Homebrew amp: 2a3 driven by single D3a stage

[mcheinstein]

Hi folks,

A couple months I stumbled across some cheap used RCA bi-plate 2a3's (matched pair at 4100 umhos), so a little weak), and decided it was time to dip my toe into DHT SET land. I am not an expert, and relied heavily on AK and other forums to help guide myself, and there was a good amount of stuff I learned that I think could be helpful for other novices building their first DHT amp. The purpose of this post is to help pass on some of the things I learned while doing this, as there were not too many prior examples. Of course, would also love any feedback out there and tweaks to consider from the experts.

Project goal
My goal with this project was to keep parts count down in the theory that the less pieces in the signal path, the greater the transparency (I know this is a very debatable point). I knew I wanted to either directly couple stages and/or have only 1 driver stage. In the end, I decided to do single stage, but de-couple driver and power stage for design executional simplicity, keep the voltages down, and cut down on power transformer needs/cost. Therefore, I embarked on creating a two stage 2a3 amp.

Building the single driver stage:
Single driver stage does cut down on the number of suitable drivers available. This makes a DHT driver somewhat out of the question. I considered using either 6SN7 or 6SJ7 drivers, but concluded I wanted more gain in the single drive stage (one very common design is having 1 6SN7 per channel, two stage driver). The ubiquitous 12ax7 wasn't considered for it's really low current capabilities, and I wasn't keen on the 12au7 for the same reason as the 6SN7. It seemed like using an EL84 pentode could also work (and I had a bunch on hand), but then I came into a pair of NOS D3a for 40 bucks, and my fate was sealed. D3a has fantastic linearity when wired in triode mode, high current drive, and high gain. Some had reported that the D3a could adequately drive 300Bs, and so it seemed like a decent bet for 2a3.

There aren't too many folks who have described their operating points using D3a as a driver for DHT. So, I took matters into my own hands and leaned heavily on https://www.trioda.com/tools/triode.html to draw load-lines. Now, where I made my first mistake trying to minimize stage distortion only using the calculator. This led me to set up my cathode resistor at 35 ohms, B+ at 295V, an anode resistor of 6.2k, with 27ma plate current (max for D3a ~33ma) and ~1V bias. When I first fired up the amp, it sounded okay, but the sound was thick and blurred. I scoured more forums, and found many people were using more like 100 ohms on their cathode, and plate currents ~10-15ma. I steadily increased the cathode resistor value (60, 80, 100, 120 ohm), and found 100 ohm to sounded the best. Then, I varied my plate resistor, and found increasing plate resistance also focused the sound and increased clarity and stereo separation. I settled on setting plate resistance at 10.4k and 100 ohm on the cathode, setting the plate current to ~14ma, bias at ~1.4V, and a plate dissipation of ~2W @ ~145-50V (~1/2 the max a D3a can take). I did do another round of testing different cathode resistor values, but preferred the 100 ohm resistor. I bypassed the cathode with an ultra low ESR e-cap. I have since rigged up a distortion analyzer (probably should have done this a lot earlier) and will be soon be making measurements of the stage distortion to see how distortion changed in real terms as operation point is changed (and probably make a few final adjustments). I'd look forward to someone schooling me on this.

For the coupling capacitors, I had a bunch of russian PIOs and two mundorf supremes hanging around, as well as a pair of miflex KPCU's that I had used in another amp previously. I tried all three, and the Miflex's won out in terms of sound. Mundorf really brightens things up, but becomes harsh after listening for a while. The russian PIOs sound pretty good, but the Miflex's give the sound that "in-the-room" feel while also retaining tons of detail. I think the russian ones lacked a little on the detail. Bang for the buck, the russian PIOs are by far the best. My opinion, YMMV.

Building the power stage
Designing this stage was pretty straight forward. Many folks have implemented 2a3 amps, and the RCA datasheet sets the standard for where to operate 2a3s at. I decided to use cathode bias instead of fixed bias to keep the power supply simple, and I didn't read any strong opinions one way or another on forums.

I kept in mind that my 2a3s were a little weaker than new, and that I would need to tweak things a bit. I also needed to tweak things for the output transformers I had. The RCA datasheet is spec'd for a 2.5k primary OPT, but I had a pair of 3k Electra-print OPTs sitting in the garage that I had from a buddy of mine. This means less power output, but also less distortion. Great. I first designed the stage very much in line with the typical 2a3 operating point: 250V on the plate, 750 ohm cathode resistor, and aimed for 60ma of current. With my weak 2a3, I was only pulling something like 30-40ma, which meant that my bias was also quite low. Playing the amp, the sound was a little thin, so I decided to tweak a bit. I upped the plate voltage to 290V, put a 1k cathode resistor. This meant I was pulling 45ma through each 2a3, producing 45V bias, and dissipating ~11W on the plate (max rating is 15W). The sound was much less thin, bass was much more there. I have kept this operating point. I am a little worried about pushing already weak tubes, but they sound good for now.

A big debate I read about while planning the power stage was DC vs AC filament supply. This is a point some people get militant about on forums (AC causes unacceptable 60hz hum, but DC sounds "sterile" in comparison). This made me first consider DC filaments, but then when i realized I could get a filament transformer for ~10-15 bucks that would power both tubes (you need separate filaments supplies for DHT tubes, if you didn't know like me when I first started the planning), I took the plunge. I was super careful about pathing the 2.5 VAC filament supply wires, and, guess what, this amp is DEAD QUIET. I use this amp with Cornwalls (>100 db/w), and there is zero hum even with my ears pressed against the speakers. I used my scope to adjust the hum balance pot for minimal 60hz, and it is dead quiet. Food for thought as you think about DC vs AC filaments. Probably a good part of my success can be attributed to the wiring/grounding (see below).

Power supply
I used PSUD II to help make a low ripple power supply. When designing the power supply, I wanted the option to create a DC filament supply from 6.3V supply, and I wanted to not spend too much on the transformer. This led me to the XPWR007 which offered 6A 6.3V, 275-0-275 at 200ma for only ~60 bucks. To keep choke costs down, I used the 159R from hammond (6H, 200ma, 150DCR). Now, there are a bunch of folks out there that insist on minimal resistance on the power supply line, but I disregarded this advice for better or for worse. I figure that I can reduce the resistance in the future if I want to try to use the same power transformer on 300Bs (maybe not the best idea, as B+ won't be high enough to drive them to full potential). When talking with Jack E of Electra-print, he helped me hone the execution for the power supply, and I took cues from him by adding smoothing capacitors in series bridged by resistors (see schematic). The wall voltage at my house is a little hot (~125-8VAC), so I ended up using the 5V secondary on the PT to buck the mains.

Plinth/chasis
I wanted a wood plinth for appearances sake, and then a metal plate to mount all the components. I was worried with getting a thick enough plate to support the weight of all the transformers without breaking the bank, and settled on 11 x 14 piece of 3/32" inch aluminum (ordered from Metals depot). To cut the holes, you need a drill press. After using a hand drill on other projects and just never getting it 100% right, I went on craigslist and splurged on a drill press for $50 where the guy threw in a bunch of great bits as well... the best $50 i spent. Now, where I screwed up is in painting the metal. Initially I thought a glossy enamel black paint might look good. I was a total idiot, didn't prime the metal, didn't thin the pain. It came out terrible. So, I flipped my drilled plate over, applied a spray primer, black matte spray paint, and then a few layers of poly spray to protect the paint and give it a nice sheen. This came out MUCH better (but the bottom where I mounted I ended up mounting the circuitry looks like a dumpster fire, see the pics). Probably the best thing would have been to get it powder coated, but alas I am a noob in the proletariat. I bought a plank of mahogany from HD, made 45 degree cuts with a mitre saw, glued it together, and then used a router to make the depression for the top plate to sit on. I stained the wood and applied a gloss poly finish.

Wiring and grounding
AC wiring is very commonly a source of serious hum, so I took great care with the filament wire paths to minimize proximity to signal and to use shielded wire for long paths. I routed the 6.3V around the outside edge in shielded wire where it wouldn't cross signal paths. For the 2.5VAC, I did the same, but cheated on one of the channels, twisting the pair of wires to cross ~2 inches above any signal wires. For the low level input signal that traverses the length of the amplifier, I used shielded cables. For grounding, I established a ground bus for the power supply, which I then grounded to the chassis in between the amplifier sections and the power supply. I then routed grounding in series to the amplifier section, which connected to the same chassis ground. I then made sure that there was only the single connection point with chassis. This made me feel pretty good about eliminating ground loops.

Conclusion
The amp sounds very nice. It is a step up from the DG-SE1 (maggie 6BQ5 SEP, which I built guided by the awesome folks on AK ) in transparency and sound stage, at least on my Cornwalls, but not so much that it puts the DG-SE1 to shame at all. I prefer this amp also to my solid state Marantz 1200 on the Cornwalls, as well as a few tube integrated amps that I have hanging around. Most of all, I had a fun time building it and I spent inordinate time reading AK, DIY audio, and other forums to help guide my build. I learned a lot about setting operating points for tubes, and I hope my struggle (esp. with the D3a's), may help some novice builder down the road if they also decide to use D3a's. What I'm going to make sure I have next time when making an amp from scratch is to have the right testing equipment (like a distortion analyzer) as I made the amp. Maybe this would have helped cut down on time fiddling to find operating points that sounded good.

And, it ain't real unless pictures. See attached for a few pics plus the schematic. Would appreciate any thoughts.

 

[zackthedog]

Well, that's splendid! A drill press is indispensable. Did you punch the socket holes or use the drill? And if so, what kind of bit did you use?

 

[mcheinstein]

Thanks for stopping by Zach! Yes I can't believe I lived without the drill press. I actually drilled the socket holes. First usedtraditional straight bits up to 1/2 inch, and the used a stepper bit to expand the holes. For the 2a3 sockets, it did take me to the last step... oops (I think it's about 1.25 inches). Maybe punching the socket holes would have been a better idea.

 

[Palustris]

Very nice build! You have an excellent layout and fine construction technique. I am certain it sounds great on Cornwalls.

I've built a number of SE 2A3 amps with a variety of high Gm drivers like the D3a so there are a few points I can make that I hope are helpful. First these high Gm tubes do oscillate if given any chance to do so. Listen for high frequency sibilance or noise; use grid stoppers right on the tube pins. I also use ferrite beads on the plate and heater pins. I am bit paranoid after hearing considerable sibilance on one of my first builds with these tubes.

There is nothing wrong with the D3a, but if you want to experiment, the 6688/E180F is very cheap and I found it to be just as good sounding. The benefit of the 6688/E180F is that you can buy several because they are cheap and match them up in the amp. With respect to the first stage, there are a couple of enhancements you can make. With your bias at -1.4V an LED will provide the benefits of fixed bias through cathode biasing. So any cheap LED like the HLMP-6000 will work and eliminates both the cathode resistor and the bypass cap: one less cap in the amp. For the plate you substitute a CCS which will flatten the load line to horizontal. I use the K&K Audio Compact Current Source Kit @ $16 each; you need two. They are very easy to mount and small enough to tuck anyplace. I use a ferrite bead directly on the plate pin so you can mount the CCS a couple inches away.

I use the E-P 3k ohm output transformers on my 2A3 builds also and don't think you can beat them without spending huge money. I use the EML mesh plate 2A3, but they are prohibitively expensive for a budget build. I would suggest that you try to find a couple more matched 2A3s when you can; the JJ 2A3-40 is the best of the "budget" 2A3s. One very important point I notice that will, in the long run, make you amp more reliable and noise free. Parallel the wiper of the hum pot with two resistors to each end of the pot. The resistors can be about 10r to 15r and the pot is generally 50r wire wound. The object is to have the cathode current run through the resistors and not the wiper of the pot which will become noisy over time with the current running through it. There will be a smaller area where the hum is eliminated, but with a scope it is easy to find. I illustrated this in the schematic below.

Finally, I can't help but advise you to eliminate your preamp: replace the 100k input grid resistor with a 100k pot and run your DAC or CD player directly into the amp. Even if for an experiment just give it a try. You can easily go back to the original setup if you don't find it is an improvement. There is a slight mistake on your schematic: the grid is a 0.0V, the cathode is at 1.4V. Don't forget that the D3a and other high Gm tubes (the 2A3 also) start to draw grid current around .5V so your actual input voltage will be <1Vrms. There is no need for any further amplification in your signal chain from the DAC.

A final note on the schematic: this is the E-P DRD amp and the schematic is 7 years old. Today I would bias the E180F at closer to -1.4V - 1.75V since the grid current makes the input voltage too narrow with a 1V bias.

 

[Palustris]

Here is an illustration of an amp with the K & K CCS as the plate load for the first tube. Note the tube is an EC86 which is another cheap tube, if you are patient. I just bought 8 for $50. It is a great tube to use to drive a 2A3: reasonable Gm and high Mu.

 

[mcheinstein]

Hey Palustris, thanks for all the thoughts. Very helpful.

I recently got E180F/CV3998s and E280Fs. I find the sound to be pretty similar as you've said yourself... and they are pretty cheap. I would second your advice about potentially using them in the future since D3a's typically sell for a bit more. Though of note is that both of those options have a little less gain than the D3a.

I did read many folks talking about LED biasing, and I did try some red LEDs that I had which produced about 1.7V bias... and I preferred the sound of the bypassed cap. I'll look out for some HLMP-6000 and give those a shot. CCS did come up quite a bit in my initial design, but I eschewed for complexity sake. Thanks for linking the boards though from K&K, looks promising and pretty simple to implement. I'll have to give those a shot in the future. Do you think the CCS improves the sound?

I have been thinking about getting a pair of JJ 2a3-40s to roll in (would also change operating point). But can't justify the price yet.. much less the mesh plates (though many folks have said they sound wonderful). You think the mesh plates are worth the $600 price tag? Super great advice on the paralleling across the wipers of the hum pot, I will be doing that this afternoon... I was a little worried about running that much current consistently through those little pots.

My preamp is passive for line level inputs. DAC comes in and then gets routed to the amp, no volume knob. I use an Emotiva XDA2 with a good stepper volume control circuit. Really, the pre is for the turntable (AR XB), which then makes it easy to switch inputs without having to mess with wires. I've tried directly into the amp vs through the passive preamp and detected no change in sound quality.

Both schematics you've posted have the first and second stage directly coupled. Have you found that directly coupled is the way to go vs cap coupled?

Thanks again for the thoughts!

 

[Palustris]

Both the schematics illustrated above are built on the same Wellborne Labs monoblock chassis that came as a kit and that I bought used. You are seeing a seven year evolution of the chassis. I would be very happy with the JJ 2A3-40, but I am happier with the EML mesh plate. Notice that the EML has an octal socket and a cathode connection derived from the mid point of the filament. I can and have cap coupled the two tubes at some point simply by removing the first PS cap turning the PS into a choke input that drops about 100V, making the B+ safe for a 2A3. In that experiment I used the Russian K40Y-9 as the coupling cap.

I felt that direct coupling was a huge improvement with the deficiency being a perceived lack of the full timber of an instrument like the violin or saxophone. Saxophone sounded much more realistic with the direct coupling. That said, I am auditioning a 45 amp that is RC coupled using a Jupiter copper foil capacitor. Listening to a saxophone quartet last night I was struck at the incredibly deep resonance of the baritone sax at its lower range: great stuff. I think that copper foil (not metalized aluminum) could be the key.

I have not used a resistor as a plate load very frequently. I experimented with the 6K5G as a driver for the 2A3 and used a resistor plate load. I liked the sound, but moved back to the high Gm tubes. Please don't discount the use of the 6CM4/EC86 as a driver. First of all, it is a real triode; it has lower Gm (than the D3a and E180F) making it easier to use, but it still has 5 times the Gm of a 6SN7; and it has a Mu of 70. Just ensure they are EC86: some clown has EL86 listed on epay as EC86. The RFT are cheap and I thought they sound great, but I also have Telefunken and Amperex. They all sound good.

 

[Palustris]

Here's a graphical illustration between a CCS plate load and a resistor plate load. Both assume an LED bias @ -1.75V. The RL has slightly better 2nd harmonic distortion figures at about 0.0% and the CCS load has about 4.8% 2H distortion. The RL has a voltage swing of 90V peak the CCS = 120Vp. The RL has a gain of = 51 and the CCS a gain = 68.5 (output V divided by input V) The RL has a plate dissipation of 1.9W the CCS = 1.17W; max dissipation is 2.2W. 16k is a standard value, but I had already drawn line.

 

[mcheinstein]

Wow this is really helpful. Thank Palustris. I had not considered the EC86, but the plate curves look awesome, and thanks for your load drawings. It's giving me the itch to experiment.

Meanwhile, I finally bit the bullet and ordered a signal generator off of the bay and got time to do some square wave analysis off the transformer 8 ohm taps into an 8ohm dummy load. I realize this amp has no feedback, so I shouldn't expect perfection. The output iron is Electra-print, so nothing to sneer at.

1st one is 1khz, looks like I have a little hf ringing at ~50 Khz not present on the raw output from my signal generator. From my reading on this and other sites that seems to be acceptable... please chime in if this you see a problem.


2nd one is 2 khz, doesn't look terrible?


3rd one is 500 hz. I understand there's going to be phase shifts as I go lower due to the OPT, which likely drives the slope on the flat portion of the square wave. Should I be concerned about this trace, or within normal ranges?


4th is 100 hz... lol.


I've tried doing my due diligence on the web of researching on this, but I'm dipping my toes into uncharted water for myself and would appreciate any help/comments.

Thanks!

 

[nerdorama]

Nice amp. I haven’t built many SE amps but have a PP 6B4G with a differential pair of 6GK5 driver tubes. They might work as a decent SE driver. They sound wonderful in PP. I based my amp on the Musical Machine amp by Audiotropic or Poindexter. There might be some useful ideas in it for an SE amp.
Have fun,
John

 

[mcheinstein]

That's interesting. I really hadn't considered / read a lot about 6GK5 tubes til you (John) and Palustris now mentioned them. Seems like a good high mu tube with decent plate current, and a similar operating point to where I am running my D3a. Where it looks different is the linearity looks a tad lower. However, big point to the 6gk5 is that the auction site has NOS tubes for <10$ a pop. I'll have to order some and play around with them... they sure are cheap enough.

Have you ever compared 2a3 to the 6b4g? It looks like the 6b4g's are essentially 6A3s with a different base, so they should sound pretty similar...

 

[PakProtector]

You know that the usual directive not to run a ccs with a pentode can deliver some solid, linear, *AND* low output impedance amplification if broken properly...
cheers,
Douglas

 

[mcheinstein]

One more update a year later.

Very nice build! You have an excellent layout and fine construction technique. I am certain it sounds great on Cornwalls.

For the plate you substitute a CCS which will flatten the load line to horizontal. I use the K&K Audio Compact Current Source Kit @ $16 each; you need two. They are very easy to mount and small enough to tuck anyplace. I use a ferrite bead directly on the plate pin so you can mount the CCS a couple inches away.

I decided to load the D3a plate with CCS instead of resistor. I'm using a 10M45S chip as my CCS set at 15mv (~180 ohm resistor in series with k pin) and anode voltage of 155 V. It's a noticeable but not huge upgrade to the sound. The biggest thing I can tell is that the stereo image is a bit sharpened and maybe the highs got a tad better.

I would say one could survive without CCS just fine, but for the cost of $2.80 / channel, you can get a 10M45S and upgrade. I'll leave the decision to you all .

Thanks again, Palustris, for your input. I'm now thinking about getting those EML mesh plates... maybe a Christmas present to myself. We'll see.

 

[mcheinstein]

More years later. I got EML mesh plates, which sound unreal. I finally did some distortion/power measurements. At this point, I've swapped out the D3a for E280Fs that just sounded so right. 10M45S CCS chip.

Thd = 0.325% at 1W, 1khz
+/-0.5db 30hz - 18khz
Develops 2.5W into 16ohm load ** I use 16 ohm altec speakers and the EP iron has a single 8 ohm tap. Luckily I still don't really find myself needing more power.

Just wanted to post because I've found relatively few looks into how much distortion these types of amps really make.