simmconn

Please check the wiring from the PSU board, through the connectors and umbilical cord all the way to the amp board. If a PSU regulator output is shorted to GND or to another power rail due to a wiring mistake, you’ll see the explosion. Check every component in the PSU as some may look okay from the outside but may have been damaged already. I know it could be heart sunken but that’s what you need to do to bring it back to life. Btw I see you didn’t use the ceramic insulators on the PSU transistors. Although it may not be the cause of the failure this time, the ceramic insulators are recommended over the silicone ones.

What’s the inter-winding capacitance look like on the DC-DC modules? That together with 300B’s mu of only 3.9, the circuit would have to work a lot harder to get the output swings the EL34 is capable of.

Some bench notes on the DIY T2 - not a builder, just a tinkerer... T2_AB.pdf

The trimmer should change the input bias current by only a small amount. The real-world devices are not as well-matched as the models in SPICE, which is why a trimmer becomes handy. Also, if you use the global NFB version of the schematic for sim, the feedback resistor will drag the input toward the output (hopefully zero) such that the change of the voltage across the input resistor is less. It would be more pronounced with the non-NFB version. Another way to look at it in sim is to monitor the input bias current directly through the 100 Ohm resistors at the base of the input transistors as the trimmer is adjusted.

I think the 10k trimmer balances the P and N arms of the input stage, hence reduces the input bias current of the amp. Too much input bias current could cause a scratchy volume pot over time. The adjustment is outside of the servo feedback loop therefore needs to be adjusted separately. I would leave the volume pot to max or disconnect the pot, adjust the trimmer to minimize the voltage across the input resistor (100k) or across the volume pot, after the amp is fully warmed up.

The good board C2M measurement doesn't look right. The VG should be higher than VS, and the Gs are tied together so the two VGs should be the same. Did you swap the G and S when filling in the table? If so, chances are you modified the VG resistor divider on the good board but not the bad board.

The equation can be made more accurate by including the voltage drop on the 100 Ohm resistor on the emitter of the PZTA42s: Vce(PZTA42) + 0.1*I(CCS) + V(offset res) + Vgs(SiC) = Vg(SiC)-V(B-) //13.5 + 0.1*17.5 + 6.8 + 3.3 = 25.35 Since you've measured all voltages except the 100 Ohm resistor, it may be the one that's off and you didn't notice. In rare occasions, the DN2540 may be 'walking wounded' with a gate leakage, in which case the voltage drop on the 50 Ohm resistor no longer reflects the actual current from the CCS. Otherwise the voltage on the 100 Ohm resistors should corroborate with the reading from the 50 Ohm resistor in the same arm, assuming the output offset voltage is near zero so the current drain through the feedback resistors is negligible.

It’s the offset pot and the resistor in series, the latter being 180 ohm or 120 ohm.

If the voltage measurements are correct, chances are the long tail resistor plus the offset pot on channel 2 is dropping too much voltage (it should be ~9V). It could be due to incorrect component values, or the idle current from the CCS is adjusted too high.

You can put together a test circuit using a trimmer pot and dial in the target current with say, 50V across the CCS and get the resistor value, something like the post here. It's okay if the actual current is a couple of mA higher than designed. It will turn to heat on the Zener diodes and won't burn a hole on your board.

You can use just about any NPN transistor with sufficient BV(CEO) and BV(CBO) rating and not too much Cob (usually means not a high current part). If the Cob is too small on the replacement part, you may want to increase the value of the compensation cap (5pf). For the 10M90S, rumor says (in its SPICE model notes) it has an internal OPAMP so it would be more complex than just a depletion mode MOSFET. In fact the actual part performs even better than the SPICE model which only has a depletion mode MOSFET and a body diode. I guess that makes the G/A/K notion more appropriate and the manufacturer can continue to charge $4 a pop.😉 You can use IXTP01N100D to replace the 10M90S in most places if you can't wait to get the amp up and running. Some resistor values need to be tuned to get to the current the circuit was originally designed for.

The long tail resistor (offset pot plus the one in series) should drop about 8.xV so Vce(PZTA42)+V(offset res)+Vgs(SiC)=Vg(SiC)-V(B-) //10+8.6+3.2=21.8 For PZTA42, Vce=10V and Ic=20mA is still not out of the woods. Just look at the DC current gain curves in the datasheet below. The non-linear region (due to Early Effect?) is not apparent in the SPICE model from the manufacturer. That's probably why it didn't catch the amp designer's attention. On the other hand, I was probably over worried about exceeding the C2M1000170D's max Vgs rating.

I’d plot the attenuation curves of the 4 individual pots and see if I can mix and match to minimize the error in the rotation range I care the most.

You won’t be able to swap the channel mapping between the amp and the phone, due to the fixed pin out in the phone cable connector. Since both phones have the same problem with the amp, it’s the amp that’s the culprit. It’s unlikely that the problem can be fixed by feeding single-ended input (RCA). I would get the amp serviced.

If the noise changes with the volume knob and appears when you turn on the source (not when you turn on the amp with no input signal), it sounds like your amp’s left channel has a problem and not your headphone. I would stop messing with the new phones and try a different pair of phones on the same amp and see what happens.

Sorry, wrong thread…

Since many of us would never be able to afford one, might as well take a guess what’s in it. A Megatron with SS CCS and 12AU7 cathode followers driving the DHT final stage?

@qqiaothere are a quite few posts in the DIY forum of this site talking about estat amp grounding that you can use for reference. Basically I’ve learned the following: 1. The metal chassis should be connected to the protective ground on the AC power inlet. This is for safety, as well as a good starting point for a hum-free amp. 2. There are two types of ground on the volume pot, the signal ground and the metal frame/shaft ground. I would keep the signal ground running along with the input cables’ signal flow, and connect the metal frame ground to the chassis (usually by tightening the collar nut with a tooth lock washer is sufficient). 3. power-line induced noise can have a few different sources, such as due to different ground AC potentials between the source and the amp; due to the power transformer leakage flux flowing thru a closed signal or ground loop; or in tube amps, the leakage flux directly modulating the signal current in the tube. The common wisdom calls for single-point grounding, either at the power filter cap, or at the audio input jack. Note that the left and right audio cables can also create a ground loop, so it makes sense to run them closely to each other inside the enclosure. 4. Some units provide a “ground lift” feature, which gives the option to connect the electrical ground to the earth ground thru an RC in parallel as opposed to direct connection. I haven’t had personal experience where ground lift is especially useful. Within the amp, everything is relative to the amp ground. The ground only becomes dirty when there is dirty current flowing from one grounding point to another. Your job is to find that current and get it out of your ground.

I wonder what’s in those black boxes marked ‘phase splitter’, 1:1 transformers? I see 4 coax cables going into them.

TL; DR: I’ve wanted to build a Blue Hawaii ever since I read about it on head-wize. I finally got around building a mini BH-BJT packed into a small enclosure like a can of sardines. Many small tweaks were done in attempt to reach the measured performance of the original Blue Hawaii. The details of the journey are recorded here. Many thanks to Dr. Kevin Gilmore for designing this legendary project. Thanks to Kerry, JimL, Joamat and many others for making it such a fun journey. KGBH1.pdf

@dip16ampthe test results are all relative (in dB). Are you sure the 0dB reference is calibrated to 1V rms when you interpret -120dB as -120dBV? Even the most advanced Audio Precision APX555B is “only” spec’d at -117dB of range plus 1uV. Besides, the limited bandwidth of the line matching transformer and the high-ish 10k resistor in combination with the input capacitance of the ADC probably make the test result look quite a bit better than it actually is.

Where are you located? We have forum members around the world, and someone close to you maybe able to help. Although the circuit is not complex, troubleshooting intermittent noise problems often require a lot of patience and decent test equipment.

PBHV2160Z is nice due to its high HFE and low Cob. I'll be using PBHV2160Z and PBHV3160Z in my upcoming GRHV build with off-board e-caps. From simulations using those transistors I learned that C5 needs to be increased to 15pf to 22pf for best phase margin, and 47n is still the preferred value for C4 that provides a low and consistent output impedance across the audio frequency range. Another interesting find is that the circuit needs about 100uF output capacitance to be stable, and increasing the value of that cap has a diminishing return in the performance, unless the cap has really low ESR (such as film caps). In case the cap's ESR is high, some additional compensation can be added to ensure stability (hundreds of pf across R7). The above findings are for the original GRHV with 400V output. YMMV if the output voltage changes.

@JoaMat Did you measure the removed part and see which transistor and which junction failed? If it takes a while to fail, chances are the power dissipation is over spec. Something else could be leaking that causes excessive current through the part. An IR camera would be handy in troubleshooting problems like this. The STN0214 may have a different Cob than the original part KSC5026, and the closed loop gain of the regulator is also different (20x vs 40x), the compensation capacitor values may need to be adjusted to maintain stability.

Yes that’s what I would do.