[Maxx134]

..
Great discussion in this thread now, honest and open, just what we need. Could be some interesting developments following this!

Great discussion?
yes..


Honest?
I don't know about that, haha!
After all, I'm accused of sayn something like my tubes are best in the world! heehee..
until I find a way to squeeze some 2a3 in there!

But at least you know where my interest are..
fixing these damn amps!


I am actually behind schedule in my mods.

SonicTrance bias mod actually made me put the switch mod on hold,
to try out fine tuning my bias points.

Also testing new driver tubes..

 

[Maxx134]

Coinmaster I have a theory as to why when you were using to different tubes in power stage,
It gave you a sonic benefit..

My theory is that in order for the DC offset to work between two different tubes,
The opamps grid control had to alter the bias in those tubes by a sonically noticeable amount.
Edit* in other words, one of those tubes bias points weren't optimal.
I am sure you already addressed this in your prototype amp.

I am finding that optimal bias maximizes soundstage and overall realism..

 

[coinmaster]

The different tubes are two sections of the balanced stage so the DC offset only applies to each tube individually.
Both sides of the load would have 0VDC either way.
 
When one tube goes positive the other tube goes negative
The diaphram in the headphones go back and forth depending on polarity which vibrates the air and creates sound.

 

 
 
 
 
So when you use different tubes the "signal" is traveling through both tubes.
 
 
 
Meaning if you use two different tube types it would travel through two different tubes which shifts the overall linearity of the signal.
 
I don't understand why one slot always sounds better than the other though. When you mix the tubes you have to try both positions because way one is always the clear winner and the other way will either sound bad or unimpressive.

 

[Maxx134]

 a less complicated solution would be to simply use a driver tube with a lower plate resistance and larger current capability. Less than 1mA is not going to drive anything at all properly.

Wouldn't running 6SN7 tubes at 10ma solve the issue? How much current is enough?

You answered it yourself.
Seems that post was not to solve, only to get a point across.

 

[Maxx134]

The different tubes are two sections of the balanced stage so the DC offset only applies to each tube individually...

Ok I am trying to understand this further..
I am thinking, although the opamps are individually working.
But we use the blue trimpots so that the differences between the two tubes or opamp sections are "zeroed".
So the two opamps are set to have 0v to each other.
Then you throw a different tube with different gain & plate & grid resistance,
and the opamp now has to adjust the tube bias at the grid to maintain this 0v difference you set between them with the blue trim pots. .

That's what I am thinking. .

 

[Maxx134]

Another thought is that the opamp reaction time is different with each tube.

And another thought is the two impedance difference between the tubes once in complete circuit with headphone.

 

[MrCurwen]

Are the output stage tubes configured so that (looking at the schem I posted), the bottom row shares an envelope, or vertically so that both phases have their own envelopes?

So basically is it so that

1) the bottoms are in same envelope and the tops are in the same envelope

2) one envelope contains one top and one bottom?

 

[coinmaster]

But we use the blue trimpots so that the differences between the two tubes or opamp sections are "zeroed".

The trimpots are to fix the error between the inputs of the opamp because they aren't perfect components.
 

Then you throw a different tube with different gain & plate & grid resistance,
and the opamp now has to adjust the tube bias at the grid to maintain this 0v difference

Yeah I've mentioned this like 10 times. It was the main reason I was skeptical about the impedance mod initially because it is essentially creating a new load line curve when you mix tubes and when you change plate resistors.
Then I realized that while the triodes load lines become unbalanced using the impedance mod, the current becomes balanced because the top triode has to take in both the current from the bottom triode and the current from the other tube so increasing the impedance of the top triode balances it out.
 

Another thought is that the opamp reaction time is different with each tube

It wouldn't matter because it's below the audio range anyway, the 100k resistor and the 1uf capacitor create like a 1.5 HZ reaction time.
 

And another thought is the two impedance difference between the tubes once in complete circuit with headphone.

Well let's assume the left triode has an impedance of 1 ohm and the right triode has an impedance of 2 ohms across the spectrum. The impedance would be 3 ohms either way.
But since they don't have perfectly linear curves I assume either triode type acts as a bottleneck of impedance across its curves shifting the curves to the lowest point of either.
 

 
 

2) one envelope contains one top and one bottom?

This
 

You answered it yourself.
Seems that post was not to solve, only to get a point across.

He told me not to use direct coupling on multiple occasions, specifically because it would be "replacing my good caps with bad ones in the power supply". He already knew I was going to be running my input tubes at 10ma. 
 
But he doesn't like me and is giving me the silent treatment so

 

 

[MrCurwen]

My insides look like spaghetti haha

Mine as well.

But spaghetti is not necessarily a bad thing!

If you route all wiring in parallel lines, like in a properly drawn schematic, you have maximum amount of capacitive coupling between wires (surface area next to each other).

Visually it is best to route wires 'beautifully', but for technical considerations there is a point to be made about simply using the shortest possible wire route and having every wire in a different angle. Of course certain spots should be avoided, such as excessive heat sources, etc.

My biggest sin regardin wiring and layout is having the output wiring coming in too close contact to input wiring. This is asking for trouble. But, you do what you can in the space you have, and then deal with suppressing oscillations afterwards.

 

[MrCurwen]

No 300B amps for you then.

No 300B amps for me, yes.

I have been toying around making a 2A3 or 300B amp for aesthetic reasons, and to sell it. I could no in way justify spending that much money for my own amp.

There simply is no benefit (for me) to these tubes. If you want linear high current (high power), you can use 6P36S (which is curvewise exactly as good as 300B) for next to nothing.

If you want directly heated magic vibes, there's 4P1L and other less known tubes. Combine that with a current buffer, and you've got the vibes and the power (much cleaner power at that).

None of my potential customer base (which is tiny, not really a base at all) care about looks, so it looks like I'm not going to build a 300B amp anytime soon.

Back before my balanced topology days, when I was doing mostly trad stuff still, I bough a bunch of classic american tubes. I have 12 globe 47's made in 1920s and 1930s. I have some 46's, and a bunch of 71A's (my favourite soundwise, also easy to heat). A whole bunch of different globe tubes, 27, 24A etc. A dozen of ST envelope 76's.

I'm going to use these globe tubes for an timeperiod authentic 1929 theremin replica for my girlfriend. One more project unfinished!

But for HIFI, no. Too expensive, too fragile (I don't have the money to buy genuine completely unused tubes from the prewar era). The new production american classic tube clones are

1) too expensive
2) too hard to heat
3) meant for too high a current to implement in my circuit without big big heat problems
4) because of 3, meant for unnecessarily high output power

I don't see a need for a 3W amp. 0.9W RMS of first class transient response (this is the catch here) power is too much for everyday use.

Seriously though, tube rolling for me is a big part of owning a tube amp. So I don't think I would want a tube amp that didn't respond to tube rolling. Takes the fun away from it IMO. It would be cheaper though, I've spent more money on good sounding NOS tubes than I'd like to admit, lol.

Yes exactly.

It's not as though these things (component selection) have to matter at least to the extent that this thread has found they do (I do believe they matter in this circuit); it's a matter of pleasing the customer base.

When you DIY an amp, even from a complete schematic taken from the internet, the connection you feel to your equipment is a heart warming thing. I am sure, that rolling components gives a small amount of that same pride, of seeing your own handiwork doing something beautifully. It connects the user to the equipment.

Your own kids' artwork is the most nice looking artwork, and the amp you helped create, even if by changing a component or two, will sound nicer than something just bought and then used as such.

 

[MrCurwen]

Honest?
I don't know about that, haha!
After all, I'm accused of sayn something like my tubes are best in the world! heehee..
until I find a way to squeeze some 2a3 in there!

Honest mistake in interpretation between two people who don't speak english as a first language!

Anyway, one thing I remembered from when I read this thread thru the first time (I'm going to do a reread when I have time) was that somebody said (again, could have been offhand) that the 6SN7 is the most linear small signal tube there is.

Splitting hairs about this issue is a pet peeve of mine; most hobbyists don't really know of that many tube types, and still the internet is full of these kinds of statements, however lightly said.

2A3 is not the best tube, or the most linear tube, it is very good but it is simply not. It's just WELL KNOWN.

Also, it suffers from the inherent problem of having two separate triodes parallel inside it. There are some obscure single plate 2A3's which cost both arms and both legs and a piece of your brain (the piece which makes good financial investments).

6SN7 is very good, I use it, but it is not the best. It is just well known and easily available.

I get that most people make these statements offhand, in a light manner, but for the advancement of our hobby I feel the need to clarify this issue.


Same with the caps. Well known and by extension well regarded by magazines, blogs and forums with users that have a very light understanding of electronics (nothing wrong with that, but it's a fact) does not mean a cap is objectively good.

If you go by the route of popularity (which components are most well regarded by consensus), then it depends a lot on which group you are listening to.

For example, there are a lot of technical forums that have gone completely bananas over the russian components (the OTK ones, not the civilian crap) in recent years. Judging from the lack of SSG or teflon enthusiasm on this thread I'm guessing you guys are simply from another faction, another tribe.

None of those things actually says anything, nothing at all, about the objective quality of the russian components. Nothing.

Neither does their price.

Saying that the selling price or the "these components are not popular in my group as of yet" is an objective measurement, is simply buffoonery.


It comes down to two philosophical routes to this question;

1) "You can't really say anything unless you've physically tried this component yourself", in which case nobody can say anything about russian components unless they have tried them themselves.

2) It is possible to have (to some degree, maybe a lot) knowledge of a components quality before trying it out yourself. The way in which to get this knowledge is to have technical understanding. For example one does not need to try out an electrolytic cap to know beforehand that it is not probably going to be as good as a high impedance node signal coupling cap as any film cap.

Which is it?

Don't say it's crap unless you've tried it, or look at the specs and then you simply cannot say it's crap.

The OTK (for military use) caps are super high quality. They are some of the best caps ever made.

For example, how many american paper in oil caps there still are on the market? Not many, for various reasons, but most of them have degenerated by now, even if NOS in storage. The russian OTK paper-in-oils (KBG for example) are actually in NEW condition still, even if made in the 50's.

The big K73 film caps (50µF / 1000V) I use in my PSU were made for industrial laser PSUs. They are still in new condition, and the specs are quite impressive.

One cannot be a subjectivist and an objectivist at the same time. Or, if one is, then one must be an agnostic consistently. Making categorical claims without

1) experience

2) technical knowledge

is simple idiocy.

 

[MrCurwen]

Another thought is that the opamp reaction time is different with each tube.

I'm going to tease you by asking you that if the opamps control the bias and furthermore the operation of the tubes to this extent, is it still A TUBE AMP?


No but seriously; You are right that a proper bias ie. op point is very very important to good operation. It absolutely is.

However I think in this amp the main point with regards to tube selection and rolling is the lack of balance.

It is in name a balanced amp; it has a balanced input and a balanced output. It is in the technical sense, as in actual electrical operation, not balanced at all.

There is nothing balancing the amp.

In most retro circuits the NFB is set up so that it helps with the balance. In old days, there was nothing practical available to force the balance; there was only

1) tube matching

2) NFB

In this amp, NFB is set up one loop per phase, so it cannot help with the balance between the phases. There is only tube matching available.

As speculated earlier, this was a smart choise because people actually do like tube rolling, a lot.

From a technical point of view, with regards to achieving the best possible fidelity, this is not very smart.

 

[MrCurwen]

If people are interested in going further into the "why caps matter in this amp" and by extension the theoretical "how could you make them not matter", I'll ramble on a bit about it. Stop me if this is off topic.

I mentioned my experience with my current workhorse amps PSU. I'll use it to demonstrate some things that can be applied to the amps of this thread.

How can one know that caps in that particular PSU don't matter? Even without auditioning all the different caps available.

Real caps have parasitic properties. They are caused by the materials of the cap (the plates, the dielectric, is it soaked in oil, what kind of oil), and the construction of the cap (planar, rolled, stacked, etc.).

These properties cause the cap to charge and discharge in different ways. A cap that is in DC steady state (disregarding leak current) is "just as good" regardless of properties, because it is not charging or discharging.

Now my workhorse amp is forced balanced. Each stage has a CCS tail with several megaohms of impedance. What this means, is that it draws the exact same amount of current all the time.

Confirmed by theory, confirmed by sim, confirmed by oscilloscope measurement.

Let us disregard the simple B+ reg (output impedance of around 20 ohms; it's basically there to set the voltage, and suppress oscillation) and just look at the PSU filtering caps.

If the amp draws the same amount of current at all times, the PSU caps do not charge and discharge with the signal.

How can they exhibit any sound?

If they were to exhibit any sound, it would mean they are charging and discharging with the signal.

Now one could ask, what about the filtering of rectified AC coming from the PSU. Well, that is another issue. But, film caps are not best for this; they produce too sharp a residual ripple. The output from the PSU filters (the ripple component on top of the DC component) looks more like a triangle wave or a sawtooth wave. With poor filtering caps, like cheap electrolytics, it looks a bit more like a sine wave, at least not as sharp.

Less headaches for the regulator and/or the actual audio circuit.

A statement that one cannot know that caps don't matter in a constant current draw amps PSU is idiotic and buffoonery. A beginner with an honest outlook would ask how can this be, or why is this.

Beginner level understanding of electronics turns the question on it's head; how could cap quality matter in a constant current draw PSU?


Now this amp in question in this thread;

Like I mentioned earlier, it is not balanced in operation. At least not very well. There will absolutely be a variable current draw from the PSU. This means the caps will have to charge and discharge in accordance with the signal. This will most certainly make them audible, they are in the signal path.

A cheaper and more sensible alternative to big film caps in the PSU would be to add even a simple 2 FET regulator. I'm not sure if there is enough voltage headroom for that (10 to 15 volts), but even arranging for that headroom is probably cheaper than big film caps.

A more invasive mod would be to add forced balance at least to the input stage. This would in practise be a CCS tail, with a negative supply (12 to 15 volts or so). This would have other benefits as well.


To clarify; the bulk of current draw delta is of course coming from the high current output followers. However if the input stage were balanced, it would no longer have the PSU caps in it's signal path.

Think of it this way; if you're the input stage, and trying to bounce up and down following a specific wave (like a friend next to you jumping up and down) as exactly as you can, would you rather jump on

1) a trampoline

2) a padded soft mat at a judo dojo

3) concrete

Number 1 is an electrolytic cap in the PSU, 2 is a film cap in the PSU, and 3 is with a CCS tail, PSU completely out of the picture.


Thinking up analogues is not always succesful. Please ask clarification if this subject is of interest.

 

[MrCurwen]

Previous rambling was about PSU caps. Let's have a quick look at signal caps.

Again, charging and discharging the caps is where the differences come about. They do these things differently on account of different materials and different construction.

Some of these properties are static, ie. they aren't dependent on the signal amplitude (or only to an extent dependent). Some are dynamic, ie. they change depending on the voltage potential between the plates, or how fast the cap is 'asked' to charge or discharge, etc.

Once again, imagine the parasitics as a pool of water, and current (charge) as water. The parasitic pool needs to be filled before the signal pool can be filled.

Now let's say we have an anode signal of 10 VPP. So now the anode potential goes up 10 V, and so does the potential between the cap plates.

The dielectric absorbs some current. This is unwanted, but it happens. The amount of current it absorbs is partially dependent on how big the potential between the plates is.

If the potential increases, the dielectric becomes hungry. It need a certain amount of current per centain amount of volts of potential increased.

This property does not care how much current is available from the thing, whatever it is, that is driving the cap. It is hungry. Not polite.

If the driving system has a very very small current capability, it cannot satisfy this hunger to full extent. The amount of water cannot fill both parasite pool and signal pool.

This is a small phenomenon, but it does affect the micro details etc.


Like I went thru earlier, a similar effect happens on the other side of the cap, and the output followers grid also.


For the charging and discharging of the cap it matters

1) what is charging the cap, and can it effectively charge the cap and it's parasitics

2) what the cap is discharging to, and can the cap discharge what the load wants/needs


So have a thing charging the caps that able to satisfy the cap, any cap, and have the cap discharge to a load that doesn't ask anything. Then the cap quality becomes moot; nobody will hear any difference.

In real world these ideal drivers and ideal loads don't exist, but there are lots of alternatives that come a lot closer than these traditional designs from the 50's.

 

[baronbeehive]

Hey guys there's some food for thought in the last 2 posts, any ideas for Implementing? I will rule myself out of it on lack of knowledge grounds!