Updated: 13/Jun/2000
As is often the case this project started with research, lots of research. I spent countless hours browsing the web, reading articles and pouring over schematics. There are a lot of very good preamp designs available and it was difficult to decide which one to start with. After several months of research I at last settled on the Loesch design as a starting point. I found a number of very favorable comments about this design. I liked the 417a used in the first stage and also the battery bias. What I didn't like was the paralleled 6DJ8 line stage and the passive volume control. But since the design was only to be a starting point these issues were easily addressed.
I decided to do a lot of experimentation with this project. No only to optimize the design but also for my own education. I firmly believe that to get the most sonic pleasure from this hobby you have to explore different topologies, try different parts, be willing to put your biases aside and most of all trust your ears. To allow experimentation I first built a prototype on a piece of wood and used it to try a number of different configurations and a lot of different components.
The first incarnation of the prototype consisted of the RIAA section of the Loesch preamp with only a minor adjustment of the equalization to allow for a 100k volume control to be used immediately after the 6GK5. For the line stage I used a 6SN7. The parts were mostly what I had on hand. The capacitors were Wima's, Vitamin Q's and some Westcap paper in oils. The resistors were Holco's and Allen Bradley Carbon comps. The prototype sounded surprisingly good. The 417a's were a bit noisy and the upper bass was a little bloated. The 6SN7 line stage sounded especially good. Very transparent with an extended airy high end.
Now it was time for the experimentation. I started by trying different caps for coupling the first stage (between the 417a and the 6GK5). I figured that this position was probably the most critical. It was remarkable how different each of the caps sounded. Here are the caps I tried ranked from worst to best:
The next experiment was converting the first stage to a FET/417a cascode. I had misgivings about putting a solid state device in the signal path. But having heard such glowing reports about this mod I was convinced to give it a try. Well as I had expected adding the FET did impart a just pinch of analytical transistor sound. However, what I did not expect was the big improvement in clarity and detail, not to mention much quieter.
After reading about Michael Maloney's Loesch variation OZLOESCH I tried his trick of adding a 3v battery to the grid of the 417a. This results in the same current and output impedance (so I did not have to re-tweak the RIAA) but it puts an extra 3 volts across the FET. As Michael reported this change cleans up the transistor sound and is very musical.
Given my very positive experience with directly heated tubes (DHT) I was anxious to try replacing the 6SN7 with a pair of 26's. Well I was not disappointed. The 26 is a great sounding tube! The 26's are very transparent with wonderful midrange. There was little less air with the 26's but overall they were the hands down winner. Initially I used filtered DC for the 26 filaments. After finding a suitable transformer I upgraded to a current regulator using an LM317T adjustable regulator. As expected using a current regulator resulted in a warmer, richer sound.
After playing with different parts and adjusting the design it was time to fine tune the RIAA curve. I started by searching for published RIAA numbers to use as a reference and found three different sources that were all very different from each other. Arrgh! Fortunately, I was able to find the formulas for computing the RIAA curves and using a spreadsheet was able to obtain the correct RIAA numbers. To be sure that these numbers were right used Microsim to model a RIAA filter. The numbers from the simulation matched my computed numbers almost exactly.
Armed with a good reference I started measuring. I was surprised to find that there was a lot of error across the board and the 500hz turnover point was off by almost 2db. With several iterations of adjusting resistor values I was able to match the RIAA curve to within +-0.1db. I was surprised that I was able to hear quite small frequency response variations. As I was able to get the response closer to the RIAA curve music just sounded more right. I am not sure that my obsessive pursuit of +- 0.1db is warranted, but my experience showed that errors as small as 0.2 - 0.3db are audible. The RIAA specification also calls for a 20hz rumble filter. In my opinion 20hz is much too high. I ended up using a 8hz cutoff and if I were to start over would probably lower this value to about 5hz.
Once the RIAA curve was fine tuned I replaced the majority of the Holco resistors with Caddocks. The Caddocks sounded cleaner and smoother. Certainly a worthwhile change but the difference was not near as dramatic as with the capacitors.
Believe it or not the volume control I used for some time was a $1.99 Radio Shack dual 100k carbon pot. I had heard that these volume controls were quite good. Being skeptical I was planning to only use this control in the prototype and eventually replace it with a high quality stepped attenuator. The Radio Shack control sounded good enough that I did not replace it with the stepped attenuator for several months. The stepped attenuator (Elna Switch with Holco resistors) did sound notably better, but the Radio Shack pot came surprisingly close.
For the power supply I wanted to try using shunt regulation using voltage regulator (VR) tubes. I had not seen VR tubes used in a preamp before and had some concerns that they might be too noisy. To get sufficient voltage two 0C3's are used in series. This resulted in 217 volts for the B+. With the prototype I listened with and without using VR tube regulation. With regulation the preamp was much quieter and there were improvements in dynamics and clarity. I am on sold on VR tubes. They are simple, inexpensive, easy to use and sound great. The 470uf capacitors are generic electrolytics with paper in oil capacitors downstream from the voltage regulators. It would seem logical that the capacitor type used in the unregulated portion of the power supply would not be of any sonic significance. However, at this point I am less certain so I plan to experiment with Elna Cerafine caps in this position. The rest of power supply is a fairly standard CLCLC design that is well filtered and stable. The first capacitor was sized to set the unregulated output voltage to about 265 volts.
As with many of my other projects I used a ray tracing program Povray to create images of the preamp before starting construction. I created images of a number of different designs before settling on the final layout. Creating these images is tedious but invaluable for designing an elegant looking piece of equipment.
My initial design was for a separate power supply chassis connected by an umbilical. However, during the prototype phase I found that the having the power supply near to the preamp had no discernible effect on hum or noise. I also had some difficulty finding suitable connectors for the umbilical. So after considerable debate I decided to go with a single chassis. The chassis layout was done to allow an under chassis shield to eliminate any hum from the power supply. With a moving magnet cartridge the preamp was quiet enough that the shield was not necessary. However, when I upgraded to a 0.9mv moving coil cartridge there was a very small amount of noise. Installing a copper shielding plate between the power supply and the rest of the circuitry made the phono stage absolutely silent.
I think that the preamp sounds very good. But of course that is what you would expect a builder to say about his creation. However, I had an opportunity to compare my preamp with a superb, extensively modified Super-It (affectionately dubbed the Hyper-It). All assembled at the listening session had a slight preference for the Hyper-It. Being a moderately competitive male I felt challenged to improve on my design. I added some 0.47uf polpropelene bypasses at the power supply connection for each plate resistor, cleaned all of the connections and made a minor change to the grounding. In a subsequent listening session the tables were turned and now there was a slight preference for the my preamp. Using a current regulator on the 26 filaments and adding 3v battery bias to the 417a raised the performance by a couple more notches. These changes made a subtle difference, but there is something about getting rid of the last little bit of grundge that is very significant to musical enjoyment.
Some comparisons of the line stage were were done by direct connecting a DAC with an internal stepped attenuator directly to the amplifiers or by feeding the DAC through the line stage. As expected there was a slight loss of transparency when the line stage was inserted. However, there was also a smoother, richer sound imparted by the great sounding 26's. Depending on music or tastes it was hard to say which path sounded best.
The Loesch preamp is an excellent design that with a few adjustments and a good implementation becomes a real high end contender.
I recently completed some changes to the line stage with significant sonic results. The schematic has been updated to reflect all of the listed changes.
I changed the operating point for the 26's. I don't know what was wrong with my brain when I chose the plate load resistors for the 26's. But 18.6k was much too low a value for a tube with an 8k RP. I bumped up the plate resistor to 32k and the cathode resistor to 2.5k keeping the plate current the same. This resulted in notably better clarity and detail. I thought that the 26 was good before, but now it really shines.
I replaced the ST shaped 26's with RCA globes. It is remarkable that a tube that is electrically the same can sound so much better. The globes are more expensive and also tend to be noisier but they are well worth the cost and effort.
After very favorable results with silver interconnects I replaced all of the wires in the signal path with 30ga silver in Teflon tubing. It was interesting that the effect was on par with replacing the much longer interconnects with silver. I love the sound of silver and as exotic as it sounds it is actually quite a sonic bargain.
I had been using a 25k stepped attenuator in series with a 75k resistor for my volume control. With my 2.0mv cartridge this gave me just enough gain. This was fortunate since I already had the 25k attenuator. However, after I upgraded to a 0.6mv Clearaudio I need a little more gain. I ordered a 100k Danish Audio ConnecT (DACT) attenuator from Aloha Audio. My only reason for the upgrade was more gain and I expected little or no sonic improvement. Boy was I ever wrong. The sonic improvement was immediately obvious. The DACT sounds so smooth and rich! This was a much bigger difference than going from the Radio Shack volume control to my old attenuator built up with Holco resistors. I don't know what it is that makes the DACT sound the way it does, but after hearing one I would never consider anything else.
