioned above, or upwards of 10 smaller uf caps in parallel to reach the 800uf value. Opinions on one large vs many smaller to achieve 800uf is welcome.

Has anyone substituted this type of cap for electrolytics as the primary cap for a power supply in a tube amp, and felt it was worth it?
output transformer to the speakers. I am not sure many have any idea what the power supply would/could be to the main amp with such a complicated number of tubes, I sure don't. Yes, ohms law ....output impedance and speaker load near the same is maximum power transfer, and a lot of heat in the process. Many people leave out information, seem I did here, that confuses the original question. I brought it up since the output of the speaker transformer reflects back to the power supply....IMF, EMF and so forth.
What is complicated about (presumably) parallel push-pull? Arithmetic still works.

Amplifiers do not generally use the maximum power transfer theorem, because it leads to poor efficiency and bad bass response.

I still have no idea what it is that you are asking us to comment on, apart from your main point about a capacitor swap. I can't comment on that as I am happy to use appropriate electrolytics.


Paid Member
2009-02-11 12:58 am
I don't understand what you mean by "This amp uses 6 x 4700 @100 for the end result of +-800uf @600vdc". Are you saying six 4700 uF, 100 V capacitors are connected in series, resulting in a cap of approximately 800 uF rated at 600 V?

What's the B+ voltage of this amp? 500 V? If that's the case, I suggest getting two 2200 uF caps rated for 350~400 V operation and connecting them in series. That should take up less space than the current configuration and would leave room for a good polypropylene cap to be connected in parallel with the electrolytic cap.

If the choice is between a large, say, 1000 uF electrolytic capacitor inside the chassis or a stack of polypropylene caps totaling 1000 uF in an external chassis, I would prefer the electrolytic inside the chassis. You'll waste all the advantage of the nice film caps by connecting them through long wires (even if it's "only" a foot or so).

So my vote is for an electrolytic can with a film cap in parallel.

Or better yet, a regulated B+ supply.

That may be what I do.

But, my original question was (more or less) has anyone used such caps and how did they work out.

I know I have choices for electrolytic caps, lots of very good ones. Space conservation is not a goal, sound is.

You mention 'long wires even a foot or so" I assume you mean parasitic effects which I can see for coupling caps, but I need enlightening on the affects and effects if we are talking about a power supply. The affects and effects of printed circuit board traces add plenty of parasitic artifacts as this amp sits now. I mean....There are many amps with outboard power supplies and most are owned by the 1% or home brew by some respected names in the DIY crowd.. What do you mean here?

I can not see how a regulated B+ is going to help. Since the music rides on top of the power supply, floats along if you will, a regulated supply ( to me) will add silicon at the least or a very complicated tube power supply to handle six 6L6GC tubes and not sag or fry the rectifier tubes with to much capacitance to try to charge.
Last edited:
Long wires on coupling caps do no harm as little signal current flows. Very different for PSU caps or decouplers. Outboard PSUs create problems and are best avoided where possible. If necessary, good decoupling in the amp can offset the problems.

Remember that the 47uF caps may have a greater effect. Correct grounding and wiring is more important than component selection.
well, you have a good point there. I had thought about the last two caps in the chain and upgrading the main supply with...gosh, Nichicon gold or Super Through or even low esr CDE caps, and concentrating on a film (like Stolen or better Audyn) for the last two. The bigest problem with this amp is...what seems to be headroom, or supply on demand in gets a bid congested as if starved for power.

In never liked PCB for tube, point to point is my prefered and if i redo some oldpiece, I try to convert to a star ground if possible and practical.


Paid Member
2009-02-11 12:58 am
You mention 'long wires even a foot or so" I assume you mean parasitic effects which I can see for coupling caps, but I need enlightening on the affects and effects if we are talking about a power supply. The affects and effects of printed circuit board traces add plenty of parasitic artifacts as this amp sits now.

For a supply cap, what you need is a capacitor with low parasitic series resistance (ESR) and low parasitic series inductance (ESL). This is because the charging pulses that run from the rectifier and into the capacitor are rather large (amps in some cases). Now, polypropylene caps usually have low ESR and ESL - at least the good ones do. So your "block-o-cap" will have low ESR and ESL, but the connection from the "block-o-cap" to to the rectifier and ground will be a few hundred mOhm and probably 50~100 nH of ESL. This needs to be added as ESR, ESL for the capacitor. In addition, the connection will act as an antenna for all the charging current and is likely to create EMI.

Yes, there are amps that have a separate chassis for the supply. However, you will notice that they feed rectified (and sometimes regulated) DC to the amp. They don't have the rectifier in one box and the reservoir cap in another. And if it's a good design, the first thing on the supply line inside the amp chassis is a decoupling capacitor. Usually, using separate chassis is more trouble than it's worth.

I don't intend to start a religious war on PCB vs point-to-point wiring, but the blanket statement, "PCB = bad; P2P = good" doesn't sit well with me. There's a reason your cell phone isn't built using P2P wiring. And parasitics matter greatly for those...
A crappy PCB layout will result in crappy performance. Just as a crappy P2P layout will. A good PCB layout will result in good performance that can be repeated across millions of units made. All layouts (PCB, vector board, P2P, etc) have parasitics. Physics is annoying that way... :)

well Tom....thank you for all that. I need to sit and chew on it. Redesigning a power supply is not an easy task...that is what this is about is essence. I can not find a more complicated and data scarce info bank then capacitors. ( go find some meaningfull data on ATOM caps....the so called best there is, Vishay does not even mention leakage {yet, I have measured caps with superior numbers over ATOM} and so much more. I can not think of any component that has more voodoo and heresay linked to it than caps.

This is neither here or there (althoughi mentioned that a second chassis can go right under the main one for leads no longer than what is ther now), but still i am looking for the best sound I can get before I die of old age. I can still hear conversations from 200 feet when my grandson can't hear me sitting beside him and we are meeting eye to eye!!! I can hear the tiniest inflection that an artist's recorded on his plastic stone while he paid no attention to the engineer who steared his ship to the stars or into the rocks.

I should have been a spy for the US instead of being cursed with seeking the holly grail of music reproduction (Western Electric vs RCA) in a world where 90% or the orginal music is loss and played on a 1 inch speaker to rave reviews.
glassandlight said:
voodoo and heresay
Do you mean 'hearsay' or 'heresy'? Either could apply to the many stories told about caps.

Having a series string of low voltage caps is bound to add unwanted inductance. Much better to have a few parallel high voltage caps. You might find that electrolytics are fine when used properly! Don't worry too much about component brands while your amp has more basic errors in the PSU.
Hi Chris and DF96,

both good points...Chris there may well be a Q problem in pentode mode as hz above 6khz or so has a edge resonance as if riding on top or beside the true frequency and mudding it up (it is not there in triode). My first, and current, impression was the power supply was just to small from the get go for 6 output tubes (and it may be) and thus this grungie (but subtle) artifact was there because of that. The amp pulls twice the power plus in pentode as it does in triode.

And DF96, you point also makes sense to this same issue, in that the 'grunge' or lower quality sound I hear is quite dull compared to the triode mode or mixed triode/pentode mode. The higher frequencies seem muted and rolled off a bit and the subtle stage clues vanish, but boy can it play loud!!!

Actually, I have looked at electrolytic cap brands data sheets and find most to be lacking and no set standard as to what they reveal (Vishay is the worst on details). I found that CDE has the best specs especially for ERS and ESL, low DF and good ripple. Most don't list ESL and few list ripple, or leakage. This is the main reason I began looking at the DC LINK MKP types as they have numbers so low as to be near meaningless, 100K life, etc.

I think the reason this amp has 6 caps in series is the year 1995, and the low profile design of the chassis 'forced' the manufacturer into using this string (that and expense of high voltage caps of quality then and their size compared to now).

So DF96, I can go with one 860uf at 630vdc electrolytic, mess with the layout to have it lay sideways to fit, but I fear I will still have most if not all of the same complaints in pentode. If I go to a higher uf (be it one or two in parallel) then the size begins to compete with the MKP caps and I am right back to one major redesign for fitting (anything other than replacing all six with like kind, means remaking or removing the PCB and going point to point).

I had even wondered if some parafeed idea would work to get the power supply out of the signal path altogether.


Paid Member
2009-02-11 12:58 am
I need to sit and chew on it.

Sounds like a plan.

Redesigning a power supply is not an easy task...that is what this is about is essence.

True. It's often at least as complicated as the amp design.

I can not think of any component that has more voodoo and heresay linked to it than caps.

There's a lot of snake oil and folklore (snakelore?) in the audio business, that's for sure. And often I find that the "audiophile" capacitors come with no meaningful datasheets or measurements. You're supposed to take their "word of mouth" advertising at face value. Sorry... I don't work that way. It's very easy to stick a capacitor on an impedance analyzer and get the ESR and ESL. You can calculate the loss tangent from there. Dielectric absorption is relatively straight-forward to measure as well. If the manufacturer isn't willing to do that much, then I'm not going to buy their products.

Once you know the parasitic components of your parts, you can construct an equivalent circuit for your power supply. Throw it at a circuit simulator and see how it behaves. I'd look at the transient response and output impedance as a minimum. The simulator is nice as it allows you to try different components (say an electrolytic cap instead of a polypropylene cap) or model the effects of an umbelical cord by changing a few values on a schematic. To model the effects of wiring, I use an inductance of 1 nH per mm of wire for the ESL and get the resistance from calculations based on the wire gauge. Add, say 100 mOhm for contact resistance and such.

Once you know how the supply behaves depending on the different components chosen, you can try it out to get a correlation between simulation and sound quality. That way you can start predicting sound quality from your supply sims. Or at least have an educated guess rather than relying on trial and error.

I find that a regulated supply is the closest I can get to an ideal voltage source, so that's what I use. The output impedance of my 21st Century Maida Regulator is 50 mOhm or less across the audio band.

Last edited:
I'm not surprised you are getting poor sound. That circuit looks like another entry for the 'most creative misuse of a 12AX7' competition.

The phase splitter has far too low anode load resistors. The driver, curiously, has higher anode loads yet is being asked to drive three valves in parallel. Now I am a fan of the 12AX7/ECC83, but even I would not ask one to do that.

The NFB arrangements look unusual (to be polite!). Is this some sort of PA amplifier? It lacks the features which would mark it as either hi-fi or guitar.

Two 12AX7 stages gives too much gain so there is an attenuator between the stages.

No point in putting expensive components into such a poor circuit.
My heavens such condemnation, I basically own a door stop then.

One of Fisher's engineers (I believe it was Fisher) said "if it sounds great and measures bad you have won, and if measures great and sound bad you have lost".

Well I can say this much, in triode and even in 1/3 triode 2/3 pentode driving a pair of Soundlab electrostatic speakers, I have had several singing and musican artists that work live at the Met and Kennedy Center say that it was as if they were there 'live' listening to this "PA amp".

It has a problem resolving music in all pentode mode without becoming grainy. That is it in a nut shell. Now some might say you can not serve more than one master, and maybe this amp can not master pentode, and I should forget about pentode and enjoy what it does do in spades.

Feedback is switchable from zero to 6db and has little affect or effect except for stage depth placement. I use zero feedback.

Here ya go, see what Stereophile thought about this PA amp:
Mesa Engineering Baron power amplifier |
And, here are five pages of component reviews that included this poorly designed amp along with Levinson, Lynn, and some others to test speakers, etc. baron

While Chip Stern gushes, John Atkinson has reservations about pentode as I have...however Mr. Atkinson uses this PA amp right thru 2006 as a reference amp in his collection (including Cary Audio Design CAD-300SEI, dual-mono Mark Levinson No.333, and other high end amps) to judge other components by.

I welcome ideas and suggestions that are constructive and have merit. If you comments and or ideas have such, then print them for others to review and consider as a solution. But throwing rocks does not seem correct for these forums.
Last edited:
glasandlight this not rock throwing it is a pointed and well stated as well as know short coming of this design. For once I totally agree with DF96 that is 12ax7 does not have the current need to drive 3 pentodes as well as too too much gain in the circuit. A read of classic Mullard 5-20 or other circuits from the 50's will give you a far better amplifier . The pentode triode debate has gone on for 60 plus years. Pentodes are fine if used as a pentode . The eico hf-81 small pentode integrated amp is a vastly better design . For that matter the ampeg has a better front end to the barron .
On this board there is a vast difference of opinion on some things but the basic design and use of tube stages in well documented and to a great extent understood.
Your emotional attachment to some thing you own is understandable . You did not design it you just own it . Sorry if you ask a group of engineers and knowledgeable hobbyist about a circuit as you have you will get honest views of it merits.
Last edited: