Author Topic: Armstrong A220 design questions  (Read 239 times)

Offline El Sid

  • Trade Count: (+7)
  • AVForums Veteran
  • ***
  • Posts: 589
Armstrong A220 design questions
« on: September 12, 2017, 06:01:52 PM »
Prompted by afroaudio's awesome Armstrong restored post, I went and retrieved the Armstrong A220 schematic (see below). After having had a look, I have a question for the boffins out there.

So there is global negative feedback via R12 (75K) and C6 (47pF) from the 16 Ω tap of the OPT to the cathode of the first stage, V6a. Seems pretty standard. But there also seems to be local negative feedback from the outputs of the two phase splitter triodes (valves V7 a & b) through the 1M resistors R15 and R14 (there are oddly two resistors marked R15 - I'm talking about the leftmost one). What is the purpose of this? Is it to control the gain of that stage? Given that they are very high resistors, are they doing something other than nfb?

Please excuse the dof question.

I found this info on http://www.audiomisc.co.uk/Armstrong/200/200page3.html

« Last Edit: September 12, 2017, 06:03:54 PM by El Sid »

Offline Ampdog

  • Trade Count: (+1)
  • AVForums Veteran
  • *****
  • Posts: 6,743
Re: Armstrong A220 design questions
« Reply #1 on: September 12, 2017, 08:06:12 PM »
Concurring with compliments uttered in the previous post.

This post: What dof questions, El Sid?  (You are confused by other posts - oops!)

The design uses the classic paraphase phase inverter (an unfairly criticised topology in some quarters, I might add).  But firstly, may we rename the "leftmost R15' rather R13? -  I do not find an R13 and from the print that "5" might well be a "3". I will thus refer to it as 'R13' further on.

The classic paraphase topology consists of R14, R15 around V7b. (For those not familiar, R14 and R15 are normally equal or near so in value. This creates anode-grid unity feedback. It thus follows V7a providing a balanced inverted signal via V7b to the power stage, therefore the name 'paraphase'.) The vaues are quite high so as not to have an effect on the load resistors of R9, R10. Remember they are feeding a valve grid as input which has very high impedance if relevant at all (one often sees 5-meg resistors there). The only limitation really is that imposed by bypassing stray capacitance and the anode-grid capacitance. (It is a very good topology - provided the designer knows his job!)

Hope this helps.

When we come to R13, that is an extra voltage feedback over V7a.
Audio must be the only branch of engineering where lack of basics' knowledge is considered a superior form of wisdom. (Anon)

Offline Ampdog

  • Trade Count: (+1)
  • AVForums Veteran
  • *****
  • Posts: 6,743
Re: Armstrong A220 design questions
« Reply #2 on: September 12, 2017, 08:25:02 PM »
Follow:

Too late to add; just noticed that I did not complete commentary regarding R13! (Phone calls in between.)

R13 is not part of the described phase inverter. It is simple voltage feedback to V7a input. The amount is not simply calculated as one has to take the effective V6a output impedance into account (too lazy right now to look up component values!).

The purpose is to somewhat lower the V7(a+b) impedance feeding the power stage. (Unfortunately the popular ECC83 topology here typically requires highish values of resistors - 100K as anode load resistors not being uncommon.) The designer (Mr. Armstrong?) probably thought it wise to lower the phase inverter output impedance somewhat, fine then.
Audio must be the only branch of engineering where lack of basics' knowledge is considered a superior form of wisdom. (Anon)

Offline El Sid

  • Trade Count: (+7)
  • AVForums Veteran
  • ***
  • Posts: 589
Re: Armstrong A220 design questions
« Reply #3 on: September 13, 2017, 12:03:02 AM »
Thanks so much Ampdog! I am happy to get a response from the boffins' boffin.  :notworthy:

I did not recognise it as a paraphase phase splitter - I was only familiar with the simpler paraphase type where the output of the first valve goes through a voltage divider to act as the input of the paraphase valve. So I went and did my homework, so now understand what many call the floating paraphase.

The output impedance-reducing feedback via R13 (you are correct about its number) I would never have worked out by myself! You are right about the high anode load resistors - R9 & 10 are actually 220K, even higher than you suspected, probably making it even more of an issue and hence requiring the feedback to address the output impedance?

Of the types of phase splitter - cathodyne, paraphase and long tailed pair (that I know of), which do you prefer in your designs and why?

Offline handsome

  • Trade Count: (+1)
  • AVForums Veteran
  • ***
  • Posts: 825
Re: Armstrong A220 design questions
« Reply #4 on: September 13, 2017, 05:20:44 AM »
Is the feedback via R13 not an attempt perhaps to equalise the impedance of the two outputs' of the paraphase inverter? Without R13 V7a essentially behaves like a normal common cathode stage but V7b has a lot of feedback around it - its gain is reduced to unity - and thus a much lower output impedance.

Offline Ampdog

  • Trade Count: (+1)
  • AVForums Veteran
  • *****
  • Posts: 6,743
Re: Armstrong A220 design questions
« Reply #5 on: September 13, 2017, 03:58:24 PM »
As Handsome said.

But I shoved my laziness and redrew the schematic with component values marked. I found the output impedance of V6a (feeding V7a) to be about 10K - 15K. Calculating the feedback factor with R13 (2M2), gives 160 -220, which is substabtially too low to have any affect. That will not appreciably change the output impedances of the phase invertor; nowhere near to also making the output impedance of V7a equal to that of V7b as Handsome thought it might be.

Not to come with unsolicited critique, but I find the phase correcting network values C1 (2.2NF) - R6 (1K)  in parallel with R4 (220K) odd. Perhaps you can check in the actual amplifier for the real values? The C1-R6 step function is there to attenuate h.f. for stability with NFB - in this case if I recall correctly a rather hefty 29dB. But design should be such that this network should not need to attenuate down to inside the audio band, so as to allow full NFB over the whole of the audio region.

Calculating, one finds that the commencement of said attenuation starts at a whopping low frequency of 330 Hz, and ceases (by virtue of R6), at about 1KHz. (I wish someone will check my arithmetic! Handsome?) This renders global NFB just about useless over most of the audible band! (A C1-value of 150pF plus R6 of some 10K would be typical - but then one should of course check for NFB stability.)

Sorry for that. I sincerely  wish that there is an error somewhere. The rest of the design deserves better.

Late note:
Since I wrote about NFB, it is worthwhile to also bring the feedback resistor (R12 = 75K) and C6 (47pF) into the discussion. They form a phase-lead NFB compensation. The start of compensation by these is at 45 KHz - a very normal value for NFB amplifiers.

So again I do hope that one can solve the C1-R6 riddle!
« Last Edit: September 13, 2017, 04:08:03 PM by Ampdog »
Audio must be the only branch of engineering where lack of basics' knowledge is considered a superior form of wisdom. (Anon)

Offline Ampdog

  • Trade Count: (+1)
  • AVForums Veteran
  • *****
  • Posts: 6,743
Re: Armstrong A220 design questions
« Reply #6 on: September 13, 2017, 04:38:13 PM »
Of the types of phase splitter - cathodyne, paraphase and long tailed pair (that I know of), which do you prefer in your designs and why?

OUCH!

All of those - but each in optimal application. As the saying goes: "Horses for courses."

Cathodyne:
Simple and especially handy with high sensitivity power valves like EL84. Advantages: simplicity and guaranteed balance.

(Note; Beware of comments that state that the output impedances are unequal; top looks like an ordinary anode follower while the 'bottom' output resembles a cathode follower. Not so. Since the same current flows through both loads, the output impedances will be the same as long as the load resistors plus whatever loads them from the following stage remains equal in value and phase. Kirchoff's Laws force this.)

Disadvantages:
No gain. The previous stage has to provide full output swing. Thus, not to step on any nerves, but used straight into less sensitive power stages e.g. EL34 etc. often drives the input circuitry to the limit. A prime correct application is in the classic Williamson circuit.

Paraphase:
As said before. In circuits where gain is required and h.t. is not into the many hundreds of volts the topology giving the largest possible output swing. But I have not used it extensively. Here the channel output impedances are unequal, but that should not impose a problem within the audio range - easily corrected.

Long-tailed pair (or Schmitt):
Can render a moderate output swing and can be direct-coupled to the previous (input) stage. Popular circuit; earliest use was in e.g. Leak amplifiers, later in many others. Probably used most often by me. It has moderate inverse feedback and is highly self-balancing. Can just about easily drive output valves like EL34, KT88 and similar. (It is often used with a constant current IC 'tail' source to make extra output amplitude available. I have not tried that though - perhaps too old school to design hybrid amplifiers!)
Audio must be the only branch of engineering where lack of basics' knowledge is considered a superior form of wisdom. (Anon)

Offline El Sid

  • Trade Count: (+7)
  • AVForums Veteran
  • ***
  • Posts: 589
Re: Armstrong A220 design questions
« Reply #7 on: September 13, 2017, 05:19:54 PM »
OUCH!

All of those - but each in optimal application. As the saying goes: "Horses for courses."

Cathodyne:
Simple and especially handy with high sensitivity power valves like EL84. Advantages: simplicity and guaranteed balance.

(Note; Beware of comments that state that the output impedances are unequal; top looks like an ordinary anode follower while the 'bottom' output resembles a cathode follower. Not so. Since the same current flows through both loads, the output impedances will be the same as long as the load resistors plus whatever loads them from the following stage remains equal in value and phase. Kirchoff's Laws force this.)

Disadvantages:
No gain. The previous stage has to provide full output swing. Thus, not to step on any nerves, but used straight into less sensitive power stages e.g. EL34 etc. often drives the input circuitry to the limit. A prime correct application is in the classic Williamson circuit.

Paraphase:
As said before. In circuits where gain is required and h.t. is not into the many hundreds of volts the topology giving the largest possible output swing. But I have not used it extensively. Here the channel output impedances are unequal, but that should not impose a problem within the audio range - easily corrected.

Long-tailed pair (or Schmitt):
Can render a moderate output swing and can be direct-coupled to the previous (input) stage. Popular circuit; earliest use was in e.g. Leak amplifiers, later in many others. Probably used most often by me. It has moderate inverse feedback and is highly self-balancing. Can just about easily drive output valves like EL34, KT88 and similar. (It is often used with a constant current IC 'tail' source to make extra output amplitude available. I have not tried that though - perhaps too old school to design hybrid amplifiers!)

Thanks Ampdog! Great feedback (positive  ;D)

One of the things I want to do after my Marsamps kit is built (and I've listened to it for a while!) is play around with different phase splitter topologies - and doing exactly what you mentioned with a voltage regulator as current source in an ltp is one of them

Offline fredeb

  • Trade Count: (+13)
  • AVForums Veteran
  • ***
  • Posts: 902
  • Lost in Music
Re: Armstrong A220 design questions
« Reply #8 on: September 13, 2017, 08:29:05 PM »
Wow ! Thanks for starting this thread El Sid , and asking the right questions .

Highly informative , thank you Ampdog , I'm sure I can do with reading this a couple of times .  :thumbs:
“And those who were seen dancing were thought to be insane by those who could not hear the music.”
― Friedrich Nietzsche

Offline handsome

  • Trade Count: (+1)
  • AVForums Veteran
  • ***
  • Posts: 825
Re: Armstrong A220 design questions
« Reply #9 on: September 13, 2017, 09:01:38 PM »
@elsid Ironically in guitar amplifiers "bad" phase splitters are the way to go, as an unbalanced splitter generates more 2nd harmonic distortion.
@ampdog the less mathematics for me the better.  you are quite right though the amount of feedback is as good as none...

Sent from my SM-A500F using Tapatalk


Offline Ampdog

  • Trade Count: (+1)
  • AVForums Veteran
  • *****
  • Posts: 6,743
Re: Armstrong A220 design questions
« Reply #10 on: September 15, 2017, 03:33:54 PM »
Calculating, one finds that the commencement of said attenuation starts at a whopping low frequency of 330 Hz, and ceases (by virtue of R6), at about 1KHz. (I wish someone will check my arithmetic! Handsome?) This renders global NFB just about useless over most of the audible band! (A C1-value of 150pF plus R6 of some 10K would be typical - but then one should of course check for NFB stability.)

One of those embarrassing moments where I should be honest and come clean. The figures are wrong -  not sure what arithmetic I did that time of night  :facepalm:

{For those interested, C1 acts in parallel with [R4 in parallel with the valve internal resistance rp = 65K average]. That gives about 50K, and the roll-over frequency with C1 (2200pF) will be about 1,5 KHz, not 320 Hz! Still rather low, but I will leave further design considertions be.}

My apology!
Audio must be the only branch of engineering where lack of basics' knowledge is considered a superior form of wisdom. (Anon)