JimL

!.5k for R11 is fine. R21 is the feedback resistor. If you want the front end voltage to be around 300 volts you should decrease R19, which is the drop resistor.

I would start with changing R11, so that the plate voltage is around 240 volts. This will ensure that the driver stage does not clip. If you decrease R11 alone, that will increase the gain, as it is part of the feedback network. Increasing R12 to keep the sum of R11 and R12 the same will maintain the same closed loop gain, but it will only change the plate voltage by the amount of voltage change in that resistor alone. In other words, if you have 1 mA plate current, and you change R12 from 10k to 11.5k, that will only change the plate voltage by 1.5 volts.

Jose, If you are using R11 of 2.0k, that is why your plate voltage is too high. I suggest dropping R11 to 1.5 or 1.6k. Then if you want, you can increase R12 so that the sum of the two is 13.3k. However, R12 shouldn't affect the plate voltage to any significant degree, it is R11 that is the prime determinant. You want to aim for a plate voltage around 240 volts.

So, just a little technical explanation if anyone is interested. In the Megatron design, the driver 12AX7 and EL34 outputs operate within a global feedback loop with the gain set by the ratio of R21 (33) to R11+R12 (23+24). The operating conditions for the 12AX7 are set by the constant current loads Q1 and Q2 and the grid-to-cathode voltage which is set by R11. If you look at the tube characteristics, for a set grid-to-cathode voltage, as the plate current increases the plate voltage will also increase. However, because the plate current runs through R11, this makes the grid-to-cathode voltage more negative, which further increase the plate voltage. The result is that the plate voltage gets closer to B+, which can result in premature clipping of the driver stage. So in order to increase the plate current without increasing the plate voltage, we must decrease the grid-to-cathode voltage. This is done by decreasing R11. Note that this will also increase the overall gain and decrease the closed loop bandwidth, however, since R11 is only 1/3rd the value of R12, a 10% change in R11 will only result in a 2.5% change in gain and bandwidth. If we wish to maintain the overall gain and bandwidth, then we need to increase R12 so that the sum total of R11 and R12 are unchanged, .e.g if we decrease R11 to 2k, R12 should increase to ll.3k.

The CCS should definitely improve the Egmont as congo5 has reported previously, but you are still limited by the basic circuit - yes it's OK if you want to build the simplest, cheapest possible amp, but the basic SRX circuit is a significant step up and not that much more complex or expensive.

Weird. I was under the impression that they were assembled in a clean room, which means masks, gloves, etc. and so no finger marks. Or maybe that's just assembling the raw drivers?

Yeah, I think the Sonoma is more a "lifestyle" item than an audiophile thing. You know, young exec who owns a BMW, a full set of Danish modern furniture, and wants a stylish headphone set to display - sort of like a B&O system.

Sounds like it's even less expensive than the SRX Plus! Which means there are now two good budget amps with different sonic "flavors."

BHSE/007 MkII w/ port mod is somewhat less of a financial jump, and some of us prefer it to the 009.

The Single Power ES-1 for $23,000 USD cracked me up.?

As best I can tell, I believe you are referring to zener diodes Z3 and Z4, which are 47volt rated in the SMR600, and presumably the same in the SRM007. It would be less confusing if you didn't make up your own names for components.

FOTM = Flavor Of The Month, i.e. whatever is the latest internet favorite, generally replaced as soon as something new comes along.

In terms of price, yes. The Pontus (cheaper) and Venus (more expensive) bracket the Yggy in price. The Terminator is almost double the Yggy's price.

Sorry to start a new topic on this, but I thought if I posted it on a previous thread it would get lost. The issue with the original power supply was that it was noisy at times. I had originally measured < 2 mV power supply noise, and 3 - 5 mV noise at the amp outputs. But on re-measuring, the power supply noise was > 5 mV with the amp output noise measuring around 50 mV. In fact, the noise measured before and after the regulator was similar. Thinking the regulator was noisy, it was rebuilt with new devices, without improvement. Since the current source (CCS) in the positive supply line should block noise from the raw positive supply, the likely remaining point of entry was the negative line. Adding a CCS there did the trick - noise dropped ten-fold. However, with a CCS in both lines, the output resistors were unable to keep the B+ and B- voltages balanced – the voltages shifted towards one polarity, until the voltage across its CCS dropped to less than 5 volts, compromising its isolation and resulting in rising noise. So, the single shunt directly connecting B+ and B- had to be discarded in favor of separate regulators for each polarity, to control B+ and B- and maintain isolation. Each regulator is similar to the original circuit, using an SPX431A (the A suffix has a tighter 0.5% tolerance) IC regulator to set the DC voltage, with associated capacitors to stabilize its internal IC, reduce its output noise, and restrict its function to DC control. The shunt MOSFETs counteract AC variations and noise by compare B+ and ground (positive regulator), or ground and B- (negative regulator) via capacitors connected to gate and source. I used 23N80s, but 19NM50s should also work. Positive and negative CCS should be adjusted to pass the same current. The schematic is posted below. It is still simple enough to be built point-to-point. With the revised supply schematic, voltage initially rose above the final value, then settled to steady state in 5-10 seconds, demonstrating the slow response of the 431 to AC perturbations, and was stable to within 0.2 volts (0.03%) thereafter, with essentially no warm-up drift. With a Fluke 189 meter, unweighted noise measured between B+ and B- was generally < 0.5 mV rms, with occasional spikes up to 1 mV rms. Unweighted amplifer output noise was <3 - 5mV rms. It is stable and relatively low noise, but not particularly low impedance. However, due to the constant load of the amplifier, with its differential design and multiple current sources, the latter is less important. Notes: 1) The 10 ohm test point resistors for both CCS should be 0.1% tolerance to allow the current sources to be set as closely as possible to the same current. When set to 0.43 volts (43 mA) across the test resistors, the shunt current will be around 3 mA. 2) The shunt supply will burn about 5-10 watts while the amp is running. However, with solid state rectifiers, the shunt supply will burn 35-40 watts at turn on for about 10-20 seconds, until the tubed amp circuit starts drawing power. It is ESSENTIAL that the shunt MOSFETs have sufficient heatsinking to absorb 35-40 watts at least for a short time, otherwise sooner or later they WILL BLOW UP. 3) The 100k output resistors and 200 ohm shunt resistors at least 1 watt rating, gate resistors ¼ watt, all other resistors ½ watt. I used 0.22 μf/400V Cornell Dubilier polypropylene caps and 22 μf /630V Solen polypropylene caps for my build. 4) The resistor chain values are shown for +/-350 volts nominal. To change to other voltages, you can alter the value of the 1.8 kilohm resistors connected between the SPX431 reference and anode terminals. To calculate the resistor R for voltage V, use: V = 2.5*(248k + R)/R, or R = 248k/[(V/2.5) – 1]

The VE Enterprise E-Lite looks to be a modification of the Stax SRM-T1 but with 6SN7 outputs and constant current output loads. Also reported to have a basic regulated PS using zener diodes and MOSFET follower, which sounds like a simplified KGBH power supply. See discussion here: https://www.head-case.org/forums/topic/13089-enterprise-e-lite-electrostatic-amp/

Apparently the Shangri-La Jr. is no longer on the HFM website. V2, here we come?

Good for you Tyll! Enjoy your richly deserved retirement! Will miss your reviews and commentary, but that's selfish on my part.

So if it sounds like sh*t, the wearer can't complain.

So, this appears to be a variation on his "Zen" amp that was first published in AudioXpress a number of years ago, basically a single MOSFET amplifying device into a current source, so single-ended class A, phase inverting, around 20% efficient, 14 dB gain (5x) and 2 watts our at 1% distortion, so either efficient speakers or headphones, although maybe not enough voltage output for something like a Susvara or HE6.

The original Stax SRX schematic that the ESX was based on actually showed two heater supplies, but Singlepower's Mikhail Rothenberg was too ignorant to realize the reason for that (the input tube cathodes and heaters need to sit at zero or positive voltage whereas the output tube cathodes and heaters need to sit near B-), so he tied them together at ground. The excessive voltage between heater at ground and cathode at roughly -400 volts broke through the insulation layer between heater and cathode damaging the tube resulting in the problems described.

Why am I getting the feeling that the Koss ESP950 is a better built headphone for a less than a tenth of the cost?

This has always confused me because in KG's original KGSS Headwize article he said that he wanted the best amplification for the then new SR-007 but didn't want to pay the money for an SRM-717 - which implies that the 717 preceded the KGSS.

Headamp has them listed at $4999 but Woo Audio has them listed at $3999. Don't know why the difference in price.

Well, they only cost $3,999, not $6000, so "obviously" there had to be some cost-cutting somewhere. Obscene profit is an integral part of the business plan, you know. Hey, at least Sennheiser used honest plastic - remember those bright yellow foam ear pads and black headband on the HD-414s?

Wasn't he raving about these just a month or two ago?