6550a question for engineer?

[KJ2005]

Harmon Kardon V

can you send me scheme please?

please send to antonio<dot>tecson<at>kentech<dot>ie my yahoo mail does not work in my office.

thank you..

Hi Tony,

one more, or you can find the circuits on:

http://www.machmat.com/schema/index.htm

Rgds

 

[TonyTecson]

Hello Kim,

thanks....

 

[KJ2005]

Hi Rich Walters

Stay simple unless one has the competence for a real adventure. The long and the short; anyone beavering with power amps should automatically do the math and keep notes to avoid doing silly mistakes.
Maybe the way forward is to try these power LTP circuits with a CCS in the cathode tail and see how they (brilliantly) perform. It does seem strange that this technique wasn’t adopted on the day, some 45+yrs ago.

richj

Hi Rich,

I get your point. - On the other hand I have built at least 7 tubeamps from old circuits like Leak, Conrad johnson a.o., and without complete professional knowledge and math calculation , they all played ( with a lot of work , intuition and "feeling") up to - or better than many factory built amplifiers. All the amplifiers has been messured in frequence linearity as well as different forms of distorsion.
I have constructed one amplifier with four EL84 tubes pr channel by myself , and the result was up to standard with no hum or extended distortion , but had I reached the limit for an optimal push pull amplifier? - I don't know. 23 Watt RMS / distortion 0,5% at half power , 5 Hz - 56kHz - 3dB , good sonic sound , tight bass responce , excellent midrange and treble - BUT:

Why did the 5,5W RMS ECLL800 Lorenz pp amplifier sound better? 18 Hz - 20 kHz +- 1,5 dB , 1% distorsion at 4 Watt, ultra small o/p transformers from Saba. Was it because we only used one cap in line? The ECLL800 includes two pentodes and one triode . This tube need for gain a half ECC83 . The sound can be compared to the best 300B or 845 singleend amplifiers- Why??

This time I am going to dig deeper in math calculation , and this is the reason why I asked this: " a question for engineer". If one looks at the circuits available you find , lets say, 10 differents layouts for input-and driverstages ? may be a few more? Should be possible to find a "diamond" pp circuit out there - if not - I am going to make it by myself.

My friend you send the fax for is teacher and audio engineer at the Danish Technical University in Lyngby , and in spite this educational background he can't cover all the details or 45 years of forgotten knowledge. He works only with singleended tube amplifiers , and he has invented some new circuits never seen before with linearity up far beyond 100kHz by use of Vanderveen outputs . But this is messured - does this amp sound good??
I would like to show him what a PP amplifier can do for the music. The EF184 drivers are going to the test bench.

If my math can't get me over the hill - I will have to go the heavy way back to my friend and ask!

Best rgds
Kim

 

[sarrisk]

hi ! a long time ago i build this http://sarris.info/main/pushpull6550
may be it is interesting. uses EF86 and ECC83

cheers
costas

 

[KJ2005]

Hi Costas

hi ! a long time ago i build this http://sarris.info/main/pushpull6550
may be it is interesting. uses EF86 and ECC83

cheers
costas

I have seen your tube psu and your mono amplifiers and they looks great. The circuit of yours is close to some of the circuits in my database, and I need to find out whether a "more symmetrical" design sounds better ? - and not how it messures at first! This issue comes later.

Thank you for your reply

Best rgds
Kim

 

[Johan Potgieter]

A few belated remarks from someone who has also been in the game for several decades.

One tends to overlook some time constants with the popular view that a directly connected LTP has only one set of coupling caps "in the signal line". The cap to common from the second LTP grid is as much a time constant as one "in the signal line", thus two time constants C-wise. While this is not usually a problem because it could be made very low, it should be such. Apart from phase shift, its presence will create an unbalance at low frequencies, the very place where that cannot be afforded.

(I am not here to criticise, but in the circuit shown by Richw in his post #19, all of C6, C7 and C19 form time constants - apart from the also sometimes overlooked power supply caps. As signal return they are also in the signal line!)

Then another small point is that while it is attractive to use a CCS as the tail of an LTP, such a circuit also needs to have sufficient linearity with frequency. E.g. using "attractive" power supply regulator chips are usually out. Also, such a circuit only guarantees a constant current through both tubes summed, not necessarally equal division of such through the tubes (each triode) in case of unbalance. While we have all the wonderful modern components for a CCS, I could never bring myself to using one instead of a good resistor and some adjustment in the anode circuit. Guess old habits die hard ....

My own choice for a top quality circuit has always been the Williamson topology, despite the two time constants (as said, the "direct" coupled LTP has it too). In a high quality 100W design using the partial cathode feedback output topology (a la Quad II), I have all of four cap time constants while still getting l.f. stability. But as Richw warned, I would also advise the availability of at least an oscilloscope and signal generator (with square wave) to keep away the blues.

Best wishes on your endeavours, Kimjul. Good advice here especially from Richw.

 

[jrdmedford]

550V too high for modern 6550s

I'd not bank on being able to find tubes that can go that high..
I've heard of people telling me that certain KT88's can only
safely go to 525V.

475V is probably the safe upper limit (design center).

-- Jim

 

[HollowState]

The cap to common from the second LTP grid is as much a time constant as one "in the signal line", thus two time constants C-wise.

For those LTP circuits that are close to the output stage where bias requirements for certian tubes used are higher, then a capacitor from the lower grid to ground is a factor. However, when inversion is done early on, as in the first stage with low bias tubes, a capacitor is rarely needed. A low value resistor will do the job especially when the lower grid is used as a NFB input. But as I said earlier in this thread, it's really designers choice since there's several ways of doing it and making it work.

Then another small point is that while it is attractive to use a CCS as the tail of an LTP...such a circuit only guarantees a constant current through both tubes summed, not necessarally equal division of such through the tubes (each triode) in case of unbalance.

I used to think that way too. So please correct me if I'm wrong here. A LTP is, in essence, a differential amplifier. If it were a steady state DC amplifier only, then your statement would be perfectly true. But it's AC so it's pushing and pulling each ½ cycle. With an active current source that becomes dynamic in it's action, it wants to regulate current flow through each tube (or section) to a set value for each ½ cycle. This is why even value plate load resistors can be used for resonably (but not perfectly) balanced tubes (or sections) without further adjustments. Of course serious imbalance would be another story. Again, designers choice.

Victor

 

[KJ2005]

Hello Johan

My own choice for a top quality circuit has always been the Williamson topology, despite the two time constants (as said, the "direct" coupled LTP has it too). In a high quality 100W design using the partial cathode feedback output topology (a la Quad II), I have all of four cap time constants while still getting l.f. stability. But as Richw warned, I would also advise the availability of at least an oscilloscope and signal generator (with square wave) to keep away the blues.

Best wishes on your endeavours, Kimjul. Good advice here especially from Richw.

Thanks for your notes, and sorry for answering late.

I try to make my own summary out of all the advices stated. It seems as if your suggested method by use of a good resistor and some adjustment in the anode/plate circuit ( "the old fasion way") , still counts together with the Williamson topology. This takes me to repete the question: "would you prefer the triode or the penthode as driver for a pair of 6550a" ?
I would like to make two different amplifiers on a woodplate for test before drilling. One circuit close to TVA-1 , and another like the STA100. First of all because I have a lot of 12AX7,12AU7,12AT7,E80CC, L63(octal), EF184,6DJ8,EF86, 6J5 and 6SL7GT in stock, and I that I have to buy 6SN7GT if this is the best tube for the driverjob. I may start without any psu regulation as well as no CCS( now I know the short form) in the test configuration - also because I don't know how the 6K6 Sowter o/p will operate together with the two designs. My hope is to achieve or come close to the very "musical" sounding D76 amplifier - even if this goal is difficult to reach. I don't care too much about distortion in the first place, to me the sound is the more important issue. In work progress a lot of attention shall be dedicated to distortion.

Johan - I thank you very much for your help in this matter.

Best rgds

Kim

 

[KJ2005]

Re: 550V too high for modern 6550s

I'd not bank on being able to find tubes that can go that high..
I've heard of people telling me that certain KT88's can only
safely go to 525V.

475V is probably the safe upper limit (design center).

-- Jim

Hello Jim,

Thanks, I have two +B possibilities approx. 450V and 560V. I have to follow the advice erlier of max. 450-460V for the 6550's - the safe area for this tube.

Thanks for reminding me.

Best rgds

Kim

 

[KJ2005]

Hi again Victor

I used to think that way too. So please correct me if I'm wrong here. A LTP is, in essence, a differential amplifier. If it were a steady state DC amplifier only, then your statement would be perfectly true. But it's AC so it's pushing and pulling each ½ cycle. With an active current source that becomes dynamic in it's action, it wants to regulate current flow through each tube (or section) to a set value for each ½ cycle. This is why even value plate load resistors can be used for resonably (but not perfectly) balanced tubes (or sections) without further adjustments. Of course serious imbalance would be another story. Again, designers choice.

Victor

Yes - this was what I , without any calculation, have been wondering about.

I would like to try an active current source for the driver to deliver the exact same amount of current to both halfparts of a double tube or two single tube drivers. The design will be quite a work for me - it may be done by use of a single tube or ( don't mention the war) a transistor( we are not very pleased with the last word).

Thanks Victor - a little piece accumulated every day makes the amplifier in the end.

Best rgds.

Kim

 

[KJ2005]

CCS???

Hi could it be done this way?

rgds

 

[richwalters]

Kim,. O.M.O; my gut instinct is to go backwards and take the presets and pots out of the circuit.. This is one of the itchy points in the design field to avoid, in otherwords design for plug in swap repeatability without the need for circuit tweaking. (Keep it simple.)
The beauty with the concertina is "swap tube" performance consistency. However, on the bench I'm tempted to relash up the 2x EF184 LTP as per Radford using a small power pentode current source to tidy up the balance. I'm quite sure this would avoid the need for an expensive balancing pot in the anode load. Those users who have simulation programs with this tube in the database are in a good position to exploit the very best performance that this LTP can offer. Worth a whirl.

richj

 

[KJ2005]

One or two resistors?

. However, on the bench I'm tempted to relash up the 2x EF184 LTP as per Radford using a small power pentode current source to tidy up the balance.

richj

Hi Rich,

If there is a difference in current draw between the two EF184, where would you place the penthode in the STA100? - seems to me to be difficult, if we keep the original design. Or did I understand you wrong? How can we regulate balance if we exchange the gnd. resistor R14 1K5 with a penthode ? Don't we need a cathode resistor for the first EF184 in order to messure current? We allready have R15 180R for the second EF184.

Rgds
Kim

 

[KJ2005]

One or two resistors?

Or do you make the balance by pot P2 - 2K?

rgds Kim

 

[richwalters]

Re: One or two resistors?

Or do you make the balance by pot P2 - 2K?

rgds Kim

Have a look at this.
www.freewebs.com/valvewizard/ccs.html

read explanation regarding Current Source balance. This should put one in the picture.

Since the STA100 LTP has a low value tail resistor (1K5) this implies quite a low tail voltage circa (24V) and is asking for gain/ drift trouble. (O.M.O the original behaves poorly). The way forward is to use a pentode or triode as CS with a quiet -100V neg rail. Power amp users using fixed bias should be able to steal sufficient neg volts.

richj

 

[KJ2005]

Re: Re: One or two resistors?

read explanation regarding Current Source balance. This should put one in the picture.

richj

Thanks Rich,

Yes - This is what it's all about. I did'nt knew that a LPT with a tale of (CCS) , and that a penthode like EF86 would make that perfect balance between the two triodes in a double tube - as example (12AU7).
I have allready made a print out.

If this is the case, I understand why you wrote something like: " Still wondering why this technique wasn’t adopted on the day, some 45+yrs ago."
Would you prefer something similar for inputstage?

The cost will be at least two EF86 more for the driverstage in a stereo amp - The cost must have been the reason, during the years, for not having a CCS implemented.

One more thing for the test bench.

Best rgds
Kim

 

[richwalters]

I lashed up an LTP using EF184 with a ECL82 pentode as current sink strapped to a minus supply to improve the CCS voltage range for the pentode, only to find the ciruit went into oscillation. In some ways this was expected. As the EF184 is effectively a high slew rate tube, the socket wiring becomes ultra critical.
Morgan jones valve amps 3rd ed page 168 does mention the risk of instability with cathode followers, which this is. So there may be some merit in looking at a tamer med mu circuit, the previous on this thread using 6SN7 (or sim) in the simplest diff pair.

One could try using a transistor as a CCS; in an attempt in a more compact arrangement and avoiding a neg voltage.
However I have an old habit of not using Solid State in tube amp signal sources, just O.M.O.

richj

 

[KJ2005]

May be 6DJ8/ECC88?

Morgan jones valve amps 3rd ed page 168 does mention the risk of instability with cathode followers, which this is. So there may be some merit in looking at a tamer med mu circuit, the previous on this thread using 6SN7 (or sim) in the simplest diff pair.

One could try using a transistor as a CCS; in an attempt in a more compact arrangement and avoiding a neg voltage.
However I have an old habit of not using Solid State in tube amp signal sources, just O.M.O.

richj

Hi Rich,

At first I try to adapt one 6DJ8 as CCS , just to find out how sensible the circuit is to different frequences and stability - on a woodboard it is quite easy to change the circuit for EF86.

I try to keep the tube amplifiers without Solid State as well, except for psu fet-regulation. I don't think it is possible to hear "fet's" when they regulate voltage for gain - and drivertubes.

best rgds.
Kim

 

[Johan Potgieter]

I used to think that way too. So please correct me if I'm wrong here. A LTP is, in essence, a differential amplifier. If it were a steady state DC amplifier only, then your statement would be perfectly true. But it's AC so it's pushing and pulling each ½ cycle.

My remark regarding the CCS keeping only the total anode current constant (and your above statement), is as true for a cathode resistor (a CCS is only a high value resistor). I was comparing the 'advantage' of CCS over a resistor only.

Not to belabour, but it might be informative to examine what advantage a CCS really gives in practice. The cathode input impedance for a grounded grid circuit is (rp+RL)/(mu+1), and one can halve that for the impedance at the two cathodes of an LTP without much error.

One then finds that for a 12AT7 the value is about 800 ohm, and for a 12AX7 about 900 ohm working under popular conditions. That means that using a cathode resistor of 33K for the 12AT7 the output unbalance with equal anode resistors is only 2,4%. For 12AX7 with 68K in the cathode circuit, unbalance will only be 1,3%. (How many of us still use 5% resistors?).

Now a CCS could give 0% unbalance - if the two triodes as well as the anode loads were absolutely equal. (I rather doubt that too many twin triodes will have equality to within <2%, including aging. After that, what about the matching of the output tubes? Are they < 2%?)

Regarding a variable resistor in the anode loads, that will serve to make the outputs exactly equal more than to improve matters with a cathode resistor over a CCS, as the above figures showed. If that is an objection, it is a simple matter to add a small resistor in series with the "bottom" load resistor to get the exactitude - again pointing to what % your resistors are in the first place!

Thus my personal feeling: For the extra cost/complexity, especially with a vacuum tube as tail; why can I not feel that it is worth the trouble for this particular application? What am I missing?