[ TNT | Who we are | Listening tests | HiFi Shows | HiFi Topics | Tweakings | Inter.Views | Termoionica Applicata ]

Termoionica Applicata DIY projects

MW Pre 01 - Line Pre-amplifier with MM Phono Input

Uncommon Implementation of a Classical Design

A sophisticated grandson of the SIMPRE

Part 3
[Italian version]

Audio Tubes Filament Power Supply

The simpler SIMPRE filament power supply was expected to require an enhancement in order to cope with the hard job of powering a phono stage. Actually, while testing the phono circuit with the SIMPRE power supply I discovered that no mains buzz was audible... so the original filament power supply remained in place.
The filaments require about 1.5A (from 1.2 to 1.46A, depending on the specific tube used) at 6.3V.

HT Power Supply

A tube rectifier PSU, with a double LC filter has been used. In this way rectification is more soft, produces less spurious noise, has a relatively low output impedance and good efficiency. The filter has a very low value input capacitor, in order to achieve the high output level of capacitor input filters, while reducing peak current in the rectifier to a minimum.
As far as I remember (I designed the PSU about 10 months ago...), the capacitor value is so low as to allow powering up the pre-amplifier with a hot rectifier tube, without causing a current peak in excess of the rectifier's capability.
The ripple at the output of the first capacitor is obviously rather high, but the double LC filter following it cuts it down to a ridiculous amount (about 20uV). The whole PSU unit really works somewhere half way between an L input and a C input PSU filter.
All capacitors in the PSU are rather high quality electrolytic components (Philips and Kendeil K-05); if possible, choose automotive range components, with an operative temperature up to 105 C. These will be probably be placed by the rectifier tube, which gets really hot...
In order to get an even better power supply regulation and even lower PSU impedance at utilisation nodes, a further set of PSU bypass capacitors has been placed as close as possible to each line stage device. Note that in line stage there is no resistance between PSU output node and these capacitors the only impedance between them is the connection wire one. Special care has been taken in selecting high quality capacitors for this usage (ROE).
The phono stage has its onw RC filters placed close to each triode, with a high value resistor and a high value capacitor; the main reason for them is that a lower supply voltage was required to drastically reduce tubes noise. As a side effect, any PSU output 50Hz ripple is reduced by further 43 dB (about 0,2uV....).

Phono stage

As you know, a phono stage must not only amplify the signal, but also must have a special, nowadays standard frequency response curve named the RIAA.
For the sake of precision, in the past there were plenty of different curves, as different companies used a curve of their own. For this reason, once upon a time professional phono cartridge pre-amplifiers had either a variable frequency response or could be adapted to different response curves by changing the de-emphasis circuit, which normally was mounted on a socket.
At the present time, as only one curve is used in modern records, only few hi-fi phono pre-amplifier have such a feature.
The need for a de-emphasis curve is due to the physics of the recording-playback process and the characteristics of the recording media. In fact, all frequencies have approximately the same peak energy in a normal musical program (this is a very rough approximation, or better a nearly false statement... but let's suppose it to be true), so you need a recorded signal to cover the same dynamic range independently from its frequency.
Hence you would expect that all signals were recorded at the same level on the record. Unfortunately, the playback device output is proportional to the variation in speed of the groove modulation depth. But not the groove modulation depth itself (for the sake of precision, a stereo pickup does not really detect grove depth, but each channel is recorded on one side of the groove, within a 45 angle with respect to vertical).
From a mathematical point of view, the pickup output signal is the derivative of the groove depth signal. Now if we compare the derivative of two sinusoidal signals, with the same amplitude but different frequencies, let's say one double of the other, we find that the amplitude of a higher frequency derivative is double the frequency of the lower one. And this is a general rule.

Let's look at this another way. From a practical point of view, this means that for the pickup, to give the same output level at two different frequencies, one twice the frequency of other, the derivative of the two recorded signals (the slope of groove modulation depth) must be the same. This can be accomplished only if the amplitude (the groove modulation depth) of the lower frequency signal is the double of the higher one.

As the human ear can hear sounds from 20Hz to 20kHz, in order to get the same electric signal from a pickup the amplitude (vinyl surface depth modulation) of a 20Hz signal should be one thousand time larger that the amplitude of a 20kHz signal, which would be just a little impractical. It would reduce the 20kHz signal level under normal vinyl surface noise and require a groove depth and width (remember: the signal is recorded on a 45 degrees plane...) of the order of millimetres....
The solution used long since was to apply a pre-emphasis curve to the signal before recording, and then apply a symmetric de-emphasis curve at playback, so that the amplitude of the recorded signal can be almost independent from frequency for the same energy.
This is why a curve is necessary. Currently there are only two standard playback curves to be taken into account, which by the way only differ at low frequencies at or under the lower end of human ear sensitivity range. The two curves are RIAA and RIAA-IEC curves.
The second one, which should be the current official reference curve, differs from the first only as far as the first is not defined for frequencies below 20 Hz. The second has a subsonic filter characteristic in this area around and below 20Hz; for higher frequencies the two curves are practically the same.
From the point of view of the implementation, the phono stage design is a very simple and direct one. The configuration used is called passive RIAA, as the RIAA filtering action is completely in charge of a passive (in our case RC) network. The passive network is set between two tube active (gain) stages.
This configuration produces a couple of good things and a lot of problems. The two good things are that the passive RIAA sound is normally considered the best RIAA sound you can listen to, and that the configuration is not so difficult to set up. The major problems are:

The first three points make it necessary to choose a high gain, low internal impedance tube for the first stage... and if you look through data sheets for such a tube you'll really be surprised to find such a small number of tubes with these features.
The search leads to a couple of well-known tubes, which have relatively low internal impedance and relatively high gain: their selection too is therefore the result of a compromise. The two tubes are Obviously I chose the second one, as I like its sound, but mainly for the reasons stated above. The real problem, not completely solved up to now, has been the selection of a really good type, currently available and in production, not terribly expensive, with low microphony and low noise... as far I can see the lowest noise versions have a rather high microphony.
Both phono stages are common cathode stages with automatic bias and non-bypassed cathode resistor. They have been biased in order to get a low anode voltage and a rather high current through them. This working point has a lower internal impedance. This reduces the dependency of the RIAA filter's frequency response on the tubes characteristics. Also reduced is tube thermal noise induced by internal resistance. Even more important for tube noise reduction is the low supply voltage.
The design makes use of three slightly different working points for the three stages of the pre, avoiding non-linear behaviours of each stage. Both stages are quite delicate and sensitive to noisy tubes. In fact, a 4mV, 1 kHz input signal is amplified up to 100mV at the output of first stage.
The passive RIAA network cuts it down by about 20dB, so that we find 10mV at the input of second stage, which very roughly shows that the second stage is only 8dB less sensitive then first stage, to tube noise around 1 kHz. In fact, I had to modify the design of second stage after testing for noise problems, reducing the anodic voltage.
Anyhow, in order to have the same pre output level as above at 20Hz, an input roughly of the order of 0.4mV is required. On the other hand, in order to have the same pre output level at 20kHz, an input very roughly of the order of 40mV is required. If we suppose an even first stage noise distribution through the full audio frequency range the signal-to-noise ratio at low frequencies is far worse than high frequencies. From a practical point of view, the phono input and first stage thermal noise appears as an essentially low frequency hiss and rumble.

Second stage noise contribution on the contrary is not dependent on frequency, so that any accurate comparison of the noise contributions of each tube is not really as easy as stated above.
Before talking about the RIAA curve precision, we must talk about its components. I have adopted a solution that is apparently not as common as I thought. I used paper and oil capacitors in the RIAA network. Well, this practice seems to be widely accepted in Italy, but when I talked about this with our friend Thorsten he was really upset: "never heard of!".
The sound you obtain, at least with the components I used, is extremely fast, neat, so brilliant that you might think that frequency response is not accurate. That is until you apreciate that it is only, simply, absolutely and definitely transparent detail is notable. While it may seem crude, there is definitely not a "vinyl" sound.
I was really doubtful about this solution. But, during last summer I was able to spend, after alas a too long time, some time listening to a concert. It was apparent to me that many analogists believe the best reproduction of musical events is the smooth, rolled off, dumb sound.
These analogists use these differences to show the absolute superiority of vinyl over digital sound. Well, this is less than incorrect and is definitely inappropriate. In one word unnatural. Live music can be crisp, brilliant, ear-drilling, detailed and pinpoint in precision.
This does not mean that digital is better than vinyl, just that you must try to get all you can out of vinyl with the best, the most transparent "tools" you can put your hands on, if you really want to try to recreate the original event.
I long ago tested a few capacitors and found out these paper and oil ones were better than any other film capacitor I tested. But, I must say that this depends probably a lot on my personal taste. As for others, they might really sound inappropriately brilliant and disappointingly cold. You do loose a good deal of the typical "good old tube sound". If you like it, use film caps. I'll go on using paper and oil whenever and wherever I can (an even when I cannot, as here...).

Apart from sound philosophy, the problem with these capacitors is that they are not inexpensive (even though the ones I used are not at all very expensive) and they have a 5% tolerance. While the cost mitigates against buying some values, I was able at least to pair them in order to obtain a similar frequency response curve on both channels.
In order to get a precise curve I had been initially working on the network with a simulator, taking into account all the three stages. I decided in the end to implement the old RIAA curve, instead of the RIAA/IEC one with the better low frequency response.
The reason will be clear when we talk about the line stage. Also the loudspeakers I use (Snell K III, very fast, but with definitely not bags of bass) had an effect on this decision. Note: this is not the current standard IEC curve, so new records could have a low frequency boost...

When the prototype was ready I deliberately disregarded any frequency response verification, in order to see if it was possible to evaluate the preamp from just listening tests. Unfortunately, during my listening tests (months...) I became aware that the frequency response was probably not correct.
Instrument tests confirmed the problem, and I had to work on components to get the correct response from the filter. The values listed below are the final values I used after further listening tests, which also are in line with simulation results.
If you have access to the necessary instruments, you could check if the frequency response of your implementation is correct. Feel free to change anything to get a better adherence to either RIAA or also RIAA/IEC curve, as the values I used could be affected by the paper and oil capacitors tolerances. I had no time to make tests with different capacitors.

Note that it is almost impossible to design such a filter with a good degree of adherence to a given curve using only standard component values. It is may be necessary to use more parallel or series components.
Anyway, from a practical point of view, strict adherence to accuracy should be considered vital. In my opinion, you must definitely make the frequency response of the two channels as similar as possible, but if I must choose between a good, pleasant transparent sound, and absolute precision... well, I choose the sound.
In the end, if you want to take care of RIAA precision, you can proceed to measuring the frequency response of the circuit and changing the components in order to get as near as possible to the correct RIAA curve.
(A note for those who can make this kind of measures: the best solution is to build a reference anti-RIAA passive filter. Drive the frequency generator output test signal through the reference filter into the phono input and measure the pre-amp output value. The value should remain at the same level for any frequency).
If not, you can simply couple the capacitors by measuring them. Take into account that 10 paper and oil capacitors are required in the pre-amplifier and 6 of these are .33uF It should not be difficult to obtain two well matched pairs; obviously you will then have to mount half of each pair in the same position of each channel, and mount the others as coupling capacitors. The remaining two are 0.1uF. If you do not want to buy several of them, you could try asking your supplier to match them for you....
I used ICAR PR20 paper and oil (actually film and oil, as far as I know) capacitors, currently discontinued, as I was recently told, and almost impossible now to find, as it seems. They have 5% tolerance. I would expect an even better result with Jensen or Audio Note ones, but costs would be far higher... All resistors are Holco's.

Go to Part 4

© Copyright 1998 Giorgio Pozzoli for TNT-Audio, https://www.tnt-audio.com

How to print this article

[ TNT | Who we are | Listening tests | HiFi Shows | HiFi Topics | Tweakings | Inter.Views | Termoionica Applicata ]