Ridiculous Valve Amplifier Overhaul

Just a reminder – having accepted the Terms & Conditions on the pop-up, you have agreed that:

  1. The page is presented entirely as an opinion piece;
  2. All opinions on the page are false unless proven otherwise;
  3. You, your assigns, heirs, agents and any third party viewing the page as a result of your access to it (and for whom legal rights against the site owner are vicariously yours In Toto) forfeit all legal rights against the site owner in respect of anything published on the page;
  4. © You will not copy the page or any of its content; and
  5. If you do copy anything in breach of Condition No. 4 and publish it elsewhere, then liability is yours solely and to the outright exclusion of the site owner.

OK

 

 

March 2018 update:  These hideous monstrosities (and all other junk-pile “audiophile” power amps previously used to tri-amplify the Jamo Oriels) whilst having proven reliable for the 5 years since the rebuild detailed here, have finally been sent to the cornfield.  Replaced by the vastly superior in every way DIY solid state amps (P3A and P68), the Oriels now sound REAL!  Please never buy the brand of amplifier you see here or any other expensive fancy-talk amplifiers.  They are IMO all rip-offs.

Old page:

The power amplifiers that couldn’t quite handle their own hot air!

This tedious rant is subject to Disclaimers 1 & 2 on the home page.  It’s about valve amplifiers that repeatedly failed I guess due to poor design and eventually required the end-user (that’s me!) to add his own forced air cooling!

C-J Premier Eight-XS

Abstract (and note to audiophiles)

A decade long love/hate relationship with these amplifiers almost came to a sorry end recently and if not for my delusion about their phenomenal sonic capabilities and a certain masochistic desire to learn and repair complex things, they might have been “thrown on the tip” to quote my friend Elson.  I almost wish I had because the painstaking repair took just under three months to complete.  The page details the repairs and modifications made by me as a novice to valve amplifier circuitry and fault diagnosis.  After hundreds of seemingly logical part replacements I was getting nowhere, but Elson came to the rescue.  Without his help, the next step would have been toward the tip.

Speaking of which, any audiophiles hoping to read about Teflon capacitors, special “quantum” fuses or even Quantum Putrifiers are sure to be disappointed with the “upgrades” featured here since they’re real. 😆

Aside

Most so-called high-end audio products are built to “reviewer standards” (that’s no standards) and use certain specialised parts chosen to impress the impressionable, but often they’re a glossy chassis hiding a nightmare of ineptitude inside.

This gargantuan dichotomy became abundantly clear to me soon after the purchase of these particular amplifiers.  It doesn’t matter that they used expensive Vishay bulk metal foil resistors as a marketing ploy if they are improperly installed and sit in a “furnace” alongside cheap garbage and fail for being under-rated for their thermal environment.  Simple metal film equivalents of higher power rating are more robust, less bulky, orders of magnitude less expensive, operate just as quietly and would not have failed in the first place – especially had the PCB been laid out properly with thermal management as a design criterion!

And these obscenely expensive amplifiers were marketed as employing simple circuits.  Forget it.  This was not at all apparent to me when I first opened them and studied the circuit diagram.  They seem ridiculously complex and when (not if) they go wrong, it is nearly impossible to diagnose and find the fault amongst the absurd over-abundance of feeble components.  Several people have indeed suggested to me that the amplifiers could have been built better with more useful features with fewer than a tenth of the parts for a ten-fold reliability gain. 😆

It also seemed very clear to me that the amplifiers were not designed with serviceability as a criterion, since the solder points for almost every component are buried under other components on the opposite side of the double-sided PCB where wiring labyrinths obscure access to practically everything.  Not exactly elegant design.

Background

OK, so these were the so-called “flagships” of the range during the nineties – the “illustrious” Premier Eights.  The only significantly bulkier amplifier that I know of the same brand was the earlier Premier One – a stereo amp at around 61Kg with 12 output valves (6 per channel).  The Premier Eights with 16 output valves shared across two chassis (at around 40Kg each) probably qualify as a “sufficiently hefty” anyway.

Now with a checkered history of failures and repairs, an intermittent – yet increasingly annoying crackle developed in the right channel amplifier and this lead to a major overhaul of both.  I now know this to have been “elevated shot noise”.  Along the repair path I made many false assumptions and diagnoses, but learnt all I needed to about how the amplifiers actually work.  So it has at least been an education.

Mine are the 120 Watt “XS” versions with EL34 output valves configured for triode operation instead of ultralinear, but with the same enormous output transformer of the 275W version, so at least saturation below clipping is unlikely.

I do not know that these were ever sold in Australia.  I guess they could have been purchased on a special order at around AUD$30K.

How are they used?

Well, the output transformers have dual secondary windings, each with a centre tap, so theoretically they could be connected for 2, 4, 8 or 16Ω.  On mine, the 4Ω taps (full length of both windings in parallel) are wired directly to nominally 8Ω Eton Hexacone midrange drivers of modified actively tri-amplified Jamo Oriels.  I don’t know if this is ideal, but changing to the 8Ω taps would require changing one of the feedback resistors and I don’t know that there’d be any benefit.  No heavy bass current demands are made on the amps and clipping is never even remotely approached.  The active crossover in front of the amplifiers is an ESP LR4 with crossover points of 102Hz and 2.88KHz.  The impedance curve of the drivers (measured in free air) ranges smoothly from about 7 to about 17Ω in the pass band.  The 4Ω bass sections of the speakers have a big solid state bipolar MF2500A power amp (waiting to fail and considered to have several grave shortcomings of its own) in front, and the tweeters have a small MV-55 also with an EL34 triode output stage (also waiting to fail).

Early problems

Here are some of the earlier problems.  If you are a loyal devotee of the brand, perhap now might be a good time to look away.

Snap-links

The amps were built with “snap-links” between wobbly valve sockets and the PCB. 😡

Over the years as they snapped (just by inserting a valve!), I had to replace them one-by-one with flexible multi-strand wires like this:

Explosive polystyrene capacitors

One day I turned the right channel amplifier off at its front switch and the left amp instantly made an almighty bang!  The plate fuse had blown as well as a polystyrene capacitor that was specified without any safety margin whatsoever.  I have no photo of that other than this, which shows the two caps where one used to be (repair done by Elson with double capacitance parts in series to handle the 500V that’s across them!):

The front panel power switches were never used again.  They’re left in the ON position and the amps are powered on (and off) together with an external soft-starter.

Under-rated bridge rectifier diodes and thin PCB tracks

The under-rated FR304 rectifier diodes for the DC supply to the input and phase splitter heaters fried (twice):

I replaced them with FR607s and had to rebuild burnt PCB tracks as shown.

That’s the third (but most robust) set so far.  I probably should have looped those leads.

The noise

So those things had been done before the intermittent crackling noise developed.

I was not confident in my ability to diagnose and trace the fault – let alone poke around where there’s up to 500 Volts DC waiting to bite, so I took the amps back to Elson, who found the noise to be in a Zener diode string at the PCB top right corner – located using a hair dryer and headphones.  I was happy to do the repair myself knowing where the fault was, since Elson was busy with other things.  They form part of the regulated B1 fixed bias supply to the dual 5751 initial amplification stage.  Here they are:

So many.  So puny!  And similar looking strings over at the top left, which are in the B2 and B5 supplies to the 6CG7 phase splitters:

B2, B5?  What is he talking about?  These designations appear on the circuit diagram.  I have reproduced a couple of segments of the schematics of these long-obsolete amps below in low-resolution for non-commercial reference and assessment purposes only.  The power supplies have been described to me by people who’s opinions I don’t entirely disrespect as “abominations of design”.  I can’t dispute it:

(at far right from top to bottom are B2, B5 and B1)

Audio input, initial amplification stage and phase splitter (output stage omitted, but you can see the eight coupling caps at the right):

(At top from left to right are B1, B2 and B5)

The excessive number of input and phase splitter valves has been described to me as a nonsense.  They probably existed solely to “elevate” the model from less expensive models in the range.  It seemed all a bit stupid, but anyway…  In the meantime, I found an old service bulletin which called for the replacement of three under-rated Zener diodes, so I replace three Zeners in the area that Elson identified with the recommended higher dissipation equivalents.  Well, the amps seemed to be OK.  The noise was still there, but it seemed quieter (perhaps wishfully) so I lived with it.

Present overhaul

It was two years later and the noise got ridiculous again.  Here is a video which demonstrates the noise using an old but efficient AlNiCo valve radio speaker driver on the work bench:

Imagine explaining that to your friends!

Assuming that the problem was around those Zeners (since the hair drier temporarily eliminated the noise when directed in their general vicinity and not at any specific diode), I lifted both of the long strings (25 Zener diodes per amplifier).  These piddle-ant ½ Watt parts made up clamps of approx. 370V and 400V respectively.  I ordered more robust 1W parts which added up to the same Voltage and went about installing them.  Under-PCB access was near-impossible behind these due to a spaghetti of wires, so I had to devise a method of doing it from above without applying too much heat to the new parts in the process.

Stampets

So what is a stampet?  Well I stumbled upon them at Element14, checked the published dimensions and smiled:

All 100 were needed!

They’re really meant to be used as PCB terminals, but I figured they could take plenty of soldering heat from above to melt new solder with existing solder on the tracks below.  And it worked:

I suppose the reader is asking “why didn’t he just remove the PCB?”  Well, I was not 100% confident that this was the problem and simply did not want to disconnect the maze of soldered-on power supply and signal spaghetti which would have to be repeated for testing after each repair. 😛

Here’s one string done (the three 2 year-old 5W Zeners there on the right) :

Quick soldering secured each diode across two stampets before cutting off their leads.

And the other string:

No shorts.  Anyway, the noise was still there.  AAARGHHH! 😳 They still run hot (I check everything with an IR thermometer), but the stampets elevate them from the PCB for air cooling and will also conduct and radiate heat away which can only be a good thing.

Someone then suggested that the noise sounded like a bad capacitor, so I replaced every electrolytic in each amp.  Actually, there are only 9 in each.  Polypropylene and polystyrene are used practically everywhere.  There is a 4700μF electro’ in the DC heater supply as well as 8 small ones for the low voltage bias indication circuit, so I replaced all of those with 105°C rated parts (cheap 85°C parts were mostly used originally):

New ones in:

A new snap-in heater supply smoothing cap installed under the PCB:

The amp was tested between each step of the repair, but it wasn’t getting any quieter.

Noisy resistors?

These are on the power supplies to the phase splitters and run very hot.  Someone suggested that they might be carbon composition types:

They looked too big for their mere 2W dissipation designation, but further investigation (actually an eBay listing for identical NOS parts in an unsuitable value) found that they were indeed 2W Dow Corning metal film resistors.  They ran at about 85° above ambient with the cover plate removed!  Looking at the circuit diagram, these need to dissipate just 0.68W each, so 2W resistors might seem adequate, but they’re far from optimally cooled – too close to each other and boxed into an “oven” flanked by the phase splitter valves (heaters) and closed in by the top cover just 10mm above.  They form a 5K resistance between the eight cathodes of the phase splitter 6CG7s and audio ground.  So these were the next to go.  Check out the heat damage and pads which simply lifted away without any force:

There had been some serious long-term frying around there and those puny PCB tracks weren’t helping matters either.  And the resistor legs (previous photo) provided little or no flex to take up expansion cycles.  Also, the PCB is of the regular cheap fibreglass variety.  Surely Teflon or a ceramic PCB should be specified for such a hostile environment!  Anyway it’s browned appearance is unimportant to the circuit which could at least be repaired.  Thankfully it wasn’t so bad as to have become conductive.

I replaced the resistors with 3W metal oxide film parts and as recommended to me, formed a loop in each lead to accommodate thermal expansion cycles:

Someone else then suggested to me that the loops might create unwanted inductance.  They’re in a DC supply and even if the were in the audio path they might affect frequency response somewhere up in the stratosphere, so I ignored him.

The resistors are physically smaller than the originals, but rated for 50% higher dissipation and have more breathing space.  They also have higher voltage and thermal ratings.  They measure at around 50°C above ambient which is 35 degrees cooler than the originals!  I formed new pads with small rings of wire, and since the tracks were so thin, I bolstered them with solder as much as possible to help conduct heat away.  The underside of the PCB got more solder too:

It reduced the noise, maybe a bit, or maybe not at all since the ambient temperature of the day seemed to be the overriding factor.

The phase splitter is what has been described to me as a crude differential amplifier, or cathode coupled Long Tailed Pair (LTP).  This confused me, but after reading an ancient (1950s) text book and with help at the ESP forum, I finally grasped the concept (which doesn’t matter really since I’m not likely to design a valve amp).

The amp has outboard arrays of the same type of Dow Corning resistors.  These were in difficult to source non-standard values (20KΩ and 24KΩ) and metal film substitutes had to come from overseas.  They’re connected between the B2 and B5 (435Vdc) supplies and the plates.  They didn’t show obvious signs of overheating, but I replaced them anyway.

The circuit diagram is wrong in many respects, including this, but shows 100Vdc across these phase splitter plate resistors.  One set has an associated trim pot (for what? See below), so I measured the resistance across that lot prior to lifting them (5.349KΩ in one amp and 5.222KΩ in the other).  I wrote those down since when the new parts were installed the overall resistance would be out by a tolerance margin.

After lifting the set with the associated trim pot, I noticed this underneath:

The cooling holes are all but blocked by an under-PCB polypropylene cap which is held up against the PCB by that cable tie.  Great!

Here are the new metal film resistors installed at the B5 side:

(a little Kapton tape in case that front loop decides to move into that ground track)

And at the B2 side:

The noise remained. 😳

On the phone to Elson

So I called Elson and he remembered the circuit well.  I mentioned that lifting the 5751s did not eliminate the noise and that lifting the 6CG7s did eliminate the noise.  He pointed out that pins 1 and 6 of the 6CG7s are directly coupled to the grids of the phase splitters (correct!) so lifting the 5751s won’t let the B1 supply off the hook.  This meant that I needed to check all of the B1, B2 and B5 supplies again.

He said that the output stage could be eliminated as a suspect since lifting the phase splitter valves silenced it – thank goodness for that anyway.

Out with the scope

I am reasonably familiar with my oscilloscope and have used it often on low voltage, low current circuits, but attaching it to these extremely dangerous potentials was frightening.  Anyway, after some coaxing I gathered my nerves and did it – attaching the probes to the powered-down amplifier, standing well back and then powering up.  So here is what the output noise looks like (scope just connected across the speaker terminals first – this one is not so dangerous):

And here is what the bias voltage noise looks like at the plates (pins 1 and 6) at the inverting side of the phase splitter (very dangerous).  The sweep was slowed down for no real reason:

It’s exactly the same kind of noise.

The non-inverting side gets the B5 supply and the noise looked the same at the plate pins there.

The input valve plate pins showed an identical, but much smaller noise which I wrongly thought might have been feeding back across the direct coupling from the phase splitter grids at the inverting side, or may have indicated the noise to be originating at B1 (I didn’t know at the time, but it was close to B1 anyway).  The direct coupling between the first two stages is a complete pain!

So it seemed that the noise was somewhere in the power supplies which happen to include these two MJE13005 pass transistors and a similarly treated MJE340:

(at the under-side of the PCB)

I think the heat sinks are mounted stupidly for several reasons.  Sil-pads are thermally hopeless when new, but not being resilient they creep with age and create even higher thermal resistance across the junction if no spring washer is used (or no other spring force is applied).  Also, those 6-finger Wakefield Heatsinks might have been too small considering their confinement amidst large capacitors with no PCB breathing holes above.  Also, the chassis bottom cover plate which is perforated in some areas is not perforated directly below these heat sinks.  Although only expected to pass 20mA, with 186V clamped across C-E they seemed very likely candidates for oven-baking.

It also seemed very stupid that a sil-pad was used at all because electrical continuity was provided by the fixing bolt!  Just stupid.

New transistors and bigger 10-finger Aavid Thermalloy heat sinks were ordered.

Since there was nothing likely to short against the heat sinks and the perceived problem was thermal, I did away with any “thermal washer” and instead polished the mating surfaces using Simichrome paste on a soft cloth held flat against a glass tile, applied thermal grease and used a spring washer:

It wasn’t really that much trouble!  And of course they fitted in. 😛

62V Zeners were installed hard against the PCB and suffocating between these over-sized bulk metal foil resistors at the first amplifier stage, so I installed new ones and raised them a bit for better dissipation.  Here is one set:

The MJE340 pass transistor on the B1 supply (which needs to pass only about 6mA) was replaced and given the same heatsink treatment:

The noise remained. 😳

Quite dejected, I called Elson again (it had been a few weeks).  I mentioned the noisy voltage at the phase splitter plates which measured higher than on the quiet amplifier.  He said that in that case, the valves might not be drawing sufficient bias current and that the cathode resistors may not be properly grounded.  This made sense to me.  He said to trace the ground track from the outboard side of the new cathode resistors.  Anyway I discovered this:

Nearly all of the vias are through-plated but had these dubious-looking blobs in the middle.  So I wired through and soldered them like this:

And guess what – it didn’t help, so the grounding was fine.  It turned out later that my DMM was not reading correctly. 😳

Back to the ESP forum where I was encouraged to measure voltages around the phase splitter which (falsely) identified an unwanted DC differential of 15V across a 1MΩ tail resistor separating the grids of the inverting and non-inverting sides.  Together with four paralleled capacitors to ground, this resistor forms a low pass filter to block the audio signal (above 20Hz) to the grids of the non-inverting side.  It was suggested that this resistor might have become noisy and that lifting certain combinations of the four 6CG7s and observing the noise ground.  The resistor was a small ¼W metal film part and it was at the “infernal core”:

The hole spacing would take a modern 2W part, so I ordered some from RS.  While waiting for those, it was suggested to me that I bypass the resistor to see what happened, so I did:

The amplifier was completely silent.  I thought the test speaker had detached but no – just total silence.  The shunt placed the same AC (upstream noise) at both sides of the phase splitter (so it was temporarily not a phase splitter at all) to be cancelled by the push-pull output stage.  In hindsight, this was probably sufficient to identify B1 and/or the 5751 plate resistors as the source of the noise.  Anyway, at the time it only seemed to again confirm that the output stage could be eliminated as the cause.  With the shunt in place I attached the scope again to the plates of the 5751s which showed that there was no noise there (I am still confused by this because later on it was found that the whole problem was at the first stage, so there must have been noise there – perhaps it was just shunting to ground.  I don’t know).

After the resistor was replaced, it seemed at the time that there would be two possible outcomes – quiet or noisy.  Quiet would suggest that the original resistor had become noisy and that would be the end of it.  Noise could mean that one or more of the four caps to ground was leaking.  Here is the new resistor installed.  A modern 1MΩ 2W metal oxide film part:

The outcome was noisy, so I lifted the caps (shown below) to see what happened to the noise:

Well the noise was still there which meant that the caps were fine and the theory was out the door.  Again, it seems that my DMM was to blame (a bad workman blaming his tools 😳 )

I had been wasting a lot of time and had ordered 8 new and expensive caps.  Aaaargh! 😳

Not just dirty/corroded valve sockets surely?

Well, there was no real reason for them to be dirty or corroded.  There may have been traces of antioxidant paste down there from that which I habitually apply to the pins of valves.  Maybe the heat of the valves had turned the paste solid/dielectric?  Well, no.  I meticulously cleaned them with an interdental brush.  There was nothing down there:

So it made no difference.

Complete dejection and time to give up?

Well, that’s how I felt for a day or so, but I decided to step back from it all for a while and systematically eliminate things.  In my inexperience, my process of elimination lead me up the garden path again so I won’t bother detailing it.  Building anxiety and frustration was being fed all along by someone sticking to a theory that the problem was most likely in the output stage bias adjustment section and feeding back to the phase splitter plates via the coupling capacitors (or some other path) between those stages.  This was, it seems based primarily on an abhorrence and horror of the enormity and complexity of the bias circuit, but I couldn’t follow the logic at all and there was no noise when the phase splitter was removed, or when the 1M resistor was shunted, so I was quietly and wishfully rejecting the notion.

Back to Elson

So I called Elson again and he agreed to come over and sort it all out for me.  When he arrived, he had no fear of the thing at all.  He didn’t even power it down to poke around inside.  A loud bang from the test speaker as he lifted a live valve just made him laugh.  He was already laughing about the design of the thing.  “You know this circuit is completely ridiculous. You don’t need all of this.”  Yes Elson.  He touched components with bare fingers seemingly blasé but not – “I’ve been doing this for over 40 years and know where I can’t touch.  I only fried this finger three or four times”.   Lifting the input valves made the noise louder in the test speaker.  He smiled.  Somehow this sealed it.  So it was B1 or the bulk metal foil resistors between B1 and the plates of the 5751s.  He said he never heard resistors making a noise like that, but then again these are bizarre resistors.  He suggested that the early transistors of B1 would most likely be the culprits.

There are two “driver” transistors (obsolete Motorola MPSU10 and MPSU60) behind the final pass transistor of B1 and some 1N4007 and 1N4148 diodes as well.  The transistors anyway were on my early suspect list way back when, but I couldn’t be sure.  He said to replace those and the diodes for good measure.

It really was quite humbling and inspiring to see Elson at work.

Earlier I had done some homework re compatible and pin-out correct substitutes for the obsolete parts and arrived at MPSW92 and MPSW42 which were dirt cheap at Elemet14, but not all of the parameters were absolutely identical, so they might have caused trouble.  Then another helpful guy at ESP found a listing on eBay USA for NOS original parts, so I grabbed some.  Apparently these were very commonly specified parts for all sorts of things and nothing special.

Another long wait for parts delivery, but what I received had 1974 date codes!  Too old!  So I asked Rod Elliott about the modern MPSW92 and MPSW42 alternatives and he said they would be OK.  New transistors and diodes installed:

New MPSW42 driver

MPSW92

Two 1N4148s at right:

1N4148s

And yes – that Vishay resistor had to be lifted to solder in the transistor of the previous photo underneath it!

The 1N4007s:

1N4007s

And guess what?  Broken record:  The noise remained!  It meant one thing:  That quad of stupid 162kΩ bulk metal foil resistors was to blame.  Here are two of them at the infernal core:

Bulk Metal Foil resistors

They didn’t look heat damaged.

And the other two “over there”:

Another two

Not even spaced apart.

I asked Rod about them and he replied “Resistors can get noisy, but it’s more common for them to change value – they typically go high.  As for the bulk metal foil types, who knows what they do after many years at elevated temperature and voltage…  I’d expect just a large increase in normal shot noise could be the culprit.”

So I lifted them and measured their resistance out of circuit.  Low and behold the one directly adjacent to the right 5751 measured high (aboout 195kΩ).  This was it!

And why use a stupid unobtainable value of 162kΩ?  To be annoying?  To force you o believe they’re “special” and have to buy their stock or send the amp at great expense to them for repair?  No way I say!  A pair of 150kΩ and a pair of 180kΩ resistors will be close enough (40.9kΩ c.f. 40.5kΩ).  And if triple rated parts are used, their noise should be very low, so I replaced them with stock standard 3W metal oxide film parts costing a few cents each:

New 5751 plate resistors

other 2

And spaced them off the PCB!  I also replaced a few other metal film resistors around this stage for good measure.

By this stage the entire B1 supply up to the input valves (with the exception of the PP caps) had been replaced/upgraded.

The bulk metal foil resistors are supposedly super quiet, but the standard metal oxide film and metal film resistors will be far more reliable as installed.

Fixed!

No crackle!  The amp then had just the faintest of hum and almost no hiss whatsoever at the test speaker.  It was quieter than the second “good” amp which still had the bulk metal foil resistors in place at the input stage plates – no doubt soon to fizzle.  So of course the second amp got exactly the same treatment.

Bruised Polystyrene caps

While the amps were open and awaiting parts delivery, I decided to replace all of the remaining polystyrene caps because they were looking a bit bruised (heat damaged and brown).

Taking Elson’s queue I used series-connected double capacitance parts to raise the voltage tolerance.  Here is one set:

Polystyrene caps in series

These 100pF/630V caps (for 50pF/1260V) replace a 47pF/500V cap.

To give an idea of how puny the old ones were, here is one that I lifted alongside a new one:

puny polystyrene cap

So next time you read about “premium quality components parts throughout” consider that as an example!

Phase splitter AC balance adjustment

So it was now time to double-check all eight 6CG7s for grid leakage and emission on my trusty Sencore Mighty Mite tester, then adjust that trim pot at the non-inverting side of the phase splitter.  The valves (old Raytheon black plates) tested fine and matched closely.  For good measure I dug out the best 5751s that I had for the input stage.  It was a toss-up between Amperex and RCA Commands – both with anti-microphonic rods between the mica spacers – important for this early gain stage.  I chose the Commands.

AC balance pot

The AC trimmer is the black one.  The metal one is just one of the eight output valve bias pots.  Initially I had no idea what the trimmer was really for, nor how to adjust it properly.  It adjusts the AC balance of the splitter to ensure symmetrical signals to the output stage to minimise distortion.  It’s DC resistance measured earlier was useful only as a starting point for a proper adjustment routine using a signal generator, a dummy load and both channels of the oscilloscope.

So I built a makeshift 8Ω/150W dummy load (just fourteen 11W 2R2 resistors soldered onto zinc-plated stand-off nails):

Dummy load

It should measure 7.7Ω, but actually reads 8.0Ω due to the wide tolerance of the cheap resistors.

According to false info “out there on the ‘net”, the adjustment should be done at around 80% full power.  It’s rubbish.  I tried that and had problems, perhaps brought about by the aged output valves or my own stupidity in believing it.  So I repeated the process at a lower output level of 20W.

The adjustment routine:

  1. Make sure safety ground link is installed at back of amp.
  2. Set signal generator to a very low amplifier input.
  3. Connect dummy load to amplifier output and connect DMM (set to AC Volts) across dummy load.
  4. Power up the amp, warm up and adjust all eight output valve bias pots as normal.
  5. Turn on generator and raise output).  Power off the generator, then the amp.
  6. Connect scope probe ground clips to amplifier ground (optional since scope is grounded anyway).
  7. Connect one probe tip to the B1 phase splitter plates (right side of amp) and the other to the B2 plates (left side).
  8. Set scope Channel 2 to Invert.  Set both scope channel inputs to vertical level to align the traces.  Set both channels to AC!   Set both vertical channel gain controls to be equal.
  9. Power the amp on and allow it to warm up.  Turn on generator.
  10. Without touching the controls on the scope or the generator, adjust the AC balance pot for equal overlapping waveforms.
  11. Sweep frequency up and to observe no deviation between the traces.
  12. Power everything down.
  13. Do not remove/swap phase splitter valves!

Here are the two traces:

Phase splitter traces

That’s at 100Hz and at all frequencies that I tried and with voltages kept within reason they did not diverge.  One channel was spot on and the other was a bit off.  They now both look like this.  Both output transformers made strange “siren” sounds at higher test frequencies.

Interesting:

With the amps fixed and while playing around with the tone generator, noise across the dummy load measured “0.000V” on the AC setting of my true RMS DMM.  This occasionally hovered to 0.001V.  The scope just showed near enough to a straight line at its highest sensitivity setting of 5mV per division.  Can’t complain too much about that in a 17 year old 130W valve amp anyway.

Forced cooling

All that wasn’t going to be sufficient.  Although things were less likely to fry again, I am a fuss pot, so while waiting for parts to arrive and in the interest of seeing another 10 years or so out of these ludicrous lumps, I installed a small PC fan under the infernal core of each.  These are bolted to the perforated chassis floor.  They fit in very tightly and blow straight up.  Of course the majority of the air won’t pass through those small PCB holes.  Most will flow across the PCB underside to cool the bottom-side tracks.  The transistors and their heat sinks should catch the breeze.  It would then hopefully flow around the front and rear edge of the PCB and convect up and out via the valve holes in the aluminium cover plate.  The fans are 9dB “Silent Eagle 1000” fans and operate at just 1000rpm.

A test fit before bolting the fans to the chassis floors:

The fan pressed into the original caps, so the new ones weren’t a waste of money after all since they were thinner:

Thinner PP caps

That green/grey cardboard serves as a partial airflow barrier preventing some of the blown air from taking an easy escape route around the front edge of the PCB.

The fan couldn’t go any further back (up in previous photo) as it needs to bolt over existing perforations in the chassis floor to draw in air.

With the whole amplifier running cooler, there should be reduced thermal noise from all of its resistors and semiconductors.

Thanks again to guidance from someone at ESP, I installed a small capacitor across the motor terminals to suppress any motor switching noises which might otherwise conduct along (and radiate from) its 12V power lead:

(A bit of hot melt glue to make sure it doesn’t rattle loose)

A couple of PCB pins were installed through a pair of existing PCB vias to provide the 12V take-off from the DC heater circuit.  The series resistor is there to drop the voltage a bit:

Fan mounted to the chassis floor.  Two bolt holes were sufficient “damage” and bolts through the other two would have clashed with two of those four caps:

A bit more hot melt glue to keep it from rattling.  Actually, this was not effective.  The fan sent small vibrations into the aluminium plate, so I stripped off that hot-melt.  The fans came with rubber mounting “bolts” but they were not suitable for the thick aluminium plate, so I used the supplied rubber washers under the nylock nuts and inserted makeshift neoprene spacers between the fan and floor plate which saw to it.

The 12V heater supply is at 75V elevation (so the heater insulation rating is not exceeded), so a bit of heat shrink was added to its leads for extra insulation from the chassis.

The fan had an unforeseen benefit.  Whatever air it draws in from under the chassis must escape somewhere and the cover plate has holes for the valves (of course).  I use special coolers on the EL34s.  Prior to the overhaul these coolers measured with the IR thermometer at around 205°C.  Now the gentle flow of air up through the cover plate holes, the coolers read at around 120°C.  85°C less!  No kidding!

Taller chassis feet

The original rubber feet (left) are the same size as those of the brand’s much smaller amplifiers (like the MV-55 on the tweeters).  These new ones (right) raise the chassis off the shelf to allow more cooling air underneath the perforated bottom plate.

And that was it.  One overhauled amp:

Overhauled Premier Eight

Conclusions

You can’t expect 1W resistors to dissipate 327mW for too long without becoming very noisy if they’re boxed tightly into a furnace!  It doesn’t matter that they are expensive Vishay Bulk Metal Foil types that cost 100 times more.

As far as these amps go, they are now far more robust then when they left the factory and although there was a very faint hum from the high-efficiency test speaker after the repairs, none of that was audible from the less efficient midrange drivers in the sound system.  Anyway, whatever noise there might be, it’s below the noise floor of any recording.  A very slight blowing sound emanates from the fans, but this is inaudible from ½ a meter.

Other problems solved externally:

No soft starter

Such large amplifiers should include a soft start device to limit in-rush current on powering up.  These don’t, so I use an external soft starter, which doubles as a bucking transformer to bring Australian mains voltage down to Hong Kong voltage and these amps, together with two other similarly deficient power amps of the same brand in the system benefit from it.

No safety ground loop breaker

The amplifiers have no safety ground loop breaker and therefore produced hum when installed in a multi-amp system (even with the good double braided coax interconnects).  Mono block amplifiers sold in pairs must be installed in multi-amp systems by definition and should have safety ground loop breakers by default!  Instead, these had what I consider to be extremely ignorant and dangerous ground lift jumpers across two binding posts at the back of each chassis.  The amplifiers are not double-insulated, so anyone (including magazine reviewers!) who powers one up with the jumper lifted is nuts!  If  these amps were never sold new in Australia, this could be the reason.  All my own amps have safety loop breakers, so I made one for each of these.  Here is one (replaced by a temporary strap while testing one of the amps on the bench):

Captured garden hose sized power cables

They’re the size of a garden hose!  Don’t believe me?

Garden hose or what?

The amps should have an IEC power socket facilitating:

  1. Removal of the amps from the system without dismantling the entire room;
  2. Work on the amplifiers without having them flap all over the work bench;
  3. Use of a regular energy authority compliant power cable so your insurer can’t use it as an excuse not to pay up when your house burns down; and
  4. An end to kidding yourself that additional girth amounts to anything in an amplifier power cable!

I see their newer products have IEC sockets.

I didn’t install IEC sockets because… Well, let’s just say that after 28 valves, 136 Zener diodes, 26 capacitors, 58 resistors. 10 transistors, 10 small rectifier diodes, 6 what ever else I have forgotten and 12 or so weeks wasted in waiting for obscure parts to arrive I’m over it!

My own DIY amplifiers made from PCBs designed by others have all of these features and more.  They cost less than one 30th of the price and sound better and measure with far less distortion.

Interesting

While fiddling about with the dummy load I was encouraged to determine the output impedance of the amplifiers.  At 50Hz it is 0.57Ω.  This means that the damping factor (at that frequency) with the 8Ω dummy load is around 14 – very low.  I thought this might explain why the speakers sound so different to my solid state system, but after reading this, I doubt it.  Of course I’d have to run the test again at a midrange frequency to see what it is there, but it doesn’t matter.

The voltage gain of the amplifier is not specified.  On the basis of measurements, I calculated it to be about 31dB.  Correct me if I’m wrong, but at 355mV input, the output voltage across the dummy load was 13.   The rated power of 130W into 8Ω is about 32V.  A bit under 0.9V at the input would achieve this.

Moral 1

Don’t be fooled.  Some vastly expensive hi-fi equipment might just be a grab bag of “impressive” and unimpressive parts in a frying pan veiled by expensive casework and marketing hype.

Moral 2

Build an amplifier yourself.  This amp is better and took less time to build than these monstrosities took to diagnose!

Thanks

Another giant thank you to Elson, Rod and also to the terrific guys at the ESP forum, especially rcw and ||81!  Although I surely cannot claim to fully understand the whole circuit, I have gained a valuable education through your help and guidance.

Update

The system has been reliable for months and is sounding very good.  Do the amplifiers account for it?  Definitely not!  Anyway, they have been running well and the sound is still fine.  The actively crossed system calls on these amplifiers solely for the midrange and years of tweaking and in-room impulse response calibrations have gone into it.  So they’ll be doing their “simple” job of applying the midrange gain for a while yet.

… 6 months later – still fine. 🙂

The silly things in action:

Leave a Reply

Your email address will not be published. Required fields are marked *