Help! JLH RIAA pre-amp

[mole42uk]

I'm using a JLH modular pre-amp with his RIAA stage to connect my Shure M75EDII cartridge. The output of that cartridge is 5mV but the gain of the RIAA stage is too low. For a measured 5mV rms input at 1kHz I'm getting only 70mV out and I really need 500mV.

Can anyone tell me where I've gone wrong?

[paulsherwin]

Are you sure that the first stage is correct? The input arrangements look completely wrong. (I don't know this design though, and may be misinterpreting it - JLH did some strange things.)

[mole42uk]

I've just been checking back to the original 1969 article and yes, the circuit is as designed.

Looking at the article, he says that a stage gain of 10 is adequate for the RIAA input, and so it seems that my iteration is correct - 5mV in, 70mV out. Certainly, testing last year proved that the response curve with the components indicated is pretty close to the ideal.

The discrepancy lies in that the other inputs (tuner, Minidisc, CD) are either 500mV or 2V and so I expected that the output of the RIAA pre-amplifier would be somewhere around the 500mV mark.

[Ted Kendall]

Originally Posted by paulsherwin

Are you sure that the first stage is correct? The input arrangements look completely wrong. (I don't know this design though, and may be misinterpreting it - JLH did some strange things.)

You can hear the bee in his bonnet buzzing yet - JLH insisted shunt feedback was the only way to implement the RIAA curve properly and that the 14dB s/n penalty was a price worth paying. Since a simple additional pole at the output of a series feedback stage would give an equivalent result, his adherence to this method is a bit of a mystery.

[paulsherwin]

I would agree that a gain of 10 is much too low for a mag cart preamp feeding a line level output. Maybe he designed it for a particular amp with a more sensitive input. 1969 is quite a while ago now, and amps were less standardised than they became in the 70s.

I still don't understand the circuit, but JLH was a much more skilled designer than I am so presumably knew what he was doing.

[mole42uk]

I've been measuring the remainder of the design, in particular the ac voltage levels. To achieve an output of about 550mV which is required to drive the 1969 Class-A power amplifier design it seems that much more gain is needed in the RIAA stage.

I'm wondering if the three-transistor design can be tweaked to increase the gain, that concept goes beyond my knowledge at this time.

[Ted Kendall]

The feedback fraction is set by the ratio of R9 and R10, so changing this would reduce the level of feedback and increase output level, assuming that the amplifier has sufficient open-loop gain to maintain curve accuracy and low distortion.

[mole42uk]

I suppose it's time I got to grips with LTspice!

[Ted Kendall]

What you could do as quick and dirty check is to split R10 into 820 ohms plus a 1K preset and tap off from the slider to the feedback network.

[mole42uk]

The output only gets up to 200mV at 1kHz and the curve goes much higher at 50Hz and almost disappears at 10kHz....

[Ted Kendall]

Running out of loop gain, then, methinks. Maybe put a 20dB gain stage after it or start afresh with a couple of 5534s...

[Radio Wrangler]

Originally Posted by paulsherwin

Are you sure that the first stage is correct? The input arrangements look completely wrong. (I don't know this design though, and may be misinterpreting it - JLH did some strange things.)

You're righter than you know, Paul.

This is an issue which used to be fairly famous.

Linsley hood had a number of odd ideas and bees in his bonnet. His RIAA circuit started a slightly heated debate in the pages of Wireless World that rumbled along for a while. Linsley-Hood thought that shunt feedback would be better for an RIAA stage, the other side was represented by Hugh Walker, an old colleague of mine, and a very very good engineer from whom I learned rather a lot.

The nub of the argument was noise. JLH said the shunt feedback arrangement was quieter. Hugh did a formal mathematical analysis to show the series feedback arrangement was better. The difference is small, but definite and you might as well pick the optimum.

Looking at the JLH circuit you will see that the second transistor is used as a unity gain inverting stage. It is only doing a 2:1 impedance buffer job as well, so on the gain front it isn't exactly earning its keep. As a result the circuit is a bit marginal on open loop gain and to try to keep reasonably close to the RIAA curve closed-loop, closed-loop gain is sacrificed and it's a bit of an underperformer on sensitivity as a result.

This was all in the days before Spice was commonly available. Hand-cranked analysis by algebra was the order of the day. Hugh has a background in very low noise system design (measuring instruments) involving rationalising stage gains and noise contributions through systems to carefully optimise overall system performance.

If you want to see his three transistor RIAA stage, it's in Wireless World in his stereo mixer article. It makes a bit better partitioned use of its three transistors, isn't so tight on gain and is a little lower in noise.

It would be easy to just support a friend in an automatic, unthinking way, but the maths is there spread across the letters columns: John Linsley Hood and Hugh P Walker. If you follow the derivations through to the conclusion, Hugh's point is correct.

A later RIAA design in WW was the one Peter Baxandall (yes, him!) did with the NE5534 and included the missing pole added on the tail end. This is a thoroughly good design. The NE5534 can be bettered by a discrete design in terms of noise, but the difference is very small for a moving magnet cartridge.

So I'd just oik the JLH shunt feedback RIAA stage out and stick in either Peter Baxandall's NE5534 design, or Hugh's three transistor job.

David

(PS, Hugh was jocularly referred to as 'Wee Shuggie' by his friends - over 6ft tall and built like a rake)

[paulsherwin]

I still don't understand how it's supposed to work. What's that 47k resistor in the input doing?

[mole42uk]

Thank you David, now I can be sure that I am not missing something fundamental, just trying to make something work that isn't able to. When I tried a prototype of the JLH circuit I measured the frequency response but not the gain. With a little tweaking the RIAA response is pretty good but there's nothing to be done about the gain. Shame, I rather enjoyed putting it all togather.

[Electronpusher0]

Like Richard I built the JLH pre-amp and i confess I have not tested it other than on the bench and have not checked the amplitude of the output, must do so!

I did however model it in ltspice.
Attached is the plot of the RIAA stage with the rumble filter and again without the rumble filter.

Peter

What's that 47k resistor in the input doing?

The point it connects to is a virtual earth, it gives the cartridge the right load.

[Radio Wrangler]

Originally Posted by paulsherwin

I still don't understand how it's supposed to work. What's that 47k resistor in the input doing?

It's doing two things.

The amplifier is using feedback around the loop to try to cancel the signal voltage at the base of the first transistor. That base becomes what's called a virtual ground.

The cartridge makes some voltage and wants to see a 47k ohm resistor and so many pf as a load.

The input parts provide this, 47k and a small capacitor. The cartridge expects to see them to ground, but instead they are to a virtual ground. The cartridge can't tell the difference so it's happy.

The cartridge voltage across the 47k resistor (etc) injects current into the first transistor, and the amplifier amplifies this until the output, via the RIAA resistor/capacitor network starts to inject an equal, but opposite current into the first transistor base.

So we can say that the gain of the system, if the transistor gains are high, is equal to the impedance of the RIAA RC network divided by 47k.

So the 47k resistor is doing two jobs. It's the load resistor for the cartridge and it's part of the feedback gain-setting network.

A deeper analysis shows that the loop gain/phase (needed to be OK for stability reasons) is influenced by the impedance of the cartridge, which is inductive and has various resonances. So the designer has to go easy on the gains he can have in this amplifier and still keep it sane.

The choice of shunt feedback is one of those design decisions that seem made in order to enable claims to be made, but in reality the devil is in the details, and a proper analysis shows that the small distortion it is expected to deliver is trivial and that its disadvantages are a net home-goal.

David

[Ted Kendall]

Originally Posted by mole42uk

Thank you David, now I can be sure that I am not missing something fundamental, just trying to make something work that isn't able to. When I tried a prototype of the JLH circuit I measured the frequency response but not the gain. With a little tweaking the RIAA response is pretty good but there's nothing to be done about the gain. Shame, I rather enjoyed putting it all togather.

Well, it has many of the bits of the Walker three-transistor jobbie, but not necessarily in the right order. The classic three-transistor layout was a two-transistor gain stage with an emitter follower to buffer it from the load and drive the feedback network from low impedance. Perhaps a re-work would be less laborious than going to ICs.

[paulsherwin]

I'm glad you understand that David, because I certainly don't. It just looks like a 47k resistor in the input signal path to me. I'm happy to acknowledge my limitations though. There's a good reason why I didn't become a professional design engineer

[Radio Wrangler]

It IS a 47k resistor in the input signal path. So it should look like one.

Just accept for a moment that the amplifier works to hold the base voltage of that first transistor fixed. The cartridge signal voltage is therefore across the 47k resistor. So we use Ohms law and that tells us the current the input is injecting into (or sucking from) that base. The transistor amplifies this current and swings the output of the amplifier. The changed output voltage of the amplifier gets applied across the RIAA network (the two resistors and two capacitors bit) We can apply Ohms law to the impedance of the RIAA network (remembering that that base is playing at looking like ground as far as AC signals go) So we know the current in the RIAA network.

The trick is that the amplifier slews the output until the RIAA network current cancels the current from the cartridge (the current set by that oddly placed 47k resistor) this gives the first transistor no net input current as the two cancel.

We can say that the current in the RIAA network must therefore be equal and opposite to the current in that 47k resistor. We know the current in the RIAA network and its values, so we can calculate the voltage across it. Knowing that there is no signal voltage on the input transistor base (because it all gets cancelled) the voltage across the RIAA network is therefore equal to the output voltage.

It's one of those things you've got to look at from several directions until it clicks, Paul.

It's called shunt feedback, but you may find more explanations if you look up "Inverting amplifier" in terms of opamp circuits.

The series feedback arrangement will be found nearby called "Non-inverting" amplifier.

If you haven't twigged these two, you're missing something that could make a lot of other circuit tricks suddenly drop into place for you. It's worth persevering. Nothing difficult is involved, but the concepts are odd the first time you come across them.

David