• WARNING: Tube/Valve amplifiers use potentially LETHAL HIGH VOLTAGES.
    Building, troubleshooting and testing of these amplifiers should only be
    performed by someone who is thoroughly familiar with
    the safety precautions around high voltages.

LCR phono stage with cascode input - any obvious improvements?

Hi!

Top of my head, here are the things you could try to improve on the design:

- go for a shunt cascode instead of a regular cascode (see Rod Coleman posts on this forum about it). Would greatly improve the PSRR. Albeit it would probably require recomputing the LCR values.

Yes, a shunt cascode first stage would improve the power supply isolation, both from conducted and radiated noise. The input and output are both referred to ground, which helps, and no part of the signal current runs through the power supply capacitor, so any old electrolytic can be used without penalty.

More importantly, the Shunt Cascode's output is a current, so you can terminate this with a (example: 6.2K) resistor to ground, and the driving impedance to the EQ becomes 6.2KΩ regardless of the ri of the E88CC, with its static and dynamic variations.
 
Hi Rod,

I really appreciate the tremendous information you posted on the folded cascode, it really inspired me (I'm working on a RIAA amp design based on it, which I'll hopefully post here in a couple of weeks).

I'd like however to point out that though the output impedance is going to stay constant (determined by RL), the gain of the cascode being gm x RL, it is going to vary substancially over the lifespan of the tube. To me, this is the biggest drawback of the cascode.

Cheers,
M.
 
Hi M,

Glad the Shunt Cascode has got the design ideas working!

Yes, ordinary (series) cascode stages can present some variations of drive impedance (that's what I meant about static and dynamic variation) but the Shunt cascode drive impedance avoids the problem of ageing, and variations due to operating conditions.
 
To see some of the improvements that Shunt Cascode offers, here is a worked-out LTSPICE version.

The Upstairs ECC88 of the original (series) cascode is removed, and used for the output stage.
The Shunt Cascode has the 6.2KΩ output resistor that provides a very stable driving-impedance for the LCR network. The second stage loses the poor-performing cathode-follower, and uses the standard DN2540 cascode CCS. Taking the source output from this drops the small-signal output impedance from 100Ω (cathode follower) to about 10Ω.

Now, we have a RIAA amplifier with 46dB gain (1kHz, inc. EQ), very low distortion, and low output impedance.

Chokes are no longer required for the power supply, and the supply voltage can be reduced to 125V, with regulation of the supply absolutely not needed.

The two bottles are reduced to one ECC88 per channel, and heater biasing problems vanish. Also, both halves of the ECC88 are run at 90V 11.5-12.5mA - almost exactly per data sheet nominal, often a good choice in itself.

The first stage power supply cap, the two cathode bypass caps, and the extra 2nd stage coupling cap are now removed, or out of the signal path. Yes, four caps down, per channel!

If my Shunt Cascode Explorer is used for the first stage, the whole circuit can be made very neatly, and at low cost.

It is worth mentioning that the stable drive-impedance of the Shunt Cascode can be used for standard RC EQ, and this makes a superb improvement compared to most other solutions. It can also have higher gain that with 6.2K LCR, if desired. In this case, spending a little of the money on some top-flight caps (Charcroft Micas do it for me) can make a very enjoyable RIAA, that measures well.
 

Attachments

  • riaashcasc-LCRs.png
    riaashcasc-LCRs.png
    32.8 KB · Views: 485
  • RIAA.zip
    13.4 KB · Views: 81
Interesting. To put it in my context, I already have a 250V supply (a WD PSU3 on steroids) with the rated maximum of the two Hammond 157Gs which feed this and two 82s in the pre being 40mA. It also already has the lifted heater supply, but that can be modified fairly easily.

What might be interesting is trying different valves in the first stage with lower current requirements (if any are suitable from a noise perspective), and seeing what the overall gain is.

P.s. I appear to be about 15 mins away from your business address.
 
Rod,
IMHO 10R output impedance is very optimistic estimation for DN2540 CCS. :p
Rather 300-400R (which is very correct value) is acceptable.

Bela, OK - I know you will have measured it properly. Thinking about it again, I did not account for the low voltage VDS in the case of cascode pair.

But 300Ω is too much for my taste, if the load is 10K, and I might prefer a non-cascode for this reason. I will look at it again and find a solution!
 
Interesting. To put it in my context, I already have a 250V supply (a WD PSU3 on steroids) with the rated maximum of the two Hammond 157Gs which feed this and two 82s in the pre being 40mA. It also already has the lifted heater supply, but that can be modified fairly easily.

What might be interesting is trying different valves in the first stage with lower current requirements (if any are suitable from a noise perspective), and seeing what the overall gain is.

P.s. I appear to be about 15 mins away from your business address.

The STC 6BR7 (CV2135 and other CVxxxx variants) is low current and low noise, low microphony purpose built for audio. Ia=2mA, S=1.2mA/V. It is similar, but superior to the EF86. I have used it it shunt cascode for RIAA, and it works quite well, but the E88CC, E180F, PC86, and various Russian valves with anode current 12-30mA, and gm 10-50mA/V give better performance than any low-current triodes, because the gm is greater, and the current in the PNP should be reasonably high for low distortion.

In a series-cascode, the low-current valves will usually need a high anode load, making drive of the EQ section more difficult.
 
Recently finished putting together an LCR phono stage as per the attached schematic, apart from the CCS for the second stage being a 10M45 run from the full B+ rather than reduced amount(haven't got round to sorting out a setting for that on my LTSpice).

The input stage is an ECC88 cascode with the output impedance used to match the LCR network. Given that there is more gain than I need and given that the load resistor needs to stay as it is to keep the frequency response right, are there any obvious changes? The things I am pondering are LED bias of the first stage (it is set up for 2.5V with the resistor as it is) or possibly having only a partial bypass of the cathode resistor (say 220ohm bypassed with 110ohm unbypassed to add a little local feedback) as the schematic appears to give 1.5V out for 5mv/1kHz which is pretty huge. Would these be worth looking at, and is there anything else people would suggest?

As a background, the project was to have a look at what an LCR stage would be if it was made to fit in with the World Designs valve kits.

I'm late posting to this, but I'm not a fan of cascodes, you are just making a pentode of the two triodes, and the input triode has a very low plate load. Why not just use a pentode?

The output is still a plate loaded resistance, which feeds a RIAA filter circuit without a low Z buffer. Filter calc needs defined source and load impedance.

It will work but is it the best option? Old school tube circuits were cheap, as components were expensive. Using an big choke/inductor in the power supply is a waste, these high voltages can easily be filtered/regulated with a HV transistor + LM317 and filtered with massive capacitance, effectively getting rid of all noise out of the power supply.

Some diode rectifiers can generate switching noise, so it may be a good idea of resistive, or capacitive, bypassing of these to reduce the switching noise. Check for jaggies on the scope and/or use your ears. Sharp on/off transitions can excite resonances from the leakage inductance in the AC mains transformer. If you can do an FFT of the noise you may want to try making an LC filter tuned at those frequencies. Not generating the noise is better than filtering it. A typical common mode choke used on 120V AC lines may be useful here. This would also be useful for filtering incoming AC line noise. Switching power supplies are everywhere, so some filtering cannot hurt on the AC input. Some HF noise may couple thru the capacitive primary/secondary leakage. A resistive loaded plate circuit will pass half of the power supply noise to the output, so the power supply is as important as you amplifier circuit itself. I though I was crazy to use W.W. resistors in plate loads until I saw the same in a Jensen microphone tube transformer preamp appnote. Don't be afraid of wire-wound resistors of high value. Inductance problems can be an issue for lower values, so keep an eye at those. W.W. noise is better than metal film.

I would prefer a constant current load for a triode gain stage, easy to do with an LM431 and a HV MOSFET. The gain and temp stability of the LM431 is great for a very constant current. You can extract max gain from a tube at the most linear part of the plate/grid curve, a horizontal line on the chart. Just pick a current on the chart where the tube have the most linear plate current spacing. (Don't bother with pentodes for CCS. ) You will get maximum, and signal independent gain from a triode with a CCS, that is not possible with a resistive load, but if the voltage output is small, you will not really use all the available voltage swing anyway, so a resistive load should be OK for a low level input.

CCS noise may be an issue. I have to make some measurements. CCS will also isolate the gain stage from the power supply.

Using a microphone transformer for a moving coil input is not a bad idea, the voltage gain is noise-less. The signal is only maybe 0.5mV so you do not need a big core. A mike transformer should be OK. HF roll-off is included for free. Bass response may be an issue with some transformers, so keep an eye on that.
 
Part of the reason for a cascode was a thought ofvseeing what could be done with the most common tube types still in current production. As mentioned previously, the output impedance of the cascode is used to match the LCR network.

The current setup is a Cadenza Blue (0.5mV/1kHz) through Sowter 8155s with appropriate loading, so 5mV at the phono stage input. Noise levels are currently unobtrusive and below what a cheap Hantek scope can pick up, using Kiwame resistors throughout.
 
To see some of the improvements that Shunt Cascode offers, here is a worked-out LTSPICE version. <snip>
Nice work!
Why does not the RIAA signal source generate a voltage output?
How do you get one diode to give 2 voltage drops?
What is the point of placing a voltage regulated shunt on the cascode output? Signal passing thru a BJT? I guess I don't get the topology.

Big 18.8H chokes, iron core?
 
Last edited:
Nice work!
Why does not the RIAA signal source generate a voltage output?
How do you get one diode to give 2 voltage drops?
What is the point of placing a voltage regulated shunt on the cascode output? Signal passing thru a BJT? I guess I don't get the topology.

Big 18.8H chokes, iron core?

The inverse RIAA block is configured for small-signal ac analysis. You can right-click on the block and edit the schematic underneath. Add a SINE signal, if you want to run .tran directives.

Diode: make multiple series instances of the same component: CTRL-right-click on the diode. In the field "Value2" you'll see N=2. change this to give other multiples, or delete the field, to revert to one diode.

The PNP transistor functions to fix the anode voltage of the valve, regardless of signal current. So it's actually just functioning as an emitter-follower: passing-through the collector current, while holding the emitter voltage as tightly as can be. In fact, the tighter the hold on this voltage, the more strictly the valve is operated with a vertical load-line, and very tight control limits distortion caused by anode-voltage variation.

Ultimately: a voltage-regulator is absolutely necessary to get the best performance from this circuit. A shunt regulator is helpful especially where an MC cartridge is feeding the system: big LP-surface clicks can introduce wideband and large amplitude pulses that must not derail the amplifier.

With large-signal versions of Shunt Cascode for driver applications, the shunt regulator improves the measured performance, compared with use of series regulators or buffered dividers; & they can also be easily built with low noise.
 
Hi Rod
"Ultimately: a voltage-regulator is absolutely necessary to get the best performance from this circuit. A shunt regulator is helpful especially where an MC cartridge is feeding the system"

Could an 85A2 voltage reference fed by a low noise cascoded CCS, say a 2Sk107bl and a DN2540 be enough?

Silviu
 
This is a combined transistor-tube design so its neither fish nor meat.

It looks very sophisticated but if one hasn't accepted the fact that tubes can do for sound some things that transistors never could, than its all a beginners work.
A nice output trannie would add to the qualities but that would cost more than the whole circuit when its of excellent quality. So this will be just another mediocre sounding preamp design as thousand others. Study the real great designs of the past, nobody has ever done it this way. For a good reason.
 
hello Silviu

The striking voltage of the 85A2 is 115V, so the supply voltage would need to be ~240-250V for a 170V base voltage.

The collector current swing for 10V (zero→peak) out into 10K is 1mA peak, and the base current excursion is probably no more than 10µA, so it would be kind of OK for RIAA, despite high output impedance (~500Ω at 3mA).

I would think that there would still be problems:

- the noise level from cold-cathode tubes is probably (much) too high for RIAA stages;
- RC noise filtering would mean electrolytics in a sensitive position;
- the very high voltage overhead means heat and stress for the upper DN2540 that we would be better without;
- the circuit really needs to be on a properly laid-out (and small-size) PCB anyway. The gain is so high, that the use of wiring (or non-optimal board layout) for the critical sections is likely to result in oscillation.
 
Hi Rod,

Well I wanna order two boards from you but I do not know what to use in that critical place and I don't know if it's gonna be that critical because I intend to use them in a fully balanced manner.
Would fully balanced be a good approach noise wise for the additional +4dB and the total elimination of even harmonics?

Silviu
 
Hello Silviu

The Shunt Cascode explorer kits have a shunt regulator built in, so you do not need to add anything here.

The harmonics are tiny, so there should be no need for balanced - but I may try it sometime, and measure.
Beware that balanced version would add noise from the extra tubes and devices of the extra half!