Last modified: 17 Jan 2012

HIS PHONO PREAMPLIFIER for magnetic pick-up cartridge (of vinyl record player/turntable) is my 'slightly improved' version of a design by Robert Danielak – for the original see his web page Bob's Vacuum Tube Phono Preamp – so I do not claim any originality for it. It is, however, the best sounding phono preamp, valve or otherwise, I've heard that I can remember, and is what I am now using in my own system...

The design has the following main features:


1.  The use of Octal instead of miniature (Noval) valves (with benefits beside that of simply avoiding the over-heating problems of miniature types, see later);
   
   
2.  Parallel-twin triode input stage for low noise;
   
   
3.  Passive RIAA equalisation, but which is split into two separate networks, each handling a separate part of the spectrum;
   
   
4.  No following grid bias resistor to shunt the primary network in parallel, which in my experience severely compromises the low-frequency response;
   
   
5.  Not least, elegant simplicity, with only two capacitors in the signal path.

History (sort of...)
The design was originally developed from RCA's phono preamplifier for magnetic pick-up, below, from their RCA Receiving Tube Manual, 1966, Figure 25-17 :–

Compare this with the following :–

Above is Richard Brice's phono preamplifier as first published in Electronics And Wireless World, June 1985. The Maplin Newton was based on this, but far less sophisticated, by main reason of cheapness, and in fact more closely resembled the RCA design.

However Brice's design does have some flaws:


1.  There is too much wiring around the input stage, with increased risk of hum pick-up, exacerbated by the size of the 10nF capacitor to the input of the first cathode-follower, which if a polystyrene type (which it should be for best audio quality), has an unavoidably large surface area. I tried this once and the degree of hum pick-up was unacceptable.
   
   
2.  More seriously, the 1 Meg grid bias resistor for the second ECC83 shunts the output side of the equalisation network, making the lower bass end about 6dB lower than it should be, so that the bass response always sounds 'too weak'. This is easily demonstrated by comparing a vinyl disc against a CD version of an otherwise identical recording. Hindsight, of course, is a wonderful thing, and suggests that this resistor is better moved to a position before the network, at the junction between the first 1µF capacitor and the 250k resistor 'Re', which would solve the problem.
   
   
3.  The circuit is somewhat over-complex, with no less than two cathode-follower stages, and involving 5 capacitors in the signal path, most of which are electrolytics, unless you use very large polypropylenes, with increased risk of hum pick-up/interference, and lack of space to accomodate them.

According to the RIAA standard*, 20 Hz from the cartridge should be 20dB down relative to that at 1 kHz, exactly a voltage ratio of 10, same as up the other way from 1 kHz to 20 kHz (+20dB). I have exhaustively recalculated the network ratios at 1 kHz then 20 Hz, simulating how it attenuates an actual signal at these different frequencies, but could never get a full 10:1 ratio between 20 Hz – 1 kHz. Is more like 5:1, hence the 6dB drop in bass.

One way of curing this is to make it into a negative-feedback network with lots of open-loop gain, then it's no problem at all – however, as World Design also point out, NFB EQ'd amps can have a slew-rate problem, which is why passive is preferred. What's particularly interesting about our Danielak type circuit (diagram at top of page) is the use of a 6SN7 with a relatively low-value anode resistor to get a low-impedance drive to the second (bass part) EQ network, without any need for an extra cathode-follower stage to do this job.


Circuit Description
Both triodes of V1 are connected in parallel, initially to minimize noise. (To read my short diatribe about this click here.)

Furthermore the Octal valve has an advantage over the miniature Noval type by virtue of having much larger area electrodes. The anodes of the 6SL7 are easily 4 – 5 times larger than those of an ECC83. In theory at least, a pair of 6SL7 triodes in parallel should produce much less noise than a paralleled ECC83 pair could manage.

There is another very important advantage. Two valves in parallel have half the anode resistance of a single one, but keep the same mu (amplification factor) figure, so you get more gain than is possible from a single one.

V1 has a mu of 70, but a Ri (internal anode resistance) of 55k / 2, or 27.5k. The voltage gain is a very respectable 50 times, even with the >=1 Megohms load of the following grid resistors of V2. 6SL7 gain is not as much as ECC83, but this compares well with that of a single ECC83 triode with the same anode load.

Resistors R3 & R4 act as low-pass filters, together with the grid-to-anode capacitances, to exclude very high-frequency signals to the grids, also to prevent any parasitic oscillations that might occur between the valves due to their parallel coupling. Similarly, the ferrite bead on the input line acts as an additional RF-band exclusion filter to reject any possible 'radio break-through' type interference on the input.

The main EQ network – R11, R12 & C6 – is not inserted between V1 & V2, as in the RCA and Richard Brice circuits, but after V2. Furthermore, a suitably low-impedance drive for the network is achieved by making V2 a higher-current 6SN7. This is much simpler than adding an extra cathode-follower stage.

To avoid overloading of the V2 stage at high frequencies, due to the signal input increasing with increasing frequency*, being 20dB higher at 20 kHz than it is at 1 kHz, the treble de-emphasis part of the network is moved to a position between V1 & V2 (R6 & C4 in my circuit). The result is two separate networks, each with a simpler job to do. The first (R6 & C4) deals solely with the treble pre-emphasis, and hence ensures that V2 is not over-driven with high-amplitude, high-frequency signals, while the latter (R11, R12 & C6) exclusively deals with the constant-amplitude* part of the recording characteristic. The result is a much cleaner sound.

Note R6 is 220k + 15k, a small amendment to the EQ of the original Danielak version by Jeremy Epstein. Latterly I also included his battery bias for the first stage; this is a 1-third AA ni-cad cell with solder tags.

---

* "The RIAA (Record Industry Association of America) characteristic, which is [also] embodied in British Standard 1928:1955, specifies that a record groove is cut [initially] with constant amplitude modulation, from 20 Hz to 500 Hz. Played back by a magnetic pick-up cartridge, the output signal voltage amplitude then rises at the rate of 6dB per octave, or 20dB per decade. After 500 Hz it is cut at constant velocity – [for which the cartridge output voltage is constant at all frequencies above 500 Hz. This is the original EMI standard. The superceding Decca standard is the same, but with the addition of] – "treble pre-emphasis, which comes in at about 2123 Hz. The net result is a replay curve with an average rise of just below 6dB/octave over the entire spectrum, with a small 'step' at the middle frequencies."

– Gordon J. King, Audio Equipment Tests

---

See also 78 rpm and RIAA Phono Equaliser; Disc Recording Equalization Demystified; Equalization NAB & RIAA.

Low impedance output is provided by cathode-follower V3, which is merely DC coupled to the output side of the EQ network. This means no coupling capacitor between V2 and V3, and no grid resistor to interfere with the behaviour of the network. Besides C2, the only other coupling capacitor is on the output side of V3, as C7. R14 ensures that the output is at earth potential, even if the output socket is unconnected.

One other mod I made to Bob's circuit was to reduce the value of the cathode resistor R10. This is to increase the anode current, and hence lower Ri and maximize the gain. It also reduces the anode voltages to about 100V, which is quite good enough, and makes a more modest voltage drop across R13, which lessens the current drain on the power supply. The value for C5 was chosen on the principle that, in a band-pass type RC network of this sort, capacitor reactance should be one seventh the value of the resistor at the desired band-pass frequency, in this case 20 Hz. Bob uses much bigger values, but in practice 100µF is quite big enough. Note also it is 100V working; this reduces the ESR (equivalent series resistance), on the basis that a higher working voltage also means a correspondingly higher ripple current capability.

Currently the amp is using NOS ('new-old-stock') Brimar 6SL7GT's for V1/101, with ex-War Department CV1988/6SN7GTY for V2 & V3. The signal-to-noise ratio is so good that stylus noise, albeit extremely faint, is still plainly discernible in the very quiet passages of classical music recordings, even though these are on the best-quality and seldom-played vinyl!


Testing The Frequency Response
Bob was able to calculate the theoretical frequency responses for his design, getting within about 1 or 2dB of the RIAA standard, but I don't think he was actually able to test the circuit for real. I did (or I did my version); the test rig I used is illustrated below :–

It assumes the use of an AF signal generator with an output attenuator having decade ranges. Because of course the normal input level of the amplifier circuit is so low, an additional 100:1 attenuator must be used to reduce the final input level. This has the added advantage of virtually shorting the inputs to ground through a low resistance (100Ω) to exclude stray pick-up interference and more closely emulates the connection of a magnetic cartridge.

The starting point is to establish input and output reference levels at 1 kHz. This means setting the AF signal generator to –20dB out (0dB being up to 10V maximum), and the fine control for an output of 600mV. This is reduced by the 100:1 attenuator to 6mV at the input, giving 400mV at the output – these values were chosen just for the sake of getting round numbers.

To compare the gain at the low-frequency end, the generator's attenuator is switched to –40dB (first!), then is wound down to 20 Hz. The output is then 60mV, reduced, post 100:1 attenuator, to 600µV. The amplifier's output is exactly the same as at 1 kHz, 400mV.

The high-frequency test involves selecting 20 kHz, then 0dB output from the generator (6 Volts). This translates to 60mV at the amplifier's input, and is 20dB over that at 1 kHz. This demonstrates the ±20dB ratios of gain either side of 1 kHz, at the 20 Hz and 20 kHz points; the actual outputs achieved are illustrated by the oscilloscope display photos at the bottom of this page (just to prove it...)

One last modification was the addition of Zener diodes (ZD1 & ZD2) to stabilize the V1 stage supply rail. This is because there is no other regulation and our mains supply has a habit of chopping up and down, which of course will be amplified and come out on the output.


Power Supply
To remove the main source of magnetically induced mains hum, the mains transformer is sited in a remote transformer box, and connected to the main unit via a multi-way cable (see photos). The housing includes a miniature cooling fan to prevent resistors and other components heating up too much and thereby generating extra thermal noise or other high-temperature associated problems. The steel case (ex-WAD) is 10.5" wide x 3" high, x about 11" deep.

I already had the fan, 40mm dia. with brushless motor and very quiet, provided it's on reasonably flexible mountings. The power supply uses a VT426, where only the 240V and 15V windings are used. The 15V winding provides the supply for the fan and the front panel lamp, also a stabilized 12.6V DC supply for the valve heaters. These are wired in series/parallel to match, see diagram. Smoothing of the HT supply is by twin series pi filters, using 100µF capacitors with two miniature smoothing chokes VT1314. The low DC resistance ensures maximum HT for the amplifiers (300V). Despite being unregulated, ripple is so negligible that it could not even be theoretically demonstrated by Duncan's PSU Designer.

Rear panel view showing earth post for record player or 'turntable' earthing wire

Interior view from amplifier side. Note fan at top right

There was subsequently a slight change to the power supply, only insofar that the cooling fan and the green front panel lamp now share the same supply point. This runs the fan under-volts (about 8Vdc) thus making it much quieter, and also makes the lamp not so bright.

.	Interior view from power supply side. Note multi-way power connector and miniature chokes at left, heatsink for heater regulator transistor at right

100:1 input attenuator used for testing gain at 3 main frequency points (see text)

.

Addendum – 2009 – 2010

The battery bias for V1 was scrapped and reverted to decoupled cathode resistor (1k + 100uF); the paper-in-oil cap for C2 replaced with ICW ClarityCap 100nF. The blue polyprops in the above photo (first ever used) have also been removed. Instead yellow 1uF ClarityCaps as seen in background.

Remote Transformer Box

The 'remote mains transformer box' is connected by colour-coded, shielded 8-way cable and 9-way connector block

Oscillographs

The output level at 20 Hz... ('scope display nearly too slow for the camera!)	...is (or should be) exactly the same as that at 1 kHz, but whose input level is 20dB higher (10 times)
Same again at 20 kHz with a further 20dB increase at the input. Here output level = 400mV peak ('scope setting = 0.5V/cm)