[ginger]

Ha! Breaks his arm trying to pat his own back. This is to announce that I've finally (after several years effort) designed and built an amp which I like marginally better than my 55 Nivarna Plus.

4 x KT88 Tubes driving Plitron Toroidal output transformer each channel. Swish phase splitter and driver (tubes as well of course).

So how much did it cost in parts - suffice to say, I could build another 4 off 55 Nivarna Plus Amps and still have change for the same price.

Rest assure all you AKSAPHILES are getting value for your hard earned bucks.

Cheers,
Ginger

[andyr]

Ha! Breaks his arm trying to pat his own back. This is to announce that I've finally (after several years effort) designed and built an amp which I like marginally better than my 55 Nivarna Plus.

4 x KT88 Tubes driving Plitron Toroidal output transformer each channel. Swish phase splitter and driver (tubes as well of course).

So how much did it cost in parts - suffice to say, I could build another 4 off 55 Nivarna Plus Amps and still have change for the same price.

Rest assure all you AKSAPHILES are getting value for your hard earned bucks.

Cheers,
Ginger

Congratulations, Ginger!  Is this an amp with some real muscle or is it one of these miserable little things you need to use with horns, to be able to hear anything??

Regards,

Andy

[Occam]

Ha!.....
4 x KT88 Tubes driving Plitron Toroidal output transformer each channel. Swish phase splitter and driver (tubes as well of course).
........
Cheers,
Ginger

and which of Van der Veen's abberrant topologies did you choose?
Iff'n you're using those tres 'spensive torroids, I'll assume your using CFB(modified UL) windings [or those totally separate UL windings?], and hence massive drive swings... and Herculean efforts to keep dc imbalance off the output trannie.
Wots a swish phase splitter? Concertina, paraphase, diff... curious minds need to know.

Dang! If I could get something that sophisticated and expensive to work, both my arms would be broken.
Congrats!

[AKSA]

Ian (Ginger),

Thank you for your interesting revelations about a PP EL34 tube amp to best the AKSA!  I congratulate you!

You have absolutely and completely vindicated the AKSA philosophy by actually going one better, though at far higher cost.  As you know, it was principally my outrage at the cost factor of tubes which drove the design of the AKSA.  It is interesting that you've done it with PP, not SET, technology.  The power supply issues with SET seem too problematic, with the huge current swings, which are scotched utterly with PP Class A.  And the fact you did it with EL34 is amazing;  I'd have expected nothing less than a couple of 845s, or at the least 811-3s.

Are you using a bi-phase split cathode follower with floating power supply for the phase splitter?  This, to me, is the most promising phase splitter of all;  accurate and repeatable by intrinsic design.  But you might need a separate power transformer for it........

What is the voltage amplification?  A couple of 6SN7s?  Knowing your quest for complete transparency, I'd expect something pretty similar.

Ian, are you using UL and/or feedback?  As we discussed before, with audio you either use about 4dB or about 64dB, and nothing in between!

Lastly, what sort of speaker?   A high Q design, with moderate sensitivity (the VAFs) or a low Q design, with high sensitivity, like a Fostex?  And does the toroidal OPT realize the promise of larger bandwidth without saturating with the slightest DC current imbalance?

Cheers,

Hugh

[ginger]

Phase Splitter is the garden variety diff amp extended in two directions. First each side of the diff amp is a cascode using either 6N1P or 6DJ8 providing higher gain and lower input capacitance. The diff amp is current source biased using the common mode voltage on the 2 top anodes connected to the current source transistor base by separate "summing" resistors - that is the current source is adjusted at audio frequency to ensure exact AC balance from the phase splitter. Powered from a High Voltage Regulator. The design has been arround for a while - Its based on the original "Hedge" circuit and has been used by amp manufacturers like Sonic Frontiers and many others. (Aside: CAE have PCB and parts kits for this design in mono and stereo form, intended for Dynaco Upgrades)
 
Driver is 6SN7 direct coupled cathode follower. These drive 4 KT88 in connected to a VDV2100 output transformer - ordinary Parallel Push Pull Ultralinear fixed biased to deliver 120W with no global feedback.

So Andy - yes lots of grunt coupled with a gorgeous live feel and those tingly highs we all seek.

The amp is by no means finished yet.
I want to experiment with adding about 6 dB of global feedback and then compare this arrangement with using VDV2100-CFB Output Transformer to add cathode feedback to the Ultralinear connection.

Hugh, this would be Menno's "Super Triode" configuration. I'm not there yet but it on "the plan".

I'm still using my AKSA 55 N Plus as my reference amp.

Cheers,
Ginger

[tinears02]

Hi Ginger
Would you be able to point to a schematic or description
of how you balance the two arms of the differential circuit
by feeding the anode voltage to the current source bjt?
I'm interested to see how this can be achieved.
Also, what do you 'measure' in order to determine
if it is ac balanced or not?

Thanks in advance
Tinears02

[ginger]

G'day Tinears02,
The front end circuit is pretty much identical to the one posted here for a Sonic Frontiers Amp.
http://www.diyparadiso.com/proj/sfm75.htm
Ignore his comments about SRPP, NOT an SRPP in sight. I guess if you look fast and briefly enough a cascode it might be confused for an SRPP.

leave out the coupling capacitors to the output stage, add direct coupled cathode followers to drive 2 pairs of KT88s.

I managed to emasculate the amp last night - added 15dB of global feedback, very "nice" in the original meaning of the word and really dull. This mod will come out again tonight.

I was trying to get a happy medium between the RAW beauty of the amp in Ultralinear with no global feedback and some of the much more civilized and less dramatic sound of my AKSA 55N Plus.

Plan is to switch the KT88s to triode mode for the next test.

Cheers,
Ginger

[Occam]

Ginger,

Thanks for the link to the DiyParadisio site. Benny, the proprieter, is really good folks. I must confess, I've always had a hard time wrapping my head around the Hedge circuit. There is an article on it  in an old Glass Audio that I think I'll revisit.
Are you using resistors for voltage drop for the screen feedback? I've read some references to using leds and resitors for voltage drops, (for both ultralinear and triode connections), but have never talked to anyone who has actually done it....

Like Hugh, I find the idea of a CF phase splitter has tremendous intellectual appeal,
http://www.tubecad.com/december2000/page10.html
but the requisite floating supply, as well the inherent necessity for capacitive coupling, gives your direct coupling approach great appeal and practicality.

[ginger]

Occam,
I use 330R/3W (Actually 3 x 1K/1W in parallel) in each screen connection to the UL taps.
Voltage Rating of the screen grids is the reason I went with KT88 rather than 6550, 6L6 or EL34.
Large screen resistors work well in Guitar Amps (the distortion and soft limit characteristic set in at much lower levels which means you don't need to set it on 11 to get the sound you want) BUT are'nt so hot in a HiFI Amp where you want clean transients.
As a professional design engineer I find the practice of designing in under rated parts and then trying to protect them is a seriously crap design approach. Am I saying that 90% of valve amp designers in the last 30 years have got it WRONG? Yeh! I am - no apologies offered. You wouldn't do it for a semiconductor design (It would'nt survive) BUT because valves will tolerate a bit of abuse it seems everyone want's to abuse them.
Cheers,
Ginger

[tinears02]

Hi Ginger
Thanks for the link.
I guess that the 1M ohm resistors really have to be
different to balance the amp?
FWIW, Ralph Karsten of Atmasphere uses  cathode
followers to couple to the final stage too. And he
cascodes the cathode followers for a 'cleaner' sound.

cheers
tinears02

[ginger]

NO the 1M resistors have to be equal.
The junction of the two 1M resistors (the base of the current source transistor) looks a bit like a "virtual earth" as far as AC is concerned. That is, if and only if the AC signal at the 2 anodes is exactly balanced (BUT opposite phase) there will be no AC voltage at this point. If there is any out of balance then an AC signal will be present at this point which will feed into the current source and adjust the current at audio frequencies such that balance is restored. This ensures perfect balance from the diff amp regardless of wether a feedback signal to the negative input (the grid of the lower triode of the right hand side cascode) is present or not. This is where the behaviour of the circuit differs from a standard diff amp (wether it is current source biased or not) and is why you don't see unequal anode load resistors or any of the other usual schemes for trying to achieve balance. Its a closed loop feedback system. This circuit method also works beautifully with 1/2 of a 12AX7 used for each side of the diff amp instead of a cascode BUT it doesn't have the same high frequency response due to high Miller Capacitance at the grids).
If this circuit has a fault its in the fact that it does NOT guarantee DC balance and in fact in my amp one anode sits at 204 Volts while the other is at 198 Volts - not enough difference to worry about (even given the fact I have direct couple cathode followers hanging off these anodes) BUT you might be forced into selecting well matched dual triodes particularly for the lower valves of the cascode.
There's one more hint: The two lower triodes should be 1 dual triode and the 2 top triodes should be 1 dual triode. This will ensure best DC balance and gives you the opportunity to float the upper valves heater at say +90 Volts if hum is an issue.
Hope this helps,
Ginger

[tinears02]

Hi Ginger
It sure does help, mucho gracias!

Tinears02

[ginger]

Friday night I started backing out the amount of feedback as the 15dB I had set earlier (at least partially) neutered the amp.

Step 1 was 11dB of global feedback -
Damn, I've cloned the AKSA55N Plus.
Hugh always claimed he had tried to get a valve sound from the AKSA. We can give this claim added credence since I've managed to do the reverse. - I swapped back and forth between the 55N+ and the 4 x KT88 Ultralinear with all sorts of music and could'nt pick between them (well the AKSA had a slightly tighter bottom end but otherwise the amps were spookily similar).

Started backing out the feedback again and every time I reduced it the amp sounded more "live" and more "exciting" until I got back to zero feedback. Swapping to other music selections soon pointed out the obvious faults with zero feedback. Spiderbait and Nickleback sounded stunning, Enya sounded crap. Just not enough damping.

A quick change of the output valves configuration from Ultralinear to Triode connection - This improved damping BUT was subjectively slow, I didn't like it. This stacks up with my previous experience, I've just never been a big fan of triode mode for push/pull.

Saturday - back to Ultralinear to get the lightning fast attack back and then started adding feedback again. The longer I listened and the more types of music I tried the more feedback I tended to put back in BUT all the time keeping the 11dB "AKSA Clone" level as an upper limit. Settled on 8dB in the end, I swapped between 6dB and 8dB a lot before before settling on 8dB.

Cheers,
Ginger

[kyrill]

thanx for telling us Ginger.
I like that. Actually in my inagination I was sitting next to you to listen to the subtle details as well. I have a close friend to do this kind of listening, albeit not on yr level od e-hardware design. Do you have a  female to help you in listening?
Kyrill

[ginger]

Kyrill,
Have done some more listening since. I am rapidly comming to the conclusion that ANY global feedback kills the excitment and speed of the amp. Unfortunately Ultralinear Mode alone does not give good enough damping (output impedance too high). This is suggesting that while we might only hear frequencies to 20kHz or so we can detect subtle waveshape and leading edge phenomenon associated with frequencies to perhaps 5 or even 10 times that. That is we are very sensitive to phase and edge speed which is where a lot of the spatial or imaging information is contained.  Engineer types talk about this as thinking in the "time domain" rather than the "frequency domain". As soon as time delayed information from the output is connected back to the input we loose time coherency and image and "liveness" start to be destroyed. The strange thing is that solid state amps do not seem to exhibit this to the same degree. The answer that leaps to mind is that the solid state amp is inherently faster and has smaller delay through the amplifier ("group delay") BUT this is just NOT the case. It was to experiment with these ideas that I bought the Plitron Output Transformers which have response to well over 200kHz.
So my aim now is for a zero global feedback amp (if its good enough for Halcros it should suit me). The prototype amp is back in Triode Mode zero global feedback as best balance between enough damping (just) and "excitement" (or "liveness" if you like). It will stay this way until new cathode feedback output transformers arrive from Plitron.
Have detoured this week anyway to do a restoration on an Ampeg VT-22 Guitar Amp for the local Music Shop ( I do about 2 or 3 Classic Valve Amp restorations for them each year).
As to if I have a female to help with the listening - well yes and no, Who can afford a parter AND a valve amp hobby. What I do have is a very "arty" cleaning lady who does 4 hrs each Friday. She brings her own music selections, has a set selection of test music from my collection and leaves me a report card on the latest amplifier mod. A second opinion is always valuable. Nice to get home from work on a Friday evening to a nice clean house and a fresh amp review.

Cheers,
Ginger

[AKSA]

Ginger,

I listen very carefully to your comments.  I'm sure you are dead right.  I agree with your statement on the use of GNFB on tube amps, too, but have come to these conclusions over a decade of tinkering:

1.  Tube amps are really only practical with OPTs because OTLs need lots of tubes and even more GNFB which kills proceedings somewhat, though not as much as GNFB with tube amps and transformers.

2.  The only way around this dilemma is to use transformers with a high turns ratio to deliver low Zout.  This flies in the face of reality since anything over about 10:1 is a compromise, but it does propel us towards low rp tubes such as the 6C33C, around 150R.  This enables use of a 600R:8R trafo, which is a turns ratio of just 8.66:1.  If we go to 800R we have our 10:1 trafo and this should deliver a sufficiently low Zout for most things......  Double these figures for PP, of course.

3.  Something very special happens with low mu triodes of relatively high rp, like the 211, 845, 811-3, 572-3.  These all require highish turns ratios, but if the OPT is done right, these SET amps remain, to my ears, the best, though their bass leaves something to be desired.

4.  On SS amplifiers, we either have no feedback, or lashes.  There ain't nothing in between.  Since we are talking GNFB, we are obliged to use very fast devices throughout, but must deliberately throttle back the VAS.  It is here that the compromises are most obvious to the sonics, but the low Zout and sheer drive is very appealing and something no tube amp can match.

5.  An option you might consider is a hybrid of bipolars and mosfets, using local feedback only for a power follower.  Then voltage amplification can be a tube at the front end with zero feedback.  Local feedback can again be raised to 60-70dB, like the AKSA, and correspondingly we can then achieve very low Zout.  This is clearly necessary for today's speakers, which are all designed for pure voltage sources.  I've tried this and it works extremely well, and with huge care can be done in non-switching Class AB.

Comments?

Cheers,

Hugh

[Occam]

Ginger,

Those torroid OPT wouldn't have any cathode windings by any chance?  As I recall, the dual secondaries on the Plitron (Van der Veen) trannies are dual 5 Ohms, which hooked up center tap grounded would allow you to apply cathode feedback in addition to the UL feedback. This was done by Audio Research in some of their models, as well as a relatively common mod  on the Dynacos with a ground at the 4 ohm tap and feedback taken to the cathodes via the "0" and "16" ohm taps. It worked quite nicely on the old Eico HF-86 as you could use the '0', '8' and '32' secondary taps for cathode feedback duties. Its a way of getting some additional dampening on the output stage without going global.

[Occam]

Ginger,

Those torroid OPT wouldn't have any cathode windings by any chance?  As I recall, the dual secondaries on the Plitron (Van der Veen) trannies are dual 5 Ohms, which hooked up center tap grounded would allow you to apply cathode feedback in addition to the UL feedback. This was done by Audio Research in some of their models, as well as a relatively common mod  on the Dynacos with a ground at the 4 ohm tap and feedback taken to the cathodes via the "0" and "16" ohm taps. (Sakura also did this on many of his SE designs as I recall.) It worked quite nicely on the old Eico HF-86 as you could use the '0', '8' and '32' secondary taps for cathode feedback duties. Its a way of getting some additional dampening on the output stage without going global.

[ginger]

Occam,
The PAT4006 trannies I have at the moment don't have cathode feedback windings or tapped secondary- just UL taps on the primary and a single 5 Ohm secondary winding. I started this design using (relatively cheap)Hammond 1650T output trannies and then made piecewise improvements from there. I picked up the PAT4006s cheap from someone who had them sitting on their shelf for sometime. Fitting these in place of the 1650Ts confirmed to me that Mr Partridge knew what he was talking about when he described output transformer distortions all those years ago. You need large inductive reactance (from the transformer primary inductance) compared to the valves output impedance for low distortion. The PAT4006 Primary Inductance of 530H compared to 59H for the Hammond 1650T was immediately apparent in eliminating the "covered" sound of the Hammonds (mostly 2H distortion as the output valve source impedance approached or exceeded the primary inductance impedance). That is I chose to address this with better transformers (higher primary inductance) BUT  I could just as well tackled it by reducing the output valves impedance which is what is achieved by cathode feeback and other feeback schemes.
 
The other reason for low output valve source impedance is to get a decent damping factor. It is in this area that the amp is currently lacking. I can get a big "live" sound compared to the AKSA 55N+ but after listening for a fair while it suddenly dawns on you that you are listening to speaker colouration as the amp delivers more power as the speaker impedance rises with frequency and fails to properly control the bass. When I apply enough global feedback to get this under control I have killed the ""excitement" or "liveness" of the amp to such an extent that I prefer the AKSA. So I need more damping with less global feedback. That means more local feedback via cathode feedback or similar.

Thats why I'm now ordering the VDV2100-CFB/H which has both cathode feedback windings (2 off independent 20 Ohm windings) and a centre tapped secondary so that it can be use as an alternate source of cathode feedback. See Norman Korens website (search on Norman Koren) for an interesting take on this. He uses the tapped secondary for Output Valve Cathode Feedback as you describe and also uses the 20 Ohm feedback windings to provide additional feedback to the driver stage in a Class AB2 Triode Push/Pull design using 4 off 6550s to give about 80Watts.

Hughs comment above about using valves with inherently low output impedance (before extra feedback is applied) such as the 6C33C or DHTs (Directly Heated Triodes) is also relevant to this arguement.
 
I'm trying to keep my design simple - that means Class AB1 (No Grid Current in the output valves).

I am not that keen on having output valve cathodes connected to the speakers via the secondary winding in the fashion you described. Yes plenty of folk have done it in the past BUT what happens to your speakers if you have an output valve go short or your amp suffers a bias failure. I like to design stuff that is "failsafe". Now thats just about impossible with valve circuits but at least I can design for "minimum collateral damage". That means keeping the speakers away from High Voltage.
Cheers,
Ginger

[Occam]

Ginger,

Good points....
Speakers are easy enough to fry without having to resort to double duty cathode feedback/secondary connections.  But while I've seen more than a few Acoustic Reseach CF output stages blow their tubes radomly (but regularly), I've never seen it take out the speakers.
Indeed, as separate cathode feedback winding provides a more tractable environment, but denys us the joy of discovering yet new ways to screw up.

Have you ever met anyone who has actually tried this Broskie special?-
http://www.tubecad.com/march99/page6.html
FDF?