ECC88 Cascode Questions

[Syrinx1]

Hi,

I'm thinking of experimenting with a ECC88 in cascode configuration as an input stage for a guitar amplifier. I was wondering if this would be a good way to reduce noise providing the power supply was adequately filtered? I'm not sure on this as the higher gain should improve SNR, but a more conventional gain stage using an ECC83 would draw less current, so may be quieter.

I'm also unsure about the component values I've calculated, so would appreciate any input or advice that anyone can offer.

I've tried following Merlins quick method for cascode design, http://www.valvewizard.co.uk/cascode.html

Perhaps I've not done this correctly, or the method Merlin is using doesn't suit the ECC88 very well?

I started by selecting an anode resistor by assuming a 100 - 200mv input signal, so thought a gain of ~ 200 would be fine and so calculated the anode resistor as 200/gm (12.5ma) giving ~ 16K.

Actual calculated gain is closer to 167 so I would be expecting no greater than
33Vp-p output which looks to be slightly below the point at which the circuit would clip from subsequent calculations of headroom.

Drawing the anode load line on the ECC88 grid curves for a HT of 250V, I added a vertical load line intersecting the 0v Grid curve just above the anode load line. Merlin indicates that this point, just below the Hip of the 0v Grid curve is the best point at which to bias the lower triode for guitar.

This yields a centre bias point for the lower triode of - 0.85V at 4ma allowing me to calculate a cathode resistor value of 212 ohms.

A 1M grid leak and 10nf capacitor would be used for bias of the upper triode and an 8uf cathode bypass capacitor for the lower triode so that the cut-off frequency is below the time constant formed by the grid leak and consistent with guitar frequencies.

The value of the anode load and cathode resistor seem somewhat lower than those I've seen on other ECC88 cascode circuits, and yield quite a low anode voltage on the grid of the upper triode of 35.5v and the lower triode anode of ~ 40V when being driven.

The upper triode anode voltage would be about 186V.

Thanks,

Jonathan

[orbanp1]

Hi Jonathan,

Some comments on your questions.
To be honest, I do not design, build tube based circuits, just repair such equipment if they come my way.
So consider my notes with this qualification.

I have not gone through in detail of your calculations, just modeled it.
There must be something off in your assumptions - the plate current in the simulation is about 10.1mA, way different from your data!
I did double check the ECC88 tube model, the curves from a simulation are pretty close to the curves in the Philips data sheet!
The plate voltage on the lower triode is about 105V, and 138V on the upper triode, with reference to the ground.
I have not modeled that where the circuit starts clipping, first you would want to have your plate voltages at the right places!

The 10nF and 8uF bypass capacitors give a kink in the low frequency response, a 9dB reduction to about 1kHz. Overall gain is about 42dB.
If that is what you want then it is good, otherwise 100nF and 100uF values would give a "conventional" transfer function, from 19Hz to 7.8MHz (-3db).

The tube models do not have any noise simulation data in them, so a simulation will not give any answer on that.

Another good source of info would be the "Tubes/Valves" sub-forum on the diyAudio forum.

Regards, Peter

[G8HQP Dave]

The cascode is not a low noise circuit, unless you compare it with a pentode. It will have about the same noise as a grounded cathode circuit.

Cascode advantages are high gain and low input capacitance.

Cascode disadvantages start with poor PSRR, so it needs a very clean supply rail. It also has high output impedance so you need to be careful about what follows it. Fortunately, this will be less of a problem with the ECC88 as it is a high current valve so the anode load resistor can be fairly low in value. Cascode also has higher distortion than a normal triode circuit, but this might not matter as an input stage.

[ColinTheAmpMan1]

I think one has to remember that, as far as I am aware, none of the usual manufacturers of valve-based guitar amplifiers used a cascode input stage. This may have been for economy reasons, or because they tried it and it didn't work too well. I find it difficult to believe that such designers as Leo Fender, Ken Bran (Marshall) and Dick Denney (Vox) didn't have a go at trying all sorts of configurations, but the common triode has survived.
None of the later amp designs have ventured far away from the "normal" configurations, either. As our cousins on the other side of the Atlantic say, "go figure".
Colin.

[Herald1360]

Radford used just that in the STA 100. Just copy that if you want to try it.

[Syrinx1]

Hi Peter,

Thanks so much for taking a detailed look at this and going to the effort of running a simulation - I think I must have been having a bad day. I see now that I somehow managed to get the anode load line completely wrong and also neglected the internal resistance of the valve.

Kind regards,

Jonathan

[Syrinx1]

Originally Posted by G8HQP Dave

The cascode is not a low noise circuit, unless you compare it with a pentode. It will have about the same noise as a grounded cathode circuit.

Cascode advantages are high gain and low input capacitance.

Cascode disadvantages start with poor PSRR, so it needs a very clean supply rail. It also has high output impedance so you need to be careful about what follows it. Fortunately, this will be less of a problem with the ECC88 as it is a high current valve so the anode load resistor can be fairly low in value. Cascode also has higher distortion than a normal triode circuit, but this might not matter as an input stage.

I admit that when I read about the cascode being low noise, it was in the context of it being a substitute for pentodes, which I probably should have borne in mind. A single ECC83 stage draws less anode current than a ECC88 in cascode, so flicker noise should be less (but so will be gain). I'm still unsure if after attenuation to comparable gain levels of an un-attenuated ECC83 stage,
the ECC88 in cascode might be "quieter"

Thanks

[Syrinx1]

Originally Posted by ColinTheAmpMan1

I think one has to remember that, as far as I am aware, none of the usual manufacturers of valve-based guitar amplifiers used a cascode input stage. This may have been for economy reasons, or because they tried it and it didn't work too well. I find it difficult to believe that such designers as Leo Fender, Ken Bran (Marshall) and Dick Denney (Vox) didn't have a go at trying all sorts of configurations, but the common triode has survived.
None of the later amp designs have ventured far away from the "normal" configurations, either. As our cousins on the other side of the Atlantic say, "go figure".
Colin.

Hi Colin,

Yes you're almost certainly right about that, but I want to try it just to see - can't help myself.

[Syrinx1]

Originally Posted by Herald1360

Radford used just that in the STA 100. Just copy that if you want to try it.

Will take a look.

Thanks

[Synchrodyne]

Rogers used a cascode input stage (with shunt feedback) for its Master Mk II control unit. In that case the ECC808 valve was used, which I think was an AF type broadly similar to the ECC83, whereas the ECC88 is an RF (VHF) type. I suspect that the rationale for using a cascode was to obtain lower noise than would be the case using a pentode in the same position, without sacrificing gain and without the need to go to DC heating. An alternative approach was to use a triode pair with series feedback, but then the unbypassed cathode of the input triode required that DC heating be used to avoid hum transfer, given the low signal input voltages (down to around 2 mV) that were involved. (Fairly well-known is that Radford found another way to avoid DC heating in its SC22 control unit by using a transistor-triode pair with series feedback.)

Prior to that Rogers had also used the Hedge double cascode phase splitter in its Senior Mk II amplifier, so it seemed to have some affinity for the cascode.


Cheers,

[Diabolical Artificer]

The ECC88 was AFAIK designed for cascode use. Can't speak for it's low noise capabilities but it was used extensively in oscilloscopes.

The Hedge cascode LTP mentioned by Synchrodyne is an interesting circuit - http://www.keith-snook.info/wireless...ode%20pair.pdf, Morgan Jones used it as well in "The Crystal Palace" amplifier to drive two big OP triodes, see Valve Amplifiers - Morgan Jones. There is also a Glass Audio article on replacing the front end etc in a Dynaco ST70 (thanks Terry) . There are also quite a few builds over on DIY audio that covers the cascode, so lots of source material to read up on.

I started experimenting with the cascode and cascode LTP, but didn't get very far, other things needed fixing, but intend to have another look at some point, but to be honest I'd be more than likely to use a pentode.

Andy.

[Radio Wrangler]

The ECC88 is a high current dual triode, Think of it as being like an ECC81, but more so, and less like an ECC83. The anode slope impedance is lower, and this reduces gain, cascoded or not. The high Gm helps the gain somewhat, but not enough to give high Mu. Its special feature is reduced heater-cathode capacitance, needed to make the upper stage work at RF in a cascode amplifier. The benefits of the ECC88 cascode only come in at RF where Miller capacitance starts to dominate.

The low cathode-heater capacitance and high Gm are what won it places in oscilloscopes.

A plain triode will give you noise performance no worse than a cascode.

With transistors, cascoding can offer some value. Miller capacitances are higher and non-linear (think of Ccb as a varactor). Transistor audio amplifiers are usually designed with very strong feedback, and to get flat behaviour across the audio range, stage poles need to be pushed up into the RF region to keep their phase lag low. Cascoding also decreases the consequences of the Early effect. I designed a silly audio amp back in 1980. It used as many transistors as possible, subject to each one having to do something useful. Two voltage gain stages, each one complementary twins of cascoded differential pairs, seasoned with constant current source bias. This is exactly the antithesis of the use-the-minimum-number-of-valves ethos back in the valve era (Tek scopes excepted!). I did it for the hell of it, and because a friend was developing a minimalist amplifier. Good fun! (The offspring of that minimalist amp did go on sale, with a maximalist price, while a bunch of my maximalist amps were made at HP.)

David

[G6Tanuki]

I think in a guitar-amp any benefits of a cascode over normal grounded-cathode will be minimal.

One 'trick' usable with simple grounded-cathode triodes to reduce noise is to wire several of them in parallel. At audio frequencies the noise falls by the reciprocal of the number of triodes - so if one triode on its own generates a noise-level we can count as 1 then 2 wired in parallel will generate 1/2 the noise, three in parallel; 1/3 and so on.

This was, I believe, used in some early tape-recorder replay amplifiers.

[G8HQP Dave]

Two in parallel generate half the noise power, which is 0.7 of the noise voltage.

[Syrinx1]

Thanks everyone. Lots to research and think about.
I can't wait to get something rigged up to test, but unfortunately practical experimentation has to take a back seat at the moment as my wife and I have just adopted two children.
I'm hoping my enthusiasm for tinkering is something that might get our children interested, though in these times of modern convenience I might be fighting a loosing battle.

[Synchrodyne]

Originally Posted by Diabolical Artificer

The ECC88 was AFAIK designed for cascode use.

Yes, that was so. Its primary application was as an RF amplifier in VHF TV tuners. It was a frame-grid successor to the ECC84, with higher slope and a shorter grid base. (The remote cutoff counterpart was the ECC89.)

ECC88.pdf

Notwithstanding its RF application origin, cascode circuits and “cascode” double triode valves found their way into audio circuits quite early on. Audio Engineering magazine for 1954 January included an article “Design of a Professional Tape Recorder” in which it was stated: “The input stage is a newly developed form of the cascode circuit described by Valley and Wallman and shown in Fig. 9. The use of this input stage results in lower equivalent input noise than most microphone preamplifiers commercially available. The production units have shown an input noise figure of the order-of -127 dbm. The newly developed 6BK7 tube is used. The high amplification facto, and low plate impedance of this tube make it well suited for cascode application. The input stage is capable of handling input levels at high as -20 dbm without exceeding 1 per cent total harmonic distortion. The large input signal handling capability and the low noise figure result in an ideal microphone amplifier input stage. Approximately 6 db of inverse voltage feedback is carried around the input stage to increase the level handling capabilities.”

The 6BK7 was one of the very early American double triode valves developed specifically for use as VHF TV tuner RF amplifiers. The first was the RCA 6BQ7, which appeared in 1951 as the result of RCA’s efforts since c.1948 to develop a better valve for this purpose than the triodes and pentodes hitherto used. The 6BQ7 et al were designed for use in series cascode circuits, whereas the original Wallman circuit was of shunt cascode form. I don’t think that these “cascode” valves were designed or specified for AF applications, but at an early stage they were used as head amplifiers for image orthicon TV camera tubes.

The Fairchild 240 Preamplifier of 1954 was an early consumer audio product use of the cascode circuit. Fairchild claimed “highest gain - lowest noise achieved by modifying famous radar and TV cascode input circuit to audio amplification…” It used a 12AX7 audio valve for the purpose.


Audio 195501 p.42 Fairchild 240, 260.pdf

Another couple of Audio magazine articles referred to cascode preamplifiers. In the 1955 October issue was an article “Cascode Preamp Improves Signal-to-Noise Ratio”. In this design a 6BQ7A valve was used for the cascode input stage, which had series equalization feedback from the second stage (via a cathode follower) to its input triode cathode. (It also had DC heating). The 1959 November issue had an article on a “Universal Stereo Control Center” that used a 12AY7 cascode flat input stage with equalization obtained by shunt feedback around the ½ 12AX7 second stage.

Audio Engineering/Audio magazine is available at: https://www.americanradiohistory.com/Audio-Magazine.htm.

The cascode for audio purposes had some supporters then. Evidently opinion was divided as to whether it was better to use a low-noise audio double triode or one of the VHF double triodes, the latter offering lower noise but perhaps higher, or at least more variable microphony. Not that there was a hard and fats division. For example, the 12AT7 was developed by GE primarily for TV and FM front end applications, but saw widespread audio use.


Cheers,

[Diabolical Artificer]

If I've read your post correctly Synchrodyne it contradict's David's assertion that a cascode is no better than a single triode. The paper attached states the cascode has low noise, whether lower than a bog standard common cathode gain stage I don't know.

I presume noise is more of an issue when you have to amplify low signals like from a mic or low OP cartridge, less so if amplifying line level signals. So if noise is an issue in an amplifying stage what is the best approach, valve choice or type of circuit?

A LTP differential amp for instance should provide good CMRR, though the LTP Cascode by L B Hedges isn't configured as a differential amp, does it provide any advantages apart from a lower valve count?

Andy.

[Radio Wrangler]

There are several factors to take into account. Because noise is random and occurs in many places, doing almost any extra activity on a signal tends to increase noise. In any two stage circuit it is extremely improbable that the random fluctuations generated in the second stage will cancel the independently created random fluctuations from the first stage. The only thing which can be added to reduce noise is a feedback loop.

All noisy amplifying devices exhibit differing noisiness depending on the impedance whatever is creating the signal presents to its input. Osing two devices in parallel is an old trick, but it does halve the source impedance at which minimum noise happens. In practice, you're usually stuck with a source impedance, and must choose the appropriate device, or array of devices to suit. In RF work, the optimum source Z for a device is a complex impedance with resistive and reactive parts. You can plot contours of noise figure on a Smith chart.

The tetrode and pentode have AC-grounded grids screening G1 from A. Effectively diverting the current in theMiller capacitance. Good. But G2 has to be at elevated voltage and G2 steals part of the current heading for the anode. This theft is lumpy as electrons are discrete chunks of charge. The noise content of the anode current is increased. It's called 'Partition Noise'

In a triode over triode cascode, the upper grid is at elevated voltage like G2 of the tetrode/pentode but it's not in grid current bias, so no partition, no partition noise. So here is the triode cascode advantage over the multi electrode valves.

Once you really get stuck into low noise work with valves, you're fighting a lost cause. There is rarely an engineering reason for sticking with a valve, so it has to be a matter of personal preference or marketing. (rarely, not never. There are a couple of special cases but they're not in radios or audio) unless you're just researching old techniques for historical reasons.

David

[tri-comp]

Originally Posted by G6Tanuki

One 'trick' usable with simple grounded-cathode triodes to reduce noise is to wire several of them in parallel.

Since the ECC88 is getting quite expensive this may be a costly affair.
If you feel anyway that path is worth investigating you should be aware of the Russian equivalent 6Н23П (Latin: 6N23P).
The cost of this excellent valve is far below that of the ECC88.

Attached is the data for the Russian valve scanned from a book I got in Kiev, Ukraine a month ago.
They have a VERY interesting bookmarket, The 'Petrovka'.

[G8HQP Dave]

I don't think what Synchrodyne quoted disputes what I said. Cascode will have about the same noise as a normal grounded cathode stage, because that it what it is: a grounded cathode stage feeding a grounded grid stage. The grounded grid stage will not add much noise because the current through it is set by the grounded cathode stage.

" The use of this input stage results in lower equivalent input noise than most microphone preamplifiers commercially available." This is too vague to deduce any hard information from, but we can probably assume that "most" other items used a low noise pentode. It must be remembered that a new circuit, or a new application of a circuit, is often accompanied by exaggerated claims for its performance although often couched in slightly vague terms.