My main DIY activity over the years has been preamplifiers. There have been many come and go however only this one (the basics of it) has remained for any length of time. This is not a final design but a work in progress with an aim to remove compromise as much as possible. My base design was a DIY version of the Audio Research SP8 with SP11 power supplies, all dual mono.
History and Initial Steps
I wanted a line stage replacement with one gain stage so gave the Bruce Rozenblit Line Level Preamp design a go.
The current was increased through the stage which improved things all round – more life, greater frequency extension, more ease and better detail retrieval. The ECC82 was then replaced with a 6SN7 to great effect. The final design was this circuit.
With the success of the line stage, I spyed a Glass Audio article that suggested the replacement of the “classic” 2nd gain stage and CF with a mu follower. This was implimented with good results.
An article by John Linsley Hood in ETI April 1992 provided the inspiration to try shunt feedback thereby removing overall feedback.
Many years later, a passive RIAA design was tried with excellent results which resulted in the phono second stage being a simple plate loaded 6072. The first stage has remained a mu follower and as Christopher Paul and Bill Petrowsky had done more work in this area, BP’s “head amp” stage was used.
An encounter with a CJ Premier 7 schematic provided the inspiration for the design of a 0 feedback stage buffer and HT regulator.
A design using 12AX7mu, RIAA, 6072, volume, 6SN7mu had been in use for about 8 years and was the basis for …
The start of any amplifier is the power supply.
The power supply was the result of a number of experiments – over specifying the transformer and toroid vs laminate types. The result is a massive 300VA HT transformer per side and separate heater transformer. The laminate was preferred to the toroid as a background gunge overlaid the music which I believe is a result of the wideband toroid design letting a lot of mains hash through. The laminate seemed to be purer with better solidity and dynamics.
Rather than attempt to reduce ripple by huge amounts with a big R and a big C before the regulator, it seems to sound better if the ripple is smoothed up a bit stage by stage hence the low value resistors between the capacitors. Having discovered Duncan’s Amp PSU designer, the values have been changed but not tried yet. The results as given by the program would indicate a major improvement in ripple reduction but what does that sound like?
What was tried was to get the same performance of a regulator but passively. Adding a choke input with all the same capacitors but replacing the series Rs and regulator with values from the PSU designer.
Al Marcy from the Joelist comes up with some great quotable lines. “What arbitrary requirement have you added which makes this a no-go?” was a recent one. Got me thinking real hard. A requirement for regulation I thought. Where does that come? erm …. duh….. because we know best? ’cause we always have? ’cause you can’t get reg performance passively?
Initial plans were for an HT regulator, as above, per stage but experiments lately have suggested a quieter design with better detail resolution is to RC decouple each stage from one HT regulator. I’ve tried many regulator designs over the years and have preferred the design above.
JC Morrison in his Siren Song article for Sound Practices said of regulation “… oh sure the bass gets all big and amazing …”
I don’t think I’ll be using regulation again. With the passive PSU, the noise floor, detail retrieval and just plain good sound coming from the system now is breathtaking in comparison to the regulated design.
Like the HT, it was planned to feed each tube element from its own regulator. Before that takes place, more work on the design and method for feeding the heaters needs to be carried out. Each heater will be RC decoupled from the heater regulator below. Currently they are all in parallel, RC decoupled from the reg output. The heaters are referenced to one common point floating about 1/4 HT voltage.
Think I might be trying a simple/alternative heater feeding scheme soon! As well as tube rectification for the HT which may allow a reduced transformer size especially if a choke input filter is retained.
Having removed the heater regulators and replaced them with a few RC combos, there is now even more music and less noise. Even though these regs perform better than the LM317 type, no regulators are even better still. Has made me wonder why regulation is a good thing.
The preamp power supply now looks like this.
The first part of the RIAA stage was BP’s “head amp” stage above which had proven to be a good performer over the last few years. I have always been attracted to using the 6SL7 as it has a fine reputation so it was put in place of the 12AX7. Compared to the AX7, the SL7 has lower noise, much lower distortion and a beautiful tone. Although a late manufacture chrome top JAN Philips GTY type was tried, it was inferior in all respects to a NOS 1945 JAN round plate, bottom getter Sylvania – the first stage setting the noise performance for the preamp. Unfortuanely, the SL7 is very microphonic so steps will have to be taken to isolate the board as much as possible. The double lead/MDF laminated case will go some way in achieving this.
The second stage has been a 6072 for years and although other tube types have been tried, it has remained champion. I want to try CCS load and fixed bias with diodes. Fixed bias with battery was not successful but may try again.
The Output Stage is now a single 6922 as that was the starting point for developing a single tube line stage. A constant current source load (C4S clone) substituted the plate resistor which removed the strident/cool quality I’ve heard in most applications where the 6922 has been used. Even though that stage was running at 10mA and only driving a 100k load, paralleling up both sections produced a jump in quality and didn’t seem to produce faults that would indicate the halves fighting each other. Quite the opposite in fact. A red LED for biasing the tube resulted in much better noise and hum performance and improved the sonics further. I feel this stage performs better than the 6SN7 mu stage but that is only a memory not a direct comparison.
The aim was to minimise signal path length, joins and wire. The tubes, HT capacitors and LT regulator are mounted on top of a PCB where the PCB foil is the ground plane. All other components are on the underside using only component leads for connections. The audio ground plane is a small PCB mounted in the middle – 3cm square. There is no wiring in the 0V connections to the ground plane for any component except the C4S. Heater wiring is twisted Kimber run on top and passed through holes to the tubes below thereby keeping it out of the way, seperated from the audio circuit by the ground plane.
Paralleling of capacitors or any component was to be avoided if at all possible. They tend to “spit” and smear time information when paralleled regardless of type or value differences. Bypassing of capacitors also ruins the purity of the signal even if the bypass is much better than the capacitor it is ‘helping’. The same goes for resistors but to a much lesser extent. Whether this all holds true in every position … To date, there are no bypass capacitors in system anywhere.
My experience lately would indicate that deposited film (metalised (MF) or bulk (BF) film) components have a strident ringing quality that in some systems can be unbearable. Mine is such a system but it has mystified me where this horrible noise had been coming from for years. To that end, I’m trying to remove as many metalised components as possible including MF resistors changing to WW and MF caps previously used in the PSU to Black Gates or where possible film/foil construction. In doing so, the music has returned.
The case is a veneered MDF/lead laminate glued together with epoxy. This came about after a number of experiments with materials and damping on an earlier prototype. As the preamp is now way simpler than it used to be, this case is now 50% fresh air!
How does it go?
This is early days for a new design and it shows good promise. It is the best I’ve achieved and has a number of attributes that are firsts for me. It removes a number of distortions I thought were part of vinyl replay like distortion which I had attributed to mistracking. It has also made me re-evaluate the amount of information on vinyl just when higher resolution digital comes along so I doubt there will be any contest. I guess one day digital might reach analog’s standard :-). ‘Noise’ in all its forms (either induced or as signal artifacts) is also at an all time low where I find myself straining to hear into quiet passages. Even “analog noise” is much reduced which makes me wonder whether most of analog’s perceived faults are actually in the electronics – an inability to keep up with vinyl’s information rich signal. How it sounds compared to anything else will be a mystery unless someone brings over another pre to try. At 27kg/60lb for just the audio box, it isn’t going anywhere. All in all, I’m enjoying listening to the black stuff more than I have in years.
Another of Al Marcy’s quotes “I am still finding more sound in small changes. I am a long way from what I suspect is possible. There is no hurry, no urgency, just vast promise.”
Also thanks to old buddies Richard Krebs, Rob Ferguson for design input and letting me rant on and on, LLDIYers Richard Daneilian (“… maybe a different approach is required …”), Andy Nehan for help in the RIAA stage, John Finlayson and Brian Clark for having me round for a listen, Thorsten Loesch (for the PCB board construction method) and finally the Joes, too many to mention, for the inspiration.
This for me closes a chapter for this type of design but opens up another world of possibilities – the future starts here.