GoodSoundClub - Romy the Cat's Audio Site

In the Forum: Melquiades Amplifier
In the Thread: Single-stage Melquiades vs. DHT amps

Post Subject: How many stages and why... and who makes that 45 or 2A3Posted by floobydust on: 1/20/2009

First, Adrian,

Thanks for your feedback on the newer designed 2A3 tubes... very interesting. I've looked at all 3 of these brands before but have not as of yet purchased any (I have a large collection of 45 and 2A3 of various NOS US brands). As much as I wanted to like the EML solid-plate 45 it fell short of my expectations. The inherent mechanical resonance issues and the fact that I need to alter some circuitry to use them. Even then, the sound was not as clear as my favorite NOS samples, possibly due to the mechanical resonance issues and the coloration (i.e., distortion) it adds. I've also concluded that you can't damp the glass envelope in cases such as this, the problem is the filament wires resonating and damping the glass won't do squat in this case. Still, you've given me cause to re-investigate the KR and EML 2A3, so thanks for that.

Second, Romy,

Thanks for your response also. Referring to DHT filaments... I feel that the simplicity of a basic AC heater is best provided you can balance the residual hum to inaudible levels. I don't do the same as all others however... the following are differences:

1- A very accurate filament drive using two accurate windings to form the two 1.25 volt halfs (Hashimoto power transformer)
2- A very high-quality, accurate and inherently noiseless cathode resistor (Caddock MP820)
3- A high-quality film capacitor for bypass where the value is calculated to tailored low-frequency response
4- The split balance technique described above which does result in improved hum balance with good tubes.

These all add up to a very quiet amplifier as noted in the specifications. The added complexity of a DC filament supply or a high-frequency supply will probably not make any measurable difference in signal-to-noise and thus should not have any positive effect sonically. Having a high-frequency oscillator in the middle of the amp also requires additional design choices to contain it and prevent it from being imposed on the rest of the circuit. Whenever possible, I opt for simplicity provided it can meet the design criteria.

The main component that limits low-frequency response in any SET is the OPT. For these initial designs, I used the smallest Hashimoto SE OPTs... their H-203S for the 2A3 and their H-507S for the 45 which are rated at 7-watts at 50Hz. Despite their small size they are excellent in quality, both mechanical construction, shielding and sound quality. Tailoring the low-frequency response for any amplifier should be a design factor so you don't over-run the OPT. Such is the case in my design for these. I purposely wanted to roll them off on the low end so you don't have issues with core saturation and/or field collapse within reasonable listening levels.

Some thoughts on direct-coupling.... for all of the pluses people rave about regarding sonics and such and the removal of the alleged "evil" signal coupling capacitor there are also some negative points which you have hinted at. First, DC coupling does present some additional challenges in the design and maintaining proper DC biasing of the cascaded stages. In many cases, additional circuitry and components are used to compensate for this, many with adjustable resistances to lock it into spec. So that raises one initial question... does the addition of several resistance components where they are of different construction (wire-wound pots, cermet-film pots, wirewound resistors, metal-film, etc.) and/or regulator circuits to maintain DC biasing compensate for the removal of one single coupling capacitor? Which is worse, the cause or the cure? Second, as you noted there is something different in the sound at times. One possible cause is the fact that ANY DC offset and/or low-frequency underlying signal (like record warp) can push the circuit out of it's linear operating region causing high distortions at these extremes and also push the OPT into core saturation and field collapse. This is yet another reason to tailor the low-frequency response of low output amplifiers so they are not pushed out of their linear operating region from signal anomalies.

Multiple stages versus one. Well, again, a choice that has to be made somwhere in the signal chain from start to finish. Let's start at the output stage... using my little design above, and the 45 triode. In the operating point I use, grid bias is approximately -60volts. Simple math shows that you need to be able to drive the grid at +/- 60 volts or 120 volts peak-to-peak for maximum power without driving the tube into A2 on positive swings or cutoff on negative swings. Due to the multiplying effect of input capacitance, you also need a suitable driver for the output stage. If you go thru the required math and use 20KHz and 60 volts swing you can calculate the required slew rate and minimum driver current to ensure you drive the grid properly... there's no free lunch here. The signal level and drive current must be met or you won't have a clean output stage. In a single stage design, you are completely reliant on the driving source. How many preamps can provide 120 volts peak-to-peak under these conditions? In such a design, the overall sound quality now becomes completely reliant to the what's driving it. Granted, if you design the entire system and take all of these points into consideration for design and implementation than you can manage it. If you plan on using other components... you've lost control.

As for the input/driver stage I eventually settled on, it has some excellent operating attributes. First, the driving stage (one triode of the 5814A) has ~2.8ma of current and can easily drive 150 volts peak-to-peak cleanly into the grid, meaning it can drive the 45 or 2A3 fully and at wide bandwidth as the response is flat past 40KHz. Yes, you use more drive currrent, but it is somewhat akin to killing a fly... you can use a fly-swatter or a bazooka... end result is still the same... dead fly. You can also look at it's square-wave response and note that it's crisp and clean which supports my calculations on drive current versus required slew rate and voltage swing. The input stage is the other half of the 5814A. The output of this stage is DC coupled to the driver... but, the operating points of each stage are selected to ensure it does not push the driver out of it's linear region. The input stage is biased at ~1.2ma of current and can handle a input voltage of 8 volts peak-to-peak without being driven into A2 or cutoff. Of course, this would result in pushing the driver stage well beyond it's maximum output level, but the design goal is to ensure the input stage has no possibility of being driven out of it's linear region or placing an odd load on whatever drives it, as class A2 would result is increasing current demands from the input. Another part of the design is the input bypass capacitor being chosen for a specific roll-off point, again to help reduce large low-frequency input offsets like record warp or rumble. The driver stage bypass capacitor is actually calculated for 1Hz or less as I don't want it adding anything like phase shift but simply following it's input signal. And lastly, the coupling capacitor to the output stage (it's the only one in the design) is also chosen for a specific roll-off point. In all, there are 3 time constants and they are not stacked but each is calculated for a predictable response on low-frequency. The end result was as required, 1dB down at 25Hz, which is 1-octave below the spec'd power level of the OPT.

There is also another point to consider here. As the deisgn has no feedback, having two cascaded stages which essentially match each, albeit out of phase, some of the inherent non-linear characteristics will be cancelled out. The proof is in the pudding as they say.... the specs of the amplifier validate the design choices and operating points chosen... or, you can't argue with success. I would have to admit that the only downside is the simple fact that the circuit has a fair amount of gain and can easily be driven beyond full power by any line-level source. Of course, this is something that needs to be factored into the overall system design. If this is done properly, the concept of 1, 2 or 3 stages becomes a moot point. The gain has to come from somewhere. One concept was to use low-mu tubes in each stage. This provides lower gain and lower noise in each stage and when measured results in lower distortion from input to output thru the 3 stages.

So, what am I driving with these little guys... it varies. I do have quite a few speakers at home. I have Quad ESL-63 US Monitors ,but they are far too low on sensitivty for any appreciable SPLs.... but they are gems in their own right with enough power. For single drivers I have a pair of Fostex F120A in 10-liter enclosures I built which are also on the low side, but a few dB more sensitive than the Quads. However, the Feastrex D5nf setup in solid red oak BVR enclosures does have sensitiviy around 95dB. The amplifiers are okay provided you don't push too loud. Granted, having 100dB+ speakers would be a huge plus, but I can monitor output levels with a Mytek DDD-603 and ensure the amps can never be driven into clipping (with digitial sources only). They get loud enough for some very critical listening but of course not loud enough for a symphonic orchestra level. I'm not entirely done on speaker choices yet, but have additional higher powered SETs on the design board. Overall I would certainly recommend the smaller 2-watt 45 amp for driving a high-sensitivity tweeter in a multi-amp system. It's overall linearity is excellent.

Finally, any design has to be engineered so it can be assembled and this creates yet another challenge... the physical layout of the chassis and components and the selection of each component and preparation of each. For this design, I used Caddock TF020 resistors for all circuitry on the input/driver. Caddock MK132 resistors were used for padding the hum balance pot and a Caddock MP820 for the cathode bias resistor. In the power supply Mills resistors were used, the MRA5 where non-critical (non-signal path) sections were needed and a MR-200 foil resistor for voltage drop and supply decoupling for the input stage. I opted for all Axon capacitors.... their 630V High-Volt films in the power supply, their 250V True-Cap for cathode bypasses and their 630V tin-foil and film for the single coupling cap. All signal path wire was pure stranded silver in teflon and a single isloated ground in 12ga pure solid silver. For part preparation, I actually take each component and polish the leads to a smooth shine and then "spin" a skin-tight teflon sheath on each one. This does two things: first, it remove the oxidation, crud and contamination from each component lead and two, protects it so it won't oxidize over time. This helps to provide consistent performance over time... talking years or decades, not hours or days.

Regards, KM

Rerurn to Romy the Cat's Site