Could that be from the amp getting stressed by the circuit? Or was the test circuit in place for both tests? The ceramic caps you are using- are they COG type linear caps or the more common X7R? COG caps get larger and are less common in high voltage applications. Some X7R ceramic caps are really bad, changing cap 50% over voltage. Mica caps are usually good as well. The film cap also could be a contributor. Mylar is not really low distortion, Polypropylene, Polystyrene and Teflon would be the prefered choice or a COG ceramic.
Another possibility- Input nonlinearity of the RTX. (I should test this...) Insert 100K resistors between the source and the input. Any capacitance modulation at the input transistor shows up as distortion. Some opamps and circuits (cascode) are much better for this.
The circuit was in place for both tests.
Not sure what COG or X7R is/means. I used these.
Ok, that's easy to test.
Thanks!
A loop back test from the RTX output through the atten to the RTX input, what does that show?
Also Demian's points on cap type are spot on.
Another test would be to connect both of the black wires to the same signal point. In theory you should get zero signal to the RTX but there will be some cm residue.
Also, is there a solid ground connection between the source and the RTX?
Is there an output on RTX of the input signal where you can hang an FFT analyzer on? See what the residual looks like?
Finally, try to make the setup as symmetrical as possible, hum and RF pickup in the parts may be assymetrical and result in hum/noise.
At this level, every molecule counts!
Jan
Also Demian's points on cap type are spot on.
Another test would be to connect both of the black wires to the same signal point. In theory you should get zero signal to the RTX but there will be some cm residue.
Also, is there a solid ground connection between the source and the RTX?
Is there an output on RTX of the input signal where you can hang an FFT analyzer on? See what the residual looks like?
Finally, try to make the setup as symmetrical as possible, hum and RF pickup in the parts may be assymetrical and result in hum/noise.
At this level, every molecule counts!
Jan
Class 1 or COG or NPO are the ceramics that are stable with voltage and frequency and very linear. Class 2 use "doping" to increase the dielectric constant ast the cost of temerature and voltage stability. Class 3 are just awful but cheap and big values. The caps are Class 2 which means NOT COG. They will have some distortion. Linear HV caps are not common. Here is what Mouser has in HV COG/NPO caps https://www.mouser.com/c/passive-co...tage Rating DC|~Dielectric|~Termination Style Its not a lot and they are expensive. I would avoid the multilayer caps until further notice. . .
I found some 220 pF mica caps and tried these, but the result was the same.
Tried 100 kOhm resistors in line, did not help either.
Direct loopback with RTX output at 10 Vrms and input at 100 Vrms setting gives 2nd harmonic at 0.0001%.
With the capacitor voltage divider in place and 1 Vrms input range, 2nd harmonic is 0.0003%.
I can't go to higher loopback test voltage, because the RTX output stops at 10 Vrms.
That would short the amp output. Does not seem right to me.
The chassis of the amp and RTX are both connected to PE at the same mains socket. Adding a dedicated wire between the amp and RTX chassis does not change the measurements.
Not sure, but this sounds like you're asking for a loopback. See above.
Last edited:
That would short the amp output. Does not seem right to me.
Not both amp outputs of course. Just pick one amp (or test generator) output and connect to both black wires. We're testing the atten cm performance.
Jan
Does the RTX have switchable high pass filters?
See if that lowers the indicated THD; you may be meassuring hum.
Jan
See if that lowers the indicated THD; you may be meassuring hum.
Jan
There's no high-pass to speak off (except the DC blocking switch).
There might be some residual hum at 50 Hz and multiples up to a few hundred Hz, but that dies off in the noise floor (see spectra in post #3340).
There might be some residual hum at 50 Hz and multiples up to a few hundred Hz, but that dies off in the noise floor (see spectra in post #3340).
As mbrennwa writes, the RTX6001 does not have switchable high-pass filters. That should be done in the analysis software, if you want to exclude parts of the spectrum.
The input caps are only active if you select the AC input coupling.
In DC mode the only high-pass filter is the filter in the ADC (AK5394A) at 1 Hz. In DC mode, large DC voltages may of course saturate the input amplifier in some cases, before it reaches the ADC.
The 40 pF sounds about right. The manual says 37 pF. I fail to remember if that was measured before or after the 22 pF was added for stability.
The input caps are only active if you select the AC input coupling.
Yes, as described above. You can select to use AC or DC input coupling. The inputs caps are 1 uF on each input, before the 100 kohm of the attenuator.
In DC mode the only high-pass filter is the filter in the ADC (AK5394A) at 1 Hz. In DC mode, large DC voltages may of course saturate the input amplifier in some cases, before it reaches the ADC.
The input amplifier uses low-noise JFET's with cascodes to improve the linearity and reduce the input capacitance.
The output would have to go through the USB connector. So the input signal is only visible in a digital format, after A/D conversion.
Did you fry the input of the RTX6001 or is your experience from using other equipment?
In the 2nd graph I see much higher mains spurs than in the 1st one. Normally THD sums root/square.
Does the THD calculation only look at components above the fundamental, or does it root-sum everything except the fundamental?
If the latter it could explain the higher THD with the atten in place.
Jan
That's because the attenuated signal was scaled back to the 50 VAC at the amp output. The relative contribution of the mains noise is therefore higher.
I am not 100% sure what REW does to determine the THD number, but it does not matter here. The point is that the individual harmonics measured from the attenuator are higher than in the direct measurement.
Jens- It was not the RTX. It was earlier in my career. I have been playing with HV stuff for quite a while- in particular Surge testing of power strips. 6KV at 3KA demands serious respect and will destruy almost any thing in its path if not protected. Scope input are particulary painful to replace. Even scarier are spectrum analyzer inputs (that hurt).
I may set up to emulate this test. I have a low distortion 100V out amp with wide bandwidth. Its single ended but a start. I want to confirm what the conditions to be tested are? To go over 100V rms I'll need to use the CLT-1 which gives a very low distortion 10 KHz up to 1 KV. But is a bear to bring into this kind of testing.
I may set up to emulate this test. I have a low distortion 100V out amp with wide bandwidth. Its single ended but a start. I want to confirm what the conditions to be tested are? To go over 100V rms I'll need to use the CLT-1 which gives a very low distortion 10 KHz up to 1 KV. But is a bear to bring into this kind of testing.
Would be super cool if you could try things on your end! I am not sure what a CLT-1 is or what it does. Anyway, I am looking at an estat headphone amp that can output up to 1600 V (peak-to-peak), with a bipolar output. Full details are here.
I was able to do distortion tests up to (almost) 100 Vrms into a 100 pF dummy load simulating an estat headphone, with the RTX directly connected to the amp output. However, I'd like to test at higher output voltage levels.
No, class 2 use ferroelectric materials (barium titanate and variations) which have very high permittivity but have distortion and poor stability with temperature. Doping is something done in semiconductors.
Sorry for the inappropriate use of the term. My point was that the "ceramics" are not pure ceramic. Regardless not well suited to low distortion or high stability applications.
I hope to set things up early next week. The CLT-1 https://linearaudio.net/sites/linearaudio.net/files/CB-to-ES-V1-ref-2-IMG_0001.pdf is used to spot nonlinearity in passive components like resistors.
I hope to set things up early next week. The CLT-1 https://linearaudio.net/sites/linearaudio.net/files/CB-to-ES-V1-ref-2-IMG_0001.pdf is used to spot nonlinearity in passive components like resistors.
Ah, that is true. Class II ceramics are all based on highly ferro-electric materials which usually means barium titanate and its cousins and mixtures of the same. You can't dope a substance to make it ferro-electric, its the symmetries and nature of the crystal lattice that determines that, but you can dope to vary some of the other properties or tune the dielectric response especially its variation with temperature, or to affect its manufacturability, or to work around a patent you don't want to pay for...
It seems its not widely realized how different ferro-electrics are: non-ferro-electric ceramics have dielectric constants around 3 to 12 or so, ferro-electric can be 1000 to 10000 or so. Very very different.
Some definitions would call ice a ceramic, in which case its type-I with a dielectric constant around 3. But liquid water is (depending on frequency) upto 80, and as badly behaved as type-II.