A variant of ECP Torpedo III

[Background]

Ever since acquired a pair of HD800S in March, I was in search of a better-than-decent amp to drive. Then I read this. So I finally come to ECP Torpedo III, one of the highly regarded amps on SBAF. It matches most of my criterion: punchy, resolving, tubey, relatively compact ( not DNA Stratus kind of big) and most importantly, affordable to me (~1k USD).

Having that said, T3 is no longer in stock on beezar.com. It has been improved and become T4, which is bigger in size, pre-assembled, offers superior design of circuit (NO output caps!) and uses premium components.

Fortunately, or should I say thankfully, that T3 is almost “open-sourced” to anyone. Tom has put comprehensive instructions together with schematics, board layout and even the complete BOM on his site. The fact is, they don’t need to. So In response to the act of generosity, and also respectfully imitating @JeffYoung , I want to pay my portion of R&D cost. I am hoping @dsavitsk or @TomB provide me the details/procedures as I am uncertain how they run the business (or I could just PayPal $50 to each of you).

[Preparation & Design]

Ok, below are what I have done so far, in sequence:

1). Gathering essential parts.

Output transformers are the most critical, in my opinion. If you can’t obtain these, probably you won’t be able to build a T3, at least not one that has equivalent performance. Doug still has some as of I start this thread.



The rightmost are output capacitors: Mundorf Supreme silver gold oil 4.3uF. By the book, these belongs to the TOTL series which are insanely expensive. However, being a custom-made version originally for a particular model of B&W speaker which has discontinued, these left-over parts often pop-up on eBay at a more reasonable price. The only uncertainty is whether they have the identical acoustics quality as the retailed version (one that without the “made for B&W” inscription).
*Extensive reading: Capacitance vs FR vs PF, according to Doug's simulation.




2). Creating symbols & footprint for all components included in the BOM.

What a slow, painful, yet educational process! I had no related experience before, I believe others may find it easier.

*Tools: Eagle PCB


3). Some R&D on EM wave vs noise vs placement of OPTs.

My company has few scopes and a Gauss meter, I figured it would be such a waste of chance not using them to do some measurements. What I did were:

a). connecting the OPTs to the scope, switched on the PT which was wired to a very common IEC inlet, then kept adjusting the distance between the PT and OPTs, in order to observe the fluctuations of induced noise on scope.​

Results: Unsuccessful. Initially I thought the fluctuation was solely due to the PT, but turned out the IEC was noisy enough to dominate the result! Because of this, the influence of PT is still unknown. So I decided to stop there and got mentally prepared that I may need a mu-metal shield, if I am unable to keep the components away from the IEC inlet in my PCB layout.​

b). Wiring a resistor to the PT to mimic the “operating mode”. Then hold the sensor of Gauss meter around the PT while paying attention on the reading.​

Result:
The front side of the PT had the strongest magnetic field, i.e. ~0.63mTesla. Top/Bottom ~0.40mT, left and right ~0.14mT. Along the front direction, I found the reading dropped significantly when moving the sensor away by approx.150mm.



​

4). Sketching schematic and placing components, trying to follow the official layout as close as possible.

Despite large effort, I would say it is ~85% identical to the official layout. Mainly due to the difference in footprint of each components.


5a). Adding some changes to the layout above.

Few highlights:

• Distance between PT and OPTs are same as official layout.

• Reserved space for mounting various output capacitors

• Reserved space for e-choke,MEC100

• Reserved space for gold-point stereo attenuator

• Moved RCA input away from IEC inlet and PT

• Tube bases will be mounted separately on a metal plate and wired to the PCB

• Added heat sinks to output css

• Easier installation of series & parallel plate load resistors

• Wiring cables underneath the PCB to connect between: RCA < > POT, Z-switch < > phone Jack

5b). Designing the housing.

• Dimensions: 310L x 180W x 50H approx.

• Material:
  • Concrete
  • Aluminum/brass (aged maybe?)
  • Whisky (yes! whisky! Not shown in pic below as I don’t know how to do rendering for liquid)

I am a fan of concrete. That said, I anticipate several inevitable iterations before successfully molding the housing out.

The design of enclosure somewhat echos my favorite song of Norah Jones, therefore I am naming my little project “Lonestars”. I am not so sure the plural form is grammatically and logically correct (I mean 2 stars are not so “lone”, right?), but hey I have two tubes here, so adding a “s” sounds legit.

The text inscription is likely to get gold plated/painted so that it will look a bit Nordic. Not sure yet. Nordic style designs are everywhere these days. Maybe I will improvise a little.





[Manufacturing of PCB]
TBC...
[Assembling of PCB & Testing]
TBC...
[Molding of Housing]
TBC...

[Special Thanks]

• My gratitude to @JeffYoung for the valuable advice on acquiring the PT. I was completely lost!

• All members contributed to this thread: ECP Torpedo III [indexed in first post]

[Some Thoughts]

P.s my target was to finish everything by the end of this year. But due to the absurdity of my local government, I have to stand up and join the rest to protect what we had but forcefully taken away, and also to fight for what we believe we deserve. Building this amp is the only effective distraction over the last three months. Now I just hope I won’t be locked away for a decade and be able to eventually finish this project. Sorry for talking about politics and loading these onto you guys here. I am afraid I just can’t help all of sudden.

 

I love the concrete enclosure!

Did you check with Tom to see if you can still get the output CCSes? It's not a deal-breaker if you can't as the original design didn't have them.

If you do get them, put as much heat-sink on them as you can. They run Nelson-Pass-hot. You also might want to move the heater PSU reservoir cap a little farther away from them (and closer to the bridge rectifiers). I'd also recommend a 105ºC cap there.

You could save a little complexity by getting rid of the 110/220V jumpers (unless you think you might sell it on at some point). Mine's 220V-only.

I look forward to seeing it come together.

 

Last months I checked, beezar still had some output ccs.
Just checked again, no, there are gone.

Anyway in my layout there is a dedicated heat sink, though not a big one, for each css board. It should work just fine.

Advice taken reagrading moving C3 closer to rectifier.

Thanks!

 

Just noticed a possible error: the negative terminals of your C6 variants should be connected to the ground plane, not to each other (well, not just to each other).

 

That picture has the ground plane hidden. The -ve terminals do connect to GND.

 

I am so afraid my first post is too long that it may scare people away (including Tom and Doug).
To ensure my appreciation can get to them, please allow me to quote myself and to tag
@TomB & @dsavitsk again.

 

My wife saw me typing and uploading pictures right before she made a very blunt comment that "the post is too long and no one is going to really read it. Also naming your amp "Lonestars" is a bit posturing..."

HP -99

 

Ahh... good.

But now I have a different suggestion.

Consider providing partial isolation between the PSU portion of your ground plane and the audio portion. This essentially makes the negative terminal of C6 your "star" ground and puts the hammering between the rectifier and your pi-filter "behind" it.



Cheers,
Jeff.

 

I like the material and shape of that concrete case. A standing pool of whiskey as a top plate is both insane and incredible. Please do this.

 

Thanks for the idea! It is something new to me, gotta study on this topic.

 

A whisky pool is quite obvious, right?
Indeed I have another concept to blend in the element of whisky, but I want to surprise you so let’s not spoil everything yet.

 

Neat project!

Since you're not locked into the original's footprint, you might want to consider designing your own CCS onto the board. If you have not done this before, there's a fantastic series of articles by Walt Jung on current source design with measurements.

The simplest solution is probably to use a single depletion mode mosfet. The IXCP10M45S is a good bet, and the DN2540 also works quite well but tend to be a bit less consistent if you're intent on matching them. Take care if using these-- the tab is not insulated, and will be connected to a high voltage node. A ceramic insulator (and insulated nut) is the safest bet, thought you could probably get away with mica of appropriate voltage rating.

Similarly, you might want to replace the e-choke with a capacitance multiplier. Or even with a real choke.

 

Or use the 10M45 or DN2540 to cascode a JFET:

 

Hahah I am already lost in reading stuffs relevant to grounding in pcb layouts yesterday. Now my reading list is further lengthened...and I kinda enjoy.

Frankly I don’t know if I am up to the task, given my background is completely mechanical, but I will certainly try because it is getting more interesting! Thanks for the comments

 

Apparently I need to catch up.

Btw I remember that in T3, Doug used an IXYS 10m45 plus a lm334. I did some researches months ago, and found this configuration not very common. Make me curious.

 

An adjustable current source like the LM334 is easier to use in production because you set the CCS current with resistor values.

The JFET (2SK170) sets the current based on its saturation current (Idss), so you have to use devices matched to the current you want. Easy enough if you're building one; more time-consuming if you're building in quantity. The upside is that the JFET version will have lower noise.

 

Been reading materials about ground current. Found this one which is easy to read even for me:
https://electronics.stackexchange.com/questions/389168/single-ground-plane-vs-split-planes

If what suggested are true, then ground current flow in my layout, say for example, from the -ve of C1, should flow along the yellow arrow because of the lowest impedance.



Partially splitting the plane, ie. adding a slot (the black line in pic below), redirects the ground current so it flows in the green arrow direction, likewise, because of the lowest impedance. However, this time it may interfere with the signal ground current more.




So I am thinking it could be better to isolate like the following?

 

Hi Lawrence,

An easier way to think about current is it likes to go back to where it came from. That's the negative terminal of the bridge rectifier. I *think* this is because the magnetic field in the transformer acts like a negative impedance, making it more attractive than the chassis ground.

But I also hadn't noticed your ground-plane-to-chassis-ground connection at the top. Consider moving that to the bottom. Your wall ground can also have a lot of noise on it, so better to tie that into the dirty side of your ground plane.

Cheers,
Jeff.

 

Here's my grounding system just for reference:

 

Hi Jeff,

If I am not mistaking your previous replies, the -ve terminal of the rectifier BR1 is the return path, instead of the path that leads to IEC GND. If this is the case, then I understand why you suggested adding a slot like that earlier.

Nice star ground!
Maybe I can “borrow” (hehe) few ideas from your layout. I will update mine tomorrow.