JimL

kevin gilmore said: got a batch of dn2540 and they have a 20% variance, so much easier to pick the resistor on a test circuit and stuff it in when you build the board. That replicates my findings with DN2540 - it does vary from batch to batch. I'm guessing it varies more than a 10M90S because that was designed as as a current source so they probably keep closer tolerances for that purpose whereas the DN2540 was designed as more general purpose. I recall posting on this but I can't find it at the moment. Agree that using a test jig is the way to go easiest way to do that is use a transistor socket, solder a 100 ohm resistor to the gate terminal of the socket, connect positive voltage to the drain terminal, a variable pot to the source terminal, then a 10 ohm resistor to the gate resistor and the other terminal of the pot (to measure the current) and the neg voltage to the other end of the 10 ohm resistor. It's easier to see with a drawing, but basically you measure the voltage across the 10 ohm resistor while adjusting the variable pot for the desired current (e.g. 10 mA corresponds to 100 mV). Then measure the resistance of the pot to get the desired resistance for the drain resistor - I would use about 2/3rd of the measured resistance and use the trimmer pot to get the correct resistance in-circuit. Remember that in-circuit there is only a few volts across DN2540 so you don't want to crank the voltage too high when measuring.

Has anyone had a chance to compare the SiC KGSSHV with either a Blue Hawaii or DIY T2? If it is comparable but significantly easier to build and using less energy (less heat) that would be something!

So, for the current loads, 250 ohms for the fixed resistor may be too high to get to 10 mA as sorenb noted. Using a lower value such as 100 ohms or 150 ohms should do the trick.

Very interesting news! Back to the Grace arm on Pretentious Food's Rega 3, that looks like a G707, which was a low mass fixed headshell arm that Linn used to put on their LP12 turntables before they developed their own Ittok arm. It's a very good arm, I used one on a modified AR-XA turntable, then on a modern AR turntable which I eventually sold to my sister. As far as I know she still has it and it's still making music.

So, re SRPP and stat headphones: As both John Broskie at TubeCAD and Merlin Blencowe in the AudioXpress articles referenced above have pointed out, SRPP as a push-pull design can be optimized only for a specific fixed resistor load. An electrostatic headphone is, of course, anything but that. However, there is a work-around of sorts. That is, if we use a load resistor that is significantly lower than the lowest impedance of the stat headphone, say 40 kilohms, the combined load is mostly the load resistor over the audible spectrum, so the SRPP just needs to be designed to be optimum for that load resistor. Of course this "solution" does nothing to address the other problems of SRPP for this application, such as low voltage swing, the need for separate filament windings for each top tube, and the fact that you are literally designing an amplifier to drive the load resistors rather than the headphones. Don't misunderstand me, if this circuit is like an ugly man with a hunchback and a hairy mole, I'm saying..... we can remove the mole. Or as Joe E. Brown says to Jack Lemmon at the end of "Some Like it Hot.", "Well, nobody's perfect."

You can adjust the cascode without hooking up the high voltage power supplies, by using a 15-20 volt supply (2 nine volt batteries in series will do). Connect the positive terminal of the battery to the positive PS terminal on the board and the negative battery terminal to the end of the 100 ohm resistor closest to the 100 ohm trim pot. Then put a meter across the 250 ohm resistor and adjust the trim pot until you get 2.5 volts, which corresponds to 10 mA for the current source. If you can't get to 2.5 volts, decrease the fixed resistor to 200 ohms, and adjust the trim pot until you get 2.0 volts across the 200 ohm resistor. If the voltage is always higher than 2.5 volts for the 250 ohm resistor, replace it with a resistor around 300 ohms and adjust the trim pot to correspond to 10 mA.

ervstil said: "...i cant stop thinking that his cascode CCS make sense for the KGST as well..." Actually, the latest version of the KGST board has the cascode CCS.

Insanity said: "Ok I did the following test. When I removed the input wires from both boards, without grounding the inputs, the sizzling channel started sizzling well noticeable, whereas the other channel remained almost silent. Interestingly, when I grounded the inputs from the sizzling channel it was almost quiet, with the sizzling being barely noticeable to unnoticeable. This seems to rule out the elma as a source of the problem. Does this sound more like a problem of the first amplifier stage or another stage? What parts would you start working at? lsk389 and mpsw06? below it? I don't have a scope..." I think the most likely thing is that you have a noisy or leaky input transistor. Here is what I think is happening. Remember that there is a 500k resistor between the input gate and ground. Any current leak into the gate will flow through the resistor to ground, causing noise that will be amplified. When you turn down the volume or short the input to ground that shorts the noise. As you turn up the volume that increases the resistance between the gate and ground which increases the resistor noise. Allen Wright mentioned a similar problem he had with a tube preamp design where the grid current running through the input pot produced noise. The only thing is, if this is the cause, I'm not sure why it is intermittent rather than continuous. However, I would replace the input transistor and see if that solves the problem.

Sure it can be improved. Substitute a state of the art discrete solid state front end for the IC op amps, add current source loads to the output stage - and you end up with the KGST. Or, Substitute a good high gain tube front end ifor the IC op amps, add current source loads and acurrent source cathode sink for the output stage - and you end up with the SRX Plus. Any questions?

DefQon said: "You're a bit far behind wink: " Wow, it's deja vu all over again!

A few more comments on this wonder of German engineering: 1) If I'm reading the Bing translation correctly, with 140 watts in it's producing 1000 V peak-to-peak out, which is about the same as the Stax SRM007tII that draws only 55 watts from the wall - and has a better all discrete transistor front end. 2) The plate resistors are only 26.5k, which is a lower value than almost any other stat headphone amp out there. The designer probably did it to have enough current running through the output tubes, but almost all that current goes to driving the plate resistors to get some voltage swing, which is probably the reason the thing wastes so much power. 3) LF351 op amps - really? Those were designed in the 1990s and have an open loop frequency response that rolls off above about 12 Hz -not kiloHertz, Hertz. There are way better op amps now, although I don't know what the current hot op amp is. 4) Love the ultra-fi pot metal RCA input sockets

Ooh, I like your style!

Yup, that would work, but it's a bit of overkill - not that there's anything wrong with that. Actually -15 volts would work perfectly well since it's supplying the current source for the tail of the input stage and a few volts here or there makes no difference whatsoever.

280v should give around 380-390 volts DC which should be enough depending on how much of a margin you need (e.g. how much does the voltage drop in the summer when the ACs are running full blast everywhere). In terms of current, the SRX only pulls about 40 mA for both channels, but since the KGST or mini KGBH are cap input filters you need some extra because of the high current draw at turn-on until the caps are topped off - I think the KGSSHV thread had a discussion about transformer requirements in terms of current, or you could just use the same transformer as the KGST 350V version as the voltage and current demands are quite similar, except the SRX doesn't need the 15 volt transformer - the -20v can be derived from the -350 volt supply using a 110-120 kilohm dropping resistor.

I don't know about Dr. Gilmore but I have no further plans to improve the SRX Plus circuit - it's good to go, and pongo5 has built one using the board files on Dr. GIlmore's site (its file SRX6), so the circuit board has been tested. I've discussed the reasons for the modifications in the SRX revised thread, and the reasons for using current loads in the output current requirements thread, so the technical discussion is basically complete. The only thing I haven't discussed is the shunt regulator PS - waiting for AudioXpress to get back to me on when they plan to publish my submission, but the KGST PS would work just fine. I also agree with Frank Cooter that the basic SRX is about as simple a high-performance tube circuit as there is - 3 tubes, 3 or 4 capacitors, 13 resistors and 1 trim pot. If you build the basic version point to point, it is easy convert it to a simplified Plus version by substituting current sources (on heatsinks) for the output load resistors, which is probably the biggest bang for the buck improvement.

Kevin Gilmore said: "when did the stax mafia care about cost?" So much for my chances of becoming a member of the stax mafia. Ever since Peter Aczel tried to insult the Boston Audio Society by sneering that they wanted audio nirvana for $79.95, that's been an unofficial motto of mine. Coincidentally or not, the price of an assembled Dynaco MkII, Mk III or PAS2 was - $79.95.

Tubes don't really mind a bit of extra heat, unlike transistors - there's a reason they are called thermionic valves. Cathodes are heated to around 425-600 degrees celsius, so what difference does a few extra degrees in ambient temperature make? If you don't believe me, do a Google search for Audio Research D250 amplifier picture - you'll see rows of 6550 tubes mounted horizontally with one row above the other - and those tubes generate a lot more heat than a couple of 6S4As.

Those values should be fine. In my build I actually made a test jig and measured the closest value to give the desired current, then soldered in the closest 1% value resistor. For the schematic, I took the actual values used and calculated a fixed resistor plus trimmer that would cover the values, plus some extra in case. On the subject of tubes, old tubes that have never been used, aka New Old Stock (NOS) used to be preferred by many tube buffs. Back in the 40s to 60s when tubes were the only, or the predominant, amplification device, they were made in the millions by big industrial firms like GE, RCA, Sylvania, Telefunken, Siemens, Mullard, Amperex, etc. with decades of development behind them, and used techniques that were not always documented. They were reliable and long lasting because of competition between these companies. Industrial tubes were routinely expected to last 5000 to 10,000 hours, and the Telefunken ECC83/12AX7 was legendary for having a 100,000 (that's one hundred thousand) hour life span, one reason that even used Telefunken tubes still sell at high prices. So tube buffs felt that they were not only the best sounding but the most reliable and long lasting. New production tubes generally come from China, Russia and Eastern Europe, often made on equipment sold by western firms that were getting out of the tube business, and for a long time had the reputation of having less reliability, a shorter life span and worse sound. However some of the newer production tubes appear to have good reliability and good sound, in some cases comparable or even preferable to NOS tubes. For example I have seen some favorable reviews of the new Tungsol 6SN7GTB and 12AT7 tubes, which are made by the Reflector plant in Russia. The 6S4A tubes are NOS only as they were primarily designed as TV tubes and of course nobody makes tube TVs any more so there is no demand except by fringe types like us. There are a number of reputable tube dealers on the internet. I have personal experience with Brent Jessee (NOS), Jim McShane (NOS and new production) and Upscale Audio (NOS and new production), but there are others as well. I have also bought tubes off eBay - but I check every tube I buy on a Hickok tube tester to make sure it is OK. The ultimate test, of course, is in circuit but a tube tester at least makes sure it isn't a dud.

Run them all at once and you don't need central heating!

Although, if the tube sections are well balanced, DC balance should put both sections into the same place on the plate curves, which should improve distortion.

"An educated consumer is our best customer."

Let’s finish off by looking at the output stage and how it interacts with the input stage. The original SRX used a 6CG7/6FQ7 dual triode in a simple balanced differential output. This tube has a maximum DC plate voltage rating of 330 volts, maximum plate dissipation of 4 watts per plate but a combined dissipation rating of 5.7 watts for both plates combined, and an amplification factor of 20. In the SRX this was run at a DC plate voltage of about 250 volts and 5 mA/plate for a total dissipation of 2.5 watts, which is quite conservative at 75% of max plate voltage and 44% of maximum plate dissipation. The 6CG7 was designed as a nine-pin equivalent of the 6SN7 (read: cheaper). If we want more power, we can go to the 6SN7’s bigger brothers, the 6SN7GTA and GTB, which are rated at 450 volts maximum DC plate voltage, 1500 volts peak positive plate voltage, and power dissipation rating of 5 watts/plate and 7.5 watts combined. Note the difference between the maximum DC plate voltage, 450 volts, and the maximum peak plate voltage, 1500 volts. Even if we run this tube at the maximum DC plate voltage, and drive it to its theoretical maximum sine wave output of 900 volts peak-to-peak, it is not going to spark over. This is quite unlike solid state, where even a momentary spike above the maximum rated voltage will let out the magic smoke. Substituting a 6SN7GTA will duplicate the gain parameters of the original design, but allow us to increase the power supply voltage and the standing current to 325 - 350 volts and 7-8 mA/plate, while still staying well below the maximum combined power dissipation. With the increased current, the cathode resistor needs to be decreased to about 1 kilohm. And, the RCA data sheet published in 1954 shows static plate curves going up to 650 volts with good linearity, which means that a differential stage can swing 1300 volts peak-to-peak with reasonable distortion. With current loads the Miller capacitance is about 84 pf, and two dual triodes (one per channel) need about 7.5 watts for the heaters. Another possibility is KG’s favorite small power tube, the 6S4A. This has an amplification factor of 16 = 24 dB, but has a significantly lower Miller capacitance, about 41 pf. Heater current is 0.6 amps per tube, or 15 watts total. Because this is a single triode, matched pairs are needed for each channel. Interestingly, even though the 6S4A has a higher maximum DC plate voltage of 550 volts and power dissipation of 8.5 watts, more than the 6SN7GTA, the RCA data sheet only shows 6S4A plate curves up to 470 volts, and TungSol shows plate curves up to 550 volts. What about a triode-connected EL34? With 800 volts maximum plate voltage (although Mullard lists 600 volts max for triode connection), 140 mA and 25 watts maximum plate dissipation it dwarfs the other two. But it has a lower voltage gain of 11 = 21 dB, as well as a higher Miller capacitance around 120 pf, and pulls 1.6 amps heater current, so 6.4 amps total, or about 40 watts just for the heaters. Again, matched pairs are needed for each channel. To simplify the comparison between these output tubes, let us assume +/- 350 volt supplies, 12AT7 input section with 42-44 dB gain, and a theoretical maximum 1400 volt peak-to-peak output. We will use the same slew rate criteria as in the output stage current thread, assuming full power 6 kHz sine wave as a worst case signal, and we will also assume that that the output stage uses current loads to improve function and maximize the voltage gain of the stage. For a 6SN7GTA output stage, the input stage has to deliver: Peak voltage = 35 volts Peak current = 0.22 mA Total open loop gain: 68-70 dB Open loop roll-off frequency: 11 kHz For a 6S4A output stage, the input tube section has to deliver: Peak voltage = 44 volts Peak current = 0.24 mA Total open loop gain: 66-68 dB Open loop roll-off frequency: 22 kHz For an EL34 output stage, the input tube section has to deliver: Peak voltage = 63 volts Peak current = 0.44 mA Total open loop gain: 63-65 dB Open loop roll-off frequency: 8 kHz These calculations include the current used in the input stage plate and output stage grid resistors. Notice that the 6S4A and EL34 tubes need higher voltages from the input stage for the same output because of their lower voltage gains. Also note that the differences in open loop roll-off frequency due to the differences in Miller capacitance between these tubes. Now, remember that one of the limitations of the SRX input section is the puny current, approximately 0.55 mA per side, or 1.1 mA total. Usually, amp designers keep distortion lowest in the input and driver stages, with the maximum distortion occurring in the output stage. John Broskie at TubeCad suggests that a good rule of thumb for low distortion in tube circuits is that the maximum current drive required be 20% of the total current or less. With the 6SN7GTA as the output tube, the maximum current demand on the input stage just meets this criteria, the 6S4A comes very close to meeting it, whereas using the EL34 as the output tube, the maximum current drawn from the input stage reaches 40% of the total input stage current. Next let’s look at the closed loop frequency response. Since the open loop gain rolls off at roughly 6 dB/octave above the frequencies calculated above, the closed loop frequency response will roll off when the closed loop gain equals the open loop gain. The SRX has a closed loop gain of 54 dB, so using 12AT7 input tubes, the -3 dB frequency is roughly as follows: 6SN7GTA: 55 kHz 6S4A: 90 kHz EL34: 22 kHz So the 6SS7GTA and 6S4A seem to be the best choices for output tubes, with low distortion operation of the input stage and good closed loop frequency response. The 6S4A provides 1-2 dB higher power potential and more extended high frequency response, but has double the heater current, and because they are single triodes, they need to be matched pairs for each channel. The 6SN7GTA is the easiest to drive, has significantly lower filament current and, at least in my unselected samples, close matching between tube sections, with plate voltages within 5-10 volts of each other. The measured frequency response of the SRX Plus using 6SN7GTA tubes matches the calculated response, and is nearly identical to the measured response of the original KGSS. One more example: if we use 6SN7s in the input stage and EL34 output tubes, the total open loop gain is only 58-62 dB, so there is only 4-8 dB of feedback available at low frequencies, and the -3 dB point is around 14 kHz. This is essentially the ES-X repair/restoration that spritzer reported on in 2010. Is it any wonder that he noted that “the HF does lack some sparkle and presence…?” The low feedback might also partially explain his other comments that “the midrange doesn't have the projection of the BH and is a bit thicker than it should be. The bass goes deep but is a bit too round and boomy…” His later substitution of 7F7s (6SL7 equivalents) in the input stage was “the biggest” improvement, which is not surprising since it significantly increased the negative feedback margin and bumped the -3 dB point up to around 22 kHz, bringing the circuit closer to its optimum function. We see that both the choice of input tubes and output tubes can significantly affect the function of the circuit as a whole. Sub-optimum tube choices in either stage can significantly worsen overall performance. The reason that these tube choices make such a difference is that the circuit is so economical – it uses just enough stages and just enough tubes to do the job, and no more. But that means that each tube must contribute a high gain. There’s nothing wrong with a 6SN7 as an input stage tube, or an EL34 as an output tube, but they are not the best tubes for this circuit. Thus far I’ve discussed the basic SRX output stage, now let’s move on to the modifications in the SRX Plus. In the output stage current thread, I show the benefits of using current loads instead of plate resistors. The SRX Plus does this, but in addition the cathode resistor is replaced by a current sink. This improves the differential balance just as it does for the input stage. But it has another benefit – the output tubes are now totally isolated from the power supply, and from the other channel. All they see is the input signal, the other differential tube, and the headphone load. When the current loads are high enough impedance they are virtually “invisible” to the output tubes – all the signal current goes to drive the headphones, maximizing their efficiency. This provides the most stable possible environment for the output tubes, and maximizes channel separation because the output tubes in one channel cannot “see” the other channel via the power supply. The result, quite simply, is the best performance the circuit is capable of. Now, the reason for using cascoded current sources in this design was not to have better sounding current sources, whatever “better sounding” means, it was to have non-sounding current sources. They should set the circuit parameters while being sonically inaudible. My argument is, the closer a practical current source approaches the ideal of infinite impedance with no noise, the less it can contribute to the sound. A cascode current source comes respectably close to that goal with a minimum number of parts, and allows the active device to have a voltage swing within 15-20 volts of the B+/- supplies. However, we run into a problem when we have both current loads and current sink, and that is that current can neither be produced nor destroyed. So, to provide a path for unmatched current to drain harmlessly, there is an adjustable resistor string between B+ and the cathode current sink. As far as I know this little bit of circuitry is unique. This resistor string is effectively in parallel with the current sink for AC signals, and while this does compromise the impedance of the current sink, at 220 kilohms, it is still much higher than the original 1.5 kilohm resistor, and in fact, is as high or higher than a simple 10M90S current sink. Also, adjusting this resistance varies the DC offset of the output plates, allowing them to be zeroed. And, as mentioned in a previous post, adjusting the current loads can help balance the voltage between the plates, although, with the 6SN7GTAs I have tried, this isn’t really necessary as the plate voltages have been within 5-10 volts of each other with my fixed current loads. The tubes are run quite conservatively. The 12AT7 input tubes are run at less than 30% of maximum DC voltage and less than 10% maximum power dissipation, so they are barely turned on. The 6SN7GTA output tubes are run at less than 80% of maximum DC voltage and less than 65% of maximum power dissipation, so they are coasting. NOS (new old stock) industrial tubes running under these conditions can be expected to last 5000-10,000 hours. You can further improve tube life by running the filaments at slightly lower than rated voltage, e.g. 6 volts instead of 6.3 volts, or 12 volts instead of 12.6 volts. Now, since we are using both tubes and transistors, is this a hybrid amp? Well, not in the sense of, say, the Blue Hawaii or T2, where both tubes and transistors are in the signal path. In the SRX Plus, the solid state is all current sources which are used to set circuit parameters and maximize the performance of the tubes, which do all the handling of the signal voltage and current. So this is still a tube amplifier, but with solid state support elements. If it has any sonic character it should be tube-like. A couple comments about tubes versus transistors. One of the enduring controversies in audio is whether one sounds better than the other, which I’m going to sidestep. I do think that tubes make a good match for electrostatics because they are high voltage low current devices. But there are practical considerations. First, while tubes physically somewhat fragile, they are electrically more rugged. When I was adjusting my SRX Plus, there were times when one plate on a 6SN7GTA output tube was sitting at +300 volts while the other was around -300 volts, meaning that one cathode to plate voltage was over 600 volts for a minutes at a time with no harm done, despite a maximum DC cathode to plate rating of 450 volts. Do that with a set of transistors rated for 450 volts and you’ll need to replace the remaining fragments. Second, the tube types used in the SRX Plus should not become obsolete in the forseeable future, as has been the case with a number of high voltage transistors. These tube types are used in guitar amps, which have a continuing demand that far exceeds the demand for audiophile tubes. So, is all this worthwhile? Well, the ES-X is the nearly identical circuit except using EL34 output tubes, which are less well suited for this circuit, and here’s what spritzer had to say about that amplifier in 2010: “The T2 and the Blue Hawaii are the best you can get and for good reason. They cost a lot to build (the T2 quite a bit more than any BH) but the money is well spent. Stable PSU and enough current reserve to never leave the headphones wanting. A KGSSHV run at full tilt would also qualify here. One step further down the ladder would be the current version of the ESX with a CCS loaded output stage. The amp with just the old plate resistors is very good but running the tubes properly would make a world of difference.” In my opinion the circuit is better than most. It is simple yet sophisticated, and has the inherent linearity of triode tubes. The dual differential design means it is inherently balanced, and it will stay that way. With its low parts count it is inexpensive and simple enough to be built point-to-point. Given that the output stage of an amplifier is where there is the most potential for distortion, the use of cascoded current loads means it should perform better than designs that use load resistors or inductors such as the Egmont, RSA A10, or Woo WES, or designs that use poorly designed current sources such as the eXstatA or Liquid Lightning (modest, aren’t I?). The SRX Plus lacks the sheer power of a T2, Blue Hawaii or KGSSHV but it comes respectably close – within a couple of dB - for a lot less. If those are Stereophile Class A, then the SRX Plus is Class B, “the next best thing to the very best sound reproduction,” and at a bargain price. So there you have it, a detailed analysis of the SRX Plus, an optimized version of the Stax SRX tube amplifier using solid state support. Any regulated power supply such as the KGST or mini KGBH PS adjusted to 325-350 volts should work fine with this. When the article comes out I can discuss the shunt regulated power supply, but I can tell you it’s a pretty basic supply. Happy building!

Some of us ancient ones recall when driving an Aston Martin was compared to driving a tractor - David Brown, the managing director of Aston and the DB of the DB-4, 5, 6, etc owned a tractor factory.

Personally I prefer triple blind. In medicine, that's where the subject doesn't know what he's getting, the doctor doesn't know what he's giving, and the investigator doesn't know what he's doing.

Very cool! Looks like your build has the single current source rather than the cascode current source, and a mini KGSSHV power supply?