simmconn

Thank you, Dr. Gilmore. Given the background of amir (mid-high level on the corp ladder at M$), I wouldn't expect him to be as hands-on as the lab monkeys who spend most of their days with an AP. In an earlier post on that site I saw him struggling with setting up the AP to measure the deviation from RIAA of a phono pre-amp. Apparently he's never done that before and just like many of us, don't want to read the manual. Everyone has his/her limitations and comfort zones that takes some courage and effort to step out of. To give this guy some credit, he is able to measure the conventional gears in a consistent way, disclose the data and offer some objective comparisons/analysis, at no cost to everyone who is interested. That alone is worth some praise. Besides, some data is better than no data. When it comes to discrete R2R DAC, what intrigues me is that most designs use generic logic chips to drive the resistors. No matter how well the resistors are matched, the driver chips are not. No manufactures would specify or production-test the matching of a 10 cent bus driver, bit-to-bit or part-to-part. Parts coming from the same batch could be from the same wafer and perform somewhat close to each other, but there is no guarantee. The end result is therefore pretty much on a 'best of luck' basis.

While waiting for the GU-50 sockets, I found a pair of reasonably matched EL34s, and decided to wire them in the original GG circuit for comparison. The TTC004Bs are used as drivers. Above the THD+N curve turning point at around 80V (multiply the X axis by 10 for actual voltage), the majority of distortion is 3rd harmonic. So I think the tube matching is not the major contributing factor here. The EL34 is biased at -37V, which means the driver transistors have to work really hard to push the tubes to make the output swing. The Pentode could be a better choice in a simple circuit such as the Grounded Grid. A more complex BH-style final stage would be needed to make the EL34 really shine. The triode-connected EL34 also lost on the high frequency distortion (100Vrms)

Thanks for the info! It is definitely helpful when I experiment with higher supply voltage next. I'll probably mill away some copper and apply some high voltage RTVs.

I stumbled across the Electrical Spacing section of the PCB calculator in KiCAD, which is from IPC2221A, Generic Standard on Printed Board Design. According to IPC2221A, most of our high voltage e-stat amp PCB build will fall into the B4 and A6 case. It looks like we may have a problem with the clearance. The biggest problem is with the IXCP10M90S footprint, where the conductor-conductor spacing is only 0.5mm. The SiC FET footprint is better at 1.7mm. Some minor problems here and there where the spacing between B- and other nets are about 0.5mm. I don't think anyone is conformal-coating the board after assembly. So in order to be fully compliant, especially for the (semi) commercial builders, it would be better to revise the PCB.

For the same phone, the lower the secondary winding impedance tap you connect to, the lower the output and the lower the distortion (assuming class A SE output stage). The max power transfer happens when you have a matched impedance. It’s not recommended to connect to a tap that has higher impedance than the load.

Now we are talk'in 😉. The tube channel has become really close to the SiC FET channel, both at 100V output. The performance is not yet at the Carbon level, but is pretty good for a tube amp. What was the problem with the previous mod? The simplified G2 supply didn't work well. This tube is quite sensitive to Vg2 (which may make it a good candidate for G2-drive applications, given the low Ig2). The previous circuit has Vg1 as part of Vg2, so the Vg1 variations causes Vg2 changes. The solution is to use one 150V zener diode for each tube, put it across G2 and K, and feed <1mA of current from GND using a resistor or a current source. I also added a 10uF capacitor in parallel with the zener. Without it, the output clips at around 300V but the distortion still looks good before the clipping. Now the circuit should deserve an audition. Since the main problem is the knee on the Ia curve and Ig2 curve below Va=100V, with higher B- (500V maybe?) and perhaps higher current, the tube circuit should perform even better. I'll leave that to the next episode. For now I'm waiting for the sockets from China to build the other channel.

The saga of messing with my 2nd Carbon build continues. Since the circuit is very close to the Grounded Grid, I'd like to give it a try. Being hesitant to spend big bucks on a nice quad EL34, I have been on the lookouts for a cheaper substitute. The curve of a pentode looks a lot like that of a SiC FET. I need a pentode with the following properties: The plate curve should have low kinks or no kinks at the low Va range. The lower the 'knee' the better. Low Ig2. Ig2 should be much smaller than Ia (20mA), ideally 1mA or less under the operational Va range such that Ig2 doesn't interfere with cathode drive/ cathode degeneration. The amount of negative bias needed to get 20mA at 400V should be reasonably easy to handle. High rated Va(max) and Pa(max) for using a higher supply voltage and/or idle current in the future. That means I may need to look into transmitting tubes. And the candidate is... (drum roll please) the FU-50/GU-50! The linearity looks pretty good at around Ia=20mA, from Va=100V all the way to 1kV! The bias voltage is between -20V and -25V, right around what Carbon has. The Ig2 is really low and changes very little from Va=100V to 1kV. More importantly, the FU-50/GU-50 are relatively inexpensive and plentiful. A lot of them were made in the USSR and China during the cold war era. I read somewhere that those were designed for the comm gear used in the tanks and had very little success in commercial applications. I paid less than $3 a pop from Ukraine about 15 year ago. The going price for a NOS tube should be close to a SiC FET today. Well, any tube not designed for audio can be cheap these days. However, that wouldn't stop people from chasing after the Telefunken LS50 and the east-Germany SRS-552s, I guess 😉 Adapting those to the Carbon is surprisingly easy. I removed the SiC FETs and the 20k bias resistor, replaced the two 175k resistors with a 100V and a 130V 3W zener diode for G2 supply. The Ig2 is really small and the two tubes can share one set of the zener diodes. They drop 230V from GND and set Vg2 right at 150V, with about 22V left for the PZTA42 and the offset pot. The heaters are powered by a 12.6V filament trans with one side tied to B-. I could have tied the CT but there was very little hum to worry about. Guess what, the GU-50s work right out of the box. I didn't even need to adjust the balance and offset! The measured performance is pretty decent: Although the distortion is low, the FFT does show some higher order 'pentode nastiness'. I guess the reasons being The pentode is not super linear to begin with. The transconductance of the GU50 is about 1/10th of the SiC FET. The PZT42 has to work much harder and the global NFB is less effective. Something else worth looking into I'm not yet able to seriously listen to the sound, because I couldn't find another pair of tube sockets in my stash for the second channel 😂. If you want to know how it sounds, try it! The GU50 with 400V PSU comfortably beats the KGST (below) on the frequency response and the output swing: Next to try is to use the pentodes on the KGST, or should I call it KGSP then?

@johnwmclean I changed the resistors so that the bias voltage across the 20 K resistor is 24.3V, the same level as the original circuit when powered at ±450V. The VGS of the SiC FETs are about 4.0V at idle (17.2mA) and Vce of the PZTA42 at 14.75V. @JoaMat I'm building this amp for the fun of trying different alt parts. As long as both channels have close enough performance, I can confidently move on to try something else. I didn't observe power-on Vce excursions, so I guess medium-voltage medium-power NPNs such as ttc004 or ksc2690 should work fine here.

I was able to bring the other channel to the same THD+N performance. While I was at it, I tried the current production low-noise JFET from Toshiba, the 2SK209. The first 2 from the cut tape matched really well at 5.9mA Idss at Vds=10V. I'm surprised that I was able to do it without using the microscope or magnifying glass (pardon the flux residue though). I guess I no longer need to pay the hefty price to get the LSK389 or the obsolete K170s. Regarding jacking up the bias of the SiC FET, it has its limits. The original bias resistors set the SiC FET G at 450*20/(175+175+20) = 24.3V to B- with a -450V supply. After adding 260k in parallel with the 175k Ohm resistor, I was able to set the G at 27V to B- with a -407V supply. Note that the VGSmax for the G2R1000MT17D and the C2M1000170D are 25V, and is 23V for the MSC750SMA170B. So I'd better dial it back a little. In normal operating conditions, the voltage on the 20k resistor will never apply entirely to the G-S. But it kind of tells me how intricate the original design was, with all corner cases considered.

In a previous post I mentioned the less-than-ideal performance when Carbon is powered by a ±400V supply, and I suspected that the lower Vce on the PZTA42 is the culprit. Now it's been proven. The PZTA42 being a high voltage transistor, has a non-linear region at low Vce, as the slanted curves you can see on the upper left side. With 407V on the negative rail, the transistors on my board works at Vce=8.4V and Ic=20mA, right around the knee. The global negative feedback would have a hard time correcting that non-linearity. It also explains why some people prefer setting the Carbon at a lower current when powered with ±400V supply, as it also improves the linearity of the PZTA42, albeit to a lesser degree. I guess Kevin chose the high voltage PZTA42 to deal with the power-on transients. I have a quick and dirty fix. Just bias the SiC MOSFET a little higher to give the PZTA42 more headroom. The SiC MOSFETs are biased by two 175k and an 20k at the gate. Reducing either 175k or increasing the 20k would do. The goal is to move the PZTA42 operating point to the right, well into the constant-current region (parallel lines). I would use Vce=14 to 15V. Pushing it even higher would increase the power dissipation on the PZTA42, eat into the max output voltage swing and have diminishing return. What I did was to put a 260k resistor in parallel with one of the 175k resistors. YMMV because it has to do with the operating point of the PZTA42 in your circuit, the Vgs(th) of your SiC MOSFET, etc. After the quick fix, one of the channels now measures as good as with the ±450V supply. We can see that the max output voltage is slightly less compared to with ±450V supply. The difference is subtle with the log scale, though. Now I'm continue to work on the other channel and see if I can find something else.

I only changed the LT1021 with the cheaper LT1236-10. The latter has the same performance as the LT1021 except the long-term stability which we don't need. Oh, and I used the DN2540 in TO-92 package for lower cost. Nothing is different from the original circuit electrically. If the output of the GRHV is low, chances are the passing element (SiC MOSFET) is not damaged. A series linear regulator such as the GRHV is probably the easiest circuit to troubleshoot. Jut need some electronics basics and a lot of patience...

Here is the attenuator I used (R3 and R4 are the AP input resistance, and not part of the attenuator). The R and C are rated at 500V or higher. The trimmers are adjusted with AP analog loop-back (repeat the below till both are satisfied) 1) Amplitude vs frequency response as flat as possible up to 100kHz balanced (I was able to reach ±0.02dB). 2) Set the AP output to balanced-grounded, short the AP +/- input to GND, one side at a time. Adjust the trimmers such that the response of the two attenuator arms are the same. I could use the common mode test feature on the AP to do this, but I figured the amount of work would probably be the same.

I'm glad to report that both the GeneSiC G2R1000MT17D and the Microsemi MSC750SMA170B work well on the KGSSHV Carbon. The G2R1000MT17D works on the GRHV, too. The overall performance of the Carbon is stellar. It's virtually distortion-free to about 200Vrms at the output, and maintains very low distortion up till 600V!! You want the ±450V power supply for the Carbon. With ±400V I ended up with something like the following. Still not bad but not as brilliant as the curve above. I figured the PZTA42s are not quite in their linear region because of the reduced Vce. A word of caution is that both SiC MOSFETs are somewhat 'fragile' compared to the Cree/Wolfspeed C2M1000170D, especially the G2R1000MT17D. I killed a few when matching them on my curve tracer. I guess the Left-right switch on my curve tracer doesn't guarantee that S connects first, then G and then D. No more failures after I connected a 10V zener diode between G-S on the test fixture. An expensive lesson learned😅

Just missed a Yahoo Japan auction on Buyee with tens of J74s, K170s, K117s and more, for just 1Yen starting bid (yes, 1 cent USD). I haven’t figured out how to make Buyee reminders sent in regular hours in my time zone. By default the emails arrive when I’m asleep (Pacific time). btw, I found some C3381s in a local store, BL and GR grades, looks authentic. Maybe I can open a group buy sometime.

A couple of years ago I bought some 2SA1486s from Taobao, the Chinese equivalent of eBay, for about 40 cents a piece. They all check out fine on my Tek 577 curve tracer. I got the green ones with formed leads, likely leftovers from some factories' custom orders. The same picture shows up on Aliexpress. Possibly from the same source. The same seller sold me some 2SA1968s for a bit more money. They checked fine too but looked like rescued from dumpsters. There are many sellers selling fakes, but if you know the right package, the right font, and can tell a grounded and then remarked top surface, with a little luck you can still have genuine parts for less.

Typo? Based on your calculations the current should be 0.93mA

Nice! Looks like around 2ma would give the least impact to the circuit due to LED aging 😄. Joking aside, the LED part you used doesn’t seem to have a sharp curve. If I’m not mistaken, the LED in the circuit here is expected to act like a zener diode with low dynamic impedance. Maybe it’s time to pick a different P/N.

The components will age, LEDs in particular, and accelerated under elevated temperatures. I would check the LED’s I/V curve on the curve tracer before popping it on a circuit. They could differ quite a bit in characteristics even if looked similar on the outside. If a circuit only designed less than 1ma for the LED, that’s likely starving it. I’d make sure the operating current is sufficiently past the “knee” area of the curve.

They are the latest Gerber versions but not with JimL's the latest PSU circuit. The PSU Gerber still have the problems I mentioned in my post. It's best to update the layout with the latest PSU or use the more complex GRHVs. Having said that, I still have a few extra boards left over from my build (slightly modified version of the above). PM me if you want to try it out.

That's the III/2 and II/2 spec you haven't looked into. II/III is the Overvoltage category, and 2 is the Degree of pollution. Once you find their definitions you can match them with your application and determine the proper voltage rating that applies.

I had the opportunity to buy the LSK389s in person from the guy who's selling them on ebay. They all tested good on my curve tracer with Idss in the reasonable range for most applications. I've not purchased any from the official channel such as a distributor. Whoever did that can chime in and let us know where the binning code is printed. I suspect Linear Systems had some sort of lab floor sweeping event about 2 years ago. The above is not the only source which suddenly appeared in the local market. Can you imagine someone approaching you with a gallon-sized ziploc bag full of LSK389s for sale? I was able to get a fistful of them, second-hand. Most of them checked out okay, although very few were in the preferable Idss range. They could very well be factory rejects, but they've served me well.

Google “Overvoltage category” and “Degree of pollution”

The Boonton should be measuring THD+n instead of purely THD given its vintage and price range. So if your one half has a noise problem it will also show up in the “Distortion” reading. Might be easier to figure out if your unit had the 400Hz high-pass filter.

The Boonton 1121 should allow you to do fully differential measurement up to 300V. Did you try that? I measured my unit with SYS2722 and the second harmonic was the only one visible that sits comfortably below -100dB. Too bad I didn't save the test result and the unit is too heavy to move around for retest. The only complaint I had was that 1mV wide-band noise. Compared to the Stax amps this is very good already. I don't remember any single-ended test results. Perhaps I did single-ended input but never single-ended output since that would not make sense. By the way my PSU is set at +/-450V, bias set at 17mA and all transistors curve-tracer matched where needed.

Thanks @Blueman2 for the tips. Here is the internal picture (the file attachment allowance of this forum is ridiculously small!). You can see that I had to cut notches on the chassis and modify a few other things to make them fit. Also low-profile HV film caps are used due to the transformer posts protruding from the top cover. The hum levels of the assembled amp (2.5mV to 5mV) are way below the audible threshold. For the hum to be audible, it needs to be somewhere near 50mV. The wide-band noise of this amps is quite low (<1mV). So higher S/N ratio is possible if the hum is under better control. The surplus power trans has two 6.3V heater windings. One of them was for the rectifier and has high working voltage. I used it for the 6SN7s because it is lifted to near the B- level. The other 6.3V heater winding has a lower working voltage rating which is good enough for the small tubes' heaters when lifted to about 65V.