Valve amplifier phono input

[PJL]

I am digging my 1970's hi-fi kit out of the boxes after about 25 years having recently received and restored a Beam Echo amplifier set up of Avantic SP21/DL7-35's. I now want to connect my record deck which uses a Shure V15 III.

The SP21 claims RIAA equalisation (C3/C4 selection) but I have no idea what cartridge that was intended for. It appears to have separate high/low impedance inputs but the output impedance of the source would appear to affect the feedback. Also the CR's do not match the RIAA requirements being 100uS and 6000uS.

I am just about OK with the all-in-one RIAA active feedback network in opamps where the minimum unity gain pretty much determines the gain conditions. However, the transfer function and impedances for valves introduce a whole new set of problems. My guess is a single EF86 is not going to be up to the job but it would be good to prove this. Would LTSpice be the way to go and where can I find an EF86 model (I have only played with the standard product).

Option 2 would be to abandon the RIAA compensation in the SP21 and go for a homebrew. Can anyone recommend a valve phono preamp circuit?

[GrimJosef]

I haven't looked at this in any detail, but I'd make a few points. First I think the RIAA equalisation is about correcting for the frequency characteristics of the record and is a separate issue from the properties of the cartridge. Second I think that a single EF86 was quite commonly used for a phono stage and was, at least by the standards of the time, considered 'up to the job'. Third (and you may well have done this) my experience of some of the capacitors in the tone and frequency circuits of vintage pre-amps is that over the years they can have drifted a very long way away from their original values. Unless they are of the more stable type (polystyrenes, for example) it would be worth measuring them to confirm that they are still OK.

Cheers,

GJ

[PJL]

The SP21 is a very high quality build with 5% HS resistors in critical places and 5% polystyrene capacitors. All paper capacitors, out of tolerance HS resistors and all carbon resistors (usual drift problems) have been replaced (so pretty much everything!).

Do you know of an example using an EF86 I can compare to?

[mole42uk]

Originally Posted by PJL

Do you know of an example using an EF86 I can compare to?

The Heathkit S-99 uses an EF86 in the phono circuit, I use mine with a Shure M75 EDII.

The Mullard 2 and 3 valve pre-amplifiers used EF86 for magnetic pickups.

The Maplin Newton used an ECC83:

[mole42uk]

Originally Posted by PJL

Do you know of an example using an EF86 I can compare to?

Found the Mullard circuit:

[GrimJosef]

Good news about the polystyrenes .

As far as the EF86 stage is concerned the Beam Echo circuit seems essentially to be the same as that used in the Mullard 2-valve and 3-valve pre-amps. I can't find an on-line link to the original 'Circuits for Audio Amplifiers' article, but there's a re-worked version for the 2-valve one here http://vintageradio.me.uk/amplifier/mullardpreamp.htm. Of course this doesn't use a single EF86 - it uses two. But your Beam Echo unit has a second stage too (an ECC83 in this case) and with luck they will have extracted enough gain from that to give you the sensitivity you need for the V15. I'm not surprised that they've used at least parts of the Mullard pre-amp circuit. I believe the DL7-35 is essentially a Mullard 5-20 isn't it ?

An alternative phono stage using an EF86 is the one used in the Quad 22 control unit, again with an ECC83 after it. There's a pretty thorough description of this here http://www.keith-snook.info/quad-22-pre-amplifier.html in which he points out that you may have to be a little careful with the apparent RC time constants if the circuit is loaded by Miller capacitance (the Mullard article mentions steps they have taken to reduce the Miller effect).

Cheers,

GJ

[Diabolical Artificer]

Heres a couple using alternative valves fron the EF86.

Andy.

[PJL]

Thanks, lots of interesting stuff. The DL7-35 is a pretty exact copy of the Mullard 5-20 but with additional output taps and associated feedback. They use HS 2% resistors there.

Interestingly the Mullard circuit does not seem to strictly follow the RIAA time constants unless I am missing something. I think I might have to try and LTSpice it so I can understand what is going on.

For some reason Beam Echo have substantially increased the input impedance from the Mullard design. The input impedance is also not right as it should be 47K+100pF.

At the moment I am leaning towards a separate phono stage.

[GrimJosef]

Originally Posted by PJL

Interestingly the Mullard circuit does not seem to strictly follow the RIAA time constants unless I am missing something. I think I might have to try and LTSpice it so I can understand what is going on.

If the inter-electrode capacitances are modelled properly in your EF86 element then that should take care of the Miller effect.

At the moment I am leaning towards a separate phono stage.

One of the big differences between vintage and modern magnetic phono stages is the noise level. Solid state can be essentially silent of course. But even with valves the use of DC heaters and decently quietened HT supplies can help a fair bit. You'll always be stuck with some level of inherent hiss if you use a pentode though. It may well be below the level of surface noise on all but the very best (and least played) vinyl mind !

Cheers,

GJ

[G8HQP Dave]

Just checking that you are aware that the connection between circuit CR time constants and the actual RIAA corner frequencies is somewhat non-intuitive, as the two CR networks interact. There is a famous paper by Stanley Lipschitz which covers this. He does all the common RIAA networks. However, he also points out that many people get it wrong.

[Radio Wrangler]

These EF86 circuits are all shunt feedback types, and with the rather finite gain of a single EF86, there will be a lot of interaction between the load impedance presented to the cartridge and the RIAA shaping. It would be VERY difficult to present the specified impedance to the V15 and at the same time provide the correct RIAA time constants. This is one of the arrangements that did not fare well in the Lipschitz paper.

Remember also that the load Z to be presented to the cartridge is nothing like the source Z presented by the cartridge, and that the equivalent noise resistance of the EF86 will be appreciably higher than the cartridge source impedance.

The only way of getting the noise down in a valve stage would be to use a fairly high gm low-noise triode.

David

[kalee20]

I designed a circuit, see here, one of the things I admit isn't nice is the first stage cathode is AF 'hot', but I've not found this a problem, even with AC on the heaters.

It does load the cartridge with the standard 47kΩ

[PJL]

I notice that the triode designs above both avoid the issue by using a passive approach. The phonodude uses mu-followers to lower the output impedance although I am not convinced it is necessary.

The SP21 design does have the benefit of -ve feedback which should reduce distortion but it seems given all the unknowns that I am only likely to be able to determine if the CR network has the correct values by measurement.

Looking at the circuit I assume PU1 is MM and PU2 is crystal/ceramic. I can't see why I can't put a 47K resistor from the input to earth for my matching and a suitably large capacitor to the current PU1 input circuitry for a bit of protection.

[PJL]

I wired it up without the modification and it sounds reasonably balanced to me. Gain is about right. There is a fair bit of hiss if you turn the volume up full.

To be honest I was expecting the gain to be way too low! I will do the mod and dig out the scope and do a graph when I get a chance.

[kalee20]

Originally Posted by PJL

I notice that the triode designs above both avoid the issue by using a passive approach

No - assuming you looked at the link I posted - my design uses negative current feedback in the first stage operating into the virtual earth grid of the second stage grid (to give a V to I converter) and negative voltage feedback around the second stage, to give you a voltage output overall.

It's a double triode design.

[PJL]

Hi Kalee20, it was the two prior to your post!

I have now run it through LTSpice and it gives a pretty good result. Reducing R3 to 0.8M to lower the gain at 1K improved it to bring it within 1dB with some roll-off from about 60Hz. I am not 100% confident in the EF86 simulation. Next problem will be component tolerances.

[PJL]

The roll-off turns out to be the screen decoupling. Slightly low LF gain is down to insufficient open loop gain. At the moment I haven't modelled the 2nd stage which is needed to get the anode load for the 1st stage correct. I wonder how good the LTSpice model is?

[PJL]

Added the 2nd stage and as suspected the loading is a (theoretical) problem and reduces the available LF gain further.

The Mullard design looks better as it drives another EF86 without -ve feedback and therefore has much less anode load. It also uses lower values for the feedback loop but this is not all good as even with my 47K, the input impedance presented to the cartridge will vary significantly with frequency.

I am sure I could fiddle with values to reduce the load but not sure that I should completely mess up the original Beam Echo design.

[PJL]

Below is the LTSpice results. I have tried a few different models but they are all pretty close.

The SP21 rolls-off at LF as a) the screen decoupling capacitor is too low b) The EF86 is running out of steam. The stage gain is only just good enough so fiddling is not going to help, but doubling the value of the resistors in the next stage to reduce the anode load almost returns it to the curve. It also hisses nicely even with the 'special' Mullard valves and has an indeterminate input impedance (LTSpice probably knows but I don't know how to extract it).

The Kalee20 come out roughly the same as the SP21 with some roll-off at low frequencies. I am not too sure of the mechanism here but it can be corrected so does not seem to be a gain problem. This has a buffer so the input impedance is a constant.

The Mullard is running at open loop gain at LF so it is showing a bit of a bump in the LF response. Not sure I like that, and the input impedance problems are worse than the SP21.

I haven't drawn the MM/MC phono box as it will overwrite the RIAA curve. It is within 0.24dB (usually much better) 10-20KHz. it uses a servo design which gives a sharp 18dB/octave cut-off at very low frequencies as well as reducing noise. Technically very smart but it uses ic's so obviously rubbish.

[kalee20]

Originally Posted by PJL

The Kalee20 come out roughly the same as the SP21 with some roll-off at low frequencies. I am not too sure of the mechanism here but it can be corrected so does not seem to be a gain problem. This has a buffer so the input impedance is a constant.

There is the obvious roll-off at the input (C1 / R2),this has corner frequency 16Hz.

There are also a few other LF roll-offs: C8 in the cathode of the second stage; and C2 between stages. C2 is deliberately large valued, plus it's driven by an imperfect current source so it shouldn't matter much, but whatever, the roll-off can't be any lower than R4 / C2 ie 8.8Hz - which isn't significantly lower than the input roll-off.

C1 and R2 can be omitted completely - I put them in as a crude rumble filter. C2 can be upped in value if desired.

Thanks very much for the analysis - very interesting!