Yes, it's a digital, line-level signal (input, in blue). The output is yellow. Same in all three pictures.
The 20kHz signal is stronger without a load (no speakers), and it goes away when switching off the amp. It does not change with moving the cables for my probes. It isn't present in the input. The oscilloscope also doesn't have a power supply, just a laptop with USB power. So, logically, I can confidently say it's originating inside the amp, can't I?
Below is an earlier measurement with L and R channels that shows these spikes in both channels.
Switching noise could be, of course.. I recently added another bridge rectifier to the 6.3V AC filament heating circuit, and it's not yet loaded, output wires just dangling. I didn't think it could output 20kHz... but then I come across this thread ... Could that be the source?
The issue with the left channel being a bit stronger was resolved after I calibrated my probes for capacitance.
The 20kHz signal is stronger without a load (no speakers), and it goes away when switching off the amp. It does not change with moving the cables for my probes. It isn't present in the input. The oscilloscope also doesn't have a power supply, just a laptop with USB power. So, logically, I can confidently say it's originating inside the amp, can't I?
Below is an earlier measurement with L and R channels that shows these spikes in both channels.
Switching noise could be, of course.. I recently added another bridge rectifier to the 6.3V AC filament heating circuit, and it's not yet loaded, output wires just dangling. I didn't think it could output 20kHz... but then I come across this thread ... Could that be the source?
The issue with the left channel being a bit stronger was resolved after I calibrated my probes for capacitance.
You should always have the chassis safety grounded (protective earth), however you might want to keep the amp negative away from safety ground to prevent ground loops. To keep it safe you could use a 'ground lifter' basically a rectifier bridge and parallel resistor between amp ground and safety ground (chassis) to allow some voltage diff between the grounds, but not enough to make it dangerous. You could also use some small caps from input ground to chassis for RF interference.
I did some experiments on this amp today. Wanted to play around with the resistor values for the 'splitter' so I put in some resistors and a trimmer instead of the 30k resistor so I could vary it while measuring distortion.
With equal values (27k), the 2nd and 4th harmonics was significantly higher, while the others seemed to stay more or less the same. With higher resistance the 2nd decreased, and came to a minimum around 31k. I kept increasing it a little bit more, but nothing seemed to happen. 30k was very close to minimum H2.
I also wanted to try different bias to see what happened. The result was not so much. It seemed that H2 increased and H3 decreased a bit and with higher bias (lower resistor values), but not significantly. Since the distortion is H3 dominant, I was hoping to be able to reduce H3. My guess was that the H3 comes from the P/P tubes, but I'm not sure, and I was not able to do much about it.
I also found a problem with one channel on the amp. There were some small 'blips' moving around on the rising flank depending on output level. They are not visible below 4Vpp though (4ohm tap and load). They showed up as pretty nasty high order distortion on the FFT. I found that swapping the output tubes between channels moved this problem. There also seems to be less gain with the problematic tube pair. I ordered 4pcs for 20Eur from Audiophonics (and I had a voucher so I got them almost for free), so I hope I can sort this out by swapping them around and measuring some more.
'blips'
Trimmer+resistors for testing:
With equal values (27k), the 2nd and 4th harmonics was significantly higher, while the others seemed to stay more or less the same. With higher resistance the 2nd decreased, and came to a minimum around 31k. I kept increasing it a little bit more, but nothing seemed to happen. 30k was very close to minimum H2.
I also wanted to try different bias to see what happened. The result was not so much. It seemed that H2 increased and H3 decreased a bit and with higher bias (lower resistor values), but not significantly. Since the distortion is H3 dominant, I was hoping to be able to reduce H3. My guess was that the H3 comes from the P/P tubes, but I'm not sure, and I was not able to do much about it.
I also found a problem with one channel on the amp. There were some small 'blips' moving around on the rising flank depending on output level. They are not visible below 4Vpp though (4ohm tap and load). They showed up as pretty nasty high order distortion on the FFT. I found that swapping the output tubes between channels moved this problem. There also seems to be less gain with the problematic tube pair. I ordered 4pcs for 20Eur from Audiophonics (and I had a voucher so I got them almost for free), so I hope I can sort this out by swapping them around and measuring some more.
'blips'
Trimmer+resistors for testing:
The 27k 30k mismatch has been mentioned previously in the thread, post #16. I remember planning to insert matching resistors, but I think I forgot to do so actually. I did some spice simulations which showed that the 27k 30k should be closer together, post #38.
I had some minor issues with my build that turned out to be the tubes. Got some replacement tubes from ebay and the amp sounds very nice. Also, I replaced some of the hardware, jacks, volume pot etc. right out the bat, as I had better ones.
I had some minor issues with my build that turned out to be the tubes. Got some replacement tubes from ebay and the amp sounds very nice. Also, I replaced some of the hardware, jacks, volume pot etc. right out the bat, as I had better ones.
Do you know why they designed it like this..?If 6F2 is indeed like ECF82 resistor values and operating point iare not right ... not a surprise from a chinese design.
The cathodyne has unequal 27K and 30K resistors ... distortions guaranteed
I'd suggest that few people would actually adjust the ac signal balance to minimise harmonics, similar to post #166, which goes against the comment "distortions guaranteed" made in #16 and perhaps echoed in #170. Imbalance from the PI triode and the output tubes almost guarantees distortions, so the assertion that high tolerance resistors are needed is a bit cute imho.
Sure, many hi-fi amps dictate matched resistors for a cathodyne, such as in the Williamson, and that certainly minimised differences with new tubes in vintage times. Perhaps that was a past era, and a trimpot should nowadays be de-rigour, along with the expectation that people can and will trim for minimum distortion whenever a PI or output stage tube is changed.
Sure, many hi-fi amps dictate matched resistors for a cathodyne, such as in the Williamson, and that certainly minimised differences with new tubes in vintage times. Perhaps that was a past era, and a trimpot should nowadays be de-rigour, along with the expectation that people can and will trim for minimum distortion whenever a PI or output stage tube is changed.
Nice going! If you're curious it shouldn't be that hard to estimate the balance of the primaries of the OT. Connect your signal generator to the primaries of the OTs and apply a couple of ac volts at different frequencies. An ideal OT would show the same output levels on both sides of the secondaries.
I actually played around with this amp again yesterday. I got the tubes I ordered, but two of them had pins missing.. Anyway, I swapped around output tubes comparing harmonic profiles on L&R channels, and was able to get them very close. The 'bumps' on the trace is gone and gain between channels is also better. I actually ended up using one pair of the original tubes, and one pair of the new tubes, so a mix of 1old + 1new per channel. It seemed the new tubes could go a little bit higher in output before clipping.
I have reported the problem with missing pins to Audiophonics, and have good experiences with them from before, so hopefully I will get a working pair instead.
I also put a 0.1ohm 5W resistor on the filament supply, so now I can run it without the bucking transformer. The resistor gets pretty hot though..
I have reported the problem with missing pins to Audiophonics, and have good experiences with them from before, so hopefully I will get a working pair instead.
I also put a 0.1ohm 5W resistor on the filament supply, so now I can run it without the bucking transformer. The resistor gets pretty hot though..