The DIY'rs Cookbook

[johnjen]

And speaking of my BIG eyerrrrnnnnn amp.

Tonight I finished the re-wire job and just need a couple of resistorators, then later when the teething problems are sorted, I'll add a pair of input isolation line level xfrms (you'll see the the circuit board where they mount when I add pictures later).

We have come up with a workable solution for the high line voltage.
And so I'll dial in the bias currents and voltages in the middle of the swing from minimum to peak (122 to 125Vac) using my variac, with the 750VA line voltage isolation xfmr (with the multiple taps on the secondary), in mind.

IOW it's almost ready for round 2 (of V.2.5) testing, of this monster of an amp.

This will be interesting to see how it has responded to the changes to the layout and re-wire job.

JJ

 

[johnjen]

So here are a few pics of the results of the rewire job, but with a pic or 2 of the original wiring as a comparison.

When it was 1st started up after the 1st rebuild.


And a close up of the analog section


Here is the power supply diode bridge and filament supply section



And a sideways look at the bias circuits and filament runs



And the re-wire job of the analog secton including the input buffer xfmrs (2 of them) circuit board


a more oblique view including the ground buss termination



I still need to do some double checking of the wiring before I power it all up again, just because.
It will be my 3rd pass thru the circuit with the first 2 passes using a magnifying glass, looking for cold solder joints, not enough solder on a joint etc.
Then I'll bring it up slowly on the variac and monitor the various voltages (filaments for all 4 tubes, B+ for each channel, bias voltages and currents for all 4 tubes.
And do this for 116 & 117 & 118 & 119 Vac, so I have a record of those particular voltages and can compare them to what the line voltage isolation xfmr with the ≈5% voltage drop delivers to get a good idea of where I want to set the mid point of the sweet spot.

The analog section is remarkably simple with only 2 single triodes per channel which makes this entire process MUCH simpler.

JJ

 

[johnjen]

And I realized I've run out of 'suitable' power cables, well that and I wired up my step down line xfmr which needs a little better iec connector, so I'm buying the materials to make up 5 more ac power cables.

They will be made from 14awg solid copper, THHN UL listed wire used in pairs for the hot and neutral for an effective 11awg.
And thus far the total cost is ≈$130 which is for all the connectors and 300' of wire of which my 5 cables will use less than half.

And after building this amp I REALLY like solid wire.
Yeah its stiff and can be unwieldy BUT it makes for a very solid connection, and it sounds superior to stranded while it 'locks in' the gear they are connected to, which is one aspect to hardwiring I seek, which in turn is why I like using xlr connectors, they don't wiggle, they are locked in.

And I'll probably cryo these power cables which adds ≈ $150 for all of them, which equates to a total cost of about $55/cable.

Here is the xfmr all wired up and ready to get plugged in.





JJ

 

[johnjen]

So I lit her up and watched as the pretty lights came up, wait a minute 2 of them are dead, shut it all down and build another led matrix that IS working properly.

Then fire it up and take a few measurements at 115Vac and 118Vac.
All in all not to bad but the filaments are running 'hot', even at these reduced line voltages.

So round 3 is underway to ameliorate the important voltages (B+ for the 45's and the 6J5 tubes, and the filament voltages as well).

I had to build a 'scaffold' of sorts for the new parts and added another tier to the 'telephone pole' in the filament section.

But the rally good news is the hum has been hugely reduced, enough so that it is quite listenable now, but the voltage variations are stifling that 'Magic in the Mids'.
But the very bottom end is still alive and well, which is still a very good sign.

Round 3 is going to lower those key voltages using strategically placed resistors.

Here is the 'telephone pole' with it's new 3rd tier.


When this 3rd tier and the adjustments to B+ for the 6J5's is 'populated' I'll post more pics.

JJ

 

[johnjen]

Ok so I lied.
Here are a few more pics…
The 'telephone pole' from a different angle.


And the changes to the 45 bias circuitry and the scaffolding for the voltage reduction circuitry of the B+ for the 6J5 tubes.
I have to fit 2) 5w resistorators and be able to add 2) filter caps in the gap between the 2 chokes, right above the bias circuits located down close to the top plate.



And I'll be adding 6) 1w resistorators inside the 'telephone poles' bottom 2 tiers, which has the same vertical spacing as the top tier to the middle tier, just to add some perspective.



And when I rewire this section, I'll clean up a few remaining 'messes', which should also help reduce hum and noise still further.

And the thought plickens.

JJ

 

[johnjen]

This is the 'cleaned up' version of the 'telephone pole'.


And the newly added B+ voltage drop for the 6J5 tubes.



Next up will be to power this amp up, measure the voltages again, and then listen to the results.

JJ

 

[johnjen]

Well, a good deal of the tube magic has returned but further tweaks are in order.
The filament voltages still need a bit more balancing, and a few other details explored.

But now the amp is 'listenable' at least enough for me, and I gave it a 1hr session to listen for SQ and shifts and changes as it warmed up and settled in.
The hum is majorly reduced, but still a bit imbalanced (the left channel is stronger than the right).

But the Magic in the Mids is peeking out from around the corner and is beckoning me to pursue.
And who could deny such a sirens song…
hahahahahahahahahahahaahahahahaha

I just made the 1st of a set of 5 new power cables.
The wire showed up and I made a 1m cable that I'll use for the new amp once it can run without any life support (ongoing monitoring and fussing).
It went much quicker and easier than the last time I made sets like this, mostly because I didn't weave the wire sets together, and just twisted them.

And they, while a bit stiff, are nowhere near as stiff as I thought they would be.
Which means it will be easy to set them in place, where they will stay.

And the line voltage reduction program (LVRP, just kidding ) has moved into Phase II as a smaller step down xfmr is inbound.
The original 750VA unit pops 10amp fuses like popcorn and 5 amp fuses splatter themselves ALL over the inside of the glass tube.
So I have a few circuit breakers inbound as well.
Besides 750VA is almost 4 times more than I really need in terms of the actual load it will see.
So a smaller capacity (250VA) version is inbound which is a much better fit in terms of 'sizing the xfmr' to the load.

More later

JJ

 

[gefski]

Ahhh...we have ignition! (MUSIC)

 

[johnjen]

So a status report.
The amp is about to undergo a through electrical examination with circles and arrows etc,
I'm configuring the wiring to the new line voltage drop xfmr, and have found a way to lower the line voltage even further down into the 115Vac range.
Which in turn means I'll need less voltage drop on the B+ and filament supply's to get them into their sweet spot.

More later.

JJ

 

[johnjen]

The parts have arrived for the next round of refinements and tweaks.
I'll probably get them installed this weekend and take measurements after everything is back up and running.

I've managed to narrow and lower the 'window' of the line voltage that feeds this amp.

From the previous window of 115 to 119Vac, down to 115 to 117Vac so the mid point of these ranges drops from 117 to 116Vac and the spread has dropped from 5Vac to 3Vac which further translates to a B+ shift down from 15Vdc, to 9Vdc.

At least that is what the math on the back of the napkin tells me.
The measurements will let me know how close those pencil scribbles actually are.

And granted these small amounts of voltage changes seem like they should be a non issue, except that this amp has NO voltage regulators, and except for fixed load resistorators there are no other means of voltage reduction nor stabilization.

Which in turn means that the more stable and steady the line voltage can be, the better the amp can be dialed into, and stay in the sweet spot.
Which, it turns out, certain portions of the circuit have a sweet spot target window of approximately ±5%.

So once we get the voltages matched, balanced and optimized between the left and right channels, then we can map out what the next steps in the development of this amp will be.
And it should be all the more listenable, enjoyable, and inviting.

JJ

 

[johnjen]

So part 1 is done.
The voltage drop resistorators for the 45 tubes filaments have been changed from 0.03Ω (x2) (both channels) to 0.045 (x2) (left ch.) and 0.05 (x2) (right ch.)
This should drop the filament voltages down to close to the target voltage of 2.5Vac for both channels.

And the obvious question is why is this necessary in the 1st place and why are such small value resistorators needed at all.

The 45 tubes are (DHT) Directly Heated Triodes, which means the filaments are the cathodes as well, which in turn means if the voltages aren't matched the 2 channels won't operate the same.
Which can really screw up the 'Magic in the Mid's' that tubes are known for.

And these tubes run on 2.5Vac for the filament voltage, and a 5% window of optimal operation means the voltages must stay within 0.125Vac (an 1/8th of a volt) on either side of 2.5Vac or between 2.375 and 2.625Vac and ideally right at 2.5Vac.
But as important and perhaps even more so, they must be matched, regardless of where in that 5% window the voltages actually are.
And generally, the low side of the voltage window (the -0.125Vac from 2.5Vac) is 'worse' than being on the high side of 2.5Vac.

So these mighty tiny tweaks which can seem to be insignificant when the whole amp runs on 300Vdc, yet they are just as significant as dialing in the 300Vdc B+ voltage and making sure both channels are fed the same B+ voltage.

These are the old skewl means and methods that were and still are used to bring a tube amp, especially those amps using DHT tubes into optimal operation.

And note, these tweaks are not really going to be used by amps that are massed produced.
These sorts of efforts to dial in an amp like this are far to labor intensive and situationally dependent, meaning each amp would need to be tweaked for the line voltage they will run at.

So most 'modern' amps use voltage regulators which eliminates the need for any of this fussing with tiny value resistorators and the like, because the voltage is held at the specified values regardless of the line voltage.

But this approach isn't old skewl.

And these newer design approaches can have SQ effects that are unavoidably 'different' than the old skewl approach.


And we may be upping the xfmr count from our current 17 to 21 and replacing a few existing xfmrs with alternatives.
All depending upon how well the amp settles down and then into it's sweet spot.

And I may need to start using an erector set and 'fly' some of these additional xfmrs, as top plate real estate is getting scarce.
hahahahahahahahahahahahahaahahahahaha

JJ

 

[gefski]

Since you mention real estate, JJ, don't know if you're familiar with the late (died 2017) Don Garber's (Fi) functional and beautiful solutions. I've never been lucky enough to see or listen to one.

EDIT-for some user error reason, the images doubled up.

 

[johnjen]

Say there gefski…
Early on, during the initial conceptual stages, I had considered a configuration using the 3rd dimension or some variation of that for my BIG eyerrrrnnnn amp.
It quickly became evident that it would complicate the design and so we went 3d lite and use both sides of the top plate, which isn't new nor unusual at all.

And those pictures you posted are examples of creativity and 3d thinking in new and novel ways that inspires me to think about that approach anew.
His use of tiers instead of risers or standoffs, could be a nifty way of isolating and 'cleaning up' a layout with aesthetic enhancements and could reduce the internal wiring and have other benefits as well.

More food for thought for sure.

Thanks JJ

 

[johnjen]

So it occurred to me that my 2, long term projects, the Power Pulse paper and my HD800 research paper had been sitting there patiently waiting for me to pick them up and resume editing and then posting them.

So here is Part 1 of the power cable research paper.

And here is the 1st post of this series, just in case you want to catch up…
Part A Why Would I Want To Do This In The First Place?
https://www.head-fi.org/threads/the-diyrs-cookbook.781268/page-85#post-14294952

Why DO power cables and such make ANY difference to our headphone gear?

Yet another experiment focusing on power delivery and some not generally understood factors to consider.

Part A Why Would I Want To Do This In The First Place?
Part 1 Theory and Expectations Based Upon My Initial Research
Part 2 Measured Results and Observations.
Part 3 Analysis and Conclusions.

Part 1 Theory and Expectations Based Upon My Initial Research

So during my research on power delivery, including the nature of fuses and the ASCC tests, I noticed a correlation between a few details that are shared between them.

Later I focused in on those factors and then went digging even deeper and this paper is the result.
At that point it was a research project, just me trying to see in more detail, the nature of these relationships and how that related to our audio gear.

One of the consequences of this research was when I posted the results of my ASSC (Available Short Circuit Current) tests, where I mentioned that there were ways of rating fuses dealing with how much current could be delivered in a very short amount of time.
The time chosen is 8ms for these I2T tests, and this duration is significant for one very good reason.

That being, it represents 1/2 of the 60 cycle frequency we use for our power delivery, in terms of time.
The 60Hz frequency our (US) power is based upon, takes 16.66 ms to complete one cycle, so half a cycle takes 8.33ms.
One half a cycle represents the 60Hz wave form completing only a positive, or a negative voltage sweep, from zero volts to the maximum voltage (120/2 = 60 volts rms) then back to zero volts.
So there is an ≈8ms positive voltage window followed by another ≈8ms voltage window (negative).

This is also why the ASSC test uses 8ms because it establishes how much current will trip the breaker (or fuse) in just one voltage sweep from start to finish.
IOW one task of the fuse is to be able to stop the flow of current into the ‘protected’ downstream device in just one voltage sweep, in ≈ 8ms.

Now, in that ≈ 8ms window, the voltage reaches max volts at ≈ 4ms, since the sine wave is a symmetric periodic waveform.
But we actually don’t, and can’t get, current delivery for all of that 8ms window.
Why?
Well, no current will flow when the voltage is zero or is ‘too low’, for one thing, but that really isn’t the major reason.

Once the power supply is turned on and stabilizes, it has a ‘reserve’ voltage which is at a lower potential (voltage) than what the peak ac voltage that feeds the diodes, that switch on and off, that allows current to flow to the rest of the downstream power supply.
This in turn means current will only flow from the ac power source when its voltage is greater than what the ‘reserve’ voltage is, at any given time.
But as the ‘reserve’ voltage fills up after the diodes switch on, and current is delivered to ‘top off’ the ’reserve’ voltage, the diodes then turn off the flow of current as the voltage in the reserve reaches its peak voltage AND as the 60Hz wave form starts to fall off after reaching its peak voltage, after that ≈ 4ms mark.

Put another way, current will only flow into the downstream power supply when the ac voltage swing is near it’s peak (60Vrms) voltage.

And since the voltage that the ‘reserve’ has, at the start and end of this less than ≈ 8ms current delivery window, varies, this changes the start and end voltage that triggers the diodes to switch on (pass current) and off (to stop passing current).
IOW as the reserve fills up, its’s voltage rises and the diodes shut off at a higher voltage than when they switched on.
This makes the current delivery asymmetric with regards to the 60Hz (symmetric) sine wave that is feeding the power supply.

This too, is an expected result for the need of current (power) but with a limited amount of time to deliver that power.
Which in turn means the power comes into the downstream storage / filtering / regulation portion of the power supply in pulses, a series of ongoing pulses followed by corresponding ‘off’ times, followed by another ‘on’ pulse of current, and so on.

So as a consequence of this research, I figured that I would see current flow in ≈ 1/2 of that ≈ 8ms window, and peak current flow in ≈ 1/2 of that.
And as it turns out I wasn’t that far off.

Instead of the guesstimate of ≈ 4ms for the total turn on time, it measures out at ≈ 3+ms.
And that peak current flow window is ≈ 1+ms and not ≈ 2ms as I guesstimated.

And since I was motivated to pursue this line of investigation, I obtained a Tek 468 O’Scope (w/the digital capture option) and an Aim iprober 520 contact current probe, and more importantly, a plan to see where all of this would lead me.

End Part 1

Next up Part 2 Measured Results and Observations.

JJ

 

[johnjen]

So tonight I took a mess of measurements and I still need to digest the results.
But the L/R differences of filament voltages have tightened up nicely, but there is just a 0.01-2volt difference, which I can address.

Yeah that seems like crazy talk trying to equalize such small voltages, but every time I close up this gap, the Magical Midrange improves.
Those DHT triodes as I mentioned before, are truly tweako, which delights my tweako nature no end.

And pretty soon I'll be WAY deep into the resolution of my DVM's (I'm already on the edge in trying to reliably measure 0.02Ω resistorators).
And the noise floor continues to lower, which is also another good sign, and oddly enough my hum is 60Hz and not 120Hz as is the usual case.

And I think I may have figured out a major stumbling block based upon the measurements and the amps behavior under a variety of 'test' conditions tonight.
And as I am often reminded, "there is no problem that MOAR xfmrs can't solve" which will probably prove to be true yet again.

And I have nearly completed the wiring for the chain of mains line voltage step down xfmrs that then feeds this BIG eyerrrrnnnn amp.
I'm just awaiting several 2" wide rubber bands and a bit more of a stable amp that can be turned on 'normally', by just flipping the switch.
It's close now but after the filaments are dialed in and one or 2 other significant details are matched up, THEN the amp can 'come on line' so to speak.

JJ