May 3, 2016

6J5 Line preamplifier Part 2

High voltage requirements

Needless to say that a good supply, free of ripple, is essential in SE stage that have very bad CMRR. This generally dictate a large capacitor bank at the expense of transient response due to great time recovery.
As I prefer a punchy response it needs a good calculation of the filtering cells to keep ripple at least 10-5 (- 80 dB) below last stage voltage. Means 240*10-5 or 2,4 mV rms while keeping the smallest possible time constant.
Using a input choke of large value greatly helps reducing the following capacitors value. Furthermore behind the choke the ripple is very smooth and quite sinusoidal, generally 2 cells are enough to have an almost perfect continuous voltage.

Main supply




To calculate the ripple behind the choke we have to know the ripple factor   η = 1,2/L*C
 (L in Henries and C in µF) where C is the first filtering cap.
An estimation of this cap can be done if we consider a time constant not more than 15 millisecond. The choke having a DC resistance of 800 ohm, C will be 15 10-3/800 = 18,7 µF, the closest usual value is 16µF.
Then η = 0,00125 and the ripple on the first cap will be ηV = 0,00125*265 = 0,33Vpp or 0,117 Vrms. Network time constant is 13 millisecond.
A second cell with 1,2 Kohm and 16µF will leave a small 25mVpp / 8.8mVrms ripple with a 19 millisecond TC.
A last cell with 2,4 Kohm and 10 µF will leave 1,6 mVpp / 0,56mVrms and a 24 millisecond TC.
These 0,56mV AC represents a -113dB attenuation below the 240V feeding the 6J5, way better than the -80dB expected, while keeping a very fast recovery power supply. I did some tests with pulse trains and supply is rock solid.

* Time constant is a very important factor in audio, it determines the capacity of an RC system to respond to an AC signal modification and is essential in the ability to follow the sound envelope or ADSR. It is too often neglected when calculating a power supply.
In the case of AC circuits, the RC time constant tells you which signals the circuit passes through and which signals the circuit filters out. An RC circuit acts as a high pass filter which passes high frequency signals and blocks low frequency signals.
When we apply an AC current to a capacitor we are basically sticking some charge onto the capacitor and then taking it off. If we do this at a low rate, smaller than the time constant, then the capacitor has time to discharge. But if we put the charge there rapidly, at a speed much greater than the time constant, the capacitor doesn't have time to discharge. In this way, we can pass high frequency signals through the capacitor.
Remember :
-1 A given capacitor charges 99,9 % at 5 RC.
-2 As a rule of thumb, TC should increase from the first to the last decoupling cell.


No more technical considerations...

Construction

Like for the line preamp just a few parts but of high quality.
An RCore transformer for low overall magnetic loss, high efficiency (just bettered by the toroidal) and slim design to fit my chassis.

A rectifier, good capacitors (German F&T dual 16µf and a couple of Siemens 8µF MP/JS + 2µF Sprague PIO to make my 10µF last decoupling cap), a few resistors, a filtering choke and quality sockets.
Add some hardware and plenty of time....and you are done.


In color with Neotron 6J5G




For chassis I used 2U professional grade steel ones, 3 layers blue gray paint, very well suited for tubes with lots of openings for optimal air flow. They provide good shielding and are super rugged. I just add some screen printed front and rear plates for better appearance and way better WAF.




Listening report


The 6J5 and its relatives are too numerous for an extensive test. I will give my impress upon the most interesting ones I have in stock. Anyone wanting to build this simple BUT "great performer on any kind of music" preamp will easily find a wide range of readily available tubes.
All the tubes listed below are up to the task, some are more involving or more waaahou !! on a specific material, but a very few simply made me listen to the music instead of searching the preamp qualities and / or shortcomings.
Most need a few burning hours to deliver there full color.

6J5V Radio Technique
Best of all. Just say evident whatever the music. Delightful triode, vivid, natural with lots of air, detailed to the extreme but never tiring. This is the 6J5 "just sit and listen". These PTT tubes were intended for very long life service, each tube came with a numbered biographical sheet to keep records for 15 years!

6J5G Neotron
Would be my second best tube. A bit more brilliant with slight grain on voices, still stays very detailed and involving. Due to incestuous relationships between tube makers Neotron's were possibely made by Visseaux (with slight construction differences). I have some tubes bearing both Visseaux and NT labels. So who made what ?

L63 / CV1067 GEC Straight Bulb
Nice sound, would say polite, lacks the details of the RT's or Neotron's, no grain but a smaller sound stage that make music more intimate.
By comparison the L63 GEC ST (see part 1) have a bigger and wider sound but are surprisingly very slow tubes that give the feeling of wrong tempo.

6J5WGT Raytheon
I did not like much this tube despite some remarkable ability on voices. Muddy sound with grain, tight and punchy bass but not as accurate as the other ones. Huge sound stage, lush presentation BUT if you like Billy, Ella and Louis or country singers with hoarse voices, this tube gonna make you cry.

Last of the batch 7193 National Union
I had this tube for decades and it is the kind of weird thing you don't want to exhibit. It is a mistake, and I regret not having foreseen the holes for the anode / grid wires. This tube is a killer by many aspects, it has the CV1067 GEC qualities with a much more authoritative presentation and a wide but realistic sound stage.

I was willing to test the 6P5G that bear an excellent reputation, unfortunately impossible to source a NOS pair at a decent price.

Like in every tube based device the rectifier plays an important role in final restitution.
Here, due to the very few parts used its impact is more than evident and the choice made will determine the preamp character. I tried all the ones that could fit, directly or indirectly heated. On the dozen I listened to, the very best for neutral but vivid presentation are Mullard GZ30/CV2748, Mazda GZ32 and STC/ Brimar 5Z4G, with a slight margin for the GZ30 in terms of bass resolution and voices transparency.




14 comments:

  1. Hi!
    May I ask what is this beautiful volume changing knobs and where can they be found?
    Thanks a lot,
    Regards,
    Maxime

    ReplyDelete
    Replies
    1. Hi Maxime, it's part of the front plate and custom made.
      Best regards.
      Fred

      Delete
  2. Came across this site by a coincidence. What a fantastic build quality on everything you make!! May I ask who is the manufacturer of the chassis you use? I really like the industrial look of them.

    ReplyDelete
    Replies
    1. Hi Tommy, thanks for visiting my blog.
      Chassis are from a little French manufacturer that make me a copy of discontinued ones I bought a long time ago. They do make a really nice but slow job. In fact, they proceed to my order when they have spare time in their scheduled production. Sub chassis and front plates are CNC made in Germany.
      Regards
      Fred

      Delete
  3. Did you use AC for the heaters?
    Ad far as I can see, the umbilical connector has five pins so that would probably mean that you run B+ and AC for the heaters in the same umbilical. I am thinking about using the same arrangement, but I am afraid that I could run into problems with hum. Any thoughts about that?

    ReplyDelete
    Replies
    1. Hi again,
      I use DC for heaters, you can see two filtering caps next to the power transformer. AC is a problem with umbilical as you can pick up 100/120Hz unless you use shielding braid. Moreover you should use a hum potentiometer next to the 6J5 without being certain to cancel it for both tubes.
      Regards
      Fred

      Delete
  4. Hello, do you eventually have some 6J5V tubes for sale or trade ?
    I have some spare NOS/NIB 6P5G RCA Victor & 6J5G Visseaux if you wish.
    Best regards, Bruno

    ReplyDelete
    Replies
    1. Hi Bruno, thanks for your request, unfortunately I just have 2 left. I sold a spare to a friend of mine who built a copy of my preamp.
      Regards
      Fred

      Delete
  5. Fred, thanks for your detailed PS post. I've learned a lot, and re-vamped my tubed PS using your calculation methods. (I had too much capacitance and too many stages).

    Regarding the increasing time constants, what is the lowest you recommend for the 1st stage and the greatest at the 3rd (or 4th) stage of filtering? And acceptable increments per filtering stage are 2 to 5 msec?

    So a series like 14-->18-->22-->26msec or 10-->15-->20-->25msec?

    And finally, my main transformer has a 230V secondary and I use a 5.6µF cap before the first choke to raise the voltage a bit. How does this affect the time constants of the filters?

    Regards, Robert

    ReplyDelete
    Replies
    1. Hi Robert,
      As a common use, TC should increase 1.5 to 2.5 times after each filtering cell. Means if the first is 10 millisecond (recommended value, not more but not less or sound becomes dry and unpleasant) the second should be 15/25 millisecond and so on. I will explain a complete CLC PS calculation in a next post. It depends upon transformer primary and secondary resistances, voltage ratio, rectifier internal resistance and LC or RC values.
      Actually, the main concern for you is not TC but ripple. With a 230V secondary and a full wave rectifier, you will roughly get 270/275V HV. A 5.6µF capacitor, which is a small value, will give a ripple value equal to Itot / C x f. Assuming a 120HZ ripple frequency (twice the mains) Vr will be around 24V (-21dB) and you will need a strong filtering to get at least -80dB ripple, -100dB would be better.
      Hope this helps
      Fred

      Delete
    2. Hi Fred,

      Thanks again for your most helpful blog and insight. I look forward to reading more.

      My overall filter is a cLCLC inside the power supply chassis (first c1=5.6µF as I wrote earlier). I'd calculated a 45Vpp immediately after this capacitor (290V at start). By the end of the cLCLC, the residual ripple calculates to 54 µVrms, which is more than 130 dB down from 290V.

      Inside the signal chassis there is another filter stage to de-couple the channels: either an RC or an LC for each channel, depending on the preamp, leaving 3.56 µVrms or 0.0129 µVrms, respectively. I hear no speaker noise from either preamp. The RC is in an 6SN7-based preamp; the LC in an 801A-based preamp (ala Thomas Mayer with a cLCLC on the filaments per channel with each C=33,000µF and loads of Lundahl chokes).

      Prior to reading your blog, I'd just added lots of capacitance for each filter stage (1,000µF here and there; the more the better!). I then realized I was messing up the timing. I did your recommended TC progression with far less capacitance and the preamps are still completely silent and sound great. The TC progression is now something like 15 msec —> 18 msec —> 22 (or 34) msec, ignoring the first 5.6µF cap.

      So again, I wish to thank you for your instructions. I feel I am finally beginning to understand how to build a well-designed power supply. After all, preamps and amps essentially only modulate the power supply; so what we're truly listening to is the power supply. How clean and well-timed it is, determines how it all sounds.

      With best wishes, Robert

      Delete
  6. Dear Robert

    I am building a one stage C3g line amplifier and will start with a bridge rectifier before switching to 5U4G tube rectification later and was wondering if can you use the same calculation for bridge rectification? or does it not make any difference?
    Best regards
    Sigurd

    ReplyDelete
    Replies
    1. Hi Sigurd, TC and ripple calculation applies to any kind of rectifier, the only difference is the very low DC forward resistance of a silicon diode compared to a vacuum rectifier.
      Hope this helps. Fred

      Delete