To achieve this goal a fairly common phono architecture was chosen, a RC coupled circuit with passive split RIAA, using commonly available and reasonably priced tubes. Since the 6CB5A amp already introduced transformer coupling, this concept was carried over to the preamp as well. At least into the line output stage. This would provide a low output impedance and the option to use a transformer volume control.
The tube choice was somewhat inspired by JC Morrison's Siren Song preamplifier which was presented in an article in the Sound Practices magazine in the 90ies. Also the passice split RIAA approach is shared with the Siren Song. But this is were the similarities end. While the Siren Song hat a differential front end, this preamp should be all single ended.
In the first stage the 6SL7 provides a lot of gain to bring up the phono signal to a reasonable level. The first stage drives the 75uS part of the RIAA network. After that a stage with paralleled halves of a 6N7 further amplifies the signal and drives the 318/3180uS network. These two stages provide about 40dB gain (20dB loss in the RIAA EQ already deducted). A final stage with a 6SN7 driving a 4.5:1 Lundahl LL1660 line output transformer amplifies the signal by another 12dB and provides a low 350 Ohm output impedance.
This is sufficient gain for MM cartridges. For MC a suitable step up transformer needs to be used. This is best directly integrated into the preamp. A suitable and very nice sounding MC step up is for example the Lundahl LL1681. Depending on the step up ratio, which can be chosen between 1:13 and 1:26. The overall gain of the preamp can reach well over 80dB, enough even for MC cartridges with very low output voltage.
This is the first version of the schematic which was built in a prototype style and tested:
This was the first version of the PSU:
In order to minimize capacitors in the signal path, all cathode resistors are unbypassed. The last two stages are DC coupled and the output stage uses an ultrapath cap. If you only have an analog front end, this circuit is optimized for this purpose. But like this it is a bit difficult to get an entry point for line level sources, since the output stage is DC coupled. We will see how the circuit can be changed to enable switchable line inputs later.
First let's discuss some further aspects of the design. How are the RIAA components calculated? The 75uS network provides a corner at 2122Hz. It gives a constant roll off of 3dB per octave from that frequency on. The filter which achieves this is a simple RC network, the resistance multiplied by the capacitance results in 75uS. But if you multiply the 200k series resistor with the 150pF, you get a different value. This is simply because the output impedance of the driving stage (this is the plate resistance of the tube paralleled by the plate load resistor) is in series with this resistance and needs to be added. Also the grid to ground resistor of the following stage plays a role. This is in parallel to the series resistance. Another component which is not directly visible but plays a role is the miller capacitance of the following stage. It is in parallel to the capacitor in the network which needs to be reduced in size accordingly. Unfortunately the datasheets of the 6N7 don't give electrode capacitance, probably because it was not intended for RF use which requires this information. So the actual cap value for the RC network was determined by measurements in circuit.
As you can see, both the miller capacitance of the second stage as well as the plate resistance of the driving stage are a significant part of the RIAA network and will influence it's accuracy. This could be seen as a weak point, but this compromise was made intentionally. One way to reduce this dependency would be to use a lower impedance RIAA network, then the miller capacitance of the second stage would be negligible. But a lower impedance network in turn would require a driving tube which has low plate resistance. Such tubes either have low gain or high transconductance. Low gain tubes are not really suitable since we want a lot of amplification in the first stage. High transconductance tubes typically tend to oscillate without suitable counter measures, therefore they where ruled out as well.
There is a series resistor to the cap to ground in the 75uS network. The purpose of this is to add the von Neuman time constant. This 4th timeconstant is not described in the RIAA standard. But it is used in the recording process. The reason is that the recording amplifier cannot increase the amplitude with rising frequencies endlessly, but stays flat beyond a certain frequency. The actual corner frequency was never really defined. It can be assumed to be around 50kHz. The series resistor stops the high frequency attenuation of the 150pF cap at a certain frequency. However this resistor cannot affect the roll off caused by the miller capacitance, which is a substantial part of the filter network. So it is debatable if it makes sense to have this resistor. I left it in, but did not do any further experimentation there.
The second part of the filter network is less influenced from it's surrounding circuits. It can be made lower impedance due to the lower plate resistance of the 6N7. The two time constants can be simply achieved by a 100k resistor in series and a 32nF cap in series with a 10k resistor to ground. 100kOhm times 32nF equals 3200uS and 10k multiplied by 32nF equals 320uS. The cap can be formed by some standard values like 22nf parallel to 10nF. With some selection of the caps they can be brought closer to the ideal value of 31,8nF. Again the output impedance of the driving stage and grid to ground resistor of the output stage needs to be considered in the calculation of the series resistor.
The schematic also illustrates the grounding approach. Each stage has it's local star ground. These ground points get connected with a heavy gauge solid copper wire. The connection between ground and the output is shown as a dotted line. This is an optional connection and can be done according to the system needs. In order to avoid ground loops it can be left open. Each channel is decoupled via it's own choke so the PSU can be shared for both channels.
Since all the tubes in the preamplifier share the octal base, the name Octal preamplifier was created. Even the rectifiers are octal base types. In total the preamp has 8 tubes (1 6SL7, 1 6SN7, 2 6N7 and 4 6AX4).
The preamplifier has been built several times in this form and works nicely. But as the suffix Mk1 in the title indicates, there is a new version on the way with some changes:
The input and output stage both share one half of a double triode between the two channels. While this is not a real problem (channel separation is an overrated parameter) it can be easily changed. The simplest way would be to just use two each of the 6SL7 and 6SN7 and wire both triode systems in each bottle in parallel. The 75uS RIAA network would require some adaption if this is done and the line output transformer would need to be exchanged for a 18mA type. But since these two are very commonly used in the audio world, I'd also like to use different, less common tubes instead of them.
For the line putput stage there is another reason for a change. The 6SN7 even working into a step down transformer provides a lot of gain of about 12dB. With most line sources having 2V RMS output, this is a bit on the high side. We want to have a sensible range on the volume control. So it will be substituted with another tube.
The Mk2 version of this preamplifier will be presented in another article. Stay tuned!
P.S.: I tried to explain all aspects of the circuit. If anything is left unclear, don't hesitate to ask.