DIY DIY Balanced Line Stage

[Jaytor]

I recently completed a new line-stage preamp build that I thought might be of some interest. I went through a few iterations of the main line-stage circuit based on listening tests, and I’m pretty happy with the final build.

This is a balanced, differential design that incorporates both a tube-based gain stage with balanced outputs and an opamp-based stage with single-ended outputs intended for driving my subwoofers.

I started this project at the beginning of 2022, so it took me a while to complete. The main reason is that I ended up implementing five different tube line stage designs before settling on the one I’m using.

I also had several other projects I was working on at the same time. During this time period, I built two speaker systems (including the ones in my main system) and three stereo amplifiers, and started on yet another preamp build.

I’ll go into more detail in subsequent posts in this thread.

 

[Jaytor]

Input Selection and Attenuation

The preamp uses a relay-switched input selector which accepts three balanced inputs and one single-ended input. I rarely connect a single-ended device. The selected input feeds a balanced relay-switched attenuator.

The input selection and attenuator are implemented using AMB-org’s delta 2 and delta1 boards respectively. I’m using two of each PCB to support balanced operation.

These boards are controlled using an Arduino with my own software for control, with a color TFT display showing the input selection, balance adjustment (if any applied) and volume setting. I have also implemented an IR remote. When the front-panel power button is pushed, the main power is applied to the tube line-stage. After 30 seconds, a 12V trigger output is enabled to turn on my amplifiers. After another 30 seconds, the output of the tube line-stage is unmuted, and the volume is ramped up to the last setting.

The switched attenuator is implemented using eight relays for each channel, providing 256 0.5db steps. I have designed the software to use a 100 step range so the lowest volume (other than mute) is 50db below the maximum volume. The 0.5db steps give me enough granularity to implement a balance control. Although I have never needed balance adjustments in my system, this preamp uses no global negative feedback so differences in tubes could cause a small channel imbalance.



I’ve chosen resistor values to provide an input impedance of about 45K ohms which will work well for my current and planned DACs.

I have spent some time over the past few years testing a variety of resistors to see if I could hear appreciable differences. The differences are fairly modest once you get to reasonable quality, but I decided to splurge on this project. I am using Vishay Z-foil resistors for the series resistors in the attenuator. Of the various resistors I have tried, the Z-foils are the most neutral and detailed, but using them exclusively can be a bit cold sounding. I found that I really liked the sound of the Audio Note silver tantalum resistors, particularly for the shunt positions. I would have considered using these for the series positions as well, but they are not available in all the values I needed, so I used them just for the shunt positions.

In all honesty, I’m not convinced the expense of these exotic resistors is worth it. The Vishay-Dale RN55/RN60 resistors sound very nice and are a fraction of the cost. It might be worth using the AN silver tantalum resistor in the final shunt position (which is always engaged), I will probably stick with the RN-series resistors for the rest of the positions in future projects.

The output of the attenuator feeds both the op-amp and tube-based line stages.

 

[paul_b]

Is that machining by Schaeffer/FPE?

 

[Jaytor]

Op-amp Line Stage

The op-amp stage is implemented using a pair of AMB.org alpha24 PCBs which utilize OPA1612, OPA1632 and OPA1611 op-amps. I found that the opamp design results in deeper, tighter, and more nuanced bass, but I prefer the tube design for everything else.

The alpha24 design is an instrumentation-amp topology. This uses a dual op-amp (OPA1612) as an input buffer on the two phases. This provides an extremely high input impedance with low capacitance resulting in almost zero load on the output of the attenuator.

The output of the 1612 feeds an OPA1632 balanced differential amplifier which provides the gain. This op-amp has both positive and negative phase outputs. These can be used to drive balanced outputs directly, but in my case I am feeding a final op-amp stage using an OPA1611 to convert balanced to single-ended for my woofer amps.



I decided to use the Neurochrome Preamp supply to provide the power for the alpha24 boards. This is a well designed ultra-low noise +/- 12V supply with enough current for these boards.

 

[Jaytor]

Is that machining by Schaeffer/FPE?

Yes. They are a bit pricey, but do good work and are fast. Since FPE is in WA, ground shipping is overnight.

 

[Jaytor]

Tube Line Stage Evolution

This is the heart of the preamp and the part I spent the most time on. When I started this project, I had planned to use John Broskie’s Octal Aikido design, with a separate stereo board for each channel. The downside of this approach is that it isn’t differential and provides no common mode noise rejection. In the end, I wasn’t happy with the way it sounded, but I think this was mainly because my operating points weren’t optimal.



My second attempt was designing my own PCB to implement a balanced, differential version of the Aikido which John published in one of his blog posts. This worked a bit better, but was still not as good as my previous preamp design. I think the two main reasons were 1) this was my first analog PCB design and my layout was far from optimal, and 2) John’s design required the use of some unusual size capacitors so I ended up mixing and matching and used some mediocre parts.



After this, I decided to try my hand at designing my own circuit and not bother with the Aikido approach. The main advantage of John’s novel Aikido architecture is significantly improved power supply noise rejection. As I will discuss later, I am implementing a low-noise regulated supply, so I don’t think the benefits of the Aikido design make up for the increased complexity for my purposes.

My next build was a differential 6N6P circuit with a balanced cascoded MOSFET source follower output stage. This worked fairly well, but I had trouble finding 6N6P tubes with matched sections. In fact, of the 40 tubes I purchased (fortunately fairly inexpensively), none of them had sections matched better than 20%. If and when I use this tube again, I will either use just one half, or use it in a configuration where I am not relying on matching.

Overall, this design worked pretty well despite the mismatched tube sections, but I thought I could do better.



My next approach was using a 6SN7 which is much easier to find with well matched sections, and has a very nice sound. I decided on a differential design with a BJT-based current sink in the tail and a push-pull line output transformer on the anodes. I liked the way this one sounded, but I thought of a few things I wanted to change after finishing this build. This will be described in another post.

 

[Jaytor]

Tube Power Supply

The tube line stage is powered by fully independent supplies for each channel, using custom Toroidy transformers fed with a Neurochrome soft-start board. This is the second set of transformers I purchased for this project since the first set had much higher voltage for the Aikido circuit. The transformers I am using are oversized with 205V 400mA and 10V 3.5A secondaries, but the B+ voltage is a bit high for my final design so I’ve had to burn off some voltage in the pre-regulator filtering circuits.

I’m using fast-recover schottky rectifiers feeding a CRCRC filter circuit. The filtered voltage then feeds a shunt-regulator based on diyAudio contributor Salas’s SSHV2 design. I had originally planned to use Neurochrome Maida regulators and had designed the enclosure to fit these boards, so I designed my own PCBs for the SSHV2 to be mechanically compatible with the Neurochrome regulator. The regulators are set to 200V DC output with the series current set to 48mA (the actual circuit draw is about 28mA, so this leaves 20mA of additional current for the shunt regulator to work well.

Here's a photo of my SSHV2 implementation.

 

[Jaytor]

Tube Line Stage - Final Design

For my fifth, and I think final, implementation, I made the following changes to my previous design.

1. I removed the input transformer since I almost exclusively use balanced sources and I didn’t think it was adding any value (and took valuable real-estate on the PCB).
2. I split the current sink in the tail (cathodes) into two independent current sinks, and connected the cathodes through a resistor. This provides two benefits - it allows the DC balance to be optimized even if the tube sections aren’t perfectly matched, and it allows the gain of the stage to be adjusted. Using a zero-ohm resistor would be the equivalent of a single current sink, but by using a small value resistor, the gain can be reduced. I’m currently using a 220 ohm resistor which results in a gain of approximately 13db.
3. I added a negative 12V supply to connect the bottom of the current sinks to (instead of using ground). This provided more voltage to work with for the current sink and significantly more headroom for the inputs.
4. I doubled up the 6SN7s - now using two sections in parallel. This halves the effective plate resistance, and thus the output impedance of the line stage.
5. I switched to a nanocrystalline core line-output transformer. This core material provides improved frequency response, better linearity, and improved efficiency.
6. I added a chassis ground circuit on the PCB. I had been using a separate PCB to connect signal ground to the chassis ground. This uses a bridge rectifier in parallel with a resistor and a cap to provide some isolation, but still provide the safety of providing a current path between signal ground and chassis ground.
7. Improved PCB layout with both inner layers coupled together as the ground plane.

The line stage is implemented as a differential pair, with each side running on a 6SN7 tube with the two sections in parallel. The anodes are connected to a Lundahl LL1660/PP Nanocrystalline transformer connected as 2.25+2.25:1.

The cathodes are connected to constant current sources (sinks in this case), split between the positive and negative phase as mentioned in my previous post. The current sinks are implemented as cascoded BJTs using LM385 voltage references. The negative side of the CCSes is connected to a negative 12V supply, which is implemented using a DC-DC converter from the filtered filament supply, feeding an LT3094-based regulator. I’m using a small choke on the input and a common-mode choke on the output of the DC-DC converter to reduce the effects of any high-frequency switching noise on the input or output of the converter.

The DC bias current through each triode section is set at 7mA, with approximately 195V between anode and cathode.

The output of the LL1660 transformer is connected to the balanced output through a muting relay, so the output is only connected after the preamp has had a chance to warm up a bit (about 60 seconds).

The 6SN7 filaments are powered by a regulated DC supply. A separate regulator is used for each tube, primarily to reduce the power dissipation in the regulator chip, but this also provides a slight reduction in coupling between the tubes.

 

[Tlarwa]

I recently completed a new line-stage preamp build that I thought might be of some interest. I went through a few iterations of the main line-stage circuit based on listening tests, and I’m pretty happy with the final build.

This is a balanced, differential design that incorporates both a tube-based gain stage with balanced outputs and an opamp-based stage with single-ended outputs intended for driving my subwoofers.

I started this project at the beginning of 2022, so it took me a while to complete. The main reason is that I ended up implementing five different tube line stage designs before settling on the one I’m using.

I also had several other projects I was working on at the same time. During this time period, I built two speaker systems (including the ones in my main system) and three stereo amplifiers, and started on yet another preamp build.

I’ll go into more detail in subsequent posts in this thread.

View attachment 85784

View attachment 85785

View attachment 85786

That is a great looking piece of equipment!

 

[paul_b]

I am curious about how you found the AC balance to behave when splitting the tail load into two CCS. I have to admit, I've never cared all that much about DC balance, but I can certainly appreciate not wanting any imbalance through those output transformers.

 

[Jaytor]

I am curious about how you found the AC balance to behave when splitting the tail load into two CCS. I have to admit, I've never cared all that much about DC balance, but I can certainly appreciate not wanting any imbalance through those output transformers.

It seems to be working fine but I have only tested with balanced sources. I suspect unbalanced sources will not work very well since I am just connecting the negative phase input to ground. But I don't have any unbalanced sources so I'm not too worried about that.

My tubes are fairly well matched with well matched sections on my Amplitrex tester.

I got this idea from Allen Wright. He used this approach in the later versions of his RTP3 preamp.

That is a great looking piece of equipment!

Thank you.

 

[Wntrmute2]

Impressive build!