You have to be careful here. If you use multiple resistors in series, the TC changes also add up and there's no difference in this respect with a single resistor.
The Bruce Hofer trick was specifically aimed at feedback dividers. For instance, if you have a 10k and 1k resistive feedback divider, and replace the 10k with 10 resistors of 1k, all resistors (assumed all the same resistors so would be identical in TC) will all change value with temperature at the same ratio, keeping the 1:10 ratio the same. But all resistors do change in value.
Jan
The Bruce Hofer trick was specifically aimed at feedback dividers. For instance, if you have a 10k and 1k resistive feedback divider, and replace the 10k with 10 resistors of 1k, all resistors (assumed all the same resistors so would be identical in TC) will all change value with temperature at the same ratio, keeping the 1:10 ratio the same. But all resistors do change in value.
Jan
Right!
And the entire point is to be aware of what parts affect performance more than others. Always attack your largest problems first. The TC of dividers is something well known in the T&M industry and directly applicable to audio (among other things).
One important takeaway missed by most is that there is no point in making one aspect of a design "perfect" while other problems will limit performance past a certain point. That's called engineering. It isn't about saving money, it is about not wasting money. There is a big difference here. Making intelligent decisions.
And the entire point is to be aware of what parts affect performance more than others. Always attack your largest problems first. The TC of dividers is something well known in the T&M industry and directly applicable to audio (among other things).
One important takeaway missed by most is that there is no point in making one aspect of a design "perfect" while other problems will limit performance past a certain point. That's called engineering. It isn't about saving money, it is about not wasting money. There is a big difference here. Making intelligent decisions.
High voltage probes are not simple. The best example is the Tek P6015 of P6013. They went through a lot of careful tweaks to get the performance. The P6013 is the most accessable since it did not use Freon or Silicone as an insulator. The compenation box shows a technique to extend the correction. The P6013 is only good to 12KV to 100KHz. The later and much more expensive P6015 is good to 24 KV or more.
If 1500V is OK then a Tek P6009 is a good choice. And large enough to keep fingers away from the HV.
For an audio analyzer I suspect making a 1 meg unity gain buffer would be a great idea. It would allow the use of scope probes and protect the input from an occassional oops. Use something like the OPA1642.
The manual for the P6013 is too big to attach so here is a download link: https://www.manualslib.com/manual/1412903/Tektronix-P6013.html#manual
If 1500V is OK then a Tek P6009 is a good choice. And large enough to keep fingers away from the HV.
For an audio analyzer I suspect making a 1 meg unity gain buffer would be a great idea. It would allow the use of scope probes and protect the input from an occassional oops. Use something like the OPA1642.
The manual for the P6013 is too big to attach so here is a download link: https://www.manualslib.com/manual/1412903/Tektronix-P6013.html#manual
Absolutely Demian!
You could only wish it was a simple resistive divider with capacitance compensation. It's important to note the input impedance decreases with rising frequency as well.
A unity gain buffer for high voltage is tricky. If it is just the potential difference, fine. But handling very high amplitude signals, which I think is the case here, that is another kettle of fishies.
You could only wish it was a simple resistive divider with capacitance compensation. It's important to note the input impedance decreases with rising frequency as well.
A unity gain buffer for high voltage is tricky. If it is just the potential difference, fine. But handling very high amplitude signals, which I think is the case here, that is another kettle of fishies.
I am pretty clueless about scope probes. Would such a passive probe allow me to see deeper into the distortion spectra of my electrostatic headphone amplifier using the RTX?
The amp has a (bipolar!) AC output of up to 1600 V peak-peak, and the typical load of an electrostatic headphone is approx. 100 pF.
I guess these probes are meant for "single ended" operation. Would I need two probes to hook up the bipolar amp to the RTX?
How much distortion would such probes add to the result?
The RTX has a balanced XLR input. Make an adapter, two BNC to XLR to measure differentially. First, you must compensate and select the probes to match each other.
This is a custom "probe' setup, so you'll have to characterize the DC accuracy and AC performance using one phase of your signal. After that you can measure your signal differentially. You'll need 100:1 probes rated for probably 3KV or better. 2 KV probes would probably do it, but as you get closer to the rated maximum on a probe, they can go a bit "funny". You need to measure at various voltages as you go up to catch this.
This is a custom "probe' setup, so you'll have to characterize the DC accuracy and AC performance using one phase of your signal. After that you can measure your signal differentially. You'll need 100:1 probes rated for probably 3KV or better. 2 KV probes would probably do it, but as you get closer to the rated maximum on a probe, they can go a bit "funny". You need to measure at various voltages as you go up to catch this.
Ok, so I need two Tek P6015 or P6013 probes for a bipolar measurement. How much residual distortion would you expect?
Oh man!
What a question!
What I would suggest is to measure a high power amplifier for higher voltage with the RTX directly, then through your probes - each one. Once that is done, try to characterise each probe at increasing voltages from known lower voltage in step up to attempt to get an idea of voltage dependence. You may find distortion reaches some inflection point at a high voltage, you just never know.
I know that certifying high voltage probes was interesting. They might be linear upto a point, then depart from a linear relationship. Some of these were designed for 30 KV + and you could see them departing as low as 1 KV.
What a question!
What I would suggest is to measure a high power amplifier for higher voltage with the RTX directly, then through your probes - each one. Once that is done, try to characterise each probe at increasing voltages from known lower voltage in step up to attempt to get an idea of voltage dependence. You may find distortion reaches some inflection point at a high voltage, you just never know.
I know that certifying high voltage probes was interesting. They might be linear upto a point, then depart from a linear relationship. Some of these were designed for 30 KV + and you could see them departing as low as 1 KV.
I am simply trying to understand if it would be worth to give it a try. Getting a pair of these TEK probes would mean a bit of money, after all.
Also, the probes are of limited use if I can't trust them at higher voltage levels (where I can't compare the results with a direct measurement without the probes).
True. But that's only part of the equation. Recall from Hofer's presentation (AES, 2015) that the thermally induced harmonic distortion is proportional to the product of the power dissipated in the resistor, the thermal resistance from the resistor to ambient, and the TCR. Dial any of those knobs to change the distortion. Multiple resistors in series would still offer lower thetaRA, so the distortion would be lower than with a single resistor.
For even better results, use multiple resistors in parallel instead of in series. Just keep in mind that the voltage coefficient of the resistors often depends on the resistor value to some degree. In the rare cases where the voltage coefficient of resistance is specified, it's often only specified for a resistance range and not for resistors outside this range.
Life is like a box of resistors. You never know what you're gonna get.
Tom
Interesting you say this. I have been trying to calculate what the result would be by partly paralleling, partly series-placed, to get to a minimum number of resistors.
IIRC 20k to 1k could be done with16 identical resistors. Saves you 5 resistors ;-)
Jan
IIRC 20k to 1k could be done with16 identical resistors. Saves you 5 resistors ;-)
Jan
Four 5K resistors in series gives 20K
Five more 5K resistors in parallel gives 1K
9 identical resistors
9 < 16
What are the other constraints which render this solution invalid?
Yes you can buy 5.00K 1% resistors, here is an expensive example at DigiKey. 4.99K is a lot cheaper.
Five more 5K resistors in parallel gives 1K
9 identical resistors
9 < 16
What are the other constraints which render this solution invalid?
Yes you can buy 5.00K 1% resistors, here is an expensive example at DigiKey. 4.99K is a lot cheaper.
Yeah, but that only addresses part of the issue. You'll get a lower thermal resistance, which will help some with thermally induced distortion, but you'd get even more bang for your buck with resistors in parallel. Say 5x100 kΩ in parallel to get to 20 kΩ. Sucks with 100x100 kΩ in parallel for 1 kΩ, though. 🙂
Getting the resistance correct is easy. Reducing distortion by throwing more resistors at the problem is a bit trickier.
I'm not sure how critical it is to have all identical resistors, though. As long as their resistances are all within the same order of magnitude. One can make 20 kΩ with, say, 20k || 20k + 20k || 20k or 30k || 30k || 30k + 30k || 30k || 30k. Both 20 kΩ and 30 kΩ are EIA E24 standard values. That would do quite a bit for the thermally induced distortion, but likely nothing for the distortion introduced by the voltage coefficient.
Tom
Hi,
maybe yo want to have a look at the Sapphire or Micsig differential HV-Probes.
I got a Sapphire SI9000 (up to +-1000V/700Vrms) used and a Micsig 20003 (5600Vdiff-max.) new.
Work fine for me.
The Micsig is a well done clone of the renowned Sapphire probes and very affordable too.
jauu
Calvin
maybe yo want to have a look at the Sapphire or Micsig differential HV-Probes.
I got a Sapphire SI9000 (up to +-1000V/700Vrms) used and a Micsig 20003 (5600Vdiff-max.) new.
Work fine for me.
The Micsig is a well done clone of the renowned Sapphire probes and very affordable too.
jauu
Calvin
That's exactly what I did (see post #3251). My results with the Micsig DP10007 were not great.
What levels of residual distortion do you achieve with your SI9000?
You're right Mark, it's 9 resistors, I remembered incorrectly. Travelling with no access to my notebook and I didn't want to do it again.
And 4.99k would be perfectly fine.
Jan
Tom, I believe the trick is to set it up so that the changes maintain the ratio in the divider. In the original setup with say 10 x 1k in series and 1k to ground (feedback divider), all resistors change with temperature and voltage but the 1:10 ratio is maintained.
BTW Nice Purifi buffer!
Jan
Although that assumes all 10 resistors have identical tempco's and voltage related distortion. For that situation, it may be worth doing a quick tempco assessment and perhaps picking a 'middle of the road' resistor for the '1' portion, and to confirm there are no gross outliers, unless the resistors come with a rated max tempco like say 10ppm/C (compared to having no spec or something larger like +/-100ppm/C).
For all intends and purposes. It will not be 100% identical, but it will be compared to different type/value resistors.
Jan