2A3 Monoblocks

[ed]

exactly why I was asking about how you'd arrived at the load figure. I see all of JB's examples have 1m but I'm sure it's explained in the text......

more reading.........

[Cressy Snr]

I've just found another bit of info from the Hawthorne audio site, by Eddie Vaughn.

Very interesting stuff, particularly the bit I highlighted that gives good info on why the UL cascode is so good at swinging big volts without nasty clipping. The clean, firm sound Vaughn refers to is exactly what I get with the E88CC UL cascode driver stage.

It is also interesting that at the end, he refers to the 6922 UL cascode as being a particularly good driver for the 2A3. I'd go further and say that that particular pairing is a match made in heaven.


"Like all cascodes, the UL cascode performs best with high Gm triodes, which means significant plate current. That means a good slew rate, and when coupled with its output Z of roughly 1*Rp, you can see that it "technically" makes a good driver for RC-coupled power tubes having significant Miller capacitance to contend with. The problem with using a cascode to drive most tubes in SET operation is that a B+ voltage the power tube likes usually isn't satisfactorily high enough for the cascode stage.

The maximum output voltage swing capability is only slightly less than a standard cascode having no NFB, but the lower gain means you'll need a hotter input signal to achieve it. Because of the particular way a cascode operates, the max voltage swing in either direction is well less than the bias voltage multiplied by the net voltage gain at the top triode's plate. For example, let's say you have a 6922 cascode and a 6922 grounded cathode where the cathode bias voltage of both is 2.5V (we'll use 2.5V for ease of calculations).

The grounded cathode stage may have a net voltage gain of around 30, and if properly designed will not cut off before the input signal equals twice the bias voltage. So, with a 5V p-p input signal (twice the bias voltage), you'd get a theoretical maximum negative voltage swing of -75V, derived from a voltage gain of 30 multipled by the 2.5V bias voltage.

OTOH, the cascode's negative half-cycle input signal voltage may only be perhaps 50% (if that) of the bias voltage before the stage reaches cutoff. Even before it does, the swing reaches down into the highly non-linear, curved "knee" at the bottom of the characteristic curves. This is why I'd said in my little paper that a cascode's distortion skyrockets even well before it actually achieves its maximum voltage swing, therefore rendering a certain percentage of its potential unusable. Operating a cascode at high plate currents does help *somewhat* with this, BTW.

This has to do with what's called the input overload voltage, or "Vth." You usually see this term applied much more frequently to MOSFETS than tubes (where it's most often overlooked). Simply put, the Vth is the negative half-cycle input signal voltage where the tube cuts off. If it's less than the bias voltage, the stage cannot attain a plate voltage swing in that direction equal to the bias voltage multiplied by the stage's net voltage gain; it cuts off first.

Gm and Vth are interactively and inseparably locked, so that typically the higher the Gm, the lower the Vth. This limits the maximum potential voltage swing of very high Gm triodes like the 5842 or 6C45 in a grounded cathode stage, where they are extremely linear over a certain range of their voltage swing, but then go to pot all at once. Remember, cascodes operate by utilizing virtually all the bottom triode's Gm, and therefore operate best with high Gm tubes like the 6922. So, like the mega-Gm types 5842 and 6C45, cascodes suffer from low Vth.

This is a major reason why you never see high Gm tubes in a cascaded grounded cathode stage, with one's plate feeding directly into the other's grid. The second tube would be cut off on the negative input signal half-cycle before any real amplification could take place. One of the most common types you see cascaded is the 6SN7, whose low Gm and mu make it a natural for that duty. A 6SN7 grounded cathode stage's Vth may be far in excess of twice the bias voltage, leaving the output voltage swing extremely linear all the way up to its maximum output. A "soft" clipping low in high order harmonics will occur, versus the ultra-high Gm triode's very unpleasant clipping.

The cascode's low input overload voltage characteristic is compounded by the fact that cascodes are normally used with quite high plate resistor values, so you have two mechanisms at work to produce more high order distortions, a characteristic reason for the "harder" sound. But, the UL cascode escapes this by two things unique to it. One, the NFB loop sharply cuts the harmonic distortions. Unlike most PP amps, where you have several reactive elements in the NFB loop and therefore time smears/phase shifts to help create high order harmonics and introduce intermodulations, the UL cascode's NFB loop contains nothing but a single resistor and its leads, and suffers from none of those maladies.

The second saving grace comes is a higher Vth. If you run the current high enough, a 6922 UL cascode's Vth can meet or even actually exceed the bias voltage. And, the very clean, firm sound could even be a plus when operating tubes like the KT88 in strapped-triode SE operation, which have very high 2nd harmonic and therefore a warm and tubey sound. You *can* operate the EL34 at lower current and a different-from-normal loading and get much lower 2nd harmonic, but at the expense of higher 3rd harmonic and up (yuck!).

This is an inescapable fact, that triode-configured 25 and 35 watt power pentodes and beam power tetrodes have around 3 to 4 times the 2nd harmonic of a large DHT like the 300B or 2A3, output watt for output watt. They also have significantly higher 3rd harmonic and up, which for all intents and purposes are almost nonexistent at normal listening levels in DHTs.

The often overlooked reality is that while DHTs like the 300B and my beloved 2A3 are extremely linear, they also have very low mu and require a lot more drive voltage than a triode-strapped EL34, KT66, 807, 6550, KT88, 6CA7, etc. So, less distortion created by the output stage tends to be offset by high distortion created by the preamp stage in supplying the extra drive voltage swing.

What you need is an extremely linear driver stage with good slew rate and low output Z. The example of an EF86 feeding into a 12B4 into a 300B is just such a critter. A very carefully tweaked 6922 UL cascode may also work well here, especially with the 2A3, which is easier to drive than the 300B. And, it would also work for driving a triode-strapped, indirectly heated power tube, at least in theory.

Eddie"


I would urge anyone who has a 2A3/6B4G amplifier to try the E88CC/6922 UL cascode driver circuit for themselves. They will not be disappointed, even with the one I came up with inadvertently.

[Paul Barker]

Steve I can't wait to hear whether you have made a variable anode load and a variable U/L // G3 bias. Because although you have come up with a great sound there is every possibility that you will be able to blend for different tones/clipping/distortion properties to suit your mood/music variables.

[Cressy Snr]

Steve I can't wait to hear whether you have made a variable anode load and a variable U/L // G3 bias. Because although you have come up with a great sound there is every possibility that you will be able to blend for different tones/clipping/distortion properties to suit your mood/music variables.

Well....how can I not try that. and how can I not bring the variable set-up to Owston

[Paul Barker]

My thoughts exactly.

[Cressy Snr]

Now I've got the UL cascode front end I'll have another play with the fixed bias setup for the output stage tomorrow.

Should be interesting.

[Cressy Snr]

Hmm got a bit of a problem here

I have converted one amp to fixed bias and am running at the standard 2A3 op points at around 57mA, 250V, with -44V grid bias.

It works fine as you would expect....that is until I turn the volume up beyond a certain point.

Now this "certain point" is well bloody loud...........but

as we approach this level, nasty distortion sets in, go beyond that and the whole thing collapses, everything goes quiet and distorted.

The amp has to be left 15 minutes before it will play again.

It sounds like catastrophic blocking to me as the amp enters grid current on the positive peaks, shutting the whole amplification process down. However,

I'm not going to jump to any conclusions just yet as it sounds really rather good in fixed bias mode.

On the other hand, it is no good if the whole bloody lot collapses at Owston where we all know that our gear needs to be capable of high volumes under arduous circumstances.

It'll have to be back to cathode bias if I can't solve this.

Sounds fab either way so I'm not about to cry over it. I would like to know what is going on though.

[Laurence]

What about a hybrid half of the bias with auto the other half with fixed. I am fairly certain this was done somewhere on a classic circuit, I thought it was the ST70 but when I just looked it up there is only a 15 ohm resistor on the cathode, which I wouldn't have thought made a difference. But I am certain people used to say it sounded like fixed bias (which at that time was believed to be best sounding bias).

[Mike H]

Any idea what the o/p power is just below "the problem" level?

Does seem like forward grid conduction, charges up the 220nF then makes the fixed bias VERY negative DC, and which also takes an extremely long time to recover!

[Mike H]

Appendix ~ by a very approximated guestimate I am making it about 1.5W r.m.s. where grid peaks at 40V.

I.e. swinging between -84 and -4.

[Cressy Snr]

Any idea what the o/p power is just below "the problem" level?

Does seem like forward grid conduction, charges up the 220nF then makes the fixed bias VERY negative DC, and which also takes an extremely long time to recover!

Thanks Laurence, Mike

@Mike

I have had a meter monitoring things with a music signal going through and
just as you say, beyond a certain level the grid bias goes so negative cuts off the valve.

Working the output power out when it does this, by putting a sine wave in from my sig gen I'm getting 4.2W rms before it breaks down

Looks like a simple case of the 6B4G bottoming out with simply no more left to give. Then the residual charge sits on the coupling cap for 15 mins, meaning the amp won't play until the cap discharges.

Never heard blocking before, but it sure is a nasty phenomenon.

I read somewhere that cathode bias is more forgiving of this sort of thing
but gives slightly less output power, whereas fixed bias gives a bit more power but when it goes, it goes all at once.

It's probably a case of getting it all adjusted right to minimise the chances of this happening; easy enough when you can dial in the HT and the bias level.

Thank goodness for scopes and sig-gens

[Mike H]

I'm very curious why it needs 15 mins to recover, that's a helluva time constant for a cap to discharge

4.2W, that's pretty good tho.

[Cressy Snr]

I'm very curious why it needs 15 mins to recover, that's a helluva time constant for a cap to discharge

4.2W, that's pretty good tho.

Well that was just an arbitrary time I left it to cool down before firing it up again. Sorry I did not make that clear.
I left it about a minute in the first instance and it still refused to play properly, so decided to let it cool for a quarter hour so the actual discharge time was probably a couple of minutes
I'll be a bit more specific in future

[Mike H]

Next thing could be tried then if you got a high impedance DMM, put it on the grid bias and watch what it does.

[Cressy Snr]

Next thing could be tried then if you got a high impedance DMM, put it on the grid bias and watch what it does.

It moves around slightly until we approach the threshold of blocking, then it goes waaay negative and there we are......... shutdown.