Mr Idler's Valve LCR phono mission

[Thermionic Idler]

With the DAC project coming to its conclusion and having fitted my DH3S cartridge from Northwest Analogue, my attention is back to the LCR phonostage, to see if the Millett can be improved upon.

This sort of follows on from my previous thread on the subject, but between the end of that thread and the start of this one I decided to revert back to a design that follows these specs:

• We use my existing Lundahl 1:32 SUTs as they match perfectly with the DH3S and are in line with manufacturer recommendations

• Valves will do all (or at least 90%) of the active amplification

• Don't be afraid to use solid state for unity gain / buffer stages

• I bought the Sowter 1280 LCR units a little while back so I'm committed to making them work now.

Hence the new thread with a more generic title. After some research I discovered a couple of possible alternatives to the D3a which are much more affordable with plenty of availability, it seems.

Some time back I found this blog post where the guy had worked around the 600R impedance problem with a step-down and step-up transformer either side of the LCR network. Having pondered this for some time, I've decided that's too many transformers in the signal path for my liking. However my interest was piqued by his choice of input pentode - he used an E83F, which, it turns out, are plentiful and affordable.

Another input pentode came to my attention in the manual for this very expensive LR stage. It turns out they use the Russian 6J11P which has similar numbers to the E280F - these are currently VERY cheap although I do wonder whether the EM/IA manufacturers buy large batches and select from those - at the price being charged that wouldn't surprise me.

Anyway, I've bought a few of both to try out.

Moving on to how we drive a 600R impedance - I've decided to sit a unity-gain op-amp between the first valve and the LCR network to solve that issue, the coupling cap between the two can then be a small value with the op-amp's output directly coupled to the LCR stage. So as a basic block diagram, something like this:



The signal through the op-amp should be around line-level pre-EQ, which is optimal for minimum THD according to page 1 of this datasheet. Given I already have op-amps in my linestage that are inaudible (to my ears at least) to me this seems the most elegant solution to the driving impedance issue.

Post LCR, as you can see from the diagram, I don't know yet. Currently under consideration:

• Plain ole' triode stage followed by either Allen Wright's SLCF circuit or a simple op-amp buffer

• Aikido circuit maybe?

Some may see the presence of op-amps in the circuit as a compromise, but I've been all around the houses with this and it seems to me that any solution is going to involve some sort of compromise - you just need to pick one and run with it. I'm perfectly happy with the transparency of the OPA* series chips in the two linestages I've built, I see no reason not to use them here where they resolve a significant sticking point.

[Thermionic Idler]

I think I'm going to go with Aikido for stage 2 - it's an off-the-shelf solution that should work perfectly for this application.

The E34F in the first stage, according to the Direct Heating Triode blog, gave a gain of about 160, or 44dB. We know we lose about 21dB in the passive EQ, and a standard Aikido configured with ECC88 / 6DJ8 valves (of which I have a decent stash already) gives about 23dB gain with less than 300R output impedance (fine for the short distance to the linestage). A bit similar to Andrew's / Simon's C3g phono recently detailed on these pages.

44 - 21 + 23 = 46dB gain, about +6dB over the 40dB I get from the Millett stage, which is a good thing as I could probably use a little more headroom in the system.

[Thermionic Idler]

Mr Idler changes his mind again.

So today I was re-reading the Linear Audio article about Erno Borbely's 'Proteus' JFET current-input stage again after driving myself mad trying to figure out the least-worst compromises for the pre-RIAA stage, and I think that with a 3 ohm cartridge, this idea is really worth exploring further. I cannot fault the SUT's I'm using currently but if there is an extra layer of transparency to be had...

I noticed early on in the article that they presented two proposals for the first stage (pre-RIAA) - one inverting, one non-inverting. They decided to go with the latter for their complete design, and that meant that an inverting op-amp was required to apply negative feedback to reduce the input / output impedances.

Apparently this was because their editor wasn't keen on the idea of users needing to reverse their cartridge terminals, but my immediate question was why can't stage 2 (post-RIAA) be an inverting amplifier as well, and then phase would be correct at the output?

Which got me thinking - the requirement for post RIAA is simple - anything that gives 44dB of quiet, flat gain, effectively a microphone amplifier, it doesn't have to be solid state as well. The Glassware (Broskie) Aikido cascode + cathode follower circuit, configured with ECC88's, fits that requirement like a glove.

Not only that, it also inverts phase, which means I can actually use the inverting Proteus stage at the front and apply whatever local feedback is needed with a straight piece of wire instead of an opamp.

So, what we have here is a sort of JFET / valve hybrid that brings together a current sensing MC input, LCR EQ and valve final stage. I don't think those technologies have ever been combined in a phono stage before, so this will be interesting:

[Proteus inverting current sense front-end] ----> [LCR] ----> [Aikido cascode].

[Thermionic Idler]

Argh! Rant alert.

So I'm re-reading (again) the Linear Audio Proteus article about the JFET current input front end (which I sadly can't share verbatim because it's paid-for content). And once again I am frustrated (admittedly it's partly down to the fact I've not been able to get the spice model to work).

The guy has picked grounded gate JFETS to do the current amplification - i.e. for the first stage that the MC cartridge 'sees'. Because of course he has. To get the input impedance down low enough, he's paralleling 4 pairs of 2SK170/2SJ74, and they all have to be matched to 0.5mA of the Idss because of course they fucking do .

It's all very well pontificating about this in 2013 which was still well after Toshiba stopped production of the devices in question. It's now 2023 and it seems to me that the only way to achieve this without completely bankrupting yourself buying multiples of the Linear Systems reissues that cost over £7 a pop (that's if you can even get hold of the bloody P-channels), is to go down the diyaudio store route, where I'm still probably looking at over £200 expenditure for 2 matched octets of the devices required plus shipping, and god help me if I damage one of them in the construction / testing process.

And... breathe.

Thing is... the article explains that grounded base BJT's have a lower impedance if you run them at sufficient current, to the point where you only need one or two pairs to get close to the ideal 'dead short'. Yet there is absolutely no explanation as to why, despite this, he's gone forward with using JFETS in the design. It seems to me that if BJT's are lower impedance, they perfectly match the use case. And far more available, and affordable.

Full disclosure - outside of op-amps, I don't understand solid state circuitry as well as valves. So what am I missing?

Pondering whether to strike out on my own and do this with BJT's instead.

[Nick]

I am only guessing, but I would think it may be they went with jfets (other than there were free Linear Systems giveaways with one issue of Linear Audio) to avoid covering the same ground as the Paradise phono stage.