Sunday, August 5, 2012

Radiotron SE2A3



Revisiting the classic pentode driver
Radiotron's answer to the WE91A
Right before the end of the 20th century, a friend brought over a Japanese built pair of WE91A clones sporting top of the line Tango X-3.5S output transformers, matching chokes and power transformers, input tubes were original WE310As and 50s date code WE 300Bs, WE274B rectifiers, Allen-Bradley carbon and NOS wire wound resistors, Vitamin Qs for coupling and can type paper in oil caps for the power supply. The amp was huge easily dwarfing my JEL SE300Bdx monoblocks and the quality of construction and wiring was impeccable. Although this WE91A clone had great qualities in the midrange my friend was convinced my amps sounded better overall. Does that mean my amp design is absolutely betterDefinitely not! Only in the context of my system - Denon DL103R/SME 3012/Garrard 301 feeding Tamura TKS83 MC transformers into a JEL deluxe preamp, JEL SE300Bdx mono blocks driving a pair of Altec 755As in open baffles - since I optimized the signal chain so that it sounds good according to my standards. I told my friend that his decision to purchase those amps should be based on how it sounds in his system not on what he heard in my listening room.

That amp made a lasting impression and reinvigorated my interest in the classic pentode driver topology which has a lot of esteemed followers and practitioners. After my friend left that night I pulled out Joe Roberts WE91A article in Sound Practices Issue No. 1, reviewed my notes on the 6SJ7-->300B I briefly prototyped - slight upper midrange glare through the 755As, the top end was a bit rolled off with a tubby bottom. I was thinking....could it be the pentode 6SJ7 being a later compact version that kills the magic? Remember that even in the 90s WE310As were already unobtaniumSo I thumb through the pages of the RC-12 tube manual in search for a close equivalent and found the 6SJ7's grandfather, the 6C6 which has similar electrical specifications and bears a physical resemblance (BIG plate and grid cap) to the WE310A. The 6V filament is a convenient bonus. By that time, I've maximized the potential of all my triode input/driver circuits for the 755A/OBs and Altec 2-way. I pursued other interests and kept the 6C6 project in the back burner...


!!!WARNING!!! 

The voltages found in this circuit can be lethal, build at your own risk!!!



At the beginning of the year (2007) I was looking for a winter project that will not entail chassis building. I reread Joe Roberts article "I never met a 2A3 amp that I didn't like" SP15 and noticed a familiar looking SE2A3 amp schematic using a 6C6 pentode driver from the Radiotron, 3rd edition. He had good things to say about this circuit and it looked easy to graft this topology on my long-time SE2A3 test mule with Tango U808s. I took inventory of my tube stash and was frustrated to find just a single RCA 6C6, a couple of 6-pin sockets but needed 1/4" medium sized grid caps.

One chilly morning I downloaded an email from an internet buddy Sergey Didkovsky from Ukraine. We have exchanged emails a few years back when he was building the JEL classic 300B circuit. He was telling me about a pentode driven SE amp that he thinks sounds better. This was a fortuitous moment because the clairvoyant Sergey was talking about a Radiotron derivative. In the course of our correspondence he pointed out a quick source for additional 6C6s and grid caps with several links and attachments of Japanese and European DIYers' single pentode driver stage schematics including tips on operating points and parts tweaks.

I started my listening evaluation with this circuit and sent my impressions to Sergey for comments:
"Try please, to a little increase the cathode resistor (up to 1,2-1,5Kohm) and reduce anode resistor (to 75k-82k). It will allow to increase Ua up to 150V and to receive ratio Ua:Ug2 = 2:1 ratio. The sound should become more "interesting and old". That is especially good for jazz. Still mine 5 cents....use good NOS Oil capacitors in G2 circuit. I use TCC VisconolI think, oil caps in the grid - adds a bass. Reduction of the anode resistor - improves the hi range a little."
His use of adjectives "interesting and old" prompted me to get as close to the suggested 2:1 plate to screen grid suppressor voltage ratio as it seems to be a key ingredient for his recipe. So far I have achieved a 1.8:1 ratio and was still hearing a slight upper midrange glare that I originally attributed to the 6SJ7 years ago. The BIG plate = FAT midrange I was expecting from the 6C6 was already there but the glare did not disappear until I replaced the plastic 2uf cap with can type 2.8uf paper in oil caps for the screen grid (suppressor) and as Sergey said it did improve the bass. Maybe the paper in oil decoupling cap will give the same results with the 6SJ7....
"Thanks for your patience, not looking at my very bad English language...."
There is a saying that music is a universal language, Sergey apologized about his English but I assured him that his communication skills were fine. Certainly driven by his passion for good sound, I had no problems understanding his tips and it was quite evident that he has done some serious research, listening and is blessed with good ears!

With these thoughts in mind I did not argue and simply looked up to him as a mentor. In this revised circuit I captured the 2:1 ratio, to be exact 1.98:1 with 152V on the plate drawing about 1.8ma. of current. I changed the grid resistor from a 270k to 330k because not only does this make an easier load for the driver stage but from experience the sound becomes warmer as the grid resistor value is increased but only to a certain extent. If you exceed the maximum 500k grid resistor value specified in the tube manual for a 2A3 you might find the plates running away cherry red! So check your tube manual before messing around with this tweak on other triodes. A 220k will give more focus and detail but lower than that value the driver stage will have limited headroom. It will clip prematurely from not having enough voltage swing to drive the 2A3 to a full 3.5W output since the grid resistor is effectively in parallel with the plate load resistor.
Radiotron SE2A3 Power Supply

I am grateful to Sergey for sharing his experience saving me time and effort to build a classic pentode driven SE2A3 amp!




100Hz - both channels

1kHz - both channels

10kHz - both channels

Radiotron derived SE2A3 - 6C6 pentode driver, Tango U808
Power output = 3.8W per channel before clipping
Frequency response = +/- 1dB from 30hz - 15khz @ 1W
( - 3dB @ 20hz & - 5dB @ 20khz)

100Hz - top trace  = signal generator

1kHz - top trace = signal generator

10kHz = top trace = signal generator

JEL SE2A3dx - 76/6SN7 input/driver, Tango U808
Power output = 4.5W per channel before clipping
Frequency response = +/- .5dB from 20hz-20khz @ 1W
( - 3dB @ 50khz)

100Hz - both channels

1kHz - both channels

10kHz - both channels

JEL Simple 45 - 1/2 6SL7 [or 6SF5], Tamura F475
Power output = 2W per channel before clipping
Frequency response = +/- 1dB 20hz-20khz @ 1W
(- 5dB @ 30khz)
Compared to my favorite topologies in driving power triodes, this R-C coupled 6C6 pentode has slightly narrower bandwidth and as shown in the square waves, it has the slowest rise time and softest leading edge at 1khz and 10khz (in theory a more vertical rise time and sharper leading edge = faster transients and better high frequency extension) and pretty much equal at 100hz (less tilt = better bass response) with the Simple 45 having a slight edge but that uses a different OPT. To put things in perspective a 30hz-15khz +/- 1dB bandwidth is not a slouch if you refer to the table below.

I believe that good engineering aided by good ears is an important factor in building a musical sounding amplifier and cannot forget a post I read several years ago at a forum "...nice people build good sounding amps, bad sounding amps are designed by annoying ---..." and so it goes...kudos to the RCA engineer(s) in the 1930s who designed the Radiotron SE2A3, it is a musical amp!



32 comments:

  1. Is the 2.8 pio value critical? I'm building this amp but can't find a source for 2.8 pio cap.

    ReplyDelete
  2. IMO, 2.8uf is the minimum value. If I had more space, I would've used 5uf or greater.Sorry for the delay in reply, I've been busy.

    Happy Listening!

    JE

    ReplyDelete
  3. Hello, I like the design of the amp you always do a good job on the write up too. I have a question if you had a 2.5 filament transformer with a center tap could you drop the 50 ohm hum pot and just hook the 100uf and 880 ohm reistor straight from ground to the transformer center tap and then have the other leads to the filament pins of the 2a3? would there be a large diffrence would it affect anything? or would just no longer be a hum adjust? without a hum adjust is there a large ammount of sonic impact? What if i were to run it as a headphone amplifier (is that even possible?) Im sorry for all these questions.

    ReplyDelete
    Replies
    1. If there is negligible hum then you can dispense with the hum bucking pot. The pot is there to null the hum if the transformer center tap is not absolutely centered.

      Good luck!

      JE

      Delete
  4. I have just built this amp but with some testing it seems that the gain/output power is kind of low to power 91db efficient speakers to a normal listening volume i must turn the volume knob to maximum from line input. I have tested voltages and all seems to be +-6 volts within your measurements. I have swap tubes with others and that did not seem to solve the issue. I am however using 6sj7 in place of the 6c6, also no hum-buck pot im going straight from the center tap of the filament transformers. If you could help much would be appreciated.

    ReplyDelete
    Replies
    1. The 6C6>2A3 or equivalent pentode input/driver circuit is what I would consider a "highish" sensitivity design requiring less than 1 volt from a typical line level source e.g. CD player, to drive it to full output. However without having your amp in front of me, my suggestion will only be mere speculation.

      Since you are using an octal based 6SJ7, make sure you converted the pin outs of the 6SJ7 to match the 6C6 in my schematic so that you are actually measuring the proper voltages from the proper electrodes.

      Hope this helps!

      Delete
  5. For JEL SE2A3dx - 76/6SN7 input/driver which was the input sensitivity for 4.5w before clipping?

    ReplyDelete
    Replies
    1. About .4V to maximum power output...

      Delete
    2. For my thoughts on distortion meters, please read paragraph three of this link - http://jelabsarch.blogspot.com/2012/07/hammond-125ese-part-2.html

      Delete
  6. Hello!
    Between 2A3 Radiotron and 2A3/76/6SN7 which sound better? Which it is the sensitivity for Radiotron at 3.8W? Thank you!

    ReplyDelete
    Replies
    1. Hi Claudiu,

      If my memory serves me right the sensitivity of the 6C6>2A3 is about .7V to full output.

      Everything I wrote at the beginning of this post re: 6C6>300B vs. 76>6SN7>300B are applicable to the 2A3 version(s).

      Happy listening!

      JE

      Delete
  7. Hi! A question maybe just silly, but... Why cathode resistor is 25W? If the current is 64mA shouldn't be 5W enough?

    I have with my a pair of 6B4G; is a good tube for this circuit?

    Best,

    Martin

    ReplyDelete
    Replies
    1. I've never tried the 6B4G, but I'm sure it is just as good since it is essentially a 2A3 with octal base and 6V filaments. I used a 25W because I had a good stock of those then. A 5W will work in that position but I'd be more comfortable with at least a 10W cathode resistor.

      JE

      Delete
  8. Hello JE. I'm hoping you are still checking this thread. I am wondering if you could explain a little about the Ua:Ug ratio(I don't really know what that means or how to calculate it) and the changes you made to the circuit to accomplish that? Are the changes just to the 6C6 tube circuit or to the 2A3 circuit as well? Thanks in advance.
    Dave
    WntrMute2@aol.com

    ReplyDelete
    Replies
    1. It's simply the screen voltage (Ug = 76V) being approximately half the plate voltage (Ua = 135V) on the 6C6 as shown in the revised schematic. Note the resistor and cap values are different from the original Radiotron.

      JE

      Delete
    2. Thanks very much. That's very helpful. What about your change from the 270K grid resistor to 330K? What significance does that play in the overall performance. I have a 180K resistor in that spot currently.
      Great blog!

      Delete
    3. I found the ideal compromise between definition and warmth in the 330k value. Lower value usually gives more detail and cooler sound. This is where you can season to taste within tube manual limits, IIRC, don't go over .5meg.

      Delete
  9. Sweet. Thanks once again. I may try a bit higher value.

    ReplyDelete
  10. Hello JE,

    Im ready to build the 6c6 2a3 se amp. Will a 5k se Opt will be good as I don’t have the 3.5k se opt ... thanks

    ReplyDelete
    Replies
    1. I've tried the 5k configuration on the U808 opts in this amp and the sound was a bit too lean while the textbook 2.5k was a bit too warm. The best tonal balance (to my ears) between definition and warmth was at 3.5K.

      Hope that helps!

      JE

      Delete
  11. Hi JE, noob question incoming. Why .22uf for the coupling capacitor 6C6>2A3 when the original Radiotron diagram calls for 0.05uf?

    ReplyDelete
    Replies
    1. Simple answer is, I prefer the sound using a 0.22uf. ;)

      Even if the original circuit topology is retained, the operating points have been "seasoned" to my taste with the help of Sergey. Thus, the resistor and cap values have been altered.

      Delete
  12. Hi, a 1nf & 1nf capacitors in series with an 18K R from plate of 2A3 to the control grid if the 6C6, pin 3 ( positive feedback ) will extend the freq response to approx 20Hz-30KHz -1dB

    Not from the secondary + speaker terminal to the cathode of the 6C6 !!

    ReplyDelete
    Replies
    1. Although this mod might improve the metrics of the amp, I'm skeptical about the sonic benefits of applying positive feedback. During my Dyna PAS hacking days, the first thing I did to improve the phono section was to remove the positive feedback resistor. To my ears, PFB is more detrimental to the sound than conventional global negative feedback. I recently cloned the Shure M65 phono preamp and also deleted the PFB resistor.

      IMHO, this amp is a classic benchmark design from the 1930s! If wider bandwidth is required, a new design would be my preferred approach.

      Delete
  13. Hi ,I note you have used the Tango U-808's, I have just ordered a pair of HO-808 MK2 from HATT Audio Japan, but the post is surface mail due to Covod problems . HATT Audio set up production in Japan with a range of drop in replacements for the TANGO range which stopped production . The New HO-808 M2 is 10Hz-80 KHz -2dB so an improvement. I understand they reverse engineered many of the old Tango,s
    http://www.single-ended.com/2A3/archive/hatt-audio.pdf

    ReplyDelete
    Replies
    1. I've read about the HO-808 in MJ Audio Tech + Stereo Sound Tube Kingdom but have no ear/hands-on experience. Let us know how they perform in your amp project.

      Thanks and happy listening!

      JE

      Delete
    2. How was your experience with the HATT Audio output transformers? I am considering ordering a pair to build this project also. Cheer, Jeremy

      Delete
  14. Isn’t 2A3 dissipation a little hot with these values? I’m calculating ~15.8 watts (320-52)*(52/880) where data sheet max is stating 15w. Am I missing something?

    ReplyDelete
    Replies
    1. I surmise that the 15W tube manual maximum rating was for the early and now hard to find ($$$s) single/monoplate 2A3 which are a lot more delicate. When I built around this tube in the late 90s for a friend, I set them at 12W plate dissipation.

      However, I've operated bi-plate 2A3s from 15-18W plate dissipation since the mid-90s with no adverse effects. The bi-plates are rugged!

      Delete
    2. Thanks. I have built the simple 45/2a3 and really enjoy it. I want to built this next, but wanted to do something different and try monoblocks. Can I just omit one channel and expect voltages to be ok? I assuming I may have to trim B+ a little at a minimum with less current demand, but unsure if the rest will just work out.

      Delete
    3. The voltages in the schematic is based on the power transformer I used. So the B+ will vary depending on the ratings of the power transformer(s) you'll be using (stereo or mono).

      Delete