Design Philosophy

Part One

Explaining a design philosophy necessarily means going into some technical detail. I have tried to keep this to a minimum but some of it is unavoidable. The easiest way to describe the design philosophy is to expand on the general principles listed on the Home Page.

No Semiconductors in the signal path

This is fairly self- explanatory. The object of an all tube design is for the signal to pass only through tubes and passive components. So, there are no semiconductor constant current sources or bias LEDs in cathodes. The only semiconductor devices are in the power supplies.

No Global Negative Feedback (NFB)

A microphone pre-amplifier needs to have a variable gain, typically over a range of 60dB, in order to accommodate the varying input signal level it is likely to be fed. With op-amps it is quite simple to vary the gain of a microphone pre-amplifier from 0dB to 60dB simply by altering the amount of NFB.

NFB has some useful properties but its big downfall is stability, mainly at very high and very low frequencies. Achieving stability at high frequencies is relatively straightforward using a dominant pole technique, but it does have the disadvantage of raising the level of higher order harmonic products as the feedback reduces as the frequency rises above the dominant pole. This effect can be minimised by ensuring the dominant pole is above the highest frequency of interest (usually 20KHz). This can be achieved in tube circuits but it is rarely achieved in op-amps most of which have a built in dominant pole at a few Hertz.

It is a different story at low frequencies. It is straightforward to ensure the feedback network operates down to dc in semiconductor designs. This means stability at very low frequencies is not a problem. With tubes, however, this is very hard to achieve because of the widely different input and output dc voltages to which the feedback network must be connected. It is possible to design a tube pre-amplifier with NFB operating down to dc but it is difficult (but not impossible) to also achieve a satisfactorily low minimum gain ( a minimum of 20dB being typical).

Another problem with tube pre-amplifiers is achieving a high enough open loop gain. With op-amps it is easy to vary the closed loop gain from 0dB to 60dB or more by altering the feedback. With a tube circuit it is difficult to obtain an open loop gain of 60dB which effectively limits closed loop gain to about 40dB if the distortion reducing properties of NFB are to be gained.  So a tube pre-amplifier with stable NFB is only likely to have a gain range from 20dB to 40dB. On balance, global NFB has little to offer in tube pre-amplifiers and some severe drawbacks either in stability or gain range.

Designs are built from standard Class A circuit blocks

The most commonly used circuit block in CTC designs is the mu follower. This topology allows a tube to achieve close to its intrinsic distortion level. With an appropriate choice of tube very low distortion can be achieved at normal operating levels (less than 0.02% at 0dBu) without the use of global negative feedback and the distortion is principally second harmonic with the higher harmonics falling off rapidly.

Channel gain is controlled by switching in pads at appropriate points in the signal path

A typical CTC channel consists of an input pad, a transformer (which provides 20dB gain), a 25dB gain stage, and a variable pad followed by another 25dB gain stage as shown in the figure below.


With the pads switched out the channel has a gain of 70dB. With both pads at maximum attenuation they introduce a loss of 60dB so the overall channel gain is reduced to 10dB. By setting appropriate values in the pads any channel gain from 10dB to 70dB can be obtained. For a 0dBu output, the second 25dB gain stage will always have an input of -25dBu and its distortion will be fixed. Depending on the input level and the pad settings, the input level to the first 25dB gain stage (and hence its distortion) can be varied. For best noise performance, the output of the first 25dB gain stage should also be at 0dBu which means the second pad must be set to give a 25dB loss. With no other pad in circuit, this gives a channel gain of 70 - 25dB = 45dB. So, if the input signal is high enough to need less than 45dB gain, you can choose to drive the first stage harder and introduce some ‘tube tone’. The table below shows the typical distortion versus output level for the first 25dB gain stage and the corresponding minimum input level required.

 

 Level (dBV) Distortion (%) Minimum input level (dBV)
 0 0.02 -43
 +6 0.04 -39
 +12 0.08 -33
 +12 0.2 -25
 +26 0.4 -19

 

Note: the level figures in the above table are in dBV (0dBV = +2.2dBu)

With a -19dBV input and no input pad, the first 25dB gain stage output is +26dBV and so produces about 0.4% harmonic distortion which is almost entirely second harmonic (all other harmonics are at least 20dB lower). Switching in a 20dB pad at the input lowers the output level to +6dBV and lowers the distortion to 0.04%. Altering the second pad preserves the output level.

Passive Mixing is used as it avoids NFB

NFB can be used to provide a virtual earth mixing point which has the benefit of excellent channel separation (low cross-talk) provided the open loop gain is high enough. Virtual earth mixing with tubes is inappropriate because:

  • Tubes cannot achieve a high enough open loop gain to achieve low cross-talk.
  • Achieving low frequency stability is difficult.

Op-amps have plenty of open loop gain and so can provide a very good virtual earth. However, virtual earth mixing with op-amps is not without its problems.

  • Bus capacitances, from all the cabling and PCB tracks between the channel amplifiers and the virtual earth amplifier, appears across the input of the virtual earth amplifier. This has the effect of eroding the phase margin and can lead to instability and even oscillation. This capacitance reacts against the feedback impedance to to increase the closed loop gain at high frequencies. Even a few pF is enough to tilt up the closed loop response well within the open loop parameters, threatening instability. In a real mixer with cables from many channels, hundreds of pF may be present. This makes ensuring the required phase and response characteristics very difficult. Sometimes a small series limiting resistor can be added to to define just how much this unwanted gain can rise, but this is at the expense of the ‘virtual earth’ now being determined by this resistor (which rather defeats the object).
  • The positive input of the virtual earth mixer op-amp is connected to ground. It therefore amplifies the signal present at this input, namely ground noise. You might think this is no problem if the virtual earth op-amp is arranged to give unity gain. Unfortunately, the gain as far as the positive input is concerned is N+1 where N is the number of channels connected. For a 32 channel mixer this gain is 33 or about 30dB. This effectively makes the virtual earth mixer much noisier than expected, a problem that can only be corrected by care and attention to the overall grounding scheme.

I am indebted to Steve Dove and his excellent series of articles on mixer design in Studio Sound for the above insights.

Needless to say, passive mixing has neither of these problems.

All modules are hard wired in screened enclosures

Tubes are high impedance devices. As such they are susceptible to stray magnetic and electric fields. For this reason it is essential they are contained in what is essentially a sealed metal box. You might expect this to cause temperature rise problems due to the heat from the tubes being unable to escape. However, a prototype channel module containing three 25dB gain stages generates less than 17 watts of heat. Tests have shown that, in a sealed die-cast box, this causes the internal temperature to rise by no more than 25°C. Even in an ambient of 30°C, the internals of the channel module would not exceed 55°C. A prototype channel in a die-cast box has been powered continuously for over three weeks and has become only luke warm to the touch.

 

Design Philosophy (part two)

Mu followers are excellent building blocks for relatively simple mixers and they are ideal for small mixers with a minimum signal path. However, their drive capability is limited so AUX sends and direct outs are not really really possible (you can have one or the other but not both). Their other disadvantage is their gain is fixed so you need attenuators at several points in the signal path to manage the overall gain. To address these problems a new design with variable gain and a high drive capability was needed. Both these features can be provided by negative feedback (NFB) so it was necessary to investigate how the limitations of NFB in tubes could be overcome.

To provide additional drive capability usually means increasing quiescent current. However, the mu follower is a single ended topology which is partly what limits its drive capability. Changing to a push pull topology would double the drive capability with the same quiescent current. Push pull usually means transformers but instead we can use a variant of the mu follower called the SRPP which achieves the same end without a transformer. This provides the required drive capability but has less gain than a mu follower and a lot more distortion. So some extra gain is needed along with some negative feedback to reduce the distortion in the SRPP.

For the extra gain a triode stage is added before the SRPP. We now need to close the loop, maintain stability and make the gain variable over a reasonable range. To maintain stability we must first close the loop at dc. We can then alter the NFB to vary the gain but we can only do this over a limited range before the amount of NFB leads to instability. That is why many of the classic tube microphone preamplifiers like the REDD 47 only allow their gain to be varied over a small range. The solution is the vary the gain of the first stage at the same time as the NFB is changed such that the amount of NFB remains nearly constant and stability is never an issue. This has already been achieved in the famous Telefunken V76 amplifier but even that did not manage to close the loop at dc.

However, because the mu follower and the SRPP both need elevated heaters for proper operation, it is possible to raise the cathode voltage of the first stage which allowed the NFB loop to be closed at dc. The resulting preamplifier has a gain variable from 6dB to 40dB, has a near constant 20dB of NFB and is unconditionally stable. As the desion uses one and one half double triodes, a pair of preamplifiers is built on a single Eurocard PCB. One preamplifier is fitted with an input transformer and switches for phantom, phase, 20dB pad and mic/line switching as well as a rotary switch providing gain setting from 26dB to 60dB making it into a self contained microphone preamplifier with excellent drive capability. The second amplifier is used as a post fader EQ gain make up and bus/direct out amplifier. The PCB is capable of driving a pan pot and several AUX sends as well as providing EQ and fader gain make up and a direct output. It forms the basis of all current channel amplifiers.


Comments