New to Forum, Throwing Out Some Thoughts

Design and use of the various types.
Lynn Olson
Posts: 16
Joined: Thu Jun 23, 2005 5:25 am
Location: Colorado, USA

New to Forum, Throwing Out Some Thoughts

Post by Lynn Olson »

Hi folks, this is a wonderful forum, and I'm especially glad to hear Dave Slagle's finest when I recently visited the "Colorado Boys" a couple of weeks ago. That 75TL amplifier is really something, it sets a high standard indeed. In fact, hearing it made me think about my "Karna" amplifier recently ...

The amplifier as it sits now is sort of a Fred Flintstone design, massive brute-force isolation between sections. The schematic is at:

http://www.nutshellhifi.com/Karna.gif

However ... the part I'm not so happy with is the ECC99 input section. Almost all the sound of the amplifier is coming from there - by measurement, the majority of the upper harmonics (5th on up) are emanating from this section, with the PP Class A 45's and 300B's providing spotless performance. This is also the case with audition, too ... the amp's character changes dramatically depending on the input tube.

I suspect the limitations of the first interstage might be part of what's going on, since the low-power tube has to drive the capacitance of the IT as well as the PP 45's. So ... here's the throwing-out-ideas part of the post ... maybe a center-tapped choke with direct coupling might be an easier load with more bandwidth. Here are two flavors I want to try:

http://www.nutshellhifi.com/Karna-alternate-a.gif
http://www.nutshellhifi.com/Karna-alternate-b.gif

Note the CT choke is really a sort of autoformer. Unlike other schemes to derive SE -> PP from a choke, though, this is merely a "phase improver" that cleans up the balance of the first stage - like a CT choke, though, the CT has to be connected to a low impedance in order to realize the "see-saw" balancing function, which I want.

I agree with what Dave posted in another thread ... using a CT choke to derive PP, you get really terrible phase match as well as not-so-great bandwidth. A little vector math reveals that 3 degrees of phase shift is equal to 5% gain mismatch in terms of PP cancellation of even harmonics. The last thing you want is a frequency dependence of harmonic structure, and using CT choke to split phase is going to do this at both ends of the frequency spectrum.

But ... using a CT choke (if you're lazy, the primary of a PP IT) as a "balancer", or more accurately, a current mirror, re-balances the input pair, assuring matched drive for the PP 45's, and also reduces even-harmonic distortion, which I want. (The residual 5% or so imbalance in all the stages gives an almost perfectly even harmonic decay out to the 10th in this amplifier, so that's a characteristic I want to leave alone.)

The big difference between alternate "A" and "B" is the degree of isolation between the input and driver stage, as well as possible issue of voltage compliance in the current source as the B+ voltage wanders up and down through a range of 40 volts or so.

Anyway, thoughts, comments, brickbats, whatever are welcome - and a big thanks to Dave Slagle for a great forum and terrific-sounding iron!
Lynn Olson
Bud
Posts: 2
Joined: Thu May 12, 2005 5:32 pm

Post by Bud »

While working out the choke situation:

You might want to tank those ECC 99's in favor of the mil version 7119's. Sound quality is quite a bit superior in my amps.

Might also try paralleling two 7119's. I am sure there are many modeled reasons not to do this, but again, using a small plate triode to shove enough plus side electrons to drive the DHT's may be asking for more than one tube can provide. Just a thought, as that solution dropped the intelligibility floor for microdynamics (weight and heft of an orchestrsa for example) rather amazingly here. Your heater transformer will tolerate the load.
Bud
dave slagle
Posts: 2085
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Re: New to Forum, Throwing Out Some Thoughts

Post by dave slagle »

Lynn Olson wrote: when I recently visited the "Colorado Boys" a couple of weeks ago.
it seems you are now one of them :-)... I was going to start referring to them as c3, but maybe it should be c4??
I suspect the limitations of the first interstage might be part of what's going on, since the low-power tube has to drive the capacitance of the IT as well as the PP 45's.
is it the capacitance, or the idea that the capacitances form side to side are imbalanced? Bud brings up an interesting point with more drive from a parallel 7119 and that would solve a matched capacitance problem, but as you know, there is no way to fix a mismatch from phase to phase.
Here are two flavors I want to try:
ahhh the proper use of a CT choke :-) now how to design that device :shock:
Note the CT choke is really a sort of autoformer.
not to pick nits here, but "my" definition of an autoformer requires primary and secondary currents to be shared in a single winding. I think many people mistake how a device is drawn with how it behaves.
this is merely a "phase improver" that cleans up the balance of the first stage - like a CT choke, though, the CT has to be connected to a low impedance in order to realize the "see-saw" balancing function, which I want.
agreed, and my guess is it is simply not a "phase screwer upper" :-) to qualify that, it will do wonders in the bass to match things up, but at high frequencies where leakage and capacitance domitate, the symmetrical choke is your best bet to keep things straight.
using a CT choke to derive PP, you get really terrible phase match as well as not-so-great bandwidth.
this is a topic i would love to discuss further at another time. I think it needs to be addressed by what it is. One half of the circuit is cap coupled to the grid of one tube through a grid choke which acts as the primary for a 1:1 inverting transformer that feeds an out of phase signal to the other grid. That said, i would design the device as the best damn inverting 1:1 i could. One last point. the bandwidth on the non-inverting half is near perfect and adding a perfect signal to a screwed up one (from the inverted half) gives you a simple additive result which may be a better thing than adding two "slightly" screwed up signals that are not the same where you get sum and difference results. (end of rant)
The last thing you want is a frequency dependence of harmonic structure, and using CT choke to split phase is going to do this at both ends of the frequency spectrum.
agreed, but how do we explain the people who love the sound?? the only explanation i can come up with is adding a near perfect spectrum to a really screwed up one is better than adding two slightly different minorly screwed up ones. (oops i guess that wans't the end of that rant)
or more accurately, a current mirror
nicely put!

i see that i went a bit overboard with the banter on this and to keep things clear, i'll add my "suggestions" in another post so things will be easier to track.

dave
dave slagle
Posts: 2085
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Post by dave slagle »

The simple design for your CT choke is well known. Use a split bobbin and wind 1/2 clockwise and the other ccw and this will match capacitances between the halves. This approach will provide tight coupling at low frequencies through the core, and as frequency goes up the core will become less effective at coupling and at some point the common cored chokes will start behaving as discrete ones.

I have to believe that using a core that supports the full audio band is a good move here. listening tests seem to confirm this, yet my crude measurements don't. (take that fwiw)

given your situation, i would start with a simple split bobbin reverse wound design with this one twist. i'd do it with some bifilar wire on each side so it could be wired as a choke or an IT. On paper they should have nearly identical performance and the IT version could drop into your existing amp for a reality check. Then you could hear the same device in the DC version for an apples to pears comparison.

i think the first step is to determine if its the IT causing the problem or the imbalance between phases of the IT and this experiment should shed some light on the situation.

dave
Bud
Posts: 2
Joined: Thu May 12, 2005 5:32 pm

Post by Bud »

Good suggestion Dave.

Another wrinkle is provided by Nat Grossener where he describes a dual well construction with cross (X) connected windings.

I have been exploring this with PP primary windings ( first coil parralleled with outer coil on other well) , going up a guage for the outer winding to match drive voltage/DCR and sandwiching a series split secondary across them. Gary Pimm has been doing testing and they do seem to work well.

Point is that they are available for insertion as Dave Slagle suggests. Only cavaet seems to be a sensitivity to surrounding EMF fields, imposed or reflected. Certainly enveloping the core or coil in a ferrous container is out.

I have attached Gary's curve traces. In my own IT's, to this plan, they sound as good as the PP outputs. They also use .014 ga. M6 EI core so you can make even picky IT's work well with common materials.
Bud
Attachments
SmallInterstageFRhighsourceZ,terminatedsecondarys.zip
Here are all of the files. Gary did experiment with loading to good effect as you will see.
(648.08 KiB) Downloaded 1422 times
IslandPink
Posts: 56
Joined: Thu May 12, 2005 8:01 am
Location: Denbigh, North Wales

6N6P

Post by IslandPink »

Hi All,

I like the look of the new circuit ideas. I remember something similar from the 'Aurora DC' options a year or so back . I watch with interest from the sidelines on the magnetic developments .

This comment is reassuring :

"The residual 5% or so imbalance in all the stages gives an almost perfectly even harmonic decay out to the 10th in this amplifier, so that's a characteristic I want to leave alone"

I've just got a batch of six 6N6Ps ( James D's fave ) to try out, they're on my desk here at work ( you can't leave a parcel like that at home, unopened ) so I'll post impressions later this week on how they compare to ECC99 and 5687s . I may even be able to try parallelled sections as Bud suggests .

...however, I can see some 45's will be required soon, I can't put it off for ever .

Mark
Lynn Olson
Posts: 16
Joined: Thu Jun 23, 2005 5:25 am
Location: Colorado, USA

Low Distortion

Post by Lynn Olson »

One of the more surprising results was comparing distortion going in to the PP 45 stage and the spectrum coming out. There was no difference - and this is a Gary Pimm measurement with a noise/distortion floor of -140dB. If the PP 45's had any distortion at all, it was hidden behind the ECC99 comb spectra.

Now - by contrast - NOS 2A3's raised the distortion floor about 5 to 10dB, so yes, the measurement was sensitive enough to visualize what different drivers were doing. Triode-connected 6W6's were similar, maybe a bit worse at 10-12dB rise. And small-tube IDHT's were on the order of 20dB worse, so there was definitely an order of merit here.

Part of the reason I'm throwing out ideas is that it'll probably take me 'till after the Rocky Mountain Audio Fest to get back in the swing of playing around with the Karna amplifier. (So Dave, don't send any parts or stuff out here for a while yet.)

So the to-do list for the Karna is investigate different input tubes, play around with an easier-to-drive topology for the same tube, and probably get some Guido Tent current drives for the 45 and 300B filament power. I agree with the earlier posts that it takes a pretty meaty tube to drive the interstage + Miller capacitance of the DHT tubes; that was probably the biggest problem with the Aurora (speaking to you here, Island Pink), where the SE input tube just couldn't do a good job of driving all that capacitance. It seemed to really magnify tube coloration to an unacceptable degree.

This brings up a related point about SE vs PP design. I've heard several really great SE amps with RC coupling - the Ongaku and Reichert being notable examples from ten years ago, with vivid, lively, dynamic sound - colored maybe, but never dull or boring.

By contrast, every RC-coupled PP amp I've heard is boxy, closed-in, and flat-sounding. Not even particularly musical or compelling, with the Williamson being the worst. But - when you transformer couple - PP is utterly changed, becoming quick, lively, fast, and very colorful sounding. For some reason, RC coupling is far worse in PP than SE.

My best current guess is that transformer coupling, by re-summing the PP output at every stage, provides a clean, low-distortion feed to the following stage grids.

By contrast, conventional RC-coupled PP amps are really a pair of SE amps in a balanced (or bridged) circuit, with the summing deferred until the final summation in the output transformer center-tap. This implies that a fair amount of distortion has built up by the final stage, and re-summing happens far too late to do any good - indeed, it might make things worse in terms of generating new harmonics.

This is obviously a wild-assed guess, but experience has shown the harmonic characteristics of all-trans-coupled PP is radically different than conventional circuits, and sounds radically different as well - as different as SE is from PP. I feel the transformers are acting as current mirrors, a well-known distortion-reduction technique in the op-amp world. But in order to be a good current mirror, the secondary of the IT has to have exact phase match (better than 3 degrees) all through the audio band and well above, otherwise the distortion spectrum will have a strong frequency dependence.

In other words, by using IT's - or real current mirrors - throughout a PP design, the build-up of excessive high-order harmonics can be controlled. Don't think SE is free of excessive high-order harmonics; cascade enough stages, with enough high-order distortions, and SE can be as dirty as any PP design. Hmm, well, imagine a 12AX7 RC-coupled to a 6DJ8 RC-coupled to a triode-connected 6550. Colorful maybe, but I don't think it would be very transparent sounding.
Lynn Olson
Paul Barker
Posts: 59
Joined: Sun Feb 20, 2005 10:20 pm
Location: UK

Post by Paul Barker »

Yes I'll go with that analysis of SE. You blend your load resistors, and caps (Rikens and Mundorf Silver Gold for instance), vary your topologies and select your valves for the sweetness of sound that suits the rest of your system. If you get a good dht valve line up you can make your heart melt. Once you haven't done anything majorly wrong, choice of valve for each stage makes the greatest impression on the sound, most often the first valve of the three (I haven't built a two stage amp with sufficient gain to beat a three stage yet, the high transconductance valves it necessitates do not sound as good as the medium or low tranconductance). Note about terminology; it doesn't really matter what our adjectives, we describe improvements which affect our satisfaction level that's all. We don't have to dismiss a statement because it doesn't incorporate mainstream hifi coined mnemonics (i.e. PRAT). These words say nothing, the fool whoi follows a person because of the poetic way he can write about his amp is a bigger fool than the one who doesnt try anything for himself properly before giving it the chop on a forum.

Can I just say your speaker design is absolutely tops. I just love the Ariels for highish powered amps, I find the base is perfectly adequate when you get the right amp behind them. From lower powered amps like 3 watts they are still good speakers but the base is light. The woofers seem to need a certain drive before they flip into a dimension where they start to rock. I was able to demonstarte this quite effectively by first playing the r less c less px25 amp throught he Ariels, the traces of low base were there and all other frequencies sounded excellent, but when I put the 212 output stage on the speakers were transformed. All who were there heard it, it's a massive improvement.

All of us who spend fortunes on getting se right, know that ecc83's ecc88's and triode connected beamers are completely out of the question at any stage. Some of us have given up on high transconductance valves.
dave slagle
Posts: 2085
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Post by dave slagle »

My best current guess is that transformer coupling, by re-summing the PP output at every stage, provides a clean, low-distortion feed to the following stage grids.
I think you are right on the money with this one. If this is indeed the case, then simply loading a cap coupled stage with a CT choke, and possibly adding a CT grid choke to the following stage might go a long way to getting a cap coupled PP amp so sound "right" I'll defer to you on actual experience here, and all I can offer are some thoughts and experiences based on my few attempts at PP amps.

I feel the transformers are acting as current mirrors, a well-known distortion-reduction technique in the op-amp world. But in order to be a good current mirror, the secondary of the IT has to have exact phase match (better than 3 degrees) all through the audio band and well above, otherwise the distortion spectrum will have a strong frequency dependence.
This brings up a topic that needs to be addressed. Do you think the current mirror needs to be wide bandwidth, or must the phase relationships be matched?

I'll start with the phase. When you say exact phase match, do you mean primary to secondary, or secondaries wrt each other? Matching the phase and bandwidth between the stages is really quite easy, just use two discrete symmetrical designs on a common core. At low to mid frequencies, the current mirror will behave as expected due to the common core. At some frequency based on the core type and thickness, the discrete windings will start to decouple from each other (increased leakage) so essentially the current mirror aspect of the design will disappear as frequency goes up. It is important to note that even though the coupling becomes poor, the phase relationship can still remain near perfect.

In order to improve the current mirror behavior of the device as frequency goes up, you simply need to couple the two windings to each other by proximity. The baggage that comes with this of course is the increased capacitance associated with reducing leakage. In your case where you have a PP signal driving PP grids, the capacitances can still be matched (ala grossner) but the overall value will be higher.

So essentially I think we need ot come to some conclusions on what is actually important in this situation and then design towards that goal. The only "ears" on experience I have heard of wrt this is a guy on the asylum who played a bit with some common cored grid chokes for balanced drive. It started from him measuring a simple split bobbin reverse wound choke by feeding a signal into one side and looking at the output from the other side. Needless to say it was butt-ugly, but he was testing it in a manner quite different than he was using it (both sides driven). Realizing that leakage was the issue for his (incorrect) measurement method, a simple interleave of the windings improved things drastically (and also whacked the capacitnace.) When it came to listening tests, the simple split bobbin sounded best as a grid choke, and the interleaved one was better for the "phase splitter" Interestingly enough, when properly measured for the appropriate situation, the ones that his ears preferred actually measured best in each case.

Based on this one case (talk about an large survey) I draw the preliminary conclusion that for the situation we are talking about (class A both sides driven) that it's possible that two common cored discrete devices might be the best choice of compromise. The important factor here is Class A. When you get into the discussion of closely coupling each half of a PP winding, the references always seem to be about it being a need for class B operation. Last time I checked, Class B (or AB for that matter) was a 4 lettered word.

In other words, by using IT's - or real current mirrors - throughout a PP design, the build-up of excessive high-order harmonics can be controlled.
I'm not sure how to ask this, but here goes. If we assume that the high order numbers we are talking about still fall within the audio band does this mean we should target the fundamental with the current mirror? If the 5th harmonic is at 25K is that as much of a concern as the 7th of a 1K signal? (does the current mirror have to work for both the fundamental and the harmonics?)

I'm not even sure how to test for the current mirror behavior in a transformer. Could it be as simple as placing a 50 ohm signal generator on one winding and a 50 ohm load on the other and checking the sines wrt each other as frequency is swept. In previous tests I have been unable to see the difference in core materials (m-6, nickel, amorphous) wrt high end frequency response. I know the difference is there and have to believe that a better core will make a more effective current mirror as frequency goes up. I just would like to put my mind at ease by confirming this with measurement.

dave
Lynn Olson
Posts: 16
Joined: Thu Jun 23, 2005 5:25 am
Location: Colorado, USA

More Thoughts about Phase

Post by Lynn Olson »

I'm not too worried about weird square-wave response as it passes through the IT or CT choke; some people fetishize and absolutely *must* have perfect 100kHz square waves, but most of those people are working with solid-state amps with a fair amount of feedback. Obviously, an unstable amp is undesirable, and ripples in the square wave are a first-order indication of this. But - in an amp with no feedback, stability is not an issue, and the appearance of the square wave has nothing to do with stability at all. I really don't worry about what amounts to group-delay errors above 30kHz - it isn't an FM or color-TV transmitter after all.

But I *do* get concerned about phase match, even at quite high frequencies, like 100kHz. I think HF symmetry is an essential quality for any balanced circuit, even at frequencies far above the audio band. Why?

Out of band HF distortion is a major concern for me; I am convinced, at an intuitive level if nothing else, that IM terms crossmodulate with each other at very high frequencies and generate beat products down in the audio band. So I like to go to some trouble to assure a very high slew rate, very generous current-delivery, and distortion as low as possible at high frequencies. For me, that correlates subjectively with open, unstressed, and relaxed sound, with little or no "electronic" coloration, while at the same time offering vivid instrumental colors.

By contrast, when I listen to limited-current, slew-rate-restricted equipment, I hear a constrained, closed-in sound, which sound very unnatural and "electronic" to me. The main culprits here are high-distortion driver circuits and slew-limiting in phono preamps that can't properly drive the reactances in the RIAA equalizer. It's gotten to the point where Gary Pimm and I can hear these colorations in the first few seconds of audition, which doesn't always make us the most desirable audio-guests.

So I'm concerned about build-up of capacitance, which unfavorably loads down the input or driver stage, but am equally concerned about phase matching at quite high frequencies, certainly well above the audio band. As mentioned up-thread, a little drawing with vectors reveals that 3 degrees of phase error is equal to a 5% gain mismatch, which is a typical value for a pair of tubes. Since the exactness of the match controls the level of *all* the even harmonics, phase wandering is going to have a major effect on spectral colorations. As far as gain-matching the tubes goes, I don't worry about it that much, since 3 to 5% match is very typical, and mismatches greater than that will start to affect DC balanced unfavorably. Needless, there's no frequency-dependent term in the gain-match, so there's not going to be any concern with odd spectral colorations.

If I had to put things in order, it would go like this:

1) Phase match from 10Hz to 100kHz, preferably (much) better than 3 degrees from 15Hz to 50kHz. This is phase match into the grids of the following stage, not the load on the preceding stage. The quality of the grid drive is critical to the sound of the entire amplifier.

2) Relatively low additional capacitance burden on the preceding stage. Another 50pF per triode section is too much.

3) Good current-mirroring capability from 500Hz on up.

4) Uh, oh yeah, good sonics.

I should mention in passing I usually assign a sort of inverse Fletcher-Munson weighting to spectral colorations and errors. The ear is 30dB more sensitive to distortion in the 1 to 5kHz region than it is around 50 to 100Hz, so I weight importance of distortion and noise mostly around this region. That's why the Ariel, Amity, and Karna sound the way they do- the design are oriented around minimizing problems in the critical 1 to 5kHz band, and give less weight elsewhere in the spectrum.

P.S. Yes, I essayed a little joke in the previous post, trying to think of a intentionally badly designed SE amp, choosing the worst possible parts I could think of. A 12AT7 would probably sound as grim as a 6DJ8 driver, just in a different way. If you were really demented (names of famous high-end designers come to mind here), you could drive the triode-connected 6550 with a 12AX7 for a super-dull jukebox sound - with a nice little screech right on top for the ding-dongs that think "screech" is equal to detail.
Lynn Olson
dave slagle
Posts: 2085
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Re: More Thoughts about Phase

Post by dave slagle »

Lynn Olson wrote:I'm not too worried about weird square-wave response as it passes through the IT or CT choke
nor am I, with the caveat that nothing "too" odd happens near the audio band.
Out of band HF distortion is a major concern for me
my guess is no magnetic device will fix this once broken, so the goal is not to "break" it.
and distortion as low as possible at high frequencies.
i have to believe that having a core that actually acts as a core at high frequencies has to help. There are those that say "above 1Khz the core isn't needed" and conventional measurements looking only at bandwidth will prove that. Just yank the core from a 10K square wave trace and see little or nothing happen to the trace! How people get from that experience to "the core doesn't matter" is beyond me.
preamps that can't properly drive the reactances in the RIAA equalizer.
people always seem to concentrate on the inductive part of the reactance and ignore the capacitive part. It's like they take the simplified LF model and extend it to HF.
I don't worry about it that much, since 3 to 5% match is very typical, and mismatches greater than that will start to affect DC balanced unfavorably.
to wander off the topic just a bit, i also feel that an intentional airgap is mandatory and luckily this nicely compensates for a little DC balance drift. Remember even order (asymmetric) distortions are the same as DC on the core :-)

1) Phase match from 10Hz to 100kHz, preferably (much) better than 3 degrees from 15Hz to 50kHz. This is phase match into the grids of the following stage, not the load on the preceding stage. The quality of the grid drive is critical to the sound of the entire amplifier.
the two chamber bifilar should do this easily and allow for plenty of inductance even with a gap that will handle 10-20% current imbalance. I do feel this is an important factor. you need some wiggle room with the operation of the tubes, otherwise you get neurotic wondering if things have drifted.
2) Relatively low additional capacitance burden on the preceding stage. Another 50pF per triode section is too much.
i assume you mean shunt capacitance that the tube must drive. the bifilar approach has a large series C but that, as far as i can tell is a non-issue. There are those that claim to not like the sound of bifilar iron, but i consider many of those reports anecdotal at best.
3) Good current-mirroring capability from 500Hz on up.
ouch... i'm not sure i see this happening much outside the top end of the audio band without interleaving the windings which opens up the other "cans o' worms"
4) Uh, oh yeah, good sonics.
this is actually the easiest one since it cuts to the chase. it should help prioritize 1-3. How many "ideal" designs have you heard that sound like crap :-)

BTW, let me know if you want to move this to a section with your name to keep it concise and allow it to splinter in any direction you choose.

Dave
Lynn Olson
Posts: 16
Joined: Thu Jun 23, 2005 5:25 am
Location: Colorado, USA

Right Here is Good For Now

Post by Lynn Olson »

I've been enjoying all the back-and-forth this thread has generated, thanks again for the great (and very lively) forum.

Sorry about the highfalutin' non-hands-on approach, but the recent move to Colorado, the completion of our new house in late August, followed by the preparations for the Rocky Mountain show are going to things exciting around here for the next several months - and focused away from audio except for the little thought-pieces like the current thread.

It has been great meeting the Colorado Boys - I guess you're right, I'm one of 'em now, what with the CO drivers license and a house under way. These guys are real live wires - one day after I remarked the Loesch preamp sounded like it had slewing/overload problems, a call is made to Steve Bench in San Diego, three days later, the SBench preamp is *here* in Colorado being auditioned by one of the boys, and the following weekend, I got to hear it in three different systems, all at a high plane of performance (all three systems had vinyl setups with Schroeder Ref tonearms, pretty fancy if you ask me). These dudes don't fool around - pretty fast company around here. The 75TL amp, in particular, is a really something else.

The Amity/Aurora/Karna have developed more or less as think-pieces, Gedanken if you will, and the sonics were just sort of an afterthought, something I do more towards the end of project. That means folks who follow in my footsteps are taking a risk, since I sometimes abandon an entire line of thought, like the Aurora, which just didn't pan out for me. The marginal performance of the first stage in the Aurora, with its ridiculously high sensitivity to tube-type and coupling cap, showed me that the circuit was just too touchy and parts-sensitive.

This point is easy to misunderstand - it was transparent, all right, but most of the input tubes just sounded dreadful - and these were fine tubes in the 6J5/6C5 family, which by rights should sound damned good. And they didn't - only the ultra-rare British exotics sounded acceptable. Granted, the WAVAC HE833 uses astonishing tubes as well - NOS WE 437A's, Genelex KT88's, and of course the famous 833 transmitter tube - but I doubt the design would sound unlistenable with more mundane tubes.

The finger of suspicion with all of these amps - the Amity, the Aurora, and now the Karna - always seems to come back to the input stage, which has been a little annoying and frustrating. The DHT part is always spotless, ultra-transparent, dynamic, colorful, filled with emotion, you name it. It delivered. But the input stage - argh!

Switch to a different input and you have a completely different amp, with almost no sonic resemblence at all. I've even been tempted to go the Dark Side and come up with a studio-type solid-state line driver just to deliver those first ten volts cleanly to the driver. And even though I designed the Raven linestage, I'm not really a fan of it, since to me it sounds just like a lot of other linestages, maybe a little less colored, but nothing all that amazing. I almost feel like I am forced against my will to build a quiet, noisefree DHT line stage, something I really don't want to do, considering the massive obstacles that have to be overcome with this sort of design.

Part of the appeal of the direct coupling is reducing the stress on the first stage, making its life a little easier, and maybe getting away with tubes that don't have an ultralow Rp for a change. It would be nice to select tubes for sonic charm, not just low Rp and other characteristics.

Hmm - I guess what I'm looking for is a buffer+current mirror, which could lead to all sorts of over-complex circuits - maybe a temptation that should be avoided. Then again, it could be as simple as a PP cathode follower, loaded by a CT choke or IT on the cathode side. Hmmm ...
Lynn Olson
shinebox
Posts: 88
Joined: Fri Jan 28, 2005 9:44 am

75TL

Post by shinebox »

Allo Lynn,
Just wondering - anywhere we can get more details on this 75TL amp?

Many thanks,
shines

PS As for a solid state buffer stage - well, a certain cat tried all sorts of stuff and the only thing he found that was truly transparent was sand, so may well be worth a try...
IslandPink
Posts: 56
Joined: Thu May 12, 2005 8:01 am
Location: Denbigh, North Wales

More misc

Post by IslandPink »

Hi Shines , Lynn et al ..

Ditto, would like to hear more about the 75TL amp .

As regards the Wavac 833 amp, there's an interesting account from Romy , about his encounter with the Wavac amp at CES 2005 . Not sure if we can glean much from this, in the system context, but it reminds me of some of the problems I've experienced getting small-signal info through transformers - made me wonder about the interstages in the Wavac . Of course it's difficult to know with Romy ...

http://www.goodsoundclub.com/Forums/Sho ... PostID=484

MJ
dave slagle
Posts: 2085
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Post by dave slagle »

Sorry about the highfalutin' non-hands-on approach
No need to apologize. Part of the goal of this place it to "lend" a few extra sets of hands. I have a fairly long train commute each day so I have time to type and do paper designs and the little spare time I get at home to actually wind and measure is put to good use. If I then had to build and listen to this stuff, I'd never get anything of value done. In my pre-k days (k is for kids) I formed my basic design beliefs and feel I can now rely a bit heavier on the experiences of others to make better use of my limited time.
and focused away from audio except for the little thought-pieces like the current thread.
These thought pieces are very valuable to me since they deal on a more conceptual level and if the concept isn't right why bother with anything else?

Part of the appeal of the direct coupling is reducing the stress on the first stage, making its life a little easier, and maybe getting away with tubes that don't have an ultralow Rp for a change. It would be nice to select tubes for sonic charm, not just low Rp and other characteristics.
Funny you should say that. For some time now I have been a fan of low Rp Gm tubes because they make high bandwidth magnetic devices easy. The 6C45, 437 etc. can also often replace two active stages.. So it has been a necessary compromise in my designs for the past several years. (for the 6C45 I'll keep it at about 100mv input, never in a position where it could see 2V) I guess I just accept it as a necessary evil.

dave
IslandPink
Posts: 56
Joined: Thu May 12, 2005 8:01 am
Location: Denbigh, North Wales

Thoughts

Post by IslandPink »

Hi Lynn ,
You could put a warning at the top of your page :

“Danger - Thought Experiment�

This project has raised a few very interesting issues . One area I’m thinking a lot about right now is something Bud Purvine said recently about plate area, ‘brownian motion’ and other such stuff . I’ve always had a feeling that you can’t successfully get a 'musical‘ signal through a large component ( transformer or capacitor ) if the tube ( I’ll use the American convention here ) isn’t BIG enough . I noticed this first with phono amps - some quality caps , big ones, don’t sound good in phono amps where the tubes ( and plate areas ) are small . I also always found Russian teflons a bit disappointing , even in power amps , lacking a bit of life , although their mid and upper signal level performance was very clean and neutral . I always had a suspicion that small signal levels were ‘getting lost’ in the cap , and they are VERY big for their values .

Now in the case of the Aurora , there are sticking points where the input splitter and the interstages are used . If you go to bigger tubes , like Lynn has for the drivers ( 45s ) then it gets much better , now I suppose this could just be on the basis of driving the 300B grids more successfully, but I think a big tube like a 45 may also be more successful in sending small signals through the interlayers and core of the interstages, too . - I wonder if this was what drove Sakuma to keep working on those big strings of DHTs and interstages ( 845s in a phono amp ?! ) .

Bud has suggested that larger plate area and size of electron cloud ( space charge ? ) has a definite advantage in overcoming the surface charge effects and randomness at low levels that are present in dielectrics, in eg. transformers . I wonder if this ( more electrons available ) has some connection with the special qualities you get from thoriated-tungsten filaments, which I guess have more electrons hovering around, waiting to be directed ?

I hope this doesn’t take us too far off the specific discussion about the circuit and the choke design, but I can see why you might want to ‘revisit’ the idea of direct-coupling the first stage . That first stage has a tough job – significant voltage gain, needs low distortion over a fairly large swing, enough plate area to deal with the interstage and 45 grids … not to mention well-matched halves if it’s a double triode . Getting rid of that first interstage could free-up a few design parameters.

I could see myself eventually building a smaller version of the Amity/Karna , or something like Gary’s 1624 amp . I think the 300B version comes down , eventually to having a serious three stage amp, like a good SE 300B , in order to get the 300B to sound really good . Unfortunately , that means X2 all the tubes for PP !

I think I’ll go off and build some SE stuff for a while , to see what makes these VV32s tick …

MJ
Lynn Olson
Posts: 16
Joined: Thu Jun 23, 2005 5:25 am
Location: Colorado, USA

Latest Thoughts

Post by Lynn Olson »

After hearing John Atwood's brilliant phono preamp, which has battery-bias on the grid side, I'm going to give try a variation on the Karna amplifier. See: http://www.nutshellhifi.com/Karna-alternate-c.gif

This will have to wait until after the Rocky Mountain show, but it should chase out some subtle cap-colorations in the front end, where they are most audible (low levels and all that, doncha know).

During the course of some long chats with John Atwood when he was visiting in the Denver area a few days ago, he mentioned that the main cause of the "battery coloration" are the varying currents they see when they are in the cathode circuit. By contrast, in an input grid circuit, the DC and AC currents are very very small, and the nonlinearities of the battery are not excited.

In a balanced transformer-coupled circuit, I have the luxury of using the battery on the CT of the secondary, instead of between the grid and the input, as in John's SE RIAA preamp.

What's not shown on the schematic is that it's probably desirable to use two separate 5687's, with the non-used section tied to ground. Another clever thing John does is his "Cool-Swap" technique, where the active section of tube socket 1 is triode 1, and the active section of socket 2 is tube section 2. Swap the tubes, and you've also swapped active sections on both tubes, thus doubling tube life, while keeping the unused section cool. This also effectively increases the usable plate dissipation compared to the more normal practice of using both halves at the same time.
Lynn Olson
dave slagle
Posts: 2085
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Re: Latest Thoughts

Post by dave slagle »

it would be interesting to try a CT choke in place of the input transformer. Maybe for kicks just try the secondary of the transformer and see what happens. You will not get the step-up, but you will have much better drive ability

depending on the turns required, the CT choke could also be wound as a step-up autoformer to allow for some extra gain if needed. (a 6dB boost switch could be added to select the tap)

dave
Raj Gupta
Posts: 37
Joined: Fri Feb 11, 2005 2:19 am

Post by Raj Gupta »

LynnO:

I'm very interested to hear that you, too were impressed with John's phono preamp. I had gotten an email from Chris Boettcher saying that that pre transformed vinyl into a completely new medium - pretty strong praise!

Anyway that piqued my interest and I wrote to John and started thinking what design ideas I could steal from him to try. I had been thinking about an Aikido phono but it's hard to get enough gain out of that topology without cascading, and I decided to put that on hold for a time.

When I opened your attachment, I thought, "Gee that's a lot of iron for a phono preamp! I guess whatever is wrong with Thomas Mayer is rubbing off on Lynn!" But then I noticed that it had no RIAA equalization and it dawned on me that it was probably a power amp :)

Here's what I'm thinking about trying: 6N1P grid-battery biased, direct-coupled via the RIAA EQ into 6SL7, with a gas reg keeping the 6SL7's cathode up at the necessary voltage. Three sockets for a stereo phono preamp, and not a whole lot to wire up. Should keep me busy for a little while anyway!

-j
Attachments
6N1P Phono.jpg
6N1P Phono.jpg (15.49 KiB) Viewed 41922 times
shinebox
Posts: 88
Joined: Fri Jan 28, 2005 9:44 am

phono scheme

Post by shinebox »

Allo J,
Something very similar is on my project list (around about #10 though...aargh). Couple thoughts:

1) I've never seen anyone use zero bias for the input stage with tubes and I'm wondering why; an MC will laugh at the impedances involved, MM should be fine too. So long as the dc grid current is minimal, 'course. I was hoping to try these puppies at about 130V, 0 bias:
http://www.nj7p.org/Tube1.php?tube=7768

At 0V bias with any tube, you might actually get the gm the manufacturers were claiming too...


2) My worry is that the VR tube will be too noisy - after the RIAA network loss, your signal won't be much higher than where it started, and those VR tubes have noise of the order of a mV I think?

OTOH if you can get it to work, should be rather splendid. The reference tubes like (what is it, 85A2???) may be a better bet - see if you can find that old tubecad article which lists a number of them.

If it doesn't work, the solution for dc coupling is prob diff pair. Then you can take the output of either phase for polarity switching.

Or, how's this for a crazy idea - make the second stage a screen driven pentode (you might have to drive it from a cathode follower though or risk loading down the RIAA etc)?

Anyway, are the schemos John's phono on the web somewhere?

best,
shines
Raj Gupta
Posts: 37
Joined: Fri Feb 11, 2005 2:19 am

Post by Raj Gupta »

John's preamp is the Artemis Labs PH-1 and info can be downloaded here:

http://www.artemislabs.com/docs/PH1_Manual_1.pdf

There are no component values on the schematic but that's not going to stop US.

Rgearding 0 bias, how much DC current would it take to toast an expensive MM cartridge? Maybe it's a s simple as that? Also, I have 0 experience designing with 0 bias, so I guess that might be other people's reason?

VR noise - well, you are allowed to bypass them a little, so maybe I'll have to do that. I just thought it might sound better than an RC combo for bias.

I spent some time working on it today, it's actually pretty far along already. I'll put up a new scan with comonent values once I see what I have in the drawers for the RIAA caps. I'm hoping for .033 and .011, I've found the .033's alrerady, I'm keeping my fingers crossed that I can kludge up something for the other value if I don't actually have .01's and .001's.

-j
Dave Cigna
Posts: 96
Joined: Sun Jan 23, 2005 10:58 pm

Post by Dave Cigna »

zero bias?

Isn't that a battery on the grid of the input tube?
shinebox
Posts: 88
Joined: Fri Jan 28, 2005 9:44 am

Post by shinebox »

Hey DC,

It was me who measured zero bias as a way of circumventing the biasing conundrum (not to mention reducing noise etc). Different conversation, ie shines going off on a tangent as usual...

Long time, hope all's well with you.

Best,
Chris
Raj Gupta
Posts: 37
Joined: Fri Feb 11, 2005 2:19 am

Post by Raj Gupta »

We're totally threadjacking here, too.

-j
Guest

Post by Guest »

More like forumjacking... I'll be quiet now but look forward to hearing how you fared.
Cheers,
shines
Lynn Olson
Posts: 16
Joined: Thu Jun 23, 2005 5:25 am
Location: Colorado, USA

Not Hijacked Yet

Post by Lynn Olson »

Still reflecting on the sonics of John Atwood's preamp, particularly how battery bias in the grid is a lot more benign than the same thing in the cathode - due to much lower current flows.

Even though I'm the guy that made a lot of noise about thinking of the current loops in a circuit, it was John that took it a step further and made me think of *how much* current each of those loops has inside them. The output and cathode side obviously have a lot, but the grid-side loops have only very small amounts - and the potential distortion mechanism of caps and batteries (which are very similar devices) may be more of a current modulation than a voltage effect. Good question what's really going on inside these things.

Anyway, after the Rocky Mountain show comes and goes, here are some variations on the Karna amplifier I plan to try:

http://www.nutshellhifi.com/Karna-alternate-a.gif
http://www.nutshellhifi.com/Karna-alternate-b.gif
http://www.nutshellhifi.com/Karna-alternate-c.gif

As for phase splitting with center-tapped chokes, I dunno. The overall operation of a PP/balanced circuit is exquisitely sensitive to the phase relationship between the two halves - 3 degrees of phase drift, barely visible on a scope, is equal to 5% gain mismatch, enough to have a substantial effect on the harmonic spectrum.

At a grosser level, when the CT choke shifts a full 180 degrees away from the desired action (either zero or 360 degrees of phase shift), you actually get a zero in the overall system gain, not the most desirable outcome either. And a CT choke is absolutely guaranteed to do just this at a low enough frequency, or a high enough frequency where capacitance dominates. Worse, the instability of phase angle coming out of each end of the CT choke is going to be level dependent.

I'm aware transformers do weird things as well, but the two halves of the secondary usually get in trouble together, instead of drifting apart. Anyway, I have no problem using a CT choke as a phase improver, or looking at it another way, a current mirror. As long as the CT is at a low impedance (relative to the Rp of the plates of the preceding section), there should be a useful gain in balance for the following stage.
Lynn Olson
IslandPink
Posts: 56
Joined: Thu May 12, 2005 8:01 am
Location: Denbigh, North Wales

Batteries

Post by IslandPink »

Lynn,

Agreed on the battery bias observations. I've never liked the sound of batteries ( NimH or NiCad) in cathode circuit, either in a driver or phono amp stages, but they sound very good in the grid circuit , and help avoid a big electrolytic . I currently have something like the Loesch phono circuit, with the second stage biased by a 3V battery and voltage divider instead of a grid leak resistor .
Funny the Artemis Labs circuit should pop up again, I saw it recently on a thread by Thorsten . He pointed out you can parallel a small film cap with the battery for extra transparency . I tried it in the phono , but the change ( or improvement ) was too close to call . It would be interesting to try in something with a more substantial voltage/current swing, perhaps .

Mark
Lynn Olson
Posts: 16
Joined: Thu Jun 23, 2005 5:25 am
Location: Colorado, USA

The source of colorations

Post by Lynn Olson »

I was not expecting much of John's preamp, mostly due to a pre-existing prejudice against 12AX7's and cap-coupling on the output. Well, I was surprised - the preamp is about as good as Steve Bench's wild design - the "Colorado Boys" had the opportunity to compare the two directly against each other on several different systems, and the differences came down to a sort of flavor difference, with no clear-cut favorite.

Both are very very good, and lot better than the usual audiophile stuff. Both are also free of the annoying and highly colored preamp slewing that sounds so much like cartridge mistracking (but isn't).

John walked me through how he gets away with the 12AX7 - it's in the middle of the preamp, is *not* driving any reactive loads (such as the RIAA network, the output, or a feedback network), and sees only a very high value plate load and the grid-resistor for the much more muscular 5687 output stage. It's basically a pure voltage amplifier, well buffered from the ugly reality of the RIAA equalization network and driving the output cable. The simple MOSFET voltage regulators sound quite good as well, something I didn't expect, but they're in a location where the high (and nonlinear) capacitance of the MOSFET doesn't matter much (it would be a different story in a current source, where you need bootstrapping techniques a la Gary Pimm).

Anyway, listening to the preamp got me thinking about the "coloration budget" of a circuit. If you're using parts that are colored-sounding, put them in a location where the colorations don't matter much, or where circuit conditions don't allow them to arise in the first place. Batteries have very low DCR, but are just as colored as caps - but unlike caps, don't have frequency-dependent colorations. So placing them in the grid circuit keeps audio-frequency currents to a minimum, in the microamp or less region, thanks to the high impedance of the first-tube grid. Note this would *not* happen in a power-tube grid, which does flow current, and in a very nonlinear way, too.

Placing the shunt caps between the primary CT and the cathode minimizes the usual strong PS cap + cathode bypass cap coloration, but it's still there to some degree. The coloration is kind of turned inside-out in a balanced circuit, since all you hear is the coloration that is generated by circuit imbalance and odd-harmonic distortion. If the circuit were perfectly balanced and free of odd-harmonic distortion, no current would flow through the capacitor, but in the real world, these error terms do flow through the cap.

You don't hear the direct and obvious coloration of a SE circuit, where the character of every part is right there for you to savor. You hear the error-term coloration instead, which fairly benign so long as the part coloration is low. Well, that doesn't happen, so you have to be extra-careful about the parts quality of anything in the common-mode circuit. That's why Teflon bypasses are useful, and I expect the combination of battery bias (in the grid circuit) and shunt regulation should be more transparent and less colored than a shunt capacitor.
Lynn Olson
dave slagle
Posts: 2085
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Post by dave slagle »

hey lynn,

i was hoping to come up with a more eloquent reply, maybe someday I will but anyway here is the short version.

when i mentioned the "CT choke as a phase splitter" I was referring to its use as a "current mirror" on the output of a transformer in much the same way two resistors can be used to derive a balanced signal from a 1:1 transformer.

By no means was I referring to the far more common (and mislabeled) procedure of calling a transformer drawn as a "CT choke" a stand alone "autoformer" phase splitter. I have tried and tried to strike up a discussion on the merits of this, because the measured performance between the individual phases is radically different yet so many folks say it sounds great. The only way i can seemingly connect the measured with what is heard is to make note of the fact that one side of this arrangement has perfect responses while the other is totally screwed up and the output is the sum of this. The key here being the sum of a screwed up signal with an ideal signal nets you a somewhat less screwed up signal but the sum of two randomly (albeit less) screwed up signals might be worse.

i agree 100% with your current mirror concept of the CT choke, and feel that may be the important factor for large signal devices (transformers at low frequency) but at small signal, the capacitances become the primary evil. (i am referencing balanced operation here where leakage can be ignored) and it's not a matter of small capacitances but matched capacitances. Of course small capacitances will push the phase mismatch up in frequency which simply means another solution to this problem is to keep the iron at low impedance so the phase crazies happen above a meg or so :-)

just following the gedanken lead and threadjacking is encouraged since we can always make a new topic, thread or forum if the 'jacker isn't just jackin'

(sorry about that one)

dave
Bas Horneman
Posts: 19
Joined: Wed Jul 27, 2005 10:15 am
Location: The Netherlands

Post by Bas Horneman »

Hi Lynn,

My first post on this forum, I must admit that I am a bit intimidated by all the "names" on this forum. With regard to the input stage of your Karna, I was wondering why you don't simply choose the brute force solution (or rather the a lot of cash solution) maybe even the "natural"solution. And that is to just use another DHT on the input as is part of your philosophy? Such as the WE remakes 101d or 205d by TJ?

Regards,
Bas
To infinity and beyond!!!
Post Reply