This is a story about the first phono preamp of Grimm Audio, the PW1. I have designed quite a few phono preamps in my time, between 10 and 20 I guess,  and using all types of technology: transistors, fets and tubes. I also analyzed and modified three or four times that number of commercial phonos for friends and clients. Every new one is a serious challenge, be it when designing a new one myself or when improving a commercial one brought in by a client. It is a well known fact among designers that coming up with a really good phono preamp is no sinecure. For a line stage, for instance, despite the fact that there are many different ways to do it and a host of choices have to be made, ending up with a well performing result is really not too difficult. Even a power amplifier is to a certain extent easier than a phono stage.

So what is so difficult about a phono preamp?

Well, for starters, the phono signals coming from your MM or MC cartridge are very tiny. Most MM cartridges output a nominal 5 mV, but that is for music of average loudness. And at 1 kHz. But at 20 Hz the cartridge’s output is ten times smaller: 0.5 mV which equals 500 µV (microvolt, one millionth of a Volt). For comparisons, most loudspeakers work with up to 30 Volt. But we’re not there yet. Music is not always loud, in very quiet passages it can be 60 dB softer. 60 dB is 1000 times, so now we’re talking 5 µV and 500 nV (nano = one billionth) at 1 kHz and 20 Hz respectively. And with MC cartridges the situation is worse as their output is usually 10 times and in some cases 30 times lower than that from MM pickups. Which means that in the case of MC you’ll have to deal with worst case signal levels of around 300 nV: 0.0000003 V. Again, this is at 1 kHz; for 20 Hz add one more 0! Dizzy already?

To get a better idea how small this is, let’s look at meters instead of Volts. Remember the voltages used to drive a loudspeaker, i.e. up to 30 V?

Now let’s go to meters: 30 m high is the height of a 6 story office building. Now assume there is a human hair lying at your feet as you stand before that building. The thickness of a human hair is about 100 µm, but to describe a low level MC signal you would be looking at 1/100th to even 1/300th of a hair’s thickness! Which even under a microscope would be hard to see. This comparison makes it clear that it is all too easy to corrupt these very very very fragile signals.

So yes, a phono input has to be very low-noise, in the case of an MC input: extremely low-noise. Even for MM inputs the gain must be enormous, at least 60 dB (1000 times) at 20 Hz, and 40 dB (100 times) at 1 kHz. In the case of MC add 20 to 30 dB, which means you’ll end up with something like 90 dB @ 20 Hz (30,000 times!) and 70 dB @ 1 kHz (3,000 times), necessary to create signal levels at the phono pre’s output that a line stage can handle.

But then why is the signal from your cartridge ten times softer at 20 Hz than at 1 kHz? And it doesn’t end there: at 20 kHz the input signal level is ten times louder than it was at 1 kHz! The short answer is: this was done to make disc cutting easier. Also to end up with a better signal-to-noise ratio when retracting the music signals from the relatively noisy surface structure in the spiral groove of an LP. In 1955 record companies agreed on a standard for this, proposed by an institution called the RIAA. As a consequence, if you want your music to come out of the phono preamplifier with a normal, i.e. flat frequency response it is necessary to add a correction network called RIAA equalisation.

Which takes us to the next problem. It is not too difficult these days to calculate a correct RIAA network, accurate to say within ±0.1 dB. Many textbooks offer the right formulae and several internet sites offer a spreadsheet-like calculus. However, I have seen quite a few examples over the last few years of amps with serious RIAA errors, say several dBs off at some areas of the frequency band. I even came across Chinese designs with deviations as large as 10 dB at the frequency extremes which is totally unneccessary, totally unacceptable, and totally ridiculous.

But even if you do get the frequency correction right, it matters a lot how you realize the RIAA correction in your circuit. There are many ways to do that and all of them lead to different overload margins and output noise performance performance. And sound. Let me sketch a few options.

By now I still haven’t even covered half of the problems that pop up when designing a phono amp. I won’t go into all the details but let me just sum up a few more topics on which choices need to be made, and potential problems exist: active components (transistors, fets, opamps, tubes), passive components (capacitors, resistors, switches), power supplies (local and remote), distortion, printed circuit board layout, overload margins of up to 26 dB, … Anyway, you get the picture.

So how to tackle these problems? On the one hand, there is the purely technical, scientific approach. Strive for minimal noise, minimal distortion and maximum overload margins. This is the road followed by many hi-fi companies. Their specifications look very reassuring, impressive even. But often this is the result of brute force, not elegance. And in my experience this does not offer a guarantee for good sound, although some hit it right and end up with a refreshingly sounding result, I frankly admit that.

Another approach is the one often seen with tube equipment manufacturers. The designer will have to accept slightly more noise, certainly more distortion, but is blessed with ample overload margins so there will never be tearing distortions due to clipping. Accurate RIAA curves can be a problem though, especially with sample-to-sample tube differences. The specification picture may look somewhat less convincing, but it is accepted knowledge that you’d have to be a total idiot as a designer to make tubes sound bad. So that’s a big plus too. And some tube phono amps sound absolutely gorgeous, though usually at a price.

So what did I do when designing Grimm Audio’s phono preamp? Because of the seemingly endless number of choices that need to be made when starting from zero, you’ll have to follow a few hunches (and hope you’ll indeed end up where you wanted to go). You just cannot try all possibilities, right? So, for starters, I decided not to use bipolar transistors in the first part of the signal chain: in my experience too crude for very low level signals. Subsequently I made a choice for two parallel routes: opamps and fets. And in each case I tried to avoid over-asking, over-stressing the active components. And I paid a lot of attention to the power supplies. After more than one hundred hours of study, measurements, prototyping and listenings tests we decided to drop the opamp itinerary. Despite all the effort it didn’t hold up to the 3-fets-per-channel approach that slowly had crystallized in a parallel universe. The latter was simply more seductive, more natural, and so musically involving. Here a picture of the first PW1 prototype, finished 2017-2018. 

Over the years I have increasingly become a supporter of the KISS principle: less is more. In doing that, you’ll have to go with the flow. I mean, you’ll have to accept the item you use, be it a fet or a transistor or an opamp, for what it is. Because, once you start to try and correct its peculiarities, you’ll be applying force and complicating things which may, and probably will, lead you away from your goal. So the approach chosen is almost like identifying with a fet and looking up with it at the power supply, and looking down to how and where it attaches to the ground rail. Similarly, when crawling inside the RIAA network you’ll be able to look around and get ‘feel’ if it ‘sits’ comfortably in the surrounding circuitry. Which, I hasten to add, does not mean that you would throw overboard what science and electronic theory and datasheets have taught you. On the contrary, you’ll need all the technical insight and data you can get, to know where you are and what you’re doing!

As an example of which I found it necessary to make one exception to the KISS rule: I let one bipolar slip in! Why? To eliminate two unwanted properties of fets that would be detrimental to the sound quality (yes, fets are nice but far from ideal, alas). But not through force, the bipolar is used here in an almost passive function (called cascode, no signal gain) and hence not messing with signal quality.

Yes, you could say my approach is to a certain extent like Zen. I remember when someone gave me a book called “Zen and the Art of Motorcycle Maintenance” by Robert Pirsig, it must have been in 1981 or so. I read it and at first I understood not one iota of what it was about. Despite the fact that I had been riding motorcycles since 1970 (always Italian!). But my narrow-minded Western scientific training was still the dominating factor in my thinking, both about technology and life in general. It took a few more years of life experience, dealing with frustration, pain, sadness and death. And then, by the end of the eighties, I picked up the book again and it was crystal clear! You had to try and see your motorcycle as a living being, treat it with love. And riding the bike became different too: I did no longer just sit on it and control it, I became one with my motorcycle during each ride.

So yes, some electronic components that you use in hi-fi may not do exactly what you want. But if you accept that and work in the line of their strengths and gently address their weaknesses , you can end up with a beautiful result. As one Chinese dentist once said: “Don’t fight the pain, pain will fight back. Go with the pain, pain will get confused” (cit. Doeschka Meijsing, De Tweede Man, 2000).

So why this approach? Because I want a sound that is as natural, as alive as possible. And this is where KISS leads the way, for me at least. In my experience, the more complex you make your circuitry, even if this is done to eliminate the circuit’s harmonic distortion or whatever, the greater the risk that you will hear the electronics involved. So while other amplifier circuits may provide you with a tool that dissects a recording down to every technical detail (useful in studios, I admit), in my designs I firstly go for conveyance of emotion, the warmth of a human voice. I do keep an eye on harmonic distortion but rather on its character (I avoid uneven harmonics) than that I would strive for a very low percentage. And I go for the seductive tone of a violin, the air around voices and instruments (which restore their 3D reality), the width and depth of the stereo image. If I get all that, it makes me feel overwhelmed, and happy.