Can anyone explain the tech of £25,000 Audiophile Pre-Amp from Mend It Mark?

[TimInCanada]

I'll admit right off, I know squat about high-end audio, but did watch with curiosity the video by Mend It Mark, and perhaps more skilled people can help explain this thing.

Backstory: Mend It Mark was given this pre-amp to fix and put a video on his Youtube channel.  Tom Evans, the builder of the device, made a copyright claim and Youtube took down Mark's video.  Of course, the video has now been reposted many times by others.  Here's one copy:   It's also at Archive.org at https://archive.org/details/the-gbp-25-000-pre-amp-that-went-wrong-tom-evans-mastergroove-sr-mk-iii-rjbp-fsfzi-i  Tom Evans apparently wanted to charge the customer £6,000 to fix it, the customer said to send it to Mark, then Tom Evans told Mark he wouldn't be able to fix it.  That worked about as well as you'd expect.

The construction is awe-inspiring, as in "shock and awe":


Mark reverse engineered it and made his own service manual.  Each channel has this block diagram:


Blocks 2,3,4,5 and 6 are dual rail power supplies.  Blocks 7 and 8 are eight op-amps in parallel as a preamp and block 1 is a buffer amp for the line out.  Power comes from a separate power supply:


In the background is a HP 8566B or 8568B spectrum analyzer for scale.  Mark initially thought this power supply contained a big transformer, but his reverse-engineered schematic (sorry, this was the best resolution I could grab) shows it's just a bridge rectifier and linear regulator:


I'm thinking about the power budget.  Just order of magnitude analysis.  Starting at the output stage, there are two op-amps (with the part numbers sanded off) and two LM334 current sources with 10R setting resistors.  From the TI datasheet, that looks like about 7 mA.


(Interestingly, in this configuration the datasheet says "The LM134 makes an ideal remote temperature sensor because its current mode operation does not lose accuracy over long wire runs. Output current is directly proportional to absolute temperature in degrees Kelvin,")

Let's say that IC2 in the output stage is a opa2134 dual audio op-amp and everything else is an op07, all with quiescent current (per amplifier) of 4 mA.

In the output stage, line output on IC2a is of the order of 1.228 Vrms into 600R, so let's call that 2 mA out. Both IC2a and b have 7 mA current source loads, and let's say IC3 only draws quiescent current.  That's 3 x 4 mA + 2 x 7 mA + 2 mA = 28 mA at +/-15 V.

The preamp stage is 8 single op-amps with negligible load, so 32 mA.  So we're up to 60 mA per rail to do the actual amplification.

Mark isolated the Block 5 power supply board and powered it disconnected from the preamp board:


That's 302 mA load on the negative rail and 211 mA on the positive.  So let's call the quiescent load of each of the supply modules an average of 256 mA per rail.

Now adding up by the blocks:

Block 1, line amp, draw is 28 mA at +/- 15V => 0.84 W
Block 6, its supply, draw will be 28 + 256 = 284 mA at +/- 21.5V => 12.21 W
    heat generation = 12.21 - 0.84 = 11.37 W

Blocks 7,8, preamps, draw is 16 mA at +/- 15V => 0.48 W each
Blocks 4,5, their supplies, draw is 16 + 256 = 272 mA at +/- 21.5V => 11.70 W each
    heat generation = 11.70 - 0.48 = 11.22 W each

Block 2, feeds Block 6, draw will be 264 + 256 = 520 mA at +/- 28V => 29.12 W
    heat generation = 29.12 - 12.21 = 16.91 W

Block 3, feeds Blocks 4,5, draw will be 2 x 272 + 256 = 800 mA at +/- 28V => 44.8 W
    heat generation = 44.8 - 2 x 11.70 = 21.4 W

The +/- 28V external supply will deliver 520 + 800 = 1320 mA.  Unfortunately I can't make out the values in the schematic, and there are three parallel resistive paths between the rails.  Assume these create a negligible additional load.  Someone check me on this, but for 1320 mA DC out, and this power supply being just a linear regulator, wouldn't that mean the mains load will be 1320 mA rms?

Mains power will be 230 V x 1.32 A = 303.6 W, to provide about 1.8 W to the actual amplifiers.

Heat generation would be 303.6 - 29.12 - 44.8 = 230 W.  No wonder it's so massive even without a transformer -- that's a lot of heat to dissipate.

Forgive me for being a little slow, but wouldn't a transformer isolate high frequency noise from the mains, then a low noise linear regulator bring the noise on the rails down to microvolts rms, which after the >100 db PSRR on the op-amps would make the effect of any power rail noise unmeasurable?  All with a whole lot less heat thermocycling the actual audio components?

I'll just throw out one more question: what's the deal with those pcb 'shields'?  Some kind of capacitive coupling mechanism?  Waste heat homogenizers?  Solar flare reflectors?  Antigravity dampeners?

Maybe I'll sign off now....

Tim

[themadhippy]

230 V x 1.32 A = 303.6 W,

but those 1.32A wont be at 230v,just because no transformer was  shown it dosnt mean there isn't one

[Gyro]

The power supply box (containing the transformer) is shown in the video. The supply to the pre-amp is already +/- DC rails. Mark indicates the pre and post repair voltages... something like +/-15V when working irrc.

There's already a thread on this video take-down fiasco with several comments on the qualities (or lack of) of the secondary regulation scheme used...

https://www.eevblog.com/forum/other-blog-specific/25-000-audiophile-pre-amp-repair-video-by-mend-it-mark/

[magic]

It's a survival of the fittest thing.
No one would drop £25k on a small box running on a few watts.
You have to come up with something ludicrous to catch those rare big fish with deepest pockets.

As for shielding, it doesn't seem to be doing jack shit besides demonstrating that it wasn't necessary in the first place, since the unit works and customers are happy. Effective shielding would need to enclose the whole circuitry and go all the way to the input jacks, that stuff here is just a gimmick like spoilers on a 1.2l car.

"Expensive end" audio is mostly cargo cult engineering.

[soldar]

[daqq]

You know, it'll be quite hilarious when out of spite the Internet Residents completely reverse engineer, analyze and improve the design of a 25k$ Bullshit Containment Device.

[TimInCanada]

230 V x 1.32 A = 303.6 W,

but those 1.32A wont be at 230v,just because no transformer was  shown it dosnt mean there isn't one

Mark's schematics seem to be pretty thorough, and the one called External Power Supply shows a lot of components with values (unfortunately illegible in the video) and no transformer.  As he created the service manual to expose Tom Evan's design, I'd expect any transformer would be shown.

[themadhippy]

As he created the service manual to expose Tom Evan's design,

No he didnt,he made the service manual to aid him in repairing the thing.You can see it use lm317/lm337 regulators ,if you shove more than around 37v up em they tend to get upset and stop working ,so what do you think there going to do with a few 100 volts chucked up em? and also look at the ground configuration ,without an isolating transformer all them grounds are going to be referenced to the mains, thats gonna give you a tingle every time you lift the tone arm.

[tszaboo]

You know, it'll be quite hilarious when out of spite the Internet Residents completely reverse engineer, analyze and improve the design of a 25k$ Bullshit Containment Device.

When I saw that schematics, I seriously considered to make a better version with 4 opamp in parallel, and sell it at cost. Without Tantalum. Just to piss Tom off.
OP: Those LM334s load the output of the opamp, and "put it into class A operation". At least that's the pseuodscience of it. Because opamps will have this built in already. But that's OK, you need to fill those PCBs with something.

[Analog Kid]

[OP: Those LM334s load the output of the opamp, and "put it into class A operation". At least that's the pseuodscience of it.

Don't the output stages of all opamps operate in class B-ish (push-pull) mode?

[langwadt]

You know, it'll be quite hilarious when out of spite the Internet Residents completely reverse engineer, analyze and improve the design of a 25k$ Bullshit Containment Device.

When I saw that schematics, I seriously considered to make a better version with 4 opamp in parallel, and sell it at cost. Without Tantalum. Just to piss Tom off.
OP: Those LM334s load the output of the opamp, and "put it into class A operation". At least that's the pseuodscience of it. Because opamps will have this built in already. But that's OK, you need to fill those PCBs with something.

does messing with parallel opamps offer any advantage over just using something like AD797, LT1115, AD8597 etc. ?

[magic]

[OP: Those LM334s load the output of the opamp, and "put it into class A operation". At least that's the pseuodscience of it.

Don't the output stages of all opamps operate in class B-ish (push-pull) mode?

Of course they do and you could likely see the difference on a THD analyzer, particularly at a sufficiently high gain.
Plus it's also marketing: we run it class A, wow.

does messing with parallel opamps offer any advantage over just using something like AD797, LT1115, AD8597 etc. ?

Parallel AD797 will still have advantage over one AD797, if source impedance is low enough.

[Kleinstein]

With BJT based amplifiers, there is a limited advantage from parallel OP-amps unless there is no single 1 OP-amp with a similar noise.
Already the AD797 / LT1115 need a really low impedance to get the lowest noise. There could still be some advantage when starting with super beta types as they can be better in the voltage noise times current noise metrics.

There is a problem with the current noise specs in quite some data sheets - there are actually 2 parts for a correlated and non correlated part and single number can catch it all. Also some amplifier assume only one type and give too optimistic numbers this way. So one has to take the current noise numbers with a grain of salt.

[Gyro]

230 V x 1.32 A = 303.6 W,

but those 1.32A wont be at 230v,just because no transformer was  shown it dosnt mean there isn't one

Mark's schematics seem to be pretty thorough, and the one called External Power Supply shows a lot of components with values (unfortunately illegible in the video) and no transformer.  As he created the service manual to expose Tom Evan's design, I'd expect any transformer would be shown.

The still image of the PSU schematic is rather fuzzy and I can't see what it says above 'AC IN'. Mark clearly drew what he needed to draw - presumably the PCB components rather than the transformer and all the mains chassis wiring (yes you could regard it as an omission in a full 'service manual'). The transformer is logically there because:

1. The unit would not be electrically safe without one.

2. The circuit would not work without one being present - remember this is UK/EU possibly a North American version too for a 120V outlet. Mains Neutral is tied to Earth in the supply infrastructure, either internal or external to the house. The circuit as shown will not be able produce symmetrical +/- output rails without being driven by a transformer with a center tapped secondary (AC IN, EARTH, AC IN). As shown in the video, it demonstrably does produce that symmetrical output.

3. The component values shown, would not support direct mains voltage input without immediate catastrophic failure. This is rendered moot by the above two points. It would have been useful to include the transformer secondary voltage somewhere.

As he created the service manual to expose Tom Evan's design, I'd expect any transformer would be shown.

You're on the wrong track there. Mark did not create the service manual to "expose" Tom Evan's design (that would be uncharacteristically malicious), he did it so that he could understand and repair the unit, completely different.

There is nothing inherently 'wrong' with the design - It uses grossly over duplicated basic linear regulator blocks (probably to the point that regulator noise contributes more supply rail noise than anything it is attenuating. Partly as a result of this over-duplication, it uses a vast number of Tantalum bead capacitors, which you could argue, affects reliability. You could also question the screening scheme for a sensitive low level input circuit.

The circuit clearly functions, it is just, what I would describe a 'uninspired'. There is nothing innovative about it, no clever discrete input stages or cleverly optimized PSU filtering, just massive duplication. The idea of paralleling gain stages to statistically reduce output noise is many decades old (as is pushing opamp output stages into Class-A by loading them to a supply rail). Statistical noise reduction used in many places.

This whole mess is nothing to do with "exposing" the design of the circuit as such. It is about Tom Evans's take-down of Mark's video for dubious and false 'violations' of his rights. Nothing more. Of course we can all discuss and pick holes in the design, construction etc, from our own perspective but Mark's intention, and achievement in the video is its successful repair.

[magic]

With BJT based amplifiers, there is a limited advantage from parallel OP-amps unless there is no single 1 OP-amp with a similar noise.
Already the AD797 / LT1115 need a really low impedance to get the lowest noise. There could still be some advantage when starting with super beta types as they can be better in the voltage noise times current noise metrics.

AD797 has its lowest noise figure near 200Ω, LT1028 near 100Ω.
Moving coil phono cartridges have output impedance from a few to tens of ohms.

But it's quite possible that this product was designed to be more universal and probably uses FET opamps, or maybe indeed one of those few high GBW superbeta types which became available in the 21st century.

Anyone with access to the hardware could quickly figure it out by desoldering the chips and running a few tests for input bias and noise.

[iMo]

I saw the video couple of hours after Mark had published it and what came to my mind as the first thing was "..what Mr. Evans will do now.."
Yep, there are sorts of customers who are ready to pay large money for goods and services because they think the more expensive the better. It is their lifestyle, you cannot change it. And where are those customers there are the clever sellers of goods and services, you cannot change it too..

Mark's video is not about the technicalities, but about those customers, imho..

[tszaboo]

With BJT based amplifiers, there is a limited advantage from parallel OP-amps unless there is no single 1 OP-amp with a similar noise.
Already the AD797 / LT1115 need a really low impedance to get the lowest noise. There could still be some advantage when starting with super beta types as they can be better in the voltage noise times current noise metrics.

AD797 has its lowest noise figure near 200Ω, LT1028 near 100Ω.
Moving coil phono cartridges have output impedance from a few to tens of ohms.

But it's quite possible that this product was designed to be more universal and probably uses FET opamps, or maybe indeed one of those few high GBW superbeta types which became available in the 21st century.

Anyone with access to the hardware could quickly figure it out by desoldering the chips and running a few tests for input bias and noise.

It's probably a NE5532 or a Tl072 or some other 1980 opamp he uses.
Just look at the regulators. The supposedly "ultra low noise" amplifier uses an LM317, not any of the improved versions of it.

[Messtechniker]

At that price level, I'd expect MIL spec components thoughout.

[magic]

It's probably a NE5532 or a Tl072 or some other 1980 opamp he uses.

Doubt it. Bipolar only has good noise figure over maybe one decade of source resistance, and vintage JFET opamps are a complete joke. About the only things noisier than TL072 is LM358 and many CMOS opamps such as TI's "improved" TL072H.

[Gyro]

At that price level, I'd expect MIL spec components thoughout.

Why? Unless you need operation over the full MIL temperature range, it's not going to buy you anything. Because of the way some parameters of analogue ICs are measured (the box method), MIL parts can actually exhibit more thermal drift over the normal ambient temperature range than commercial spec parts. This is normally more of an issue for things like precision references, but given the amount of time the manufacturer apparently spends matching Opamps...

[TimInCanada]

As he created the service manual to expose Tom Evan's design,

No he didnt,he made the service manual to aid him in repairing the thing.

Yeah, he said he needed a service manual, but doesn't it seem like rather a lot of work to trace every net on every board, including the external psu, draw it all up in CAD, include the values of every part that hadn't had its markings sanded off, even make board layout diagrams (at 11:43), all after he identified the external psu was operating correctly and the problem was a voltage rail on one channel was being pulled low?  I mean, dozens of dipped tantalum capacitors.  Perhaps the customer wanted to pay all that labour to get the service manual?

On the other hand, Tom Evans did tell Mark he'd never be able to fix it.

You can see it use lm317/lm337 regulators ,if you shove more than around 37v up em they tend to get upset and stop working ,so what do you think there going to do with a few 100 volts chucked up em? and also look at the ground configuration ,without an isolating transformer all them grounds are going to be referenced to the mains, thats gonna give you a tingle every time you lift the tone arm.

I agree no transformer would be suboptimal, but given the load being essentially constant, the series resistors after the rectifier could be sized to drop enough voltage.  However, at about 9:20, Mark measures the output of the unloaded psu, so I concede there must be a transformer.  Gyro also makes a great point that UK 230V power has one neutral conductor referenced to earth, so the DC ground would end up at 28V to earth.  (Here in North America, 230V is two lines at 115V to earth.)

I also made a mistake on my estimate of the preamp drawing 1320 mA at +/- 28V.  That's for each channel, not both.  Which points to about 150W going into the unit. 

[fzabkar]

Who are these audiophools who can stump up $25K for a box of BS? Most young people are flat out paying off their mortgage. In any case, how much fidelity does one need for a music source that goes "doof doof"? By the time a person acquires enough spare wealth to splurge on audiophoolery, presbycusis has already taken its toll on their hearing.

[Kleinstein]

This seems to be an amplifier mainly for a turn table. So no matter how good the amplifier one is still limited by the old style media.
Chances are this is mainly a thing for older guys having earned big money and getting nostalgic with a collection of old LPs. It is not really about actual performance, more about a good feeling.

[fzabkar]

Nostalgia makes sense, but in order for it to be authentic, you would need to dig out the old "hi-fi" equipment from that era. As for me, when I look at a turntable, I see Flintstones technology. The idea of a needle riding up and down, and side to side, in a plastic groove, producing predistorted sound that needs to be subsequently massaged by an RIAA equalised preamp, is just absurd.

[armandine2]

I see Flintstones technology. The idea of a needle riding up and down, and side to side, in a plastic groove, producing predistorted sound that needs to be subsequently massaged by an RIAA equalised preamp, is just absurd.

.. absurd yes but some see analog technology more sympathetically